UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

The climatology and meteorology of windstorms that affect southwest British Columbia, Canada, and associated… Read, Wolf Anthony 2015

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2015_september_read_wolf.pdf [ 67.63MB ]
Metadata
JSON: 24-1.0166485.json
JSON-LD: 24-1.0166485-ld.json
RDF/XML (Pretty): 24-1.0166485-rdf.xml
RDF/JSON: 24-1.0166485-rdf.json
Turtle: 24-1.0166485-turtle.txt
N-Triples: 24-1.0166485-rdf-ntriples.txt
Original Record: 24-1.0166485-source.json
Full Text
24-1.0166485-fulltext.txt
Citation
24-1.0166485.ris

Full Text

THE CLIMATOLOGY AND METEOROLOGY OF WINDSTORMS THAT AFFECT SOUTHWEST BRITISH COLUMBIA, CANADA, AND ASSOCIATED TREE-RELATED DAMAGE TO THE POWER DISTRIBUTION GRID  by  Wolf Anthony Read B.Sc. (Natural Resources), Oregon State University, 2005 A THESIS SUBMITTED AS PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREEE OF  DOCTOR OF PHILOSOPHY  in  The Faculty of Graduate and Postdoctoral Studies (Forestry)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  July 2015 © Wolf Anthony Read, 2015Abstract  ii Abstract  High-wind generating extratropical cyclones routinely strike southwest British Columbia. Improved understanding of these storms could help mitigate damage to electrical and other infrastructure. All independent windstorms from 1964-2012 were determined using the Victoria, Vancouver and Abbotsford hourly surface observations. For all qualifying events 1994-2012, storm tracks and central pressure tendencies were determined using surface maps. Storms were classified by peak wind direction and track location. Detailed synoptic and mesoscale maps were made and analyzed for eight strong windstorms. BC Hydro Corporation distribution-system power outage data from October 2005 to August 2009 was used in an analysis of tree-related line faults within a 50 km radius of Vancouver. For events with peak speed >40 km h-1, maximum wind, gust, wind direction, storm total precipitation and storm duration were used to predict line faults using linear and Poisson regression. Data from four strong windstorms was used in an hour-by-hour analysis of wind speed and line faults. There were two dominant categories of cyclonic windstorm in this region: westerly and southeasterly. Cyclone tracks had much variation. However, compared to westerly storms, southeasters tended to have a stronger northward component to their direction of motion and tended to pass closer to the study region. Most low-pressure centers weakened rapidly as they moved inland and, as a result, peak winds typically occurred at landfall. The exception is for westerly storms where the wind is dependent on an onshore pressure gradient behind the low. There was moderately strong linear relationship between peak two-min wind or five-s gust and the frequency of distribution line faults. Poisson regression models indicate the potential for widespread grid damage for wind speeds approaching the historic 12 Oct 1962 windstorm. On average, southeasters cause approximately 1.5-1.9 times more line faults than westerly windstorms. A few-hour lag in the occurrence of line faults relative to observed wind speed, perhaps due to reporting delays, was observed. This has not been reported before and could confound analyses that use daily data. The modeling done here could be expanded to forecast storm impacts to the power grid using numerical weather prediction model inputs.  Preface  iii Preface  Nearly all the contents of this thesis are the independent and original work of the author, Wolf Read. This includes figures, maps, photos and text. There are two exceptions: i) The use of four water vapor satellite photos from the U.S. National Weather Service; and ii) the use of one WRF-GFS Numerical Weather Prediction map of surface wind speed, valid for 1200 UTC 25 Feb 2012 and from the 0000 UTC 25 Feb 2012 initialization, from the University of Washington Department of Atmospheric Sciences. The satellite photos are public domain images, and permission was obtained from David Ovens and Cliff Mass to use the WRF-GFS graphic. The material in the dissertation remains unpublished at the time of this writing, with the exception of one synoptic map, two mesoscale surface maps and the peak gust map for the 2006 Hanukkah Eve Storm that are included in a paper co-authored with Professor Dorothy Reed of the University of Washington. The paper, titled "The 2006 Hanukkah Eve Storm and associated civil infrastructure damage in the Cascadia Region of the United States and Canada", was presented at the full-paper 12th Americas Conference on Wind Engineering on 19 Jun 2013 and also included in the conference proceedings. The method developed here of creating synoptic charts that integrate 30 kPa, 50 kPa and surface features, used in Chapter 3 of this thesis, was also employed in a published peer reviewed paper: Cheng, T. K., Hill, D. F. and W. Read, 2015: The contributions to storm tides in Pacific Northwest estuaries: Tillamook Bay, Oregon, and the December 2007 storm. Journal of Coastal Research, 31, 723-734.  Wolf Read 25 May 2015  Table of Contents  iv Table of Contents  Abstract ........................................................................................................................................... ii	  Preface ............................................................................................................................................ iii	  Table of Contents ........................................................................................................................... iv	  List of Tables ............................................................................................................................... viii	  List of Figures ................................................................................................................................ xi	  Acknowledgements ..................................................................................................................... xxii	  Dedication .................................................................................................................................. xxiii	  1	   Introduction ............................................................................................................................... 1	  1.1	   Southwest British Columbia and Pacific Storms ............................................................... 1	  1.2	   Windstorm Meteorology and Climatology ........................................................................ 4	  1.2.1	   High Winds ................................................................................................................. 4	  1.2.2	   Windstorms ................................................................................................................. 4	  1.2.3	   Windstorm Climatology .............................................................................................. 5	  1.2.4	   Windstorm Meteorology ............................................................................................. 9	  1.2.5	   Historical Windstorms .............................................................................................. 10	  1.3	   Windthrow ....................................................................................................................... 12	  1.4	   Power Outages ................................................................................................................. 12	  1.5	   Knowledge Gaps .............................................................................................................. 15	  1.6	   Research Objectives ......................................................................................................... 16	  1.6.1	   Windstorm Climatology Objectives ......................................................................... 17	  1.6.2	   Windstorm Meteorology Objectives ......................................................................... 17	  1.6.3	   Wind Impacts to the Power Grid Objectives ............................................................ 18	  1.7	   Approach .......................................................................................................................... 19	  1.8	   Thesis Structure ............................................................................................................... 20	  2	   The Climatology of High-Wind Storms that Affect the Lower Mainland and Greater Victoria, British Columbia, Canada ............................................................................................................. 21	  2.1	   Introduction ...................................................................................................................... 21	  2.2	   Methods............................................................................................................................ 25	  2.2.1	   Windstorm Definition ............................................................................................... 25	  2.2.2	   Anemometer, Methodological and Data Availability Changes ................................ 26	  2.2.3	   Wind Speed Averaging ............................................................................................. 28	  2.2.4	   Pressure Gradients .................................................................................................... 29	  2.2.5	   Storm Track Determination and Climatology ........................................................... 30	  2.2.6	   ETC Central Pressures and Bomb Cyclogenesis ...................................................... 32	  2.2.7	   Mean Return Intervals ............................................................................................... 33	  2.2.8	   Long-Term Pacific Climate Variability .................................................................... 33	  2.3	   Results .............................................................................................................................. 34	  2.3.1	   General Statistics of Southwest BC Windstorms ...................................................... 34	  2.3.2	   Cyclone Statistics ...................................................................................................... 42	  2.3.3	   ETC Tracks for Southwest BC Windstorms ............................................................. 47	  Table of Contents  v 2.3.4	   Windstorm Seasonality ............................................................................................. 56	  2.3.5	   Mean Return Frequency and Long-Term Pacific Climate Variability ..................... 58	  2.4	   Discussion ........................................................................................................................ 63	  2.4.1	   General Statistics of Southwest BC Windstorms ...................................................... 63	  2.4.2	   Cyclone Statistics ...................................................................................................... 68	  2.4.3	   ETC Tracks for Southwest BC Windstorms ............................................................. 70	  2.4.4	   Seasonality ................................................................................................................ 73	  2.4.5	   Return Intervals and Long-Term Climate Variability .............................................. 74	  2.5	   Conclusions and Recommendations for Further Research .............................................. 76	  3	   Detailed Comparative Analysis of Major High-Wind Storms in Southwest British Columbia and Northwest Washington ........................................................................................................... 79	  3.1	   Introduction ...................................................................................................................... 79	  3.2	   Methods............................................................................................................................ 87	  3.2.1	   Study Region ............................................................................................................. 87	  3.2.2	   Synoptic Analyses ..................................................................................................... 88	  3.2.3	   Mesoscale Analyses of Low Tracks ......................................................................... 90	  3.2.4	   Plotting Frontal Systems ........................................................................................... 93	  3.2.5	   The Special Case of Bent-Back Fronts ..................................................................... 94	  3.2.6	   Wind and Gust .......................................................................................................... 97	  3.2.7	   Windstorm Definition ............................................................................................. 100	  3.2.8	   Geostrophic Wind and Three-Point (Two-Dimensional) Pressure Gradients ........ 101	  3.2.9	   Conceptual Modeling .............................................................................................. 104	  3.3	   Results ............................................................................................................................ 104	  3.3.1	   General Windstorm Statistics ................................................................................. 104	  3.3.2	   Synoptic Analysis ................................................................................................... 111	  3.3.3	   Mesoscale Analysis ................................................................................................. 124	  3.3.4	   Comparative Analysis of Surface Response ........................................................... 149	  3.3.5	   Pressure Gradients and Surface Wind Speed .......................................................... 160	  3.4	   Discussion ...................................................................................................................... 173	  3.4.1	   General Windstorm Statistics ................................................................................. 173	  3.4.2	   Synoptic Analysis ................................................................................................... 175	  3.4.3	   Mesoscale Analysis ................................................................................................. 176	  3.4.3.1	   Broad ETC Characteristics, Lee Lows and Surface Wind Response .............. 176	  3.4.3.2	   Identification and Movement of Frontal Systems ............................................ 182	  3.4.4	   Pressure Gradients, Frontal Passages and Surface Wind Speed ............................. 189	  3.4.4.1	   Pressure Profiles and Slopes ............................................................................ 189	  3.4.4.2	   Surface Pressure Gradients, Winds and Frontal Passages ............................... 193	  3.4.5	   Comparative Analysis of Surface Response ........................................................... 198	  3.4.6	   Windstorm Conceptual Models .............................................................................. 200	  3.4.6.1	   Conceptual Model of Synoptic Evolution ........................................................ 202	  3.4.6.2	   Conceptual Model of Surface Wind Response for Southeasterly Windstorms 203	  3.4.6.3	   Conceptual Model of Surface Wind Response for Westerly Windstorms ...... 206	  3.4.6.4	   Conceptual Model for Bent-Back, or Secondary Cold Fronts ......................... 207	  3.4.6.5	   Conceptual Model for Warm Sectors .............................................................. 208	  3.4.7	   Some Forecast Considerations ................................................................................ 208	  3.5	   Conclusions and Recommendations for Further Research ............................................ 209	  Table of Contents  vi 4	   Tree-Related Line Faults on the Power Distribution Grid During Windstorms that Affect Southwest British Columbia, Canada ......................................................................................... 214	  4.1	   Introduction .................................................................................................................... 214	  4.2	   Methods.......................................................................................................................... 220	  4.2.1	   Region of Interest .................................................................................................... 220	  4.2.2	   Wind Data and Discrete Storms .............................................................................. 220	  4.2.3	   Power Outage Data ................................................................................................. 225	  4.2.4	   Modeling ................................................................................................................. 227	  4.2.4.1	   Hour-by-Hour Analysis ................................................................................... 227	  4.2.5	   Peak Wind of the 1962 Columbus Day Storm ........................................................ 228	  4.2.6	   Known Confounders ............................................................................................... 229	  4.3	   Results ............................................................................................................................ 229	  4.3.1	   Basic Analysis of the Eight Meteorological and Power Grid Variables ................. 229	  4.3.1.1	   Meteorological Variables ................................................................................. 229	  4.3.1.2	   Power Grid Variables ....................................................................................... 234	  4.3.1.3	   Pearson Correlations Among the Eight Variables ........................................... 235	  4.3.1.4	   Further Details About the Power Grid Response to the Sample Storms ......... 237	  4.3.2	   The Wind Speed Dose Response of the Power Grid .............................................. 241	  4.3.3	   Hourly Response of the Distribution Grid to Strong and Severe Windstorms ....... 252	  4.4	   Discussion ...................................................................................................................... 258	  4.4.1	   Basic Analysis of the Eight Meteorological and Power Grid Variables ................. 258	  4.4.2	   The Wind Speed Dose Response of the Power Grid Using Independent Storms ... 260	  4.4.2.1	   Southeasters Verses Westerly Winds ............................................................... 266	  4.4.3	   Hourly Response of the Distribution Grid to Strong and Severe Windstorms ....... 267	  4.4.3.1	   Lag Effects ....................................................................................................... 268	  4.4.3.2	   Relationship Between Wind Speed and Line Faults on an Hourly Basis ........ 272	  4.4.3.3	   Wind Direction and Line Fault Frequency ...................................................... 274	  4.5	   Conclusions and Recommendations for Further Research ............................................ 275	  5	   Synthesis, Conclusions and Recommendations .................................................................... 278	  5.1	   Synthesis ........................................................................................................................ 278	  5.1.1	   Windstorms ............................................................................................................. 278	  5.1.2	   Natural Climate Variability (PDO) ......................................................................... 282	  5.1.3	   Climate Change ....................................................................................................... 283	  5.1.4	   Wind-Speed Dose Response of the Power Grid ..................................................... 283	  5.2	   Recommendations .......................................................................................................... 288	  5.2.1	   Windstorm Climatology .......................................................................................... 288	  5.2.2	   Windstorm Meteorology ......................................................................................... 290	  5.2.3	   Windstorm Forecast Considerations ....................................................................... 291	  5.2.4	   Power Grid .............................................................................................................. 292	  Bibliography ............................................................................................................................... 294	  Appendices .................................................................................................................................. 305	  Appendix A: Windstorm Damage Photos ............................................................................... 305	  Appendix B: Southeasterly and Westerly Windstorms at the Vancouver International Airport................................................................................................................................................. 308	  Appendix C: Synoptic Charts For Southwest BC Windstorms .............................................. 313	  Table of Contents  vii Appendix D: Mesoscale Charts for Southwest BC Windstorms ............................................ 334	  Appendix E: Regression Model Fit Statistics for the Affect of Wind on the Frequency of Distribution-Grid Line Faults ................................................................................................. 350	   List of Tables  viii List of Tables  Table 2.1 History of anemometer and observational methodology changes relating to wind at the three study stations (Hare and Thomas 1974, Tuller 1980, EC 2009, MANOBS 2013, MOC 2013, J. F. Fleming pers. comm. 2013). Different anemometers sometimes had overlapping periods of service: During these times, the system being used for the official wind reports is not always clear. .................................................................................................................... 26	  Table 2.2 Weather datasets used in this study. ............................................................................. 28	  Table 2.3 Basic statistics of the 58 high-wind storms that affected CYYJ, CYVR and CYXX during 1994 to 2012. In cases where peak gust is not reported, the value is estimated from peak wind using a 1.3 gust factor. Estimated gust speeds are denoted in italics. ................. 35	  Table 2.4 Wind and pressure gradient data for the 1962 Columbus Day Storm and the most significant windstorms (wind ≥20 m s-1 and/or gust ≥30 m s-1 ) from the 1964 to 1993 U2A era. Gust values in italics are estimated from peak wind via a 1.3 gust factor. .................... 37	  Table 2.5 Pearson correlation coefficients using wind and pressure-related variables for all windstorms lumped together (n=58). .................................................................................... 38	  Table 2.6 Pearson correlation coefficients using wind and pressure-related variables for southeasters (n=21) and westerly windstorms (n=27). ......................................................... 40	  Table 2.7 All ETCs with a ≥6.0 hPa (100) km-1 pressure gradient that did not reach high-wind status for the period January 1994 to June 2008. .................................................................. 41	  Table 2.8 General track and central pressure statistics for the 58 events that produced high winds in the study region. Storm track designation locations are shown in Figure 2.7. ................. 43	  Table 2.9 For given track designations, the frequency of high-windstorms in the study region. Results are for the primary cyclones, excluding secondary developments, and are broken down by peak wind direction. ............................................................................................... 50	  Table 2.10 Mean high-windstorm recurrence intervals for various breakdowns of the 1964 to 2012 dataset. ......................................................................................................................... 58	  Table 2.11 Mean return intervals in years for increasing categories of wind speed and for different wind-direction classes of windstorm, 1964 to 2012. .............................................. 60	  Table 2.12 Total number of windstorms and mean return frequencies for different wind-direction classes largely based on peak wind direction at CYVR for 1964 to 2012. ........................... 61	  Table 2.13 R2 values between two PDO-related indexes and the annual frequency of windstorms per five-y interval using the data for 1964 to 2012. Results shown are for five different starting points. For all cases, n=9. ........................................................................................ 63	  Table 3.1 Peak wind with timing and direction and peak gust at CYVR, CYYJ and CYXX for the selected eight windstorms. ............................................................................................ 106	  Table 3.2a Peak gust direction, magnitude and timing for key areas in and around the region of interest for the first four (arranged by date) of the eight windstorms selected for detailed analysis. Values led by a tilde (~) are the highest available from incomplete datasets. Times followed by an "E" are estimated. ....................................................................................... 107	  List of Tables  ix Table 3.2b Peak gust direction, magnitude and timing for key areas in and around the region of interest for the remaining four windstorms selected for detailed analysis. Values led by a tilde (~) are the highest available from incomplete datasets. Times followed by an "E" are estimated. ............................................................................................................................ 109	  Table 3.3 General track and central pressure statistics for eight windstorms with either north Vancouver Island (NVI) or south Vancouver Island (SVI) tracks. .................................... 111	  Table 3.4 Maximum two-dimensional pressure gradients, with associated pressure slopes, for the eight detailed windstorms. Pressure slope averages in italics were determined from the vector components. Geostrophic wind triangle locations are in Figure 3.4. ....................... 161	  Table 3.5 Coefficient-of-determinations for linear regression between hourly pressure gradient and two-min wind speed for windstorms that landed on northern Vancouver Island (n=36 in all cases, save 46146 where it varied from 23 to 36). Six-station average compares the average wind speed among the six stations for each hour to the hourly pressure gradient. Six-station peak compares the highest wind speed among the six stations for each hour to the hourly pressure gradient. ............................................................................................... 170	  Table 3.6: Summary of wind response for the Lower Mainland and Greater Victoria windstorm conceptual model. NVI refers to north Vancouver Island landfalls, and SVI south Vancouver Island landfalls. These acronyms are followed by the peak wind direction designation for the two landfall classes (e.g. SE or W). All figures are generalizations. ... 205	  Table 4.1 Basic statistics for the five meteorological variables under consideration. The output for all storms together is shown at the top, with southeasters (SE) and westerly windstorms (W) broken out below. ........................................................................................................ 231	  Table 4.2 Ratio of averages for southeasters to the averages for westerly windstorms (SE / W) for four of the meteorological variables presented in Table 4.1. ........................................ 232	  Table 4.3 Basic statistics for the three power-grid related variables under consideration. The output for all storms together is shown at the top, with southeasters (SE) and westerly (W) windstorms broken out below. ............................................................................................ 234	  Table 4.4 Ratio of averages for southeasters to the averages for westerly windstorms (SE / W) for the three power-grid related variables presented in Table 4.3. ..................................... 235	  Table 4.5 Pearson correlation coefficients (R) for the eight variables under consideration. The output for all storms together is shown at the top, with southeasters (SE) and westerly (W) windstorms broken out below. Correlations involving wind speed and power-grid variables are highlighted in bold. Wnddir = wind direction, maxwind = peak two-min wind, maxgst = peak five-sec gust, strmpcpn = storm total precipitation, linefalt = line faults, uniquelf = unique circuits affected, custo = customers out, stormdur = time between the first and last occurrence of 19 km h-1 winds also bounded by a given storm's pressure couplet. ........... 236	  Table 4.6 The frequency that a single line fault affects a given number of customers. ............. 237	  Table 4.7 For all line faults during the 119 sample storms, the frequency at which incidents fell into specific outage types. A description of the different categories is in section 2.3. ....... 240	  List of Tables  x Table 4.8 List of storms that resulted in a "major disaster" based on BC Hydro's own classification. Usually, a storm has to affect ≥100,000 customers qualify for this classification. Storms in italics may not have actually made the cut, but were likely listed as major disasters due to their occurrence immediately after a severe windstorm. Note that total line faults and customers out are for the study region and do not reflect the entire impact of a given storm to southwest BC. .......................................................................... 243	  Table 4.9 A list of the Poisson regression models considered in this study and their associated Akaike information criterion (AIC) values. A lower AIC generally means a better model.  Wind direction type separates windstorms into discrete southeasterly and westerly categories. ........................................................................................................................... 249	  Table 4.10 Using the best multivariate Poisson regression model, testing differences in precipitation and storm duration on the predicted number of line faults for the 1962 Columbus Day Storm (CDS) and 2006 Hanukkah Eve Storm (HES). The HES produced 220 actual line faults in the study region. The total for the CDS is unknown. ................... 252	  Table 4.11 Linear regression R2 for hourly line faults and hourly two-min wind speed during the main wind surge for four strong to severe windstorms at CYVR and CYXX. To test for lag effects, the time of the hourly outages is shifted back and forth three hours from the initial assumption of t + 1 h. For example, for 0 h, the line faults for 1301 to 1400 h are compared to the 1400 h wind speed, and for +3 h, the line faults for 1601 to 1700 h are compared to the 1400 h wind speed. The strongest fits are accented in bold. ......................................... 257	  Table 5.1 Proposed interior regions for windstorm track analysis, including return intervals. A station is considered interior if it resides east of the first range of coastal mountains. Station choices are limited to being land-based, long-term (operating for ≥30 y) and with 24-hour operation. ............................................................................................................................ 288	  Table E.1 Fit statistics for regressions using all 119 independent storms with peak winds >40 km h-1 that occurred during October 2005 to August 2009. ..................................................... 350	  Table E.2 Parameter estimates for the regressions in Table E.1. ................................................ 351	  Table E.3 Fit statistics for Poisson regressions using all 119 independent storms with peak winds >40 km h-1 that occurred during October 2005 to August 2009. ........................................ 352	  Table E.4 Parameter estimates for Poisson regressions in Table E.3. Model 10 includes the interactions between peak wind or gust and storm type based on peak wind direction. .... 353	  Table E.5 Fit statistics for the best-fit linear regressions on hourly line faults and 2-minute wind speed during the three strong southeasterly windstorms 15 Nov 2006, 11 Dec 2006 and 12 Nov 2007. Lag is the time difference between the end time for hourly total line faults and the reported hourly 2-min wind speed (explained further in Chapter 4). ........................... 354	  Table E.6 Parameter estimates for the best-fit linear regressions in Table E.5. ......................... 354	  Table E.7 Fit statistics for the best-fit linear regressions on hourly line faults and 2-minute wind speed during four selected strong windstorms. Lag is the time difference between the end time for hourly total line faults and the reported hourly 2-min wind speed (explained further in Chapter 4). ...................................................................................................................... 355	  List of Tables  xi Table E.8 Parameter estimates for the best-fit linear regressions in Table E.7. ......................... 355	  List of Figures  xi List of Figures  Figure 1.1 The region of interest, showing the names of key geographic features. Elevation in the inset map a rough approximation. Contours are at 400 m intervals starting at sea level (light green) and reaching 2000 m in the Pacific Ranges (medium gray). ....................................... 2	  Figure 1.2 Key population centers in the region (orange-filled circles) and some of the surrounding communities. The Lower Mainland of British Columbia is roughly outlined in orange. ..................................................................................................................................... 3	  Figure 2.1 The geostrophic wind triangle used in the calculation of pressure gradient magnitude and orientation. The two interpolation points required to make a right triangle are denoted IP1 and IP2. Some important geographical regions are also indicated. ................................ 29	  Figure 2.2 Timing of peak wind compared to the timing of peak surface pressure gradient for the 58 windstorms 1994 to 2012. Calculations based on the nearest hour. Negative values represent storms that had a wind maximum ahead of the gradient maximum. ..................... 39	  Figure 2.3  Linear regression between maximum pressure gradient and three-station average peak gust for 21 southeasters and 27 westerly windstorms from 1994 to 2012. The R2 for southeasters is 0.35 and for westerly windstorms 0.07. ........................................................ 41	  Figure 2.4 Frequency of central pressures at landfall for 50 ETCs that produced high winds at the study stations. ........................................................................................................................ 45	  Figure 2.5 Scatterplot for ETC central pressure at landfall and three-station average peak gust. The open and filled diamonds depict all high-wind generating cyclones that tracked across NVI from 1994 to 2012 and the dashed line shows the linear regression best-fit for these selected events (n=10, R2=0.45). The filled circles and filled diamonds indicate all lows, regardless of strength, that tracked across Northern Vancouver Island (NVI) from January 2008 to April 2013 and the solid line indicates the linear regression best-fit (n=50, R2=0.62). Note that two events are shared between the datasets. ......................................................... 46	  Figure 2.6 The tracks for 62 ETCs associated with 57 independent high-wind events that occurred in the Greater Vancouver and Victoria areas, BC, from 1994 to 2012. Solid-black lines indicate southeasters, dashed-black westerly windstorms, solid-gray southerly and southwest events and dashed-gray Fraser outflow events. The single pure "pushed" event has no track. .......................................................................................................................... 48	  Figure 2.7 ETC landfall frequency—the number of ETCs that crossed the boundary on the left side of the polygons—for the 19-y period. The mean return interval in y is in brackets. For the polygons along the coast, this can be considered the landfall frequency, though technically for lows entering bodies of water such as the Queen Charlotte Sound, landfall has not quite occurred. Results are for the primary cyclones, exluding secondary developments. ....................................................................................................................... 49	  Figure 2.8 The frequency of occurrence of ETCs passing through 2.5º x 2.5º quadrilaterals. Mean storm tracks are drawn for southeasterly (orange) and westerly windstorms (blue), and the List of Figures  xii far north Pacific (including the GAK category) lows that trigger high winds in the study region (tan). ........................................................................................................................... 52	  Figure 2.9 The frequency of ETC cyclogenesis initiations for storms that caused high-winds in the study region. Thirteen of the 62 cyclones started off the map. Raw numbers are shown; however, the color scheme is based on a smoothing filter described in the methods. .......... 53	  Figure 2.10 The tracks of 20 ETCs that triggered southeasterly windstorms 1994 to 2012. ........ 54	  Figure 2.11 The tracks of 29 ETCs that caused 27 independent westerly windstorms 1994 to 2012....................................................................................................................................... 55	  Figure 2.12 Frequency of high-windstorms (solid line), as measured at CYYJ, CYVR and CYXX, by month for the 1964 to 2012 period (n=132), and mean return period by month in years (dashed). ...................................................................................................................... 56	  Figure 2.13 Windstorm seasonality. Tracks of 62 ETCs that triggered 57 independent high-wind events from 1994 to 2012 separated by bi-monthly intervals during the storm season. Track types are as described in Figure 2.6. a) September to October; b) November to December; c) January to February; and d) March to April. ........................................................................ 57	  Figure 2.14 Timing and peak wind magnitude of all identified high-windstorms from 1964 to 1993 (n=74). Peak wind is the highest among CYYJ, CYVR and CYXX. .......................... 59	  Figure 2.15 The frequency of high-windstorms per five-y interval (e.g. 1964 to 1968, 1969 to 1973…) compared to the five-y average PDO (a) and NPI (b) for each half-decade from 1964 to 2012. For the PDO the R2 is 0.75 (n=9) using linear regression and for the NPI the R2 is 0.71 (n=9). .................................................................................................................... 62	  Figure 2.16 Trendlines for all five runs comparing the average wet-season PDO index to the frequency of windstorms for the same five-y intervals. ....................................................... 64	  Figure 3.1 The study region. The larger map encompasses the area examined for the synoptic analyses and the inset shows the area considered for mesoscale analyses. .......................... 88	  Figure 3.2 Weather stations used for the development of mesoscale surface analysis maps. Some stations not shown, including Qualicum Beach (VOQ), North Cowichan (VOO) and Victoria Heartland (WVV). The key stations Vancouver, Victoria and Abbotsford are highlighted in Orange. Weather station codes are actually four digits, with Canadian observation site designations being lead with a "C" (e.g. YVR becomes CYVR). Stations in the contiguous 48 United States are lead with a "K" (e.g. SEA becomes KSEA). ............... 91	  Figure 3.3 Some examples of easily detectable bent-back fronts using water-vapor satellite imagery. Shown is the 11 Dec 2006 windstorm (a), 02 Apr 2010 (b) and 12 Mar 2012 (c) just before landfall in all cases. The major 15 Dec 2006 windstorm is also shown (d), which has strong wrap-around moisture, but the exact position of the bent-back front is not as clear as with the other ETCs in this figure. Images courtesy of the US. National Weather Service and are in the public domain. ................................................................................................ 96	  Figure 3.4 Stations used in the construction of the geostrophic wind triangles. White-filled circles with black outlines are station locations. Station identifiers are listed at the end of each triangle's designation. Gray lines denote the station triads. Black lines and orange List of Figures  xiii shading delineate the right triangles used in the calculations. Triangle endpoints not touching a station are interpolation points for sea-level atmospheric pressure. ................. 103	  Figure 3.5 Tracks of eight ETCs that produced significant high-winds in southwest BC during the 1994 to 2012 period. In this study, these storms are given detailed analyses. Paths with a blue shade denote those ETCs that landed on south Vancouver Island, and tracks in red and orange are from those storms that moved onto the north end of the island. The gray tracks trace the paths of two major windstorms that occurred outside of the study period: the 12 Oct 1962 "Columbus Day Storm" (AKA "Typhoon Freda") and the 13-14 Nov 1981 "Friday-the-13th" windstorm. They are included for reference. .......................................... 105	  Figure 3.6 Synoptic analysis 1200 UTC 02 Mar 1999. Orange shading depicts the 30 kPa jet stream, with orange and brown numbers labeling isotachs in m s-1. For the 50 kPa level, black lines denote the heights in dm; upper lows and highs are also marked in black. Isotherms in ºC are indicated with white dashed lines.  At the surface: Lows and central pressures in hPa are indicated with dark blue, and tracks in light blue. Highs are indicated in red. Troughs are delineated with an orange dashed line. Key fronts are also indicated. Not all surface features are shown. ............................................................................................ 112	  Figure 3.7 Synoptic analysis 0000 UTC 03 Mar 1999. See Figure 3.6 caption for the key. ...... 113	  Figure 3.8 Synoptic analysis 1200 UTC 03 Mar 1999. See Figure 3.6 caption for the key. ...... 114	  Figure 3.9 Synoptic analysis 1200 UTC 13 Dec 2001. See Figure 3.6 caption for key. ............ 116	  Figure 3.10 Synoptic analysis 0000 UTC 14 Dec 2001. See Figure 3.6 caption for key. .......... 117	  Figure 3.11 Synoptic analysis 1200 UTC 14 Dec 2001. See Figure 3.6 caption for key. .......... 118	  Figure 3.12 Synoptic analysis 0000 UTC 12 Mar 2012. See Figure 3.6 caption for key. .......... 119	  Figure 3.13 Synoptic analysis 1200 UTC 12 Mar 2012. See Figure 3.6 caption for key. .......... 120	  Figure 3.14 Synoptic analysis 0000 UTC 13 Mar 2012. See Figure 3.6 caption for key. .......... 121	  Figure 3.15 Twelve-hourly analysis at the 85 kPa height covering 1200 UTC 02 Mar to 1200 UTC 03 Mar 1999. Light gray lines denote the lowest several height lines during the initial time-step, with black marking the second and dark gray the third. Dashed lines denote the -5ºC isotherm. ....................................................................................................................... 122	  Figure 3.16 Twelve-hourly analysis at the 85 kPa height covering 1200 UTC 13 Dec to 1200 UTC 14 Dec 2001. See Figure 3.15 caption for the key. .................................................... 123	  Figure 3.17 Twelve-hourly analysis at the 85 kPa height covering 0000 UTC 12 Mar to 1200 UTC 13 Mar 2012. See Figure 3.15 caption for the key. .................................................... 124	  Figure 3.18 Mesoscale surface analysis for 1400 UTC 03 Mar 1999. Wind speed is 2.5 m s-1 per half-barb; 25 m s-1 for a pennant. A full circle represents calm conditions. Solid isobars are drawn in four hPa increments; dashed isobars showing further detail are at two hPa resolution. ............................................................................................................................ 126	  Figure 3.19 Mesoscale surface analysis for 1600 UTC 03 Mar 1999. See Figure 3.18 caption for the key. ................................................................................................................................ 127	  Figure 3.20 Mesoscale surface analysis for 1800 UTC 03 Mar 1999. See Figure 3.18 caption for the key. ................................................................................................................................ 128	  List of Figures  xiv Figure 3.21 Mesoscale surface analysis for 0100 UTC 14 Dec 2001. See Figure 3.18 caption for the key. ................................................................................................................................ 129	  Figure 3.22 Mesoscale surface analysis for 0500 UTC 14 Dec 2001. See Figure 3.18 caption for the key. ................................................................................................................................ 130	  Figure 3.23 Mesoscale surface analysis for 1000 UTC 14 Dec 2001. See Figure 3.18 caption for the key. ................................................................................................................................ 131	  Figure 3.24 Mesoscale surface analysis for 1500 UTC 12 Mar 2012. See Figure 3.18 caption for the key. ................................................................................................................................ 132	  Figure 3.25 Mesoscale surface analysis for 1700 UTC 12 Mar 2012. See Figure 3.18 caption for the key. ................................................................................................................................ 133	  Figure 3.26 Mesoscale surface analysis for 1900 UTC 12 Mar 2012. See Figure 3.18 caption for the key. ................................................................................................................................ 134	  Figure 3.27 Comparison of bent-back front surface responses for the 02 Apr 2010, 11 Dec 2006 and 15 Dec 2006 windstorms, using data from Estevan Point, Solander Island and Buoy 46206. Wind direction (WDIR, thin dashed lines) and temperature (T, heavy solid lines) are shown. Available Buoy 46206 data (reporting is spotty at times) are included mainly due to the absence of wind direction reports from Estevan Point during the 15 Dec 2006 windstorm. .......................................................................................................................... 137	  Figure 3.28 Low center track and peak gusts for the 15 Nov 2006 north Vancouver Island ETC. Peak gusts are shown to the nearest m s-1. High-wind criteria speeds (~ ≥25 m s-1) are indicated with red-filled circles. The time indicated for the storm center positions (large circle with "X") is in UTC, with the starting day being the windstorm designation. Central pressure is indicated in kPa. ................................................................................................ 141	  Figure 3.29 Low center track and peak gusts for the 11 Dec 2006 north Vancouver Island ETC. See Figure 3.28 caption for the key. ................................................................................... 142	  Figure 3.30 Low center track and peak gusts for the 12 Nov 2007 north Vancouver Island ETC. See Figure 3.28 caption for the key. ................................................................................... 143	  Figure 3.31 Low center track and peak gusts for the 12 Mar 2012 north Vancouver Island ETC. See Figure 3.28 caption for the key. ................................................................................... 144	  Figure 3.32 Low center track and peak gusts for the 03 Mar 1999 south Vancouver Island ETC. See Figure 3.28 caption for the key. ................................................................................... 145	  Figure 3.33 Low center track and peak gusts for the 14 Dec 2001 south Vancouver Island ETC. See Figure 3.28 caption for the key. ................................................................................... 146	  Figure 3.34 Low center track and peak gusts for the 15 Dec 2006 south Vancouver Island ETC. See Figure 3.28 caption for the key. ................................................................................... 147	  Figure 3.35 Low center track and peak gusts for the 02 Apr 2010 south Vancouver Island ETC. See Figure 3.28 caption for the key. ................................................................................... 148	  Figure 3.36 ETC central pressure trends before and after landfall. Blue lines highlight two storms that did not exhibit filling for several hours after landfall. Dashed lines depict those storms List of Figures  xv with northern Vancouver Island tracks and solid lines those that moved into southern Vancouver Island. The orange line is an average of all eight storms. ................................ 150	  Figure 3.37 The tracks of an additional ten ETCs that produced either southeast or westerly windstorms during 1994 to 2012. These storms were not analyzed with the detail of the initial eight presented in the preceding figures. Paths with a blue shade denote those ETCs that landed on south Vancouver Island, and tracks in red and orange are from those storms that moved onto the north end of the island. The gray track traces the path of two major windstorms that occurred outside of the study period: the 12 Oct 1962 "Columbus Day Storm" (AKA "Typhoon Freda") and the 13-14 Nov 1981 ("Friday-the-13th" windstorm) and are provided for reference. ........................................................................................... 151	  Figure 3.38 Georgia Strait pressure slope (thick lines) compared to the bearing between CYVR and the low-pressure center (thin lines) for 14 of the 18 Vancouver Island-landfalling ETCs that produced either a southeaster or westerly windstorm from 1994 to 2012. NVI is north Vancouver Island and SVI is south Vancouver Island. The approximate time of landfall is indicated by gray shading. .................................................................................................. 153	  Figure 3.39 Distance of the low-pressure center from CYVR and two-min wind speed in six-h increments. For southeasterly windstorms with ETC landfalls on northern Vancouver Island. .................................................................................................................................. 154	  Figure 3.40 Bearing of the low-pressure center from CYVR and two-min wind direction in six-h increments. For southeasterly windstorms with ETC landfalls on northern Vancouver Island. .................................................................................................................................. 155	  Figure 3.41 Distance of the low-pressure center from CYVR and two-min wind speed in six-h increments. For westerly windstorms with ETC landfalls on northern Vancouver Island. During the study period, there were only two events in this category. ............................... 156	  Figure 3.42 Bearing of the low-pressure center from CYVR and two-min wind direction in six-h increments. For westerly windstorms with ETC landfalls on northern Vancouver Island. During the study period, there were only two events in this category. ............................... 156	  Figure 3.43 Distance of the low-pressure center from CYVR and two-min wind speed in six-h increments. For southeasterly windstorms with ETC landfalls on southern Vancouver Island. .................................................................................................................................. 157	  Figure 3.44 Bearing of the low-pressure center from CYVR and two-min wind direction in six-h increments. For southeasterly windstorms with ETC landfalls on southern Vancouver Island. .................................................................................................................................. 158	  Figure 3.45 Distance of the low-pressure center from CYVR and two-min wind speed in six-h increments. For westerly windstorms with ETC landfalls on southern Vancouver Island. 159	  Figure 3.46 Bearing of the low-pressure center from CYVR and two-min wind direction in six-h increments. For westerly windstorms with ETC landfalls on southern Vancouver Island. 160	  Figure 3.47 Coastal pressure gradients in hPa (100) km-1 for the 12 Mar 2012 windstorm based on hourly observations. Geostrophic wind triangles are arranged south to north with the warmer colors representing the southern sites. ................................................................... 163	  List of Figures  xvi Figure 3.48 Coastal pressure gradients in hPa (100) km-1 for the 11 Dec 2006 windstorm based on hourly observations. ....................................................................................................... 164	  Figure 3.49 Coastal pressure gradients in hPa (100) km-1 for the 02 Apr 2010 windstorm based on hourly observations. ....................................................................................................... 165	  Figure 3.50 Coastal pressure gradients in hPa (100) km-1 for the 03 Mar 1999 windstorm based on hourly observations. ....................................................................................................... 166	  Figure 3.51 Coastal pressure gradients in hPa (100) km-1 for the 15 Dec 2006 windstorm based on hourly observations. ....................................................................................................... 167	  Figure 3.52 The CYYJ-CYQQ-CWSK, or Georgia Strait, geostrophic wind triangle and key stations in and around the region encompassed by the polygon. Stations marked with red-filled circles are those primarily used in the analysis of wind speed and pressure gradients during analyzed windstorms. .............................................................................................. 168	  Figure 3.53 Surface pressure and wind response for the 12 Mar 2012 windstorm using the Georgia Strait geostrophic wind triangle. Solid black and dark blue lines demark 0.25 and 0.40 times the calculated geostrophic wind speed (Mg) in km h-1. The dotted black line is the pressure slope orientation in degrees. The solid gray and light blue lines indicate the geostrophic potential wind speed adjusted, generally downward, based on the relationship between pressure slope and the ideal ~125º orientation in the Georgia Strait for supporting ageostrophic winds. Dashed gray lines plot the sea-level pressure in kPa for the three stations used in the calculations. Orange diamonds indicate the wind speed at CYQQ, with the dashed orange line showing wind direction. Black circles denote the average wind speed for each hourly observation among these six stations: CYYJ, CYQQ, Entrance Island, Ballenas Island, Sisters Island and Buoy 46146; blue circles show the highest wind speed among the same six stations. ............................................................................................... 169	  Figure 3.54 Surface pressure and wind response for the 11 Dec 2006 windstorm using the Georgia Strait geostrophic wind triangle. See Figure 3.53 caption for the key. ................. 171	  Figure 3.55 Surface pressure and wind response for the 03 Mar 1999 windstorm using the Georgia Strait geostrophic wind triangle. See Figure 3.53 caption for the key. ................. 171	  Figure 3.56 Surface pressure and wind response for the 15 Dec 2006 windstorm using the Georgia Strait geostrophic wind triangle. See Figure 3.53 caption for the key. ................. 172	  Figure 3.57 Conceptual model of the "blunting" of strong ocean ETCs as they interact with the rough terrain of coastal BC. Panel a depicts a peak intensity low that is leaving upper support as it moves inland. Panel b shows a low nearing peak intensity that moves inland along with good upper support. Pressure gradients (in blue) are for 100 km. Orange dashed lines show how the low may have deepened had it remained over the open ocean. Time intervals are uneven to reflect the tendency for ETCs to slow down as they make landfall. Vertical scale is exaggerated. .............................................................................................. 181	  Figure 3.58 Conceptual model of the entrainment of a fresh pool of cold air as an ETC lands on Vancouver Island, in this case using a mesoscale map from the 11 Dec 2006 windstorm. Large arrows show the general motion of surface airmasses, with color indicating relative List of Figures  xvii temperature (blue cold, orange warm). Cold air drawn from the BC interior is carried around the low and down the rear flank, creating a secondary cold front (a). This has the effect of turning the initial cold advection field into a de-facto warm sector (b). Remnants of the original warm sector are secluded against the Cascade Mountains (c). A small wedge of air emplaced ahead of the storm remains (d). There is an implied quasi-stationry frontal boundary (e) between the colder, drier interior air and warmer, moister marine air, with the coastal mountains acting as a barrier between the two airmasses. ..................................... 186	  Figure 4.1 The study region. The large red-filled circle encompasses a 50 km radius around the Vancouver International Airport (CYVR). The land area shaded dark red denotes the region in which tree-related line faults on the BC Hydro distribution grid was examined. Surrounding weather stations are also indicated. ................................................................ 221	  Figure 4.2 For 119 independent storms with peak winds >40 km h-1 at either CYVR or CYXX that occurred from October 2005 to August 2009, the frequency of events within given wind (panel a) and gust speed (b) categories. The probability and cumulative distribution functions (PDF and CDF) are also shown. ......................................................................... 230	  Figure 4.3 For the 119 independent storms, the frequency of events for given duration categories in calendar days. Duration is largely based on the range of time that winds >18 km h-1 (80th percentile) occurred at CYVR (see methods). A storm only has to affect part of a day for the day to be counted (i.e. storms occurring within just one calendar day have durations typically <24 h). .................................................................................................................. 233	  Figure 4.4 For the 119 storms, duration of winds >18 km h-1 compared to peak wind speed (panel b) and gust speed (b). There is no relationship. .................................................................. 233	  Figure 4.5 For the BCH distribution grid within 50 km of CYVR, the number of circuits impacted by trees compared to the amount of customers affected. Data are for 119 independent storms with peak wind speeds >40 km h-1 during October 2005 to August 2009. Circles with light gray fill represent a single instance at the given coordinate: Areas with much overlap (e.g. many similar events) are reflected by increasingly dark fill. Linear and Poisson regression model estimates are shown (blue and orange respectively). ......... 238	  Figure 4.6 The number of tree-related line faults that occurred on the 131 separate circuits that were struck in the study area from October 2005 to August 2009. The total number of line faults is 2146 and total number of circuits is 491. Due to space constraints, only every third circuit ID label is shown. .................................................................................................... 239	  Figure 4.7 Frequency of storms for given numbers of line faults. Bins are labeled by the first number of their range, with 0 being 0 to 4 circuits impacted, 5 being 5 to 9 and upward. 240	  Figure 4.8 For the respective peak wind-speed categories of the 119 independent storms isolated in the study, the total line faults for all the storms that fell into a given bin along with the average line faults per storm. The number of independent storms in each bin is also shown.............................................................................................................................................. 241	  Figure 4.9 For the 119 independent storms, the number of events that fall within specific peak wind categories compared to the proportion that caused one or more line faults. .............. 242	  List of Figures  xviii Figure 4.10 For the BCH distribution grid within 50 km of CYVR, peak wind for all independent storms with wind >40 km h-1 at either CYVR or CYXX compared to number of associated line faults. Linear and 2nd-order-polynomial best-fit lines are shown (blue and gray respectively) alongside the predicted from a Poisson regression model (orange). Data are for 119 independent storms during Oct 2005 to Aug 2009. Circles with light gray fill represent a single instance at the given coordinate: Areas with much overlap (e.g. many similar events) are reflected by increasingly dark fill. Solid black represents ≥5 overlapping datapoints. Trendlines are projected to the estimated peak wind of the 1962 Columbus Day Storm for both CYVR (92 km h-1, open circles) and CYXX (102 km h-1, open diamonds).............................................................................................................................................. 244	  Figure 4.11 Comparison of various models, including linear (black), 2nd-order polynomial (blue) and Poisson regression (orange). Two simple exponential models are also included (gray): Both use as a starting point of 4.52, the average number of outages observed for a 41 km h-1 peak wind. The vertical orange-dashed line marks the peak wind for the 1962 Columbus Day Storm (CDS) estimated for CYXX. The blue box outlines the data range used in constructing the models. ..................................................................................................... 246	  Figure 4.12 For the BCH distribution grid within 50 km of CYVR, peak gust for all independent storms with peak wind >40 km h-1 at either CYVR or CYXX compared to number of associated line faults. Linear and 2nd-order-polynomial best-fit lines are shown (blue and gray respectively) alongside the predicted from a Poisson regression model (orange). Data are for 119 independent storms during Oct 2005 to Aug 2009. Circles with light gray fill represent a single instance at the given coordinate: Areas with much overlap (e.g. many similar events) are reflected by increasingly dark fill. Solid black represents ≥5 overlapping datapoints. Trendlines are projected to the peak gust of the 1962 Columbus Day Storm for both CYVR (126 km h-1, open circles) and CYXX (145 km h-1, open diamonds). ............ 247	  Figure 4.13 For the BCH distribution grid within 50 km of CYVR, peak wind for all independent storms with wind >40 km h-1 at either CYVR or CYXX compared to number of associated line faults. Southeasterly (SE, filled circles, n=49) and westerly windstorms (W, open circles, n=70) are considered separately. Linear best-fit lines are shown (SE black, W gray), with projection to the peak wind at CYXX during the 1962 Columbus Day Storm (dashed). Curves showing the predicted from separate Poisson regressions are also shown (SE black, W gray). .............................................................................................................................. 248	  Figure 4.14 The predicted for the strongest multivariate Poisson regression model (black diamonds) compared to actual observations (gray circles) for 119 independent windstorms. Variation in the predicted for a given peak wind speed is due to the interaction of model variables: Peak wind, peak gust, wind direction type (SE and W) and interactions with wind and gust, storm total precipitation and storm duration. The blue curve is an exponential fit to the predicted (R2 = 0.73). .................................................................................................... 251	  Figure 4.15 Hourly two-min wind speed at CYVR compared to the number of tree-related line faults that occurred within a 50 km radius of the weather station during the following hour. List of Figures  xix Depicted are data for four strong to severe windstorms: Panel a) is 15 Nov 2006; b) 11 Dec 2006; c) 15 Dec 2006; and d) 12 Nov 2007. ....................................................................... 253	  Figure 4.16 Average hourly two-min (sometimes 10-min) wind speed at eleven local stations, and the maximum and minimum reported wind speeds at the same stations compared to the number of line faults within the study area due to tree and branch failures during the following hour. Depicted are data for four strong to severe windstorms: Panel a) is 15 Nov 2006; b) 11 Dec 2006; c) 15 Dec 2006; and d) 12 Nov 2007. ............................................ 254	  Figure 4.17 Hourly two-min wind speed at Abbotsford (CYXX), and 10-min wind speed at Pitt Meadows (CWMM) and White Rock (CWWK) compared to the number of line faults in the study area due to tree and branch failures during the following hour. Depicted are data for four strong to severe windstorms: Panel A) is 15 Nov 2006; B) 11 Dec 2006; C) 15 Dec 2006; and D) 12 Nov 2007. ................................................................................................. 255	  Figure 4.18 Scatterplots of hourly two-min wind speed and hourly number of line faults in the study region for three strong to severe southeasterly windstorms (15 Nov 2006, 11 Dec 2006 and 12 Nov 2007). The peak wind surge (i.e. duration) for each windstorm is based on all observations ≥ 1/Euler's number times peak wind. Shown are the results for Abbotsford (a) and Vancouver (b). R2 values are for a linear fit. .......................................................... 256	  Figure 4.19 Hourly two-min wind direction at CYVR compared to the number of tree-related line faults that occurred within a 50 km radius of the weather station during the following hour. Depicted are data for four strong to severe windstorms: Panel a) is 15 Nov 2006; b) 11 Dec 2006; c) 15 Dec 2006; and d) 12 Nov 2007. ............................................................... 258	  Figure 4.20 Conceptual model showing predicted line faults following a steep s-curve, flattening out at a theoretical maximum as winds climb to an unknown but very high peak. The dashed line demarks the upper range of the estimated wind speed that causes nearly 100% tree damage in conifer stands based on windthrow modeling, and provides a guide for the upper asymptote of the s-curve (see text). Included is a scatterplot for the 119 storms used in this study (gray circles), and the predicted for the best multivariate Poisson regression model (black diamonds). ................................................................................................................ 264	  Figure 4.21 A smoothed conceptual model of an individual storm's winds (black line) and the associated line fault response (gray line). The timing of peak wind relative to the arbitrary placement of the start of a day can affect the line fault count for the day, and therefore the number of line faults associated for a given wind speed. Note that the d1 and d2 peak winds are simply switched between panels a and b, but that the number of line faults associated with the two wind magnitudes is markedly different. ......................................................... 271	  Figure A.1 An awning wrecked by high winds on the Oregon coast, 03 Dec 2007. .................. 305	  Figure A.2 Trees broken by high winds in Vancouver, British Columbia, 08 Apr 2010. .......... 306	  Figure A.3 Catastrophic windthrow of an even-aged stand in the Oregon Coast Range due to high winds, 03 Dec 2007. ................................................................................................... 306	  Figure A.4 Multiple trees were forced onto power lines during high winds on the Oregon Coast, pulling down wires and disrupting utility poles, 03 Dec 2007. .......................................... 307	  List of Figures  xx Figure A.5 High winds toppled a church steeple, which in turn fell onto and destroyed a utility pole in Bay City, Oregon, 03 Dec 2007. ............................................................................. 307	  Figure B.1 University of Washington WRF-GFS 1.33 km domain 12-h forecast valid for 1200 UTC 25 Feb 2012. Ten-meter wind speed is depicted in knots, with direction barbs and sea-level pressure (hPa). In the Metro Vancouver area, strong NW winds in the Georgia Strait are reduced just a short distance inland, likely due to turbulent drag. ................................ 309	  Figure C.1 Synoptic analysis 0000 UTC 15 Nov 2006. Orange shading depicts the 30 kPa jet stream, with orange and brown numbers labeling isotachs in m s-1. For the 50 kPa level, black lines denote the heights in dm; upper lows and highs are also marked in black. Isotherms in ºC are indicated with white dashed lines. At the surface: Lows and central pressures in hPa are indicated with dark blue, and tracks in light blue. Highs are indicated in red. Troughs are delineated with an orange dashed line. Key fronts are also indicated. Not all surface features are shown. ............................................................................................ 314	  Figure C.2 Synoptic analysis 1200 UTC 15 Nov 2006. See Figure C.1 caption for the key. .... 315	  Figure C.3 Synoptic analysis 0000 UTC 16 Nov 2006. See Figure C.1 caption for the key. .... 316	  Figure C.4 Synoptic analysis 0000 UTC 12 Nov 2007. See Figure C.1 caption for the key. .... 317	  Figure C.5 Synoptic analysis 1200 UTC 12 Nov 2007. See Figure C.1 caption for the key. .... 318	  Figure C.6 Synoptic analysis 0000 UTC 13 Nov 2007. See Figure C.1 caption for the key. .... 319	  Figure C.7 Synoptic analysis 1200 UTC 11 Dec 2006. See Figure C.1 caption for the key. ..... 320	  Figure C.8 Synoptic analysis 0000 UTC 12 Dec 2006. See Figure C.1 caption for the key. ..... 321	  Figure C.9 Synoptic analysis 1200 UTC 12 Dec 2006. See Figure C.1 caption for the key. ..... 322	  Figure C.10 Synoptic analysis 1200 UTC 14 Dec 2006. See Figure C.1 caption for the key. ... 323	  Figure C.11 Synoptic analysis 0000 UTC 15 Dec 2006. See Figure C.1 caption for the key. ... 324	  Figure C.12 Synoptic analysis 1200 UTC 2006. See Figure C.1 caption for the key. ............... 325	  Figure C.13 Synoptic analysis 1200 UTC 02 Apr 2010. See Figure C.1 caption for the key. ... 326	  Figure C.14 Synoptic analysis 0000 UTC 03 Apr 2010. See Figure C.1 caption for the key. ... 327	  Figure C.15 Synoptic analysis 1200 UTC 03 Apr 2010. See Figure C.1 caption for the key. ... 328	  Figure C.16 Twelve-hourly analysis at the 85 kPa height covering 0000 UTC 15 Nov to 0000 UTC 16 Nov 2006. Light gray lines denote the lowest several height lines during the initial time-step, with black marking the second and dark gray the third. Dashed lines denote the -5ºC isotherm. ....................................................................................................................... 329	  Figure C.17 Twelve-hourly analysis at the 85 kPa height covering 0000 UTC 12 Nov to 0000 UTC 13 Nov 2007. See Figure C.16 caption for the key. ................................................... 330	  Figure C.18 Twelve-hourly analysis at the 85 kPa height covering 1200 UTC 11 Dec to 1200 UTC 12 Dec 2006. See Figure C.16 caption for the key. ................................................... 331	  Figure C.19 Twelve-hourly analysis at the 85 kPa height covering 1200 UTC 14 Dec to 1200 UTC 15 Dec 2006. See Figure C.16 caption or the key. .................................................... 332	  Figure C.20 Twelve-hourly analysis at the 85 kPa height covering 1200 UTC 02 Apr to 1200 UTC 03 Apr 2010. See Figure C.16 caption for the key. ................................................... 333	  List of Figures  xxi Figure D.1 Mesoscale surface analysis for 1700 UTC 15 Nov 2006. Wind speed is 2.5 m s-1 per half-barb; 25 m s-1 for a pennant. A full circle represents calm conditions. Solid isobars are drawn in four hPa increments; dashed isobars showing further detail are at two hPa resolution (when present). ................................................................................................... 335	  Figure D.2 Mesoscale surface analysis for 1900 UTC 15 Nov 2006. See Figure D.1 caption for the key. ................................................................................................................................ 336	  Figure D.3 Mesoscale surface analysis for 2100 UTC 15 Nov 2006. See Figure D.1 caption for the key. ................................................................................................................................ 337	  Figure D.4 Mesoscale surface analysis for 2000 UTC 11 Dec 2006. See Figure D.1 caption for the key. ................................................................................................................................ 338	  Figure D.5 Mesoscale surface analysis for 2200 UTC 11 Dec 2006. See Figure D.1 caption for the key. ................................................................................................................................ 339	  Figure D.6 Mesoscale surface analysis for 0000 UTC 12 Dec 2006. See Figure D.1 caption for the key. ................................................................................................................................ 340	  Figure D.7 Mesoscale surface analysis for 0600 UTC 15 Dec 2006. See Figure D.1 caption for the key. ................................................................................................................................ 341	  Figure D.8 Mesoscale surface analysis for 0800 UTC 15 Dec 2006. See Figure D.1 caption for the key. ................................................................................................................................ 342	  Figure D.9 Mesoscale surface analysis for 1100 UTC 15 Dec 2006. See Figure D.1 caption for the key. ................................................................................................................................ 343	  Figure D.10 Mesoscale surface analysis for 1300 UTC 12 Nov 2007. See Figure D.1 caption for the key. ................................................................................................................................ 344	  Figure D.11 Mesoscale surface analysis for 1500 UTC 12 Nov 2007. See Figure D.1 caption for the key. ................................................................................................................................ 345	  Figure D.12 Mesoscale surface analysis for 1700 UTC 12 Nov 2007. See Figure D.1 caption for the key. ................................................................................................................................ 346	  Figure D.13 Mesoscale surface analysis for 2200 UTC 02 Apr 2010. See Figure D.1 caption for the key. ................................................................................................................................ 347	  Figure D.14 Mesoscale surface analysis for 0000 UTC 03 Apr 2010. See Figure D.1 caption for the key. ................................................................................................................................ 348	  Figure D.15 Mesoscale surface analysis for 0200 UTC 03 Apr 2010. See Figure D.1 caption for the key. ................................................................................................................................ 349	  Acknowledgements  xxii Acknowledgements  The creation of a major creative work such as a PhD thesis requires the input of many people. The project first germinated due to my supervisor, Steve Mitchell, who not only saw my potential but also helped direct me down the long and branching path to a final dissertation. Thank you for the opportunity. My wife, Carol Swan, also played a pivotal role in providing support of all kinds, including thesis editing, and also insight into the nature of the graduate student experience. And there are my children, Kesri and Orion, who provide inspiration and, through their deep curiosity, remind me that there are always new worlds waiting to be explored. Appreciation also goes to my parents, Nancy and Antonino Read-Rojas, who also provided much support during my program. Aunt Charlotte and Uncle Don Cresswell deserve much thanks, too, for their ongoing encouragement.  Many thanks go to my graduate committee members Roland Stull and Doug McCollor, both of whom provided excellent insight into my research topic. Special thanks to Felipe Hirata for helping to extract power outage data for the Lower Mainland, and providing insight about power outages. Ken Byrne deserves thanks for helping me get started with my graduate program, and also taking the time to show me the way for assisting at Fall Camp.  Other people who were a big help during my graduate career at UBC include Andrea Chan, Rosemarie Cheng, Norm Hodges, Gayle Kosh, Christine Mutia and Natasha Thompson. Thank you all for your support through the years.  Many thanks to Gwen Shrimpton and Tom Wells of the BC Transmission Corporation (now folded into BC Hydro) for providing this research opportunity and also guidance along the way. I also thank Brian Fisher, Jeff Labelle and Duncan Isberg for providing the detailed distribution dataset, and insight into what the data means. Thanks also goes to Giselle Bramwell of Environment Canada for providing much of the weather data used in this study, and to David Roth of NOAA for additional climate data.  I also wish to thank Cliff Mass for being a sounding board for ideas related to my thesis, his interest in and support of my windstorm research and also encouraging me to present my results at the Pacific Northwest Weather Workshop. I also owe thanks to George Taylor who took an early interest in my work on windstorms and provided an avenue for me to delve into their climatology at the Oregon Climate Service during my undergraduate tenure at Oregon State University—this was a springboard that ultimately led to my doctoral dissertation.  Finally, much appreciation goes to the BCTC, NSERC, and the MITACs program, all of which provided funding for this project. Thanks also goes to Steve Mitchell, Sue Watts and Bruce Larson for offering TAs, and Lori Daniels for sharing instruction of the abiotic forest disturbance course with me.  Dedication  xxiii Dedication  To my parents, Nancy and Antonino Read-Rojas.  Thank you for your support over the years.  And to my immediate family, Carol, Kesri and Orion.  Thank you for being there for me.   Chapter 1: Introduction  1 1 Introduction   1.1 Southwest British Columbia and Pacific Storms  Southwest British Columbia (BC), situated on the Pacific Coast of Canada, has a complex geography of mountains, valleys, straits, bays, islands and estuaries (Figure 1.1). Elevation climbs steeply from sea level upwards to generally one to two km in the surrounding mountain ranges, with some peaks, such as Mt. Rainier in the south (4392 m) and Mt. Garibaldi in the north (2678 m) being locally higher. Human population is largely concentrated in the lowlands (Figure 1.2). This includes Vancouver and surrounding communities along the BC and Washington (WA) border, also known as the Lower Mainland. There is another population center around Victoria, located on the southern tip of Vancouver Island. As of ~2011, there were approximately 8.1 million people living in the region including the Puget Lowlands of WA, with approximately 2.6 million people occupying the Lower Mainland and 0.3 million in Victoria (Statistics Canada 2014, US Census 2014). Intense storms moving in off of the Pacific Ocean routinely impact the major cities in the region (Read 2008, Mass and Dotson 2010). The winds and rain associated with these storms contribute to tree failures, some of which strike power lines and/or utility poles, causing disruption of electrical service (BC Hydro 2007, Hirata 2011). Sometimes the raw wind force is enough to destroy the power infrastructure (Franklin n.d., Read 2008). Power interruption causes numerous problems, including the loss of communications, failure of electrical heating systems which has lead to fatalities form carbon monoxide poisoning from improperly ventilated Chapter 1: Introduction  2  Figure 1.1 The region of interest, showing the names of key geographic features. Elevation in the inset map a rough approximation. Contours are at 400 m intervals starting at sea level (light green) and reaching 2000 m in the Pacific Ranges (medium gray).    Chapter 1: Introduction  3  Figure 1.2 Key population centers in the region (orange-filled circles) and some of the surrounding communities. The Lower Mainland of British Columbia is roughly outlined in orange. generators, disruption of traffic due to signal failures and also stopping commerce since most transactions are done with electronic devices (Liu et al. 2007, Read and Reed 2013). Such outcomes underscore the importance of constantly improving our understanding of coastal storms and how they affect the power grid.    Chapter 1: Introduction  4 1.2 Windstorm Meteorology and Climatology  1.2.1 High Winds  The U.S. National Weather Service (NWS) considers high winds to be one- to two-min average speeds ≥18.0 m s-1 (65 km h-1) and/or three-second gusts ≥25.7 m s-1 (93 km h-1) (NWS 2012). Environment Canada (EC) uses similar parameters for its wind warnings, depending on the region (EC 2015). These high wind thresholds are based on potential damage to structures (Appendix A Figure A.1), vehicles and trees (Appendix A Figure A.2) (personal communication, Tyree Wilde, NWS, 22 Jan 2004).  1.2.2 Windstorms  Storms that generate high winds result from many different atmospheric phenomena and occur over a wide spatiotemporal range (Golden and Snow 1991, Stull 2000). At the mesoscale, thunderstorms generally operate over a period of minutes to hours and can produce powerful downbursts and also tornadoes, which are perhaps the most widely-known source of intense winds at small scales. Thunderstorms and tornadoes are uncommon in coastal British Columbia (BC) (Phillips 1990, Etkin et al. 2001). Foehns and gap winds, both within the local wind category, can operate over hours and even days, and under the right circumstances attain high-wind levels (Stull 2000). In BC these localized events typically occur outside of densely populated regions, with the key exception being Fraser gap winds that sometimes cause widespread damage in the Lower Mainland and beyond (Mass et al. 1995). Chapter 1: Introduction  5 At larger scales, tropical cyclones, which generally persist over days to weeks, routinely bring intense winds to the Gulf and Atlantic coasts of the U.S. and Canada (Keim and Muller 2007). Only rarely do these storms land on the Pacific Coast, and historically only from southern California southward (e.g. Hurd 1939, Simpson and Riehl 1981). Extratropical cyclones (ETC) are the weather systems that regularly affect BC (Koeppe 1931, Hare and Thomas 1974, Mesquita et al. 2009). These storms, which can span over 1000 km and typically persist over a period of days, can bring heavy rainfall and high winds to coastal and inland regions of the Pacific Northwest (PNW), including southwest BC1 (Lynot and Cramer 1966, Read 2008, Mass and Dotson 2010). ETCs that bring strong winds overland are often referred to as windstorms, and those events that produce mean winds of ≥18.0 m s-1 (65 km h-1) at ≥2 weather stations within a specific region (e.g. Puget Lowlands) have been classified as major windstorms (Mass and Dotson 2010). Such storms can cause widespread damage to structures, trees and electrical infrastructure, sometimes with an attendant loss of human life (Franklin n.d., Lynot and Cramer 1966, Mass 2008, Mass and Dotson 2010).  1.2.3 Windstorm Climatology  Despite the importance of windstorms in the PNW, only a few case studies of these events have been done (e.g. Lynot and Cramer 1966, Reed 1980, Reed and Albright 1986, Steenburgh and Mass 1996). These examinations generally use detailed analysis of surface and upper-air maps, satellite photos when available and in some cases numerical weather model data. There has only been limited intercomparison between analyzed storms, including upper-air and surface                                                 1 For the purposes of this dissertation, Pacific Northwest refers to the region from northwestern California to southwestern British Columbia that is west of the Cascade Mountain crest. Chapter 1: Introduction  6 features. There are even fewer climatological analyses of high-wind storms, and these have been focused on WA and Oregon (OR) (e.g. Read 2008, Mass and Dotson 2010). These studies have used a wide variety of approaches, including intercomparison of storm events (especially surface features), storm track typing and summaries of all storm events in the available long-term wind record meeting arbitrary cutoff requirements. None of the climatological analyses have focused specifically on southwest BC, and only one has employed a track-typing method (Read 2008), though without consideration of mean return intervals such as done by Keim and Muller (2007) for hurricanes. Mass and Dotson (2010) separate major windstorm events into four regions from north to south in what might be considered a rough approach to track typing, but there is no consideration of exactly where the associated ETCs tracked in their list of qualifying storm events. Overall, there has been little attempt to bring upper-air features (e.g. 30 kPa jet stream, 50 kPa trough), surface conditions (e.g. peak wind and gust) and storm tracks (e.g. low center position relative to upper features, landfall location) together in a comprehensive comparative analysis of windstorms. Using 2.5º-resolution reanalysis data, Mass and Dotson (2010) did determine average 50 kPa heights, 85 kPa temperatures and surface pressures for major windstorms and the deviation from normal, but did not directly compare upper-level and surface conditions between individual events. The Pacific Decadal Oscillation (PDO) is a multi-decadal spatiotemporal bi-modal change in the Pacific climate system (Mantua et al. 1997, Mantua and Hare 2002, Whitfield et al. 2010). These changes can alter mean storm tracks (Trenberth and Hurrell 1994, Moore and McKendry 1996), prompting inquiry into the potential for long-term changes in the wind climatology of southwest BC. Tuller (2004), using seasonal and annual mean wind speeds, and pressure-triangle wind speed, found a relationship between the Pacific-North American (PNA) and PDO indexes Chapter 1: Introduction  7 for some southwest BC stations where winds tended to be slower during the warm phase, including at the Vancouver International Airport. Abeysirigunawardena et al. 2009, using long-term data from the Vancouver, indicated a PDO influence on the expected return frequencies for given wind speeds. These two studies, however, did not focus on independent high windstorms. A relationship between the PDO and a response variable, such as the frequency of high winds, does not necessarily mean that the PDO is the direct cause. It is possible that some other forcing mechanism is operating that shifts both the PDO index and the response variable at the same time (Newman et al. in submission). Also, the PDO does not appear to be a discrete oscillation, but is the outcome of the interaction between different atmospheric and oceanic phenomena, including the El Nino Southern Oscillation (ENSO) through an atmospheric bridge mechanism, stochastic atmospheric variation, Rossby and coastal ocean waves and ocean memory. Many of these factors can influence the strength of the Aleutian low, which in turn influences the direction of the PDO signal. When a relationship is found between the PDO and a response variable, further work is required to isolate the exact cause. Extreme wind return intervals have been explored in detail for a few stations in the Georgia Strait including Vancouver (Abeysirigunawardena et al. 2009) using the method of independent storms that is outlined in Harris (1999). ETC track information was not included in the analysis. Windstorm return intervals that incorporate track information have been studied in the most detail for hurricanes on the east coast of North America (e.g. Keim and Muller 2007); this is largely an area open to inquiry for BC and indeed much of the Pacific Coast of North America. In addition to case and broad climatological studies, the initiation locations and frequencies of rapidly intensifying ETCs over a large part of the Northern Hemisphere including the northeast Pacific Ocean has been studied using data from surface analysis maps (Sanders and Chapter 1: Introduction  8 Gyakum 1980). A climatology of ETCs using global reanalysis data and an automated method that tracks relative vorticity at the 85 kPa level has been done (Mesquita et al. 2009). Seiler and Zwiers (2015) also produced a comprehensive climatology of ETCs using a similar method to explore the performance of coupled model intercomparison project (CMIP5) climate models in reproducing rapidly deepening storms. These studies did not focus on high-wind generating ETCs in coastal BC, with Mesquite et al. (2009) primarily focused on the Gulf of Alaska and Seiler and Zwiers (2015) concentrating on the Northern Hemisphere, especially over the Pacific and Atlantic oceans. Global climate change due in part to fossil fuel emissions, primarily the release of CO2 gas into the atmosphere, appears to be underway and projected to intensify during the 21st century (IPCC 2013). Global climate models used to forecast the magnitude of climate change over the next several decades, such as CMIP3 and CMIP5, tend to forecast a reduction in the frequency of ETCs during the winter, with a slight polar shift in tracks in the Northern Hemisphere (e.g. Bengtsson et al. 2006, Ulbrich et al. 2009 and McDonald 2011). For specific regions, the trends can be different, with an increase in frequency expected for the Great Britain, central Europe and the Aleutian Islands. Studies on the effect of climate change on storm intensity have led to inconsistent results (e.g. Lambert and Fyfe 2006 and Zappa et al. 2013). It appears that the different projections are likely due to the manner in which intense cyclones are defined. Chang (2014) shows that for North Pacific cyclones, analysis focused on the minimum central pressure show a slight increase in intensity, whereas an examination of ETC depth relative to the surrounding pressure field shows a slight decrease. The difference is due to a projected strengthening of the mean Aleutian low, which is supportive of deeper central pressure minima over the North Pacific.  Chapter 1: Introduction  9  1.2.4 Windstorm Meteorology  The bent-back front, or occlusion, has been described as a rear-ward extension of the baroclinicity that drives a given ETC (Shapiro and Keyser 1990, Steenburgh and Mass 1996, Mass and Dotson 2010). Other terms for this feature are bent-back trough, back-bent front and bent-back warm front (Godske et al. 1957, Steenburgh and Mass 1996). In satellite photos, it can often be seen as a band of clouds wrapping over the pole-ward side of the low and arcing into the rear flank of the storm. Mass and Dotson (2010) point out that peak winds during windstorms in western WA and OR, and other regions such as the United Kingdom, are often associated with the bent-back front. They do acknowledge the role of unstable air in warm sectors as also being a contributor to peak winds. It is unclear whether bent-back fronts are the primary determiner of high winds in nearby regions such as southwest BC. Indeed, aside from discussion of bent-back fronts and warm sectors, the conditions that determine the timing of peak winds for a given location, especially in relation to the position and strength of the low-pressure center, have not been analyzed in detail. Mass and Dotson (2010) also created a four-stage conceptual model for PNW windstorms, using the 15 Dec 2006 Hanukkah Eve storm as an example: i) pre-frontal, the time of weakest winds as the storm approaches; ii) post-frontal, with increasing winds as the leading frontal systems move through; iii) bent-back trough, with the bent-back front delivering peak winds; and iv) termination, when the storm away from the region and/or weakens. This model is focused on the bent-back front being the key driver of peak winds, which may not be the case for all windstorms. Also, the model appears best applicable to windstorms that have a northeast to east track—directions that clearly bring the bent-back front inland—as opposed to a more northerly Chapter 1: Introduction  10 track up the coast that some windstorms follow (e.g. Lynot and Cramer 1966). Furthermore, due to differences in geography between the study region and western WA and OR, the model may not be entirely appropriate for southwest BC windstorms. Shapiro and Keyser (1990), using detailed analysis of a north Atlantic storm that included periodic aerial reconnaissance of atmospheric conditions, developed a four-stage "fractured-front" model that describes the developmental cycle of an ETC. In Stage I, the incipient cyclone begins to develop in a baroclinic region. Stage II is the period of frontal fracture, with the development of clear warm and cold fronts that stretch outward from the vicinity of the developing low center (called a triple-point or point-of-occlusion low in some weather forecasting circles). Stage III, called the frontal T-bone, has an obvious bent-back front reaching from the point-of-occlusion into the rear flank of the low, making the namesake shape. ETCs are typically at or near peak intensity at this time, especially with respect to pressure gradients in the vicinity of the bent-back front, though central pressures may continue to drop in the next stage. Stage IV, the warm-core seclusion phase, is typically characterized by repeat wrapping of the bent-back front around the low center. A pool of warm air is secluded in the vicinity of the low center if not at the surface then in the upper levels, encircled by the cooler air wrapping around the low. The fractured-front model provides a means of classifying the developmental stage of landfalling ETCs, something that has not been done systematically for PNW windstorms.  1.2.5 Historical Windstorms  In the weather history of southwestern BC, a few windstorms stand out above the others due to a large amount of associated damage, a wide area of affect and/or a high loss of human Chapter 1: Introduction  11 life. Two important events from the era of modern surface airways observations are the 12 Oct 1962 Columbus Day and the 15 Dec 2006 Hanukkah Eve storms. The Columbus Day Storm is sometimes called Typhoon Freda due to the fact that the ETC was energized by incorporating the fading remains of a West Pacific tropical cyclone that had been designated Freda (sometimes mistakenly written Frieda) by weather forecasters (Franklin n.d.). Throughout much of its strike zone from northwest California to southwest BC, the Columbus Day Storm set peak gust records, many of which still stand today (Lynot and Cramer 1966, Read 2008, Mass and Dotson 2010). Freda was also the most destructive PNW windstorm in the 20th century, causing extensive damage to trees and the largest single loss of timber on record for the region (Franklin n.d., Read 2008, Mass and Dotson 2010). The power grid suffered extensive damage not only from broken trees and branches, but also from the raw wind force destroying utility poles and transmission towers (Franklin n.d.). For the PNW region west of the Cascade Crest, the 15 Dec 2006 Hanukkah Eve Storm was roughly a 10-year event (Read and Reed 2013). However, for specific locations such as the Puget Lowlands of Washington (WA) the storm produced some of the strongest winds since 1962. The Hanukkah Eve Storm is particularly important because of extensive damage to trees and, as a result, the power grid from northwest Oregon into southwest British Columbia (Read and Reed 2013). The Hanukkah Eve Storm was the last in a series of three windstorms that struck over a five-day period, with each event compounding the damage from the previous storm (BC Hydro 2007, Mass and Dotson 2010, Read and Reed 2013). The significance of the series of storms suggests that power outage data including 2006 may be quite rich for an analysis of wind-induced electric service interruptions. Such data could be useful for the construction of models that help predict the outage potential of a storm before it strikes, allowing electrical utilities to Chapter 1: Introduction  12 respond beforehand—e.g. by prepositioning crews to areas identified as being the most likely to incur damage during a given storm.   1.3 Windthrow  Windthrow refers to the breakage or uprooting of trees by wind forces that exceed a tree's capacity to resist the load (Stathers et al. 1994, Mitchell 2013). The term can refer to a single broken tree, or the toppling of a large patch of forest. Tree failures that occur during routine annual wind maxima are sometimes called endemic windthrow, while catastrophic windthrow refers to the widespread tree damage that occurs during infrequent intense windstorms (Appendix A Figure A.3) (Everham and Brokaw 1996, Mitchell 2000a). Given the predominance of large trees in the PNW, which are effectively force amplifiers once they are toppled by wind (i.e. big hammers), windthrow adds to the damage potential of high windstorms (Mass 2008). For instance, broken branches or entire trees striking power lines and poles can be a major cause of electrical service interruption during windstorms (BC Hydro 2007, Hirata 2011).  1.4 Power Outages  In a power outage, electrical service is cut to one or more customers (BC Hydro 2007, Hirata 2011). A variety of events can produce an outage, including vehicles striking utility poles, equipment failure and wildlife coming into contact with lines (BC Hydro 2007). Among BC Hydro outage cause categories, trees and adverse weather are the most frequent contributors of outages. For example, during the 2006 and 2007 fiscal years these two causes respectively were responsible for 44% and 77% of lost customer hours (BC Hydro 2007). Adverse weather (i.e. Chapter 1: Introduction  13 storms) can damage power infrastructure through the toppling of trees either through snow or wind loading (Appendix A Figure A.4), mass wasting events, the raw wind force breaking utility poles and flying debris hitting electrical infrastructure (Appendix A Figure A.5) (BC Hydro 2007, Hirata 2011, Reed et al. 2010, Read and Reed 2013). Hirata (2011) found that 62% of outages in BC were due to tree-falls, with another 4% from branches. Trees can affect power lines in situations not related to adverse weather, including branches growing into contact with lines. Guggenmoos (2011) using Puget Sound Energy data covering part of western WA, reported that 95% of tree-related outages were due to branch, bole or root failure—i.e. likely due to the effects of wind rather than grow-ins. A tree or branch can impact a line without causing an outage. "No-outage" events appear to occur mostly on the service drops—e.g. the line that goes from the utility pole to a house (Duncan Isberg, BC Hydro Manager, personal communication, 30 Mar 2015). BC Hydro personnel become aware of these incidents when they are reported, usually by phone, by a customer. During situations with numerous outage incidents, such as during a big storm, events that do not cause an outage are given low priority—nevertheless, most of them have to be cleared at some point and are therefore nontrivial. Because trees and branches can strike lines without causing an actual outage, the term "line fault" is here preferred over outage to refer to these incidents—regardless of the number of customers affected.  Various methods have been used to analyze the relationship between weather data and tree-related power outages in the PNW. Hirata (2011) used logistic regression on a spatial outage dataset that covered the North Shore of the Greater Vancouver Metro District of BC. The analysis, which used a BC Hydro outages dataset from Oct 2005-Aug 2009, incorporated gridded numerical weather prediction (NWP) model data including precipitation and wind speed, Chapter 1: Introduction  14 elevation, slope, and stand variables including height and age. Guggenmoos (2011) fit a Weibull curve to data from western WA to compensate for a dataset that was incomplete, especially for higher wind speeds. Taylor and Neale (2007) used threshold analysis that included antecedent rainfall and peak wind on outage data for Victoria, BC (Taylor and Neale 2007). Looking outside of the PNW, Liu et al. (2005) used negative binomial regression and GIS-based spatial outage datasets to determine the number of outages for Carolina hurricanes using three sample storms. Liu et al. (2007) used accelerated failure time (AFT) models, a kind of survival analysis, to determine restoration times after hurricanes and ice storms in the Carolinas and Virginia. Liu et al. (2008) used generalized linear mixed modeling on spatial outage datasets for several recent hurricanes and ice storms in the Carolinas to model outages for 3 km x 3 km grid cells. Davidson et al (2003), looking at hurricanes in the Carolinas, used GIS-based spatial outage datasets and a wind-field model that interpolated peak gust speeds using available surface observations to examine the number of outages per 1000 overhead transformers. Zhou et al. (2006) used Poisson regression on wind-related outage data for Manhattan, Kansas. The first four approaches focused on several storms of the type being examined, while the third used daily weather and outage data. None approached outage modeling using a complete independent storms dataset. Only Hirata (2011) has explored power outages in the Lower Mainland region of BC, in this case with a focus on the North Shore. However, Hirata (2011) focused largely on spatial relationships and did not take an independent storms approach, save for the examination of a few extreme events. Zhou et al. (2006) or Guggenmoos (2011) also did not use an independent storms approach, but instead used daily data. Though Guggenmoos (2011) reported using independent storms for part of the study, the use of daily data instead of hourly would limit the Chapter 1: Introduction  15 effectiveness of any approach to isolate discrete storms. No study performed an hour-by-hour analysis of power outages to reported wind speeds.  1.5 Knowledge Gaps  With respect to windstorm climatology, no study has focused on determining all the high-wind events that have affected the most populated regions of southwest BC within a given period of record (e.g. 1964 to 2012). Furthermore, though there has been some use of track-typing for these storms (e.g. Read 2008, Mass and Dotson 2010), these studies did not include a comprehensive account of all storms within given track types, did not have an estimate of mean return intervals for specific tracks, and were not focused on southwest BC. Analysis of long-term trends in mean wind speed and wind return intervals in southwest BC supports an influence by the PDO (e.g. Tuller 2004, Abeysirigunawardena et al. 2009). However, these studies do not focus specifically on high windstorms. Looking at windstorm meteorology, though there are detailed analyses of individual storms that include surface and upper-air conditions (e.g. Lynot and Cramer 1966, Reed and Albright 1986, Steenburgh and Mass 1996), there has been only limited intercomparison of events, especially using both synoptic- and meso-scale analysis. Also, storm track is not controlled for when comparisons are made. A systematic examination of central pressure trends for landfalling high-wind generating ETCs has not been done. Though Mass and Dotson (2010) have suggested that peak winds during windstorms are often associated with the bent-back front, this may not be the case during all high windstorms and for all regions, something that a comparative approach could help determine. The fractured front model of Shapiro and Keyser (1990) could be used to Chapter 1: Introduction  16 catalog the developmental phases of landfalling ETCs and determine what stage is most often associated with high winds. This has not been done systematically for PNW windstorms. For power grid resiliency to wind loading, only one study has focused specifically on the Lower Mainland (Hirata 2011), and this was limited to the North Shore, not the populated regions south and southeast of the Burrard Inlet. Though Hirata's (2011) spatial approach provides good information about the contributing factors for tree-related power outages, it did not use an independent storms method, nor did it look at storm peak wind speeds from observation sites, and instead used hourly NWP averages. No one appears to have analyzed wind speeds and associated outages at a fine temporal resolution, such as on an hour-by-hour basis. GIS approaches require often difficult-to-obtain and/or assemble spatial datasets. Some of the difficulty arises from concern among public utilities about releasing private information such as home addresses. Outage data with limited spatial information is often easier to obtain, making non-spatial approaches such as Zhou et al. (2006) appealing.  1.6 Research Objectives  This research project will initially describe the windstorm climatology of southwest BC using an independent storms and track-typing approach. A detailed assessment of how high-wind generating ETCs interact with the local terrain of southwest BC and northwest WA follows, with a focus on determining the primary variables associated with the timing of peak winds and developing a conceptual model for windstorms that affect the study area. Finally, the wind-speed dose response of the power grid to strong winds is modeled for the Lower Mainland using independent storms determined from hourly weather observations, also with an hour-by-hour examination of the power grid response to selected strong windstorms. Thus, the research spans Chapter 1: Introduction  17 windstorm climatology, windstorm meteorology, and power grid resiliency with respect to wind and tree-caused damage.  1.6.1 Windstorm Climatology Objectives  The research objectives are to:  1) utilize a track-typing method to identify key landfall locations of ETCs that generate high winds in the most populated region of BC; 2) describe average storm tracks for classes of windstorms based on peak wind direction; 3) identify the origin points of high-wind generating ETCs; 4) examine the central pressure tendencies, pressure gradients and wind magnitude, looking for relationships between these variables; 5) assess seasonality, return periods, and possible variation in windstorm frequency that may be related to known long-term Pacific climate cycles.  1.6.2 Windstorm Meteorology Objectives  For selected significant Lower Mainland and Greater Victoria windstorms from the 1994 to 2012 period that resulted from Vancouver Island landfalls, the objectives are to:  1) develop a method to examine synoptic conditions in an efficient manner; Chapter 1: Introduction  18 2) examine patterns and the variability of synoptic conditions between storms at the 30 kPa level, 50 kPa, 85 kPa and surface; 3) produce mesoscale surface maps to show the isobaric patterns, fronts and surface wind response around the time of peak winds; 4) determine the tracks and central pressure trends as the ETCs moved inland; 5) provide detailed frontal positions to help understand how air mass boundaries influence surface winds among the complex coastal terrain; 6) examine surface pressure gradients and their relationship to surface winds; 7) build conceptual models for southeasterly and westerly windstorms that provide detail specific to the region of interest, including identifying the key variables that contribute to peak wind timing.  1.6.3 Wind Impacts to the Power Grid Objectives  The objectives are to:  1) determine the frequency of storms that cause distribution-grid line faults; 2) model the relationship between peak 2-min wind speed and the frequency of distribution-grid line faults, also including the interactions with peak 5-sec gust, peak wind direction, storm duration and storm-total precipitation; 3) using an hour-by-hour analysis, relate the frequency of distribution-grid line faults to hourly 2-min wind speed for specific extreme storm cases.  Chapter 1: Introduction  19 1.7 Approach  For the climatological analysis, candidate high-wind storms were selected by using 2-min peak wind, 5-sec peak gust and peak wind direction from the hourly and special surface weather observations taken at the Vancouver, Victoria and Abbotsford Airports. Hourly and special observations were provided by Environment Canada (EC). The identified storms were then classified by wind direction. ETC tracks and landfall locations were mapped and examined for trends, using maps archived by the US. National Climatic Data Center (NCDC), the US. Weather Prediction Center, the National Oceanic and Atmospheric Administrations (NOAA) Central Library Daily Weather Maps and the EC library microfilm collection (NCDC 2013, NOAA 2013, WPC 2013). Pressure tendencies were also analyzed, mainly to determine the rate of intensification of the ETCs. Monthly frequencies of high winds events were used to determine seasonality. A potential relationship between long-period climate oscillations, such as the Pacific decadal oscillation (PDO), and windstorm frequency, was also undertaken. The monthly PDO index was from Mantua (2000) and monthly NPI data from Trenberth and Hurrell (2013). For the meteorological analysis of eight candidate ETCs, synoptic surface and upper-air maps showing the development of the storms over 24 h, and mesoscale surface maps depicting the time of landfall were drawn and used extensively. The same data sources used for storm track determinations (above) were also used for the development of these maps. Additionally, satellite photos obtained from the National Weather Service (NWS) Seattle (SEW) and Portland (PQR) offices were also used (NWS SEW 2014; NWS PQR 2014). Surface observations were obtained from EC, the NOAA National Data Buoy Center (NDBC), the University of Washington, the Plymouth State Weather Center, the NOAA Aviation Weather Center (AWC) and the Emergency Weather Network (AWC 2011, NDBC 2011, PSWC 2011, EWN 2012). Tables and Chapter 1: Introduction  20 maps of peak gusts were also used to characterize storm outcomes for different geographical regions. The relationship between storm track and peak gust distribution, time of landfall and the timing of peak wind, and the passage of frontal systems in relation to the occurrence of peak wind were examined. Sea-level pressure was also analyzed via the development of pressure-wind triangles and related to measured surface wind speeds, and the evolution of ETCs as they encountered the high frictional effects that occur at landfall. Additional information from ten other ETCs with Vancouver Island landfalls was brought into some parts of the analysis, especially to aid in the development of a conceptual model for southwest BC windstorms. High-wind impacts to the power grid were studied using a distribution outage dataset from BC Hydro that covered October 2005 to August 2009 and hourly weather data from EC. For the Lower Mainland region within a 50 km radius of the Vancouver International airport, the magnitude of tree-related line faults—determined by the number of times trees or branches strike power lines during a given independent storm—was related to the highest peak wind and gust speed at either Vancouver or Abbotsford. Linear and Poisson regression models were used. An hour-by-hour analysis was also undertaken for selected windstorms to further explore the relationship between wind speed and the number of outages. For the hourly analysis, data from eleven weather stations within the region were considered, though the focus largely remained on Vancouver and Abbotsford.  1.8 Thesis Structure  This thesis is divided into five chapters. Chapter 1 provides a broad background for the thesis topic, a rationale for the study and research objectives. The next three chapters are written as stand-alone papers with a complete research-manuscript structure including an introduction Chapter 1: Introduction  21 that includes literature review, methods, results, discussion and conclusions. Chapter 2 focuses on the climatology of windstorms for southwest BC covering 1964 to 2012, with a more detailed focus on the 1994 to 2012 period. In Chapter 3, the meteorology of eight of the strongest windstorms from 1994 to 2012, all of which were due to Vancouver Island landfalls, is explored in detail. Electrical distribution resiliency to windthrow-related line faults is analyzed in Chapter 4. Chapter 5 summarizes key findings drawn from Chapters 2 to 4, and provides some recommendations for further windstorm climatology and meteorology research along with recommendations for distribution system management for windthrow-related damage and additional research avenues related to the study of power grid resiliency. Chapter 2: Climatology of High-Wind Storms  21 2 The Climatology of High-Wind Storms that Affect the Lower Mainland and Greater Victoria, British Columbia, Canada  2.1 Introduction  High winds are a recurring feature in the climatology of many midlatitude regions and are associated with a number of weather phenomena including tornadoes, hurricanes and extratropical cyclones (ETC) (Golden and Snow 1991). All these storms are capable of generating wind speeds that can cause significant damage to trees and human structures with an associated loss of human life (Golden and Snow 1991, Rappaport 2000, Ashley 2007, Ashley and Black 2008). Southwest BC, being typically under the influence of midlatitude zonal flow off the Pacific Ocean, experiences a high number of ETCs (Koeppe 1931, Hare and Thomas 1974). An examination of media accounts from two local papers (Vancouver Sun and Vancouver Province) of high-wind events that struck the Lower Mainland and Greater Victoria from 1994 to 2012 revealed many storm-related consequences. Windthrown trees often damaged structures, vehicles, power-lines, and frequently blocked roads. The raw force of the wind shattered windows, sometimes removed panes from high-rises, tore off roofing, damaged outbuildings, toppled fences and patio rails, broke or upended utility poles and either bent or destroyed signs and stoplights. Power outages disrupted traffic and contributed to gridlock during rush hour; also, significant economic losses have occurred given a retail network that is now very dependent on electricity. Electrical service failure also meant an inability to keep homes warm; loss of heating sometimes lead to fatalities due to the use of gasoline-powered generators in poorly ventilated houses. Fatalities also resulted from trees impacting occupied houses and Chapter 2: Climatology of High-Wind Storms  22 vehicles, and from people coming too close to or in contact with downed live wires. Ferry sailings were often cancelled, bridges closed, and trolley-buses shut down by trees impacting the overhead lines. High winds also disrupted air traffic, and a few planes crashed during the storms, sometimes with fatalities. Boats broke free of their moorings and either washed ashore or were swamped. Occupied vessels being operated during windstorms have been swamped or capsized and sunk, sometimes with loss of life, and requiring rescue operations in hazardous conditions. Some of these storms brought significantly low atmospheric pressure that, when the event coincided with high tide, caused flooding in shoreline areas, an inundation likely aided by the strong winds. Heavy rain is associated with some of these windstorms, with attendant mass wasting and flooding issues. Given the consequences of high-wind generating ETCs in coastal BC, it is surprising that a systematic method of cataloguing events has only been partially done. For windstorms in the Pacific Northwest of the United States, there have been a few detailed examinations of specific events (Mass and Dotson 2010). These include the 26-28 Oct 1950 series of storms (Smith 1950), the very destructive Columbus Day Storm (AKA "Typhoon Freda") of 1962 (Lynot and Cramer 1966), the 12-13 Feb 1979 windstorm that sunk the Hood Canal Floating Bridge (Reed 1980), the large extratropical cyclone that swept the Pacific Coast on 13-14 Nov 1981 (Reed and Albright 1986, Kuo and Reed 1988) and the powerful Inaugural Day Storm of 1993 (Steenburgh and Mass 1996). These storms were generally examined from a meteorological not climatological standpoint. Lynot and Cramer (1966) did try to ascertain how the Columbus Day Storm fit with other big storms of history, most notably commenting on major storms on 09 Jan 1880 and 29 Jan 1921, but did not undertake a rigorous assessment of previous events. Chapter 2: Climatology of High-Wind Storms  23 From 2000 to 2008, I conducted a more thorough climatological accounting of major windstorms with a focus on northwest California (CA), western Oregon (OR) and Washington (WA), including the analysis and presentation of surface pressure tendencies, pressure gradients, peak wind and peak gust speeds (Read 2008). A windstorm climatology by Mass and Dotson (2010) informed in part by Read (2008) includes brief summaries of significant occurrences, a climatological analysis of "major windstorms" based on the definition of two or more stations reporting 18.0 m s-1 (35 knots) or greater wind speeds, and a conceptual model summarizing the progress of a typical event. Mass and Dotson focused on events in western WA and OR. The Pacific Storms Climatology Products (PSCP) project has made available online extreme-event summaries for weather stations around the Pacific Ocean including within the Cascadia region (Kruk et al. 2013), and this includes summaries of peak wind speed. However, the PSCP application does not provide a discrete storm focus, the kind of approach that is characteristic of the hurricane record (EC 2013a, NHC 2013). Broad ETC analyses have been performed for the North Pacific, such as that by Mesquita et al. (2009), using reanalysis data and a feature-tracking algorithm that captures many ETCs of a wide range of intensities. However, Mesquita et al. (2009) do not distinguish the high-wind-generating coastal storms from routine events. There has been more interest in a systematic documentation of ETCs that affect the east coast of North America, such as nor’easters (Jones and Davis 1995, Zielinski 2002) and post-tropical storms (Kruk et al. 2010). Also, classification schemes have been developed for nor'easters; for example, Dolan and Davis (1992) developed one using relative wave power. The focus on the Atlantic region is likely due to a high concentration of people and property in the area, and perhaps the availability of a longer historical record.  Chapter 2: Climatology of High-Wind Storms  24 Tropical cyclones have been more comprehensively documented than high-wind-generating ETCs. An extensive archive of storm data is available on the National Hurricane Center (NHC) and Environment Canada (EC) websites that includes good datasets for tracks and central pressures going back over a century (EC 2013a, NHC 2013). This availability offers opportunity for detailed study, such as with spatiotemporal patterns in landfall locations and intensity (Keim and Muller et al. 2007), also known as track typing.  Hurricane tracks are largely plotted via a combination of surface analysis, satellite photo interpretation, Doppler radar and available upper-air data provided by means such as radiosondes and routine observations via aircraft (Simpson and Riehl 1981, Willoughby and Chelmow 1982, Sheets 1986). A similar integration of data is utilized by weather forecasting agencies in the creation of routine synoptic-scale surface and upper-air weather maps as a forecast aid. ETCs, however, are generally not monitored via dedicated aircraft such as hurricane hunters, a fact that likely results in less precise and accurate track determination for storms over the oceans. Though these maps are not without error, they are arguably an important resource for following the evolution and path of high-wind generating ETCs, which also makes them a useful tool for the management of risk to public safety and infrastructure posed by these storms. The goal of this paper is to provide a systematic climatology of high-wind generating ETC's that affect the south coastal area of BC. The specific objectives are to: i) utilize a track-typing method to identify key landfall locations of ETCs that generate high winds in the most populated region of BC; ii) describe average storm tracks for classes of windstorms based on peak wind direction; iii) look at the origin points of these ETCs; iv) examine the central pressure tendencies, pressure gradients and wind magnitude of these events; and v) assess the return periods and possible variation in relation to known long-term Pacific climate cycles. Chapter 2: Climatology of High-Wind Storms  25 2.2 Methods  2.2.1 Windstorm Definition   Three first-order stations were selected as the basis for determining the presence of a high windstorm in the study region: Vancouver International (CYVR), Victoria International (CYYJ) and the Abbotsford Airport (CYXX). These sites were selected because they encompass the most populated region of BC, and have a long period of hourly record going back to at least 1953. These airports are also not located on the outer Pacific Coast, which is prone to extreme winds on a frequent basis. The three sites are sheltered in part by the coastal mountains, including the Olympics and Vancouver Island ranges. Thus, when high winds do occur at one or more of these stations, it often reflects a significant storm. CYVR, however, with a long overwater fetch in the western quadrant, is favored for high wind readings during westerly storms.  The US. National Weather Service (NWS) defines high winds as when either the 1-min average wind is ≥18 m s-1 (35 knots) and or gusts are ≥25.7 m s-1 (50 kt) (NWS 2012). In this study, the high wind cutoff is lowered by 1 knot for wind, to ≥17.5 m s-1 (34 kt), and 2 knots for gust, ≥24.7 m s-1 (48 kt). This allows for a 3% error in wind measurement known for the conical 3-cup sensor used in Canada (Devine 2008), and potential rounding differences associated with the conversion of units due to the switch from reporting miles per hour to knots in the mid-1970s. Lowering the cutoff also brings additional events into the analysis. Any wind meeting one or both of the above thresholds at just one of the three stations counted as a windstorm. Once identified, windstorms were put into subcategories largely based on peak wind direction at CYVR. Categorization by wind direction is important due the complex local geography and its potential influence on wind speeds at the three sites. Maximums out of Chapter 2: Climatology of High-Wind Storms  26 40 to 90º were classified as Fraser gap events, 100 to 160º "southeasters", 170 to 200º "southerly", 210 to 250º "southwesters" and 260 to 320º "westerly". Both CYYJ and CYXX were sometimes considered in the determination of categories, especially when their peak wind speeds were higher than CYVR. Wind direction can vary considerably during the course of a storm and across stations (e.g. a storm with a peak gust from 130º at CYVR may produce a 180º peak at CYXX). Further discussion focused largely on CYVR is presented in Appendix B.   2.2.2 Anemometer, Methodological and Data Availability Changes  Historically, at the three stations, different anemometer types, sensor heights and averaging periods have been used (Table 2.1). Canada modernized many of its weather stations Table 2.1 History of anemometer and observational methodology changes relating to wind at the three study stations (Hare and Thomas 1974, Tuller 1980, EC 2009, MANOBS 2013, MOC 2013, J. F. Fleming pers. comm. 2013). Different anemometers sometimes had overlapping periods of service: During these times, the system being used for the official wind reports is not always clear. Instru-ment Type Approximate Dates of Service Anemometer Height (m) Sen-sor Type Averaging Period Obser-vation Type Read-out YYJ YVR YXX YYJ YVR YXX Wind Gust Dynes Dines?  1936-1946, 1946-1959 1944-1949  ?, 21 28 Pres-sure tube? ? ? ? ? 45B 1945-1964 1959-1970, 1970-1977, 1977-? 1949-1959 21.8 19, 18.3, 12.2 6.4 3-Cup 15-30 s Instant Manual Blink Light, Dial, Chart U2A 1964-2005 1964-1998 1959-1971, 1971-2005 10 10 23.8, 10 3-Cup 1 min Instant Manual Dial, Chart 78D 2005-2013 1992-2013 2005-2012 10 10 10 3-Cup 2 min 5 s Auto Digital WS 425G 2013- 2013- 2012- 10 10 10 Sonic 2 min 3 s Auto Digital Chapter 2: Climatology of High-Wind Storms  27 in the 1990s, incorporating the 78D anemometer, a digital cup-based system (Richards and Abuamer 2007, Nav Canada 2013). Another round of modernization is currently underway, with sonic anemometers being deployed in 2012 at CYXX and 2013 at CYVR and CYYJ (EC 2013b). For the 78D anemometer, gust is the highest instantaneous report (MANOBS 2013). However, the anemometer displays a 5-s vector-average wind speed; therefore gust appears to be a 5-s average. Wind is a 2-min average. The U2A anemometer is a short lag-time analog system with output on direct-reading dials or strip-chart recorders (MANOBS 2013). By utilizing the display or chart, wind speed was estimated by visually determining the median value in the range of high and low excursions, excluding outliers, during the observing window, with gust being the highest instantaneous value. Automation provides the most consistent wind record available. Manual wind observations are prone to certain types of error, including rounding (Tuller 1980). However, daily maximum gust on the 78D typically has been estimated from the strongest reading during the 10-min observation window each hour; therefore, despite the automation, the daily maximum gust record is not as thorough as with chart-recording U2A systems which provide continuous coverage (Richards and Abuamer 2007).  Hourly and special observations for the study stations were supplied by EC, the U.S. National Climatic Data Center (NCDC) and supplemented via the use of the Plymouth State Weather Center archives (PSWC 2013; Table 2.2). When reporting information from stations other than the main three (e.g. CWEB, KBLI, TTIW1) the same sources were used. For the 78D era, hourly and special observations were used to determine high-wind events. Hourly records capture many events, but sometimes high-winds occur for just a short period between the regular observations. For example, during the period 1994 to 2008 at CYVR, a dataset with the special Chapter 2: Climatology of High-Wind Storms  28 observations removed captured 96% of events. This is an important consideration because the U2A wind record does not contain specials from 1964 to 1976. Some windstorms from 1994 to 2012 achieved high-wind status solely through peak gust. Gust is not available in the EC hourly record for the U2A period. Peak gust, however, is available in the daily data available online (EC 2013c). CYYJ, however, has no peak gust records from January-August of 1964. Table 2.2 Weather datasets used in this study. Organization Coverage Observation Frequency Wind Observations Available EC 1953-present Hourly Wind EC 1994-present Hourly and Special Wind and Gust EC 1953-present Daily Peak Gust NCDC 1977-2010 Hourly and Special Wind and Gust PSWC 1998-present Hourly and Special Wind and Gust   Instrumental and methodological changes, and the availability of certain types of data such as hourly gust and synoptic charts from the NCDC, is what determined the main time period of this climatological analysis: the 1994 to 2012, or roughly the 78D era. For some longer-period analysis, a windstorm dataset was also created from the U2A era for the years 1964 to 1993 and is combined with the 1994 to 2012 windstorm record.  2.2.3 Wind Speed Averaging   For some examinations, the peak wind measured at each of the three stations is averaged. This is termed average peak wind. The same is done for peak gust, here called average peak gust.  Chapter 2: Climatology of High-Wind Storms  29 2.2.4 Pressure Gradients  Pressure gradients were determined using sea-level pressure (SLP) data from the three study stations. A geostrophic wind triangle, using a method described in Stull (2000), provided a means of capturing the gradient field in two dimensions (Figure 2.1). A right triangle is necessary for the calculations, requiring the use of two interpolation points due to the arrangement of the three study stations. Interpolated pressures were based on linear estimation. The calculated direction of the geostrophic wind, designated α, is parallel to the isobars and therefore reflects the orientation of the pressure gradient field. The pressure slope (see Lange 1998 for a discussion of pressure slope) is 90º counter-clockwise from α. Pressure gradients are reported in terms of hPa (100) km-1, a convenient distance that happens to be close to the  Figure 2.1 The geostrophic wind triangle used in the calculation of pressure gradient magnitude and orientation. The two interpolation points required to make a right triangle are denoted IP1 and IP2. Some important geographical regions are also indicated. Chapter 2: Climatology of High-Wind Storms  30 separation of the stations (e.g. CYYJ to CYXX is 88.9 km). Pressure gradients were generated for the entire hourly record from January 1953 to July 2008. For all windstorms between July 2008 and December 2012, the gradients were calculated over a 36-hour period roughly centered at the time of CYVR minimum pressure.  2.2.5 Storm Track Determination and Climatology   Storm tracks for 1994 to 2012 were prepared from high-resolution surface and upper-air maps digitally archived at the NCDC (NCDC 2013a) and also provided in hardcopy form in some cases. Some maps are from the Weather Prediction Center (WPC 2013) historic archive, the NOAA Central Library Daily Weather Maps (NOAA 2013), and the EC Library microfilm collection. In the case of the event 01 Jan 1997, David Roth, NOAA Forecaster, provided key surface analysis charts. Maps covering the East Pacific were typically available at 6-h intervals. For the North American and United States domains, 3-h maps were common. The drawn tracks were largely based on 6-h low positions, with 3-h positions used as a guide when available. In just a few cases only 12-h positions could be found for part of the track, usually in the region to the west of the North American map domain. ETC initiation points are based on the first occurrence of a low pressure designation on a surface map, or, in the case of the re-energizing of an existing ETC, the point of highest reported central pressure before deepening ensues.  A grid based on 2.5º of latitude and longitude provided the basis for a climatological assessment of storm tracks. The approach is similar to Sanders and Gyakum (1980) though with a higher grid resolution and simpler averaging schemes. Storm tracks inside each quadrilateral were summed to determine a frequency of events in a gridded fashion. Given that a degree of longitude diminishes in length northward, there is a difference in grid cell area between lower Chapter 2: Climatology of High-Wind Storms  31 and higher latitudes. This difference is approximately a factor of 2 between a quadrilateral at 30.0 to 32.5º N and one at 62.5 to 65.0º N. In essence, there is a disparity in area of approximately 1.5 times between the center band of quadrilaterals along 45 to 50ºN, where most of the ETCs in this study tended to track, and the ones at the extreme ends of the depicted region, 30ºN and 65ºN. Adjustments were made via correcting total counts by differences in square area, but this did not materially change spatially averaged track positions. Therefore, the raw unadjusted values are presented.   For drawing spatially averaged storm tracks, a smoothing filter is used consisting of the simple mean of the frequencies in a 3x3 grid with the tabulated value placed in the middle quadrilateral. General ETC tracks were determined using the smoothed values. The mean paths follow the quadrilaterals with the highest averages, or go between two cells where the averages are approximately equal.  A system of 'bins' that encompass prominent geographic regions is used to track landfall locations, similar to Keim and Muller (2007). However, each bin is not intended to capture every major storm that afflicted the region inside the bin, but simply mark the locations of ETCs that brought high winds to the three study stations. Bin boundaries were placed on easily recognizable landforms such as the endpoints of large islands. For large gaps in the coastal terrain, such as the Queen Charlotte Sound, a line that roughly parallels the mainland coast is drawn from the extreme ends of coastal landmasses. The bins are roughly the same length, though they were designed to be longer further north where fewer high-wind storms (for the study region) track. There are also two off-shore bins to capture ETCs that were absorbed before making landfall, and a line of bins in the BC interior to capture the position of lows that develop overland. Chapter 2: Climatology of High-Wind Storms  32 To gain a large sample for a regression analysis of the relationship between ETC landfall central-pressure and the associated average peak gust speeds at the three study stations, surface analysis maps were examined to find all lows, regardless of synoptic structure and associated wind speed magnitude, that tracked across northern Vancouver Island from January 2008 to April 2013. Selected events were also limited to track directions in the quadrant between north and east. North Vancouver Island systems were chosen because these tracks are relatively close to the study region, helping to ensure that surface winds are strongly influenced by the cyclonic structure of each storm, but are not so close as to have the complication of low centers passing right over any of the three stations (e.g. southern Vancouver Island tracks). A focus on tracks with north to east directions helps limit the effects of storm motion on wind speed—e.g. a storm tracking southeast is expected to have the southerly component of its direction vector reduce the S vector component of wind.  2.2.6 ETC Central Pressures and Bomb Cyclogenesis  Central pressures were taken from the surface maps. Cyclogenesis/cyclolysis central-pressure statistics were strictly examined for cyclones that developed over the ocean (here termed ocean cyclones) and only for those events that had enough good data to generate numbers with some degree of confidence. Some ETCs underwent most of their development cycle in under a day, preventing a genuine calculation of the 24 h rate of deepening. These events have the same value for 24 h and absolute rate of deepening. An ETC is considered to have undergone explosive cyclogenesis (also termed a "cyclogenic bomb") if it deepens at the rate of one hPa h-1 for 24 h or more (Sanders and Gyakum 1980). Due to changes in the strength of the Coriolis effect, the threshold deepening Chapter 2: Climatology of High-Wind Storms  33 rate varies by latitude. Sanders and Gyakum (1980) adjusted the bomb cutoff using sinϕ/sin60, where 60º latitude is taken to be the reference. Given that many of the cyclones in this study tracked between 45 to 55º N, 21 hPa (24) h-1, the approximate value for 50º N, is used.  2.2.7 Mean Return Intervals  A detailed examination of extreme wind return frequency using the method of independent storms with the hourly dataset from CYVR and some other stations in the Georgia Strait was done by Abeysirigunawardena et al. (2009). Other approaches, such as the Gumbel, have been utilized for high wind return frequencies, such as in the BC interior (Murphy and Jackson 1997). These approaches are not used here, largely because the focus of this study is on high-winds and discrete windstorms. The wind threshold is so great that it limits the usefulness of the independent storms approach (e.g. Abeysirigunawardena et al. used a 40 km h-1 cutoff), and the Gumbel method is typically applied to annual extremes, not discrete events. Instead, mean return intervals in terms of events per year are presented, an approach used by Keim and Muller (2007).   2.2.8 Long-Term Pacific Climate Variability  Monthly NPI data is from Trenberth and Hurrell (2013), and the monthly PDO index from Mantua (2000). Since windstorms are largely concentrated in the cold-season months, the average of the January to March and October to December index values are used to create a value for each year. Five-year averages of the annual numbers were then computed and compared to corresponding five-y mean windstorm return intervals via linear regression. Since the starting point of the five-y cutoffs is arbitrary, this analysis used five separate half-decade Chapter 2: Climatology of High-Wind Storms  34 sets: the first set starts with 1964 to 1968, with each subsequent set stepped up one year (e.g. the next set starts with 1965 to 1969). Half-decades were chosen to provide enough time to capture a significant number of windstorm events, but also be short enough to yield more than just a few data-points. Also, though a specific PDO polarity (warm/cold) tends to have dominance over intervals on the order of 20 to 30 y, there is variability in the indexes over smaller time scales (Mantua et al. 1997); five-y intervals offer some sensitivity to shorter-period fluctuations.  2.3 Results  2.3.1 General Statistics of Southwest BC Windstorms  For CYYJ, CYVR and CYXX, 58 discrete high-wind events occurred between January 1994 and December 2012 (Table 2.3). The filter for maximum wind returned 49 events, with the maximum gust filter adding an additional nine. Of the additional nine, seven occurred before the 78D became fully implemented at all three stations (1994 to 2005) and two occurred after (2006 to 2012). Between the two periods, there is a two-fold difference in the rate at which storms produced a high-wind criteria peak gust without producing a high-wind criteria peak wind. In terms of wind-directional categories, two key storm types are apparent: 27 of the storms had peak speeds from a west to northwest direction at CYVR, here termed westerly wind events, and there were 21 southeasters. The southeaster category includes two storms that produced peak winds from 90º; they are classified based on the wind direction response of lower-than-maximum winds at CYVR and the peak wind response at CYYJ and CYXX. Wind maxima for southeasterly storms are usually southerly at CYXX and southeast at CYYJ, though on occasion both stations had peak winds from the SSW to SW during southeasters. During  Chapter 2: Climatology of High-Wind Storms  35 Table 2.3 Basic statistics of the 58 high-wind storms that affected CYYJ, CYVR and CYXX during 1994 to 2012. In cases where peak gust is not reported, the value is estimated from peak wind using a 1.3 gust factor. Estimated gust speeds are denoted in italics. Peak Wind           Date (LST) Peak Wind t (LST) Peak Wind Dir Desig-nation CYVR WDIR (º) CYVR WSPD (m s-1) CYYJ WDIR (º) CYYJ WSPD (m s-1) CYXX WDIR (º) CYXX WSPD (m s-1) CYVR GSPD (m s-1) CYYJ GSPD (m s-1) CYXX GSPD (m s-1) 3-Sta AVG WSPD (m s-1) 3-Sta AVG GSPD  (m s-1) Max Pres Grad CYYJ-CYVR-CYXX (hPa [100] km-1) Max Pres Grad CYYJ-CYVR-CYXX α (º) 08Nov94 1700 SE 100 15.4 130 14.4 50 6.7 24.7 21.6 8.7 12.2 18.3 4.0 150 09Mar95 1610 S 170 14.9 210 14.9 170 17.5 23.1 24.2 26.2 15.8 24.5 4.1 185 17Nov95 2200 SE 160 15.4 230 12.3 180 12.9 25.2 19.0 26.8 13.5 23.7 7.4 256 04Dec95 0200 W 290 19.5 300 5.7 240 10.3 29.8 7.2 20.6 11.8 19.2 4.3 307 12Dec95 2300 SE 100 10.3 230 13.4 190 10.3 15.4 28.8 24.7 11.3 23.0 6.9 225 01Jan97 0600 S 190 11.3 210 16.5 200 13.9 14.9 25.7 20.6 13.9 20.4 4.3 183 01Mar97 0600 SE 140 9.8 150 10.3 120 17.5 14.9 15.4 22.6 12.5 17.6 4.5 167 30Mar97 1700 SE 150 18.0 150 12.9 180 15.4 28.8 18.5 24.2 15.4 23.8 5.6 173 03Apr97 0800 W 300 19.5 300 7.2 210 6.2 27.8 15.4 8.7 11.0 17.3 4.2 287 24Nov98 0300 SE 140 16.5 120 12.9 170 11.8 23.7 21.6 25.7 13.7 23.7 7.6 224 17Dec98 1300 W 290 18.0 300 9.3 280 10.3 22.6 16.5 15.4 12.5 18.2 4.2 287 29Jan99 1100 W 160 13.4 210 9.3 210 15.4 20.1 15.4 26.8 12.7 20.8 3.2 198 02Feb99 1600 SE 270 17.5 220 14.9 210 12.3 22.1 20.6 18.5 14.9 20.4 5.0 255 05Feb99 1700 SE 160 17.5 130 15.4 180 10.3 25.2 25.7 19.5 14.4 23.5 5.2 147 03Mar99 0600 SE 130 17.5 130 15.4 170 14.4 22.1 24.7 21.6 15.8 22.8 9.0 227 25Sep99 0400 W 290 17.5 210 9.8 250 13.9 25.2 13.4 21.6 13.7 20.1 5.7 241 04Nov00 1000 W 290 17.5 260 7.7 230 7.7 22.6 10.3 18.0 11.0 17.0 4.5 262 15Dec00 0500 W 290 22.1 260 12.9 270 12.9 24.7 20.1 26.2 16.0 23.7 4.1 265 29Jan01 1000 W 280 19.0 140 8.7 220 12.9 21.6 10.3 16.5 13.5 16.1 4.5 258 29Nov01 1100 W 280 18.5 140 9.8 250 6.2 22.6 15.4 8.2 11.5 15.4 3.5 241 14Dec01 0400 W 300 22.6 210 13.9 270 14.4 26.8 15.4 27.8 17.0 23.3 5.9 251 14Apr02 0026 SW 260 13.4 250 14.4 230 15.4 24.7 20.6 28.3 14.4 24.5 5.9 301 22Apr02 2100 W 300 18.5 280 11.8 240 9.3 25.2 12.9 13.4 13.2 17.2 4.4 273 28Oct03 1900 W 290 22.6 250 8.7 260 9.3 26.2 14.9 15.4 13.5 18.8 5.0 286 27Apr04 1300 W 300 20.1 300 10.3 190 7.2 23.1 17.0 10.8 12.5 17.0 4.9 271 09Sep05 0000 W 280 18.0 130 5.7 80 6.2 22.1 7.2 8.2 10.0 12.5 4.9 251 31Jan06 1700 SE 90 13.4 130 15.9 190 12.9 19.0 26.8 17.0 14.1 20.9 5.3 164 04Feb06 0943 SE 140 16.5 160 12.3 180 19.0 20.6 18.5 25.2 15.9 21.4 7.6 226 13Feb06 1300 W 290 18.5 320 9.3 230 8.7 22.1 14.4 14.9 12.2 17.1 3.2 279 08Mar06 2142 W 320 18.5 260 11.3 190 13.4 23.7 15.9 19.5 14.4 19.7 4.7 242 15Nov06 1300 SE 160 15.9 260 11.8 180 20.6 21.1 19.0 27.3 16.1 22.5 4.2 276 11Dec06 1400 SE 150 18.0 140 12.9 170 24.2 23.7 21.1 30.4 18.4 25.1 5.9 173 13Dec06 1100 SE 90 12.3 250 18.5 200 14.9 15.4 27.3 18.5 15.2 20.4 5.4 155 Chapter 2: Climatology of High-Wind Storms  36 Peak Wind           DD MM YYYY (LST) Peak Wind HH (LST) Peak Wind Dir Desig-nation CYVR WDIR (º) CYVR WSPD (m s-1) CYYJ WDIR (º) CYYJ WSPD (m s-1) CYXX WDIR (º) CYXX WSPD (m s-1) CYVR GSPD (m s-1) CYYJ GSPD (m s-1) CYXX GSPD (m s-1) 3-Sta AVG WSPD (m s-1) 3-Sta AVG GSPD  (m s-1) Max Pres Grad CYYJ-CYVR-CYXX (hPa [100] km-1) Max Pres Grad CYYJ-CYVR-CYXX α (º) 15Dec06 0234 W+S 290 23.1 270 15.4 180 18.5 26.2 21.6 25.2 19.0 24.3 7.8 258 05Jan07 2237 W 300 23.7 220 11.3 190 15.9 31.4 20.1 19.5 17.0 23.7 5.5 256 09Jan07 1500 W 280 21.6 210 13.9 230 19.0 27.3 20.6 24.7 18.2 24.2 4.1 281 12Nov07 0939 SE 130 17.0 140 15.9 180 22.6 21.1 24.7 30.9 18.5 25.6 5.4 155 14Jan08 2000 W 300 19.5 270 9.8 260 12.3 24.7 14.4 18.5 13.9 19.2 5.3 275 13Nov08 0800 W 290 19.0 170 10.3 250 10.3 23.1 17.5 14.9 13.2 18.5 4.4 269 21Nov08 1846 SE 160 12.9 150 9.3 160 19.0 15.4 14.9 23.1 13.7 17.8 4.0 179 15Dec08 1200 E 80 6.2 60 14.9 50 20.1 8.2 20.1 26.8 13.7 18.4 2.5 187 29Dec08 1400 W 270 19.5 260 9.8 190 15.4 23.1 15.4 18.0 14.9 18.8 3.4 263 20Mar09 1515 SW 220 10.8 250 12.3 250 18.0 12.9 17.0 23.1 13.7 17.7 3.9 253 23Oct09 2100 W 270 17.5 130 7.2 200 6.7 20.6 11.3 11.3 10.5 14.4 4.9 251 05Nov09 1700 SE 160 14.9 160 10.8 180 18.5 15.9 15.4 26.2 14.7 19.2 3.1 250 16Nov09 0250 SE 150 12.3 150 10.8 180 17.5 17.5 14.9 25.2 13.5 19.2 3.3 190 16Nov09 2340 S 230 11.3 220 12.3 180 20.1 15.9 18.5 28.3 14.6 20.9 3.5 171 18Jan10 0400 SE 140 18.0 130 15.9 170 12.3 22.6 23.1 17.5 15.4 21.1 5.5 157 02Apr10 1400 SE 140 19.0 140 13.9 180 18.5 24.7 23.1 22.1 17.1 23.3 7.4 231 08Apr10 0200 W 290 21.6 260 9.8 240 12.9 25.7 17.0 15.9 14.8 19.5 5.1 253 15Nov10 2000 W 260 16.5 250 15.4 230 12.9 19.5 25.2 19.5 14.9 21.4 4.8 262 22Nov10 1900 E 40 7.2 50 12.3 50 17.5 10.8 19.0 25.7 12.3 18.5 4.7 64 02Mar11 1014 SW 230 14.4 130 13.4 180 20.1 18.0 21.1 26.2 16.0 21.8 5.1 168 11Nov11 1411 W 300 18.5 300 10.8 230 11.8 27.8 19.0 17.0 13.7 21.3 6.6 269 25Dec11 1226 SW 220 6.7 230 10.8 170 19.0 12.9 16.5 26.8 12.2 18.7 5.1 250 10Jan12 0100 W 290 18.5 300 5.7 200 5.1 22.1 7.2 6.7 9.8 12.0 4.5 253 25Feb12 0900 W 280 19.0 140 9.3 210 11.3 22.6 14.4 14.9 13.2 17.3 5.9 256 12Mar12 0900 SE 140 18.0 130 18.5 180 11.3 24.7 26.2 17.5 15.9 22.8 6.0 160  westerly windstorms, peak winds occurred from a variety of directions at both CYXX, ranging from E to S to SW and W, and CYYJ, ranging from SE to S to W and NW. One event, the major 2006 Hanukkah Eve windstorm, qualifies for both the west and south categories given that high winds occurred out of the S at CYXX ahead of the low followed by winds of even stronger magnitude out of the W at CYVR. This windstorm excluded from most summaries and Chapter 2: Climatology of High-Wind Storms  37 comparisons of southeasterly and westerly events, though its inclusion in either category would not materially change the results. Of the remaining nine storms, four are southwesterly, three are southerly, and two are Fraser outflow events. Given the limited number of examples in these three categories, they do not lend themselves to thorough analysis like westerly and southeasterly windstorms. The southwester category has one storm, 14 Apr 2002, that produced maximum winds out of 260º at CYVR. The decision to put the event in the southwest bin instead of west is based on peak wind out of the 250º at CYYJ and 230º at CYXX, with higher maximum speeds at these two stations than at CYVR. This storm is one example of a borderline case and a reflection that these categories are imperfect. Table 2.4 Wind and pressure gradient data for the 1962 Columbus Day Storm and the most significant windstorms (wind ≥20 m s-1 and/or gust ≥30 m s-1 ) from the 1964 to 1993 U2A era. Gust values in italics are estimated from peak wind via a 1.3 gust factor. Storm Date Peak Wind (m s-1) Peak Gust (m s-1) Peak Wind Direction (º) Max pres gradient (hPa [100] km-1) CYVR CYYJ CYXX Avg CYVR CYYJ CYXX Avg CYVR CYYJ CYXX 12Oct62 25.0 21.4 24.7 23.7 35.0 27.8 40.3 34.4 160 110 180 11.2 15Dec64 18.1 13.4 22.4 18.0 23.3 19.2 33.6 25.4 320 50 50 5.5 01Jan65 15.6 13.4 19.8 16.3 19.7 18.9 31.4 23.3 140 160 180 4.5 05Oct65 10.7 8.0 22.0 13.6 16.7 8.3 28.6 17.9 180 270 160 2.9 03Dec65 8.5 6.7 22.0 12.4 12.5 9.7 30.0 17.4 180 140 160 4.4 29Jan67 14.3 8.9 22.4 15.2 22.8 17.8 24.7 21.8 290 270 180 4.0 05Dec67 15.6 15.2 17.0 15.9 25.8 30.3 24.2 26.8 180 160 180 8.5 27Feb72 12.5 18.1 12.5 14.4 19.2 30.3 17.8 22.4 220 230 190 4.0 30Mar75 21.2 10.3 11.2 14.2 30.0 17.8 15.6 21.1 280 290 240 6.8 18Dec90 6.7 12.3 20.6 13.2 9.7 18.1 31.9 19.9 70 20 50 4.6  The maximum 2-min wind speed among all the storms is 24.2 m s-1 in the case of CYXX during the 11 Dec 2006 event, and the maximum 5-sec gust, 31.4 m s-1, occurred during a windstorm on 05 Jan 2007 at CYVR. Among the 58 windstorms, 45 produced high winds at just one of the three study stations and the remaining 13 affected two. None affected all three stations Chapter 2: Climatology of High-Wind Storms  38 with high winds. However, looking before 1994, one event did cause high-wind criteria speeds at all the study sites: the 12 Oct 1962 storm (Table 2.4). For comparison with 1994 to 2012, 1964 to 1993 had a peak wind speed of 22.4 m s-1 on two occasions and a peak instant gust speed of 33.6 m s-1, all occurring at CYXX. Table 2.5 Pearson correlation coefficients using wind and pressure-related variables for all windstorms lumped together (n=58). Variable Peak pressure gradient α of peak pressure gradient CYVR peak wind CYYJ  peak wind CYXX peak wind CYVR peak wind 0.23 0.53  -0.22 -0.36 CYVR peak gust 0.31 0.46 0.87 -0.17 -0.36 CYVR peak wind direction -0.15 0.78 0.65 -0.55 -0.43 CYYJ peak wind 0.32 -0.51   0.42 CYYJ peak gust 0.37 -0.48 -0.22 0.91 0.35 CYYJ peak wind direction -0.10 0.69 0.36 -0.30 -0.28 CYXX peak wind 0.05 -0.35    CYXX peak gust 0.16 -0.27 -0.36 0.43 0.86 CYXX peak wind direction 0.01 0.64 0.52 -0.20 -0.23 three-station peak wind -0.03 0.17    three-station peak gust 0.19 -0.15    three-station avg peak wind 0.33 -0.16    three-station avg peak gust 0.47 -0.20    Peak pressure gradient  0.01     Peak pressure gradients determined from the hourly SLP readings of CYYJ, CYVR and CYXX (Table 2.3) ranged from a rather low 2.5 hPa (100) km-1 during a Fraser outflow event to a maximum of 9.0 hPa (100) km-1 during a strong southeasterly windstorm. Southeasters had an average peak pressure-gradient of 5.6 hPa (100) km-1 (n=21, range 3.1 to 9.0, SD=1.7) and westerly windstorms 4.7 hPa (100) km-1 (n=27, range 3.2 to 6.6, SD=0.8). The six highest gradients are associated with southeasters—though the 2006 Hanukkah Eve Storm, if included in the westerly windstorm category, would bring a data point into this upper range. Chapter 2: Climatology of High-Wind Storms  39 For all 58 high-wind events, the geostrophic wind direction for the maximum pressure gradient, α, ranged from 64 to 307º (Table 2.3). For southeasters, the range was 136 to 276º with a non-vector-based average of 194º (n=21, SD=39.7), corresponding to a 104º pressure slope. Westerly windstorms had a range of 241 to 307º with an average of 265º (n=27, SD=16.1), corresponding to a 175º pressure slope.   Figure 2.2 Timing of peak wind compared to the timing of peak surface pressure gradient for the 58 windstorms 1994 to 2012. Calculations based on the nearest hour. Negative values represent storms that had a wind maximum ahead of the gradient maximum.  Correlation between maximum pressure gradient and peak wind, peak gust, average peak wind and average peak gust shows weak to no relationship, with R-values typically below 0.4 and sometimes close to zero (Table 2.5). When considering these correlations, keep in mind that peak wind and maximum gradient do not always occur at the same time (Figure 2.2). Thirty-one storms had peak winds occur within two h of pressure gradient maximums with a strong peak at zero to one h, but there is also a wide range, from -12 to 27 h. Peak wind direction compared to the orientation of the pressure gradient field (α) have some of the strongest correlations for all Chapter 2: Climatology of High-Wind Storms  40 stations. Interestingly, α had stronger correlation with peak wind and gust than the strength of the pressure gradient. CYVR also showed some relationship between peak wind speed and peak wind direction. When peak wind direction is isolated, some R2-values for wind speed and pressure gradient comparisons improved (Table 2.6). Maximum pressure gradient and three-station average peak gust among the 21 southeasters had an R2 of 0.35 (n=21), shown in Figure 2.3. Interestingly, for the 27 westerly windstorms, the R2 for the same variables is 0.07 (n=27). Table 2.6 Pearson correlation coefficients using wind and pressure-related variables for southeasters (n=21) and westerly windstorms (n=27). Variable Southeaster R for Max Pres Gradient Westerly Windstorm R for Max Pres Gradient CYVR peak wind 0.41 0.12 CYVR peak gust 0.38 0.28 CYYJ peak wind 0.40 0.16 CYYJ peak gust 0.51 0.12 CYXX peak wind -0.13 0.21 CYXX peak gust 0.06 0.21 three-station peak wind -0.04 0.12 three-station peak gust 0.07 0.30 three-station avg peak wind 0.27 0.21 three-station avg peak gust 0.59 0.26  It is important to keep in mind that the pressure gradients, and all statistics, are strictly for high-wind storms that fit the selection criteria for this study. Many non-high-wind generating storms have produced gradients ≥2.5 hPa (100) km-1. Using a cutoff of 6.0 hPa (100) km-1, 12 independent ETCs reached or exceeded this threshold from January 1994 to July 2008. Of these storms, four did not produce high-winds at the three key stations (Table 2.7). However, all of these events did generate wind gusts above 20 m s-1 at one or more of the observation sites. The Chapter 2: Climatology of High-Wind Storms  41 16 Jan 2000 storm triggered severe winds south of the region, including at Bellingham (KBLI), just 63 km to the southeast of CYVR, where a gust to 29.3 m s-1 occurred.   Figure 2.3  Linear regression between maximum pressure gradient and three-station average peak gust for 21 southeasters and 27 westerly windstorms from 1994 to 2012. The R2 for southeasters is 0.35 and for westerly windstorms 0.07.  Table 2.7 All ETCs with a ≥6.0 hPa (100) km-1 pressure gradient that did not reach high-wind status for the period January 1994 to June 2008. Event Peak Wind (m s-1) Peak Gust (m s-1) Peak Wind Direction (º) Max pres gradient (hPa [100] km-1) CYVR CYYJ CYXX Avg CYVR CYYJ CYXX Avg CYVR CYYJ CYXX 31Oct94 10.8 15.6 7.2 11.2 18.1 23.6 12.2 18.0 200 250 210 6.8 17Oct96 11.9 12.8 7.8 10.8 17.5 21.7 9.7 16.3 90 120 90 6.2 16Jan00 15.0 12.8 12.8 13.5 19.2 20.6 23.1 21.0 140 130 190 6.8 16Nov02 11.9 8.3 11.4 10.6 15.0 13.9 21.1 16.7 140 240 180 6.1  Chapter 2: Climatology of High-Wind Storms  42 2.3.2 Cyclone Statistics  Of the 58 storms, 57 were triggered by ETCs directly affecting the study region and one by the migration of strong high pressure over the Pacific into the region, creating a "pushed wind" event (Lange 1998 describes pushed winds). Typically the lows were discrete occluded systems that reached the coast at or just past peak intensity. Most ETCs decayed as they tracked inland, apparently filling in response to ageostrophic surface flow through mountain gaps, coastal inlets and valleys. A few storms maintained strength or actually intensified despite interacting with rugged coastal terrain. These ETCs tended to have an open-wave form with the low center near the intersection of the warm and cold fronts (i.e. approximately Stage I of Shapiro and Keyser 1990), sometimes called triple-point or point-of-occlusion lows. A few of the high-wind-generating lows were weakly defined, secondary-spinup lows along persistent baroclinic bands (i.e. atmospheric rivers). Occasionally, lows completely fell apart before strongly affecting the study region, with a resulting "beheaded front" being the trigger for high winds. Incoming high-pressure cells played a critical role in the resulting windstorms by creating a pushed wind. Some pushed events were aided by the development of a low in the central or southern British Columbia interior that helped enhance and maintain a strong west to southwest pressure slope over southwest BC. A few windstorms were triggered by spontaneous low formation just off the coast, with the ETC rapidly tracking inland in just a few hours—an important phenomenon since these small and quickly-developing depressions can be difficult to forecast. Six of the ETCs tracked in a north direction when at closest approach to the study area (Table 2.8). A majority 24 ETCs tracked in a north-northeast to northeast direction, with 17 tracking east-northeast to east and nine following east-southeast to south-southeast paths. The   Chapter 2: Climatology of High-Wind Storms  43 Table 2.8 General track and central pressure statistics for the 58 events that produced high winds in the study region. Storm track designation locations are shown in Figure 2.7. Storm Event (LST) Storm Track Desig-nation General Track Direction at Closest Approach to CYVR ETC Min Central Pres (kPa) ETC Central Pres at Landfall (kPa) ETC Max Deepen-ing Rate (hPa [3] h-1) ETC Max Filling Rate (hPa [3] h-1) ETC Max Deepen-ing Rate (hPa [24] h-1) ETC Absolute Central Pressure Change (hPa) 08Nov94 NEPAC NE 96.8 NA 8 10 40 44 09Mar95 NVI NE 96.4 96.9 7 8 22 22 17Nov95 NVI NNE 98.8 98.8 4  15 15 04Dec95 OLY ENE 98.6 99.2 4  18 20 12Dec95 OLY NE 95.3 96.1 6 11 33 33 01Jan97 SQCS ENE 95.9 96.6 12 7 18 23 01Mar97 SQCS ESE 97.4 98.6 5 5 24 38 30Mar97 SVI NNE 98.7 99.2 6 5 20 25 03Apr97 KEAK E 97.7 98.9 6 15 13 25 24Nov98 SVI NNE 96.2 97.8 10 11 31 35 17Dec98 NHG ENE 99.8 101.2 4 1 18 20 29Jan99 SQCS N 98.5 98.5 4 2 8 13 02Feb99 NHG E 98.0 98.3 3 2 5 5 05Feb99 GAK NNE 94.9 NA 7 4 30 49 03Mar99 SVI NE 96.8 97.6 15 9 38 45 25Sep99 NVI E 99.2 99.2 6 6 11 12 04Nov00 CBCIN SSE or E 99.6 NA     15Dec00 SVI E 99.1 99.1 6 4 11 11 29Jan01 NVI E 99.2 99.2 3 3 12 16 29Nov01 SVI ENE 97.1 98.5 5 5 27 29 14Dec01 SVI ENE 97.9 97.9 6 1 8 8 14Apr02 NHG SE 98.1 98.3 7 6 14 14 22Apr02 PUSHED NA NA NA     28Oct03 SEAKS SE 98.3 100.8 5 5 24 31 27Apr04 SBCIN E 98.6 NA     09Sep05 EWA E 99.2 NA     31Jan06 SHG NNE 96.3 96.3 8 6 13 19 04Feb06 NVI NE 97.0 97.3 14 12 38 38 13Feb06 SBCIN SSE 102.3 NA     08Mar06 SQCS NNE 97.6 98.5     15Nov06 NVI E 97.6 98.2 7 7 17 20 11Dec06 NVI NE 97.8 97.8 8 9 9 18 13Dec06 NHG ENE 95.7 96.9 7 7 18 41 15Dec06 SVI NE 97.4 97.6 7 6 20 32 05Jan07 CBCIN E 99.3 99.3     09Jan07 NQCS SE 98.9 100.9 7 1 23 33 Chapter 2: Climatology of High-Wind Storms  44 Storm Event (LST) Storm Track Desig-nation General Track Direction at Closest Approach to CYVR ETC Min Central Pres (kPa) ETC Central Pres at Landfall (kPa) ETC Max Deepen-ing Rate (hPa [3] h-1) ETC Max Filling Rate (hPa [3] h-1) ETC Max Deepen-ing Rate (hPa [24] h-1) ETC Absolute Central Pressure Change (hPa) 12Nov07 NVI NNE 96.7 96.7 9 4 31 39 14Jan08 SEAKS N 98.2 98.6 7 9 13 18 13Nov08 SEAKS ENE 97.0 97.4 4 5 11 11 21Nov08 SHG NNE 97.6 97.6 10 8 29 29 15Dec08 NWCA SSE 99.8 100.5     29Dec08 SVI NE 99.8 99.8 4 5 6 6 20Mar09 CBCIN NNE 99.6 100.2     23Oct09 KOAK N 96.1 96.1     05Nov09 GAK NE 94.4 NA     16Nov09 NVI NE 99.2 99.7 3 4 8 12 16Nov09 SQCS NNE 98.0 98.0 4 6 13 13 18Jan10 NQCS N 97.2 98.0 6 4 18 19 02Apr10 SVI NE 97.6 97.7 10 6 33 36 08Apr10 SHG NE 99.8 100.0 2 1 8 16 15Nov10 SHG ESE 100.0 100.1 6 5 16 16 22Nov10 OLY SE 99.1 99.8     02Mar11 SQCS N 96.2 98.2 9 11 27 46 11Nov11 SEAKN SE 98.8 99.0 6 3 16 16 25Dec11 SVI NNE 100.7 100.7     10Jan12 NQCS NE 100.0 100.2 3 9 9 10 25Feb12 NQCS E 99.5 99.6 5 2 11 18 12Mar12 NVI N 96.2 96.2 8 10 41 49  remaining storm tracked nearly due south and then banked sharply to the east when at closest approach to the Lower Mainland, putting it in an unusual category. The deepest central pressure among the ETCs, 94.4 kPa, occurred with the low located well offshore. Landfall central pressures for 50 ETCs ranged from 96.1 kPa upwards to 101.2 kPa (Figure 2.4), with a mean of 98.5 kPa (SD=1.3). The remaining ETCs did not make landfall. Of the 50 land-falling ETCs, 16 were at maximum depth when the low reached the coast and an additional 15 had central pressure within 5 hPa of minimum. Chapter 2: Climatology of High-Wind Storms  45 For 45 of the ETCs, ocean cyclones with good data, maximum 3-h rates of deepening ranged from 2 to 15 hPa (3) h-1, with an average of 6.5 (SD=2.8). Maximum rates of filling, usually post-landfall, were similar in magnitude, ranging from 1 to 15 hPa (3) h-1 with an average of 6.0 (SD=3.3). Note that this is the peak rate of intensification and weakening, not the average rate for each ETC. Also, for all the statistics relating to central pressure, standard deviations are generally quite high. This likely reflects, in part, significant uncertainty in central pressure determinations over the Northeast Pacific. Using the adjusted cutoff for one Bergeron, 21 hPa (24) h-1 (i.e. 0.875 hPa h-1), it is clear that many high-wind events are triggered by storms that do not reach explosive cyclogenesis status. Of the 45 ocean ETCs, only 16 deepened at the average rate of 0.875 hPa h-1, or more, for 24 h. Twenty-one deepened ≥21 hPa during their entire lifetimes. Some of the ETCs went Figure 2.4 Frequency of central pressures at landfall for 50 ETCs that produced high winds at the study stations. Chapter 2: Climatology of High-Wind Storms  46 through much of their development cycles in under a day, allowing too little time to calculate a pressure change over 24 h. Of the top five windstorms 1994 to 2012 based on three-station average peak gust, only two qualified as cyclogenic bombs. The strongest bomb event among the 45 ETCs exhibited an intensification rate of 41 hPa (24) h-1, or about 1.9 Bergerons.  Figure 2.5 Scatterplot for ETC central pressure at landfall and three-station average peak gust. The open and filled diamonds depict all high-wind generating cyclones that tracked across NVI from 1994 to 2012 and the dashed line shows the linear regression best-fit for these selected events (n=10, R2=0.45). The filled circles and filled diamonds indicate all lows, regardless of strength, that tracked across Northern Vancouver Island (NVI) from January 2008 to April 2013 and the solid line indicates the linear regression best-fit (n=50, R2=0.62). Note that two events are shared between the datasets.  Chapter 2: Climatology of High-Wind Storms  47 There appears to be a relationship between ETC central pressure at landfall and the average of peak gusts for the three stations—when track location is controlled for. For the case of the 10 high-wind ETCs that tracked across north Vancouver Island (NVI) between 1994 and 2012, the coefficient-of-determination between central pressure at landfall and three-station average peak gust is 0.45 (Figure 2.5). For a broader examination, all lows that tracked across NVI were identified for the time period January 2008 to April 2013. This returned 50 discrete events, including two high-wind storms present in this study. The average low tracking across NVI had a 99.7 kPa central pressure (Range 96.1 to 102.1, SD=1.3) at landfall, depressed the CYVR barometer to 100.7 kPa also at landfall (Range 97.6 to 102.8, SD=1.0), resulting in a 1.0 kPa pressure gradient (Range 0.05 to 2.95, SD=0.62) across Vancouver Island that contributed to a three-station average peak gust of 14.4 m s-1 (Range 6.7 to 22.6, SD=3.6). For the 50 NVI lows, a linear regression between central pressure and three-station average peak gust resulted in an R2 of 0.62. The line has a steeper slope than the one that fits the 10 high windstorms.  2.3.3 ETC Tracks for Southwest BC Windstorms  Given a sparse network of regular reporting stations, ETC center positions over the Pacific are not exact. This appears so even with the aid of satellite, ship and buoy observations. A quick comparison (not shown) of a few ETC track determinations for some of the storms in this study between two different meteorological agencies, the National Oceanic and Atmospheric Administration (NOAA) and EC, revealed that cyclone path determinations can differ by 1º of latitude—even higher for longitude—and in some cases can approach 1.5º. Track agreement appears to get better near the coast, where regular observing stations including coastal buoys aid in the placement low centers. Chapter 2: Climatology of High-Wind Storms  48 A wide variety of ETC tracks have produced high winds in southwest BC (Figure 2.6). Even though 58 discrete windstorms have been identified during the study period, some 62 tracks are shown. This is because five windstorms, mainly westerly events, were supported by two separate cyclones in scenarios where a decaying low "hands off" energy or frontal systems to a newly developing low just downstream. The pushed event does not have an ETC track. A broad swath of the Northeast Pacific is covered by ETCs that produced high winds in the Lower  Figure 2.6 The tracks for 62 ETCs associated with 57 independent high-wind events that occurred in the Greater Vancouver and Victoria areas, BC, from 1994 to 2012. Solid-black lines indicate southeasters, dashed-black westerly windstorms, solid-gray southerly and southwest events and dashed-gray Fraser outflow events. The single pure "pushed" event has no track. Chapter 2: Climatology of High-Wind Storms  49 Mainland and Greater Victoria region. Such storms have landed from CA to Alaska's (AK) Kodiak Island. Nevertheless, tracks tend to cluster fairly close to the region of interest. There appears to be at least one strong constraint: The ETCs generally tracked to the north of, or right over, the study region's latitude. There are just four exceptions. Looking at landfall points (Figure 2.7, Table 2.9), the south Vancouver Island (SVI) and NVI bins each contain 10 tracks, converting to a mean recurrence interval of about one high-  Figure 2.7 ETC landfall frequency—the number of ETCs that crossed the boundary on the left side of the polygons—for the 19-y period. The mean return interval in y is in brackets. For the polygons along the coast, this can be considered the landfall frequency, though technically for lows entering bodies of water such as the Queen Charlotte Sound, landfall has not quite occurred. Results are for the primary cyclones, exluding secondary developments. Chapter 2: Climatology of High-Wind Storms  50 Table 2.9 For given track designations, the frequency of high-windstorms in the study region. Results are for the primary cyclones, excluding secondary developments, and are broken down by peak wind direction. Storm Track Designation SE W S - SW NE - E Total Olympic Landfalls 1 1 0 1 3 South Vancouver Island 4 5 1 0 10 North Vancouver Island 7 2 1 0 10 South Queen Charlotte Sound 2 1 3 0 6 North Queen Charlotte Sound 1 3 0 0 4 South Haida Gwaii 2 2 0 0 4 North Haida Gwaii 1 2 1 0 4 Landfalls Dixon Entrance North 0 6 0 0 6 Gulf of Alaska and NE Pacific Lows 3 0 0 0 3 BC Interior and Eastern Washington 0 5 1 0 6 Northwest California 0 0 0 1 1 Pure Pushed Events 0 1 0 0 1 Total 21 28 7 2 58  wind storm every two y for each region, and one event per annum for the entire island. Another 10 events tracked through the Queen Charlotte Sound, with another eight moving across Haida Gwaii. Combined, Vancouver Island, the Queen Charlotte Sound and Haida Gwaii account for 38 of the 62 tracks. AK landfalls, plus storms that moved into the GAK and decayed before landing, and one event that moved into the Northeast Pacific, looped off of NVI and then tracked south just off the coast and became absorbed before landing, account for another 10 high-wind storms. The two GAK lows produced southeasterly high winds in the study region. Both of these storms were particularly intense, with sub-95.0-kPa central pressures, providing some evidence that deeper lows tend to have larger high-wind fields, or at least a further reach when interacting with coastal terrain. Looking to the south, three events tracked into the Olympic Peninsula (OLY) and one low tracked far south into Northwest California. One of the OLY lows, 12 Dec 1995, went on to track across SVI after moving through the Olympics. Two of the southern storms were Fraser outflow events, following paths that are not entirely unusual for strong Chapter 2: Climatology of High-Wind Storms  51 easterly gap winds. The final 10 ETCs developed in the BC interior and generally tracked east to southeast. The majority of these events, eight, produced westerly windstorms, with the remainder being southwesters.  With the exception of ETCs in the Gulf of Alaska (GAK) and northeast Pacific bins, which contain cases of very intense systems, southeasters are more likely to occur relative to westerly windstorms for tracks that are closer to Vancouver Island (Table 2.9). For the three bins SVI to South Queen Charlotte Sound (SQCS), 12 southeasters and eight westerly windstorms occurred, for a (SE/W) ratio of 1.5. For the next three bins going northward, North Queen Charlotte Sound (NQCS) to North Haida Gwaii (NHG) tracks, four southeasters occurred relative to eight westerly events, a ratio of 0.5. Landfalls from the Dixon Entrance (DXE) north all resulted in westerly windstorms. The interior BC lows are also more likely to generate westerly gales, this being due to the fact that the pressure slope is typically favorable for onshore flow on the southwest BC coast when a low is inland over the province. Looking at the 20 storms that landed on Vancouver Island, of the six ETCs that tracked east to east-northeast at closest approach to the three key weather stations, five produced westerly windstorms (Tables 2.8 and 2.9). Of the 13 cyclones that tracked northeast to north-northeast, nine produced southeasters, with an additional two resulting in south to southwest events. Only two produced westerly high winds.  A concentration of storms moved through the latitudinal band encompassed by 47.5 and 52.5º N, with 50.0 to 52.5º N capturing the majority of events (Figure 2.8). Twenty-one windstorms tracked through the quadrilateral just to the northeast of Vancouver Island, with 19 ending up just west of the island. The number of high-wind storm tracks drops off dramatically to the south and southeast of Vancouver Island, and more slowly to the north and northwest. Chapter 2: Climatology of High-Wind Storms  52  Figure 2.8 The frequency of occurrence of ETCs passing through 2.5º x 2.5º quadrilaterals. Mean storm tracks are drawn for southeasterly (orange) and westerly windstorms (blue), and the far north Pacific (including the GAK category) lows that trigger high winds in the study region (tan). A key storm formation region is evident over the Northeast Pacific to the west and southwest of the study region (Figure 2.9). ETCs forming in this area typically followed a classic development cycle, likely getting their start in the diffluent eastern side of the mean 50 kPa trough axis, located in the vicinity of ~135 to 155º W depending on the ETC event, and then tracking northeast with the upper steering currents toward southwest BC to finally land at or past peak intensity. Many southeasters had their origins in this area of the Pacific, along with some southwesters and a few westerly windstorms. Another area of significant high-wind-triggering Chapter 2: Climatology of High-Wind Storms  53 cyclogenesis is found near Vancouver Island. This zone reflects the occasional development of cyclones just off of the coast and also inland, where lows sometimes form in the lee of the BC Coast Range. Most of these cyclones do not reach explosive cyclogenesis status as they typically degrade upon encountering steep coastal terrain or the Rocky Mountains. Many of the systems that began in this area produced westerly high winds, though the list includes at least one southeaster and both of the Fraser gap events. A third, minor region is located in the GAK and   Figure 2.9 The frequency of ETC cyclogenesis initiations for storms that caused high-winds in the study region. Thirteen of the 62 cyclones started off the map. Raw numbers are shown; however, the color scheme is based on a smoothing filter described in the methods. Chapter 2: Climatology of High-Wind Storms  54 reflects ETCs that generally track east to east-southeast into northern or central BC, bringing westerly windstorms in all cases.   Figure 2.10 The tracks of 20 ETCs that triggered southeasterly windstorms 1994 to 2012.  Southeasterly windstorms tend to follow a recurving path starting around 45º N at 150º W, sagging to 40º N around 140º W and then heading northeast toward Vancouver Island (Figures 2.8 and 2.10). A lower-frequency subset originates further south, around 30 to 35º N and follow a persistent northeast direction to Vancouver Island. Both of these mean pathways are shown in orange. While the trackways are only approximated in Figure 2.8, the infrequent intense GAK lows that trigger southeasterly windstorms tend to follow a path east to northeast. In contrast to Chapter 2: Climatology of High-Wind Storms  55 the southeasters, the main westerly windstorm track does not show much recurvature (Figures 2.8 and 2.11). Though the origin of the mean path is similar in location to the southeasters, these "west-jet" cyclones tend to head east to east-northeast toward the Queen Charlotte Sound. A lesser path north of the main corridor represents storms that develop in the GAK and track east to southeast into the BC interior; and, as with the southeast storms, GAK lows sometimes trigger westerly windstorms usually by carrying a strong frontal system into southwest BC.  Figure 2.11 The tracks of 29 ETCs that caused 27 independent westerly windstorms 1994 to 2012.   Chapter 2: Climatology of High-Wind Storms  56 2.3.4 Windstorm Seasonality  For the 132 windstorms during the 1964 to 2012 period, there is a strong seasonal pattern of monthly frequency (Figure 2.12). Windstorms predominate at the height of the cool season: 84 events occurred during the three months November to January. When the two datasets are separated, windstorm occurrence for 1964 to 1993 peaked in December and January with 20 events each, about one month behind the November peak of 15 events during 1994 to 2012. Looking at track positions and wind direction category for 1994 to 2012, the September to October period had few windstorms, and all were westerly events (Figure 2.13a). This low frequency is also evident in the 1964 to 1993 dataset, with just five early-season events of which  Figure 2.12 Frequency of high-windstorms (solid line), as measured at CYYJ, CYVR and CYXX, by month for the 1964 to 2012 period (n=132), and mean return period by month in years (dashed). Chapter 2: Climatology of High-Wind Storms  57 two were westerly, two southerly and one southwesterly. The November to December period was quite active, with a wide variety of track and wind direction categories (Figure 2.13b). January to February and March to April were similar (Figure 2.13c and d). Aside from a possible favoring of westerly storms in the early period, there appears to be no clearly identifiable seasonal trends in track position, direction or with wind-direction category.   Figure 2.13 Windstorm seasonality. Tracks of 62 ETCs that triggered 57 independent high-wind events from 1994 to 2012 separated by bi-monthly intervals during the storm season. Track types are as described in Figure 2.6. a) September to October; b) November to December; c) January to February; and d) March to April. Chapter 2: Climatology of High-Wind Storms  58 2.3.5 Mean Return Frequency and Long-Term Pacific Climate Variability  Though 58 windstorms during the 19-y from 1994 to 2012 suggests a mean of three high-wind storms each year (Table 2.10), or an interval of approximately four months between events if seasonality is not considered, this does not mean that such storms occur with such regularity. This is evident in the extreme time intervals between events. During 1994 to 2012, as much as 600 d (1.6 y) passed between high-wind storms (03 Apr 1997 to 24 Nov 1998) and as little as 10 h (two on 16 Nov 2009). The greatest number of events for a given calendar year is eight in 2006. The year 2006 also contains a sequence of three high-wind storms in five days from 11-15 Dec, or 1 storm every ~48 h, with the last event being the strongest. Interestingly, a similar sequence occurred from 01-05 Dec 1967, with the final storm being the strongest and indeed nearly producing high winds at all three stations (Abbotsford fell 0.5 m s-1, or 1 kt, short on gust), marking a rare storm, and suggesting that multiple-storm strikes in a short period are a recurrent feature of windstorm climatology in the region. When the intervals are broken down by peak wind direction, up to 623 d (1.7 y) has passed between westerly windstorms and a much Table 2.10 Mean high-windstorm recurrence intervals for various breakdowns of the 1964 to 2012 dataset. Time Period Mean Recurrence (no. y-1) Mean Recurrence Interval (y) 1964-2012 2.7 0.37 1964-1993 2.5 0.41 1994-2012 3.1 0.33 1964-1973 4.5 0.22 1974-1983 1.4 0.71 1984-1993 1.5 0.67 1994-2003 2.4 0.42 2004-2012 3.8 0.26 Chapter 2: Climatology of High-Wind Storms  59 longer 2,526 d (6.9 y) between southeasters. The low number of southeasters occurred during 2000 to 2005, a period that overall had reduced high-wind frequency.  Figure 2.14 Timing and peak wind magnitude of all identified high-windstorms from 1964 to 1993 (n=74). Peak wind is the highest among CYYJ, CYVR and CYXX.  The 1964 to 1993 dataset captures 74 separate high-wind events (Figure 2.14). Divided over 30 years, this results in approximately 2.5 events per year, somewhat less than 1994 to 2012. Note that without the special observations from 1964 to 1977, there is probably some undercounting in the earlier record. The overall average for the 49 y 1964 to 2012 is 2.7 windstorms y-1, or about one event every 0.37 y without consideration of seasonality (Table 2.10). For the decadal breakdowns, the mean recurrence went from 4.5 events y-1 during 1964 to 1973 to a low of 1.4 events y-1 over the next decade. Then, from 1984 onward there appears to have been a slow "recovery" approaching the 1964 to 1973 windstorm frequency by the 2004 to 2012 period. Note that the last interval only has nine years. In terms of mean return intervals for specific magnitudes of wind (not gust), the frequency of events drops dramatically across a narrow band of peak wind ranging from 17.5 to 25.0 Chapter 2: Climatology of High-Wind Storms  60 Table 2.11 Mean return intervals in years for increasing categories of wind speed and for different wind-direction classes of windstorm, 1964 to 2012. Windstorm Category Peak Wind Magnitude among CYVR, CYYJ and CYXX (m s-1) Total Number of Events in 49 y Mean Events Per Year Mean Years Between Events Example All Windstorms            Catastrophic ≥ 25.0 <1 <0.02 >49.0 12 Oct 1962 Severe 22.5-24.9 6 0.12 8.2 15 Dec 2006 Strong 20.0-22.4 14 0.29 3.5 15 Nov 2006 Endemic 17.5-19.9 76 1.55 0.6 02 Apr 2010 Moderate ≤ 17.4 >76 >1.55 <0.6 14 Apr 2002       Westerly           Catastrophic ≥ 25.0 0   None Severe 22.5-24.9 3 0.06 16.3 14 Dec 2001 Strong 20.0-22.4 7 0.14 7.0 09 Jan 2007 Endemic 17.5-19.9 38 0.78 1.3 09 Sep 2005 Moderate ≤ 17.4 >38 >0.78 <1.3 08 Jan 1975       Southeasterly           Catastrophic ≥ 25.0 <1 <0.02 >49.0 12 Oct 1962 Severe 22.5-24.9 2 0.04 24.5 11 Dec 2006 Strong 20.0-22.4 1 0.02 49.0 15 Nov 2006 Endemic 17.5-19.9 24 0.49 2.0 02 Apr 2010 Moderate ≤ 17.4 >24 >0.49 <2.0 31 Jan 2006  m s-1 (Table 2.11). Moderate windstorms with speeds < 17.5 m s-1 are the product of routine winter weather system and typically occur several times during the wet season. The endemic windstorm has peak winds in the range of 17.5 to 19.9 m s-1 (~60 to 69 km h-1) and is what occurs approximately once each year. Strong windstorms, with maximum winds ranging from 20.0 to 22.4 m s-1 (~70 to 79 km h-1) occur approximately two to three times a decade. Severe Chapter 2: Climatology of High-Wind Storms  61 windstorms with peak winds in the range of 22.5 to 24.9 m s-1 (~80 to 89 km h-1) occur approximately once a decade. Only one windstorm within the modern surface airways record produced winds of  ≥ 25.0 m s-1, and was the 1962 Columbus Day Storm which occurred just before the 1964 to 2012 era utilized here. The Columbus Day Storm generated maximum winds of 25 to 25.5 m s-1 around 2330 LST on the CYVR U2A strip-chart trace, perhaps higher at CYXX based on a faster peak gust (Table 2.4). The widespread damage wrought by the storm (Franklin n.d., Lynot and Cramer 1966) lends the name to this category: catastrophic. Owing to a higher number of events, westerly windstorms among the various peak wind-speed categories have shorter return periods relative to southeasters. The higher number of strong to severe events, 10 verses three, reflects in part CYVR's exposure to westerly winds. Given the favored direction, it is interesting that no westerly wind reached catastrophic category in the 49 years of the study, or even going back to 1953, though one event on 20 Feb 1960 came close and produced maximum winds of 24.6 m s-1 on a 45B anemometer. Table 2.12 Total number of windstorms and mean return frequencies for different wind-direction classes largely based on peak wind direction at CYVR for 1964 to 2012. Windstorm Type Number of Events in 49 y Mean Events Per Year Mean Number of Years Between Events Example Westerly 54 1.1 0.9 14 Dec 2001 Southeaster 45 0.9 1.1 11 Dec 2006 South 14 0.3 3.5 01 Jan 1997 Southwester 12 0.2 4.1 02 Mar 2011 Fraser Outflow 7 0.1 7.0 18 Dec 1990  Looking strictly at the wind the direction category for all high-wind storms (endemic to catastrophic), southeasterly and westerly windstorms occurred about once a year (Table 2.12). Chapter 2: Climatology of High-Wind Storms  62 Southerly and southwesterly storms occurred roughly every four years. Fraser River Valley outflow windstorms had a mean recurrence of about once every seven years. The windstorm record shows some marked trends. First, out of the 22 high-wind events at CYVR during 1964 to 1993, all were westerly events save for two: 19 Jan 1964 and 05 Dec 1967, both southerly windstorms. This is in contrast to the occurrence of seven southeasterly high-wind events at CYVR in the shorter interval from 1994 to 2012. There is also a noticeable reduction in the frequency of high-wind storms starting after about 1973 and especially between 1976 and 1990 (Figure 2.14, Table 2.10). The entire decade of the 1980s only had eight high-wind events. Indeed between 16 Dec 1982 to 08 Jan 1986, some 1,119 d (3.1 y) passed without a windstorm. After a second high wind event in Jan 1986, another 1,107 d (3.0 y) passed before the next. The 1980s contrast sharply to the 1964 to 1973 period when 45 windstorms occurred, a difference of a factor of 5.6. Instrument changes are, to a degree, controlled for by limiting this comparison to the U2A anemometer era.  Figure 2.15 The frequency of high-windstorms per five-y interval (e.g. 1964 to 1968, 1969 to 1973…) compared to the five-y average PDO (a) and NPI (b) for each half-decade from 1964 to 2012. For the PDO the R2 is 0.75 (n=9) using linear regression and for the NPI the R2 is 0.71 (n=9). Chapter 2: Climatology of High-Wind Storms  63 The periodic reduction in high-windstorm frequency may be related to the PDO. A linear regression between the five-y windstorm frequencies from 1964 to 2008 relative to the five-y average NPI for the months January to March and October to December has a positive slope and an R2 of 0.71 (n=9) (Figure 2.15). Slipping the five-y intervals forward one year provides five different, but related, sequences that yielded coefficient-of-determinations of 0.40 to 0.71 (Table 2.13). Regression between the PDO index and the frequency of high-wind storms during the same period yields a negative slope and an R2 of 0.75 (n=9), with a range of 0.63 to 0.75 among the five sequences. The trendlines for the five-y intervals (Figure 2.16) show that changes in windstorm frequency even appear to follow shorter-period (here five-y averages) variation of the PDO index. Table 2.13 R2 values between two PDO-related indexes and the annual frequency of windstorms per five-y interval using the data for 1964 to 2012. Results shown are for five different starting points. For all cases, n=9. Starting Five-y Interval NPI R2 PDO Index R2 1964-1968 0.71 0.75 1965-1969 0.58 0.66 1966-1970 0.71 0.64 1967-1971 0.40 0.67 1968-1972 0.54 0.63  2.4 Discussion  2.4.1 General Statistics of Southwest BC Windstorms  When using the peak wind and gust filters on hourly and special observations to isolate the windstorms, more events were identified via gusts in the earlier period of the 1994 to 2012 record, suggesting higher ratios of gusts to wind. A methodological and/or instrumental change Chapter 2: Climatology of High-Wind Storms  64 in wind measurement is suspected of being the cause. The 78D became fully operational at CYVR in 1998, and at CYYJ and CYXX in 2005 (Table 2.1). Of related note, the 1964 to 1993 period had a lower all-time peak wind coupled with a higher peak gust than 1994 to 2012. The higher gust ratio implied for 1964 to 1993 also likely reflect the use of different instruments as well as nonstandard anemometer heights at some stations. No windstorm from 1964 to 2012 brought high winds to all three stations. However, such an event occurred just a few years earlier: on Columbus Day 1962 (Table 2.4). The use of 45B anemometers with their differences in observational methodology relative to the U2A and 78D along with nonstandard sensor heights, and the absence of a gust record at CYYJ confounds this determination. However, taken at face value, the three-station average peak gust of 34.4 m s-1 is 7.6 m s-1 higher than any windstorm from 1964 to 2012, suggesting that even if the speeds were adjusted for instrument changes the storm would still stand out. Measured wind speeds in  Figure 2.16 Trendlines for all five runs comparing the average wet-season PDO index to the frequency of windstorms for the same five-y intervals. Chapter 2: Climatology of High-Wind Storms  65 Western WA and OR, on different anemometers and with generally lower sensor heights (Read 2008), are consistent with the readings from the Canadian stations. The remnant latent-heat energy and upper-level vorticity of Typhoon Freda, and perhaps anomalously warm sea-surface temperature coupled with a strong thermal gradient in the storm formation region, likely contributed to the intensity of the 1962 ETC (Namias 1963, Lynot and Cramer 1966). Also, the storm quickly tracked (~25 m s-1) north-northeast up the Pacific Coast, reaching peak intensity of at least 96.0 kPa off of Cape Blanco, OR, and maintained this depth up to the latitude of Astoria. By the time the ETC center passed just east of Tatoosh Island, WA, it had experienced some weakening to 97.0 kPa. Continuing on a path across Southern Vancouver Island, the degrading but still intense storm center passed very close to the three study stations, bringing the steepest pressure gradient in the 60 y 1953 to 2012 (Table 2.4). The fast northward storm motion likely gave a boost to already strong pressure-gradient-driven S to SE surface winds. Isallobaric analysis, not performed here, might help capture the effects of ETC motion on surface wind speeds. The combination of a rapid northerly track and the low-center passing very close to the region-of-interest with enough intensity to bring extreme pressure gradients among the study stations is rare. In the 1964 to 2012 period, just three strong ETC events followed a track similar to the Columbus Day Storm: 14 Nov 1981, 12 Dec 1995 and 16 Jan 2000 (Read 2008). Of these three, only the 1995 storm reached high-wind status at the study stations, this despite all bringing major winds just to the south such as in the Greater Seattle Area. None of these storms appear to have moved as fast as the 1962 event. The 1981 and 2000 storms tracked further west, keeping their centers further away from the region of interest, and the 1995 storm tracked perhaps a little Chapter 2: Climatology of High-Wind Storms  66 further east and appears to have brought broader core, with a more bowl-like sea-level pressure profile, right on top of the study region. For the 58 events 1994 to 2012, the generally weak correlation between pressure gradient and surface wind speed (Table 2.5) is not surprising given that surface winds are influenced by many factors including the interaction of airflow over complex terrain. This includes the tendency for lee troughing in the lee of the Olympics (Reed 1980), which depending on atmospheric conditions can strongly influence the local pressure field in the northeast quadrant of the mountains—the very region where the pressure gradient triangle is located (Figure 2.1). Also, there are errors from the method used to reduce station pressures to sea-level (Mass and Dempsey 1985, Benjamin and Miller 1990). The coefficient-of-determinations (e.g. Figure 2.3) in this study are similar to Benjamin and Miller's results for the Western United States. Also, the low correlations (Tables 2.5 and 2.6) are likely related to strictly including events that produced wind speeds near the top of the range for the sites, limiting the range of wind speeds that are under scrutiny. Interestingly the pressure gradient orientation (α) for the sample windstorms has a higher correlation with peak wind and gust than the actual strength of the pressure gradient (Table 2.5). Allowing for the limitations of the method (e.g. only looking at peak pressure gradient and not the progress of individual storm pressure gradient fields over time), this suggests that once pressure gradients reach a certain level, the orientation relative to the surrounding terrain is more important in determining the magnitude of peak winds than the ultimate strength of the pressure differential, probably within limits. The high correlation between peak wind speed and direction at CYVR (Table 2.5) is likely the result of the long over-water fetch in the westerly quadrant. Among the wind-direction Chapter 2: Climatology of High-Wind Storms  67 categories, westerly windstorms are most likely to have the highest wind speeds at CYVR, and this angle happens to be higher up on the compass rose (~270º) than the next key category, southeast (~135º), supporting a positive correlation. The correlation between peak wind direction and α suggests that the pressure gradient orientation, or pressure slope, has a strong influence on wind direction even amid complex terrain during the steep gradients associated with windstorms. Nevertheless, the average α readings for the two main wind direction classes of windstorm do not align with what might be considered the ageostrophic-wind ideal based on the orientation of the key mountain ranges that surround the stations: approximately 215º (pressure slope 125º) for southeasters, which had an average α of 194º, and 35º (pressure slope 305º) for westerly windstorms, which had an average of 265º. This suggests that the terrain has a more complex influence on wind direction than the simple geographic arrangement of the surrounding mountains. Note that, for westerly windstorms, a 35º α puts the low to the southeast of the study stations and that the low centers of most windstorms tracked to the north, perhaps offering some explanation for the poor agreement between α and the ageostrophic ideal. Also, the gradient orientation does not need to be ideal for a particular wind direction to occur. A southwest pressure slope with just enough tilt relative to the surrounding terrain is enough to trigger NW winds down the Georgia Strait (Lange 1998). Another consideration is that the triad of stations is at the confluence of the Puget and Georgia Troughs, where the mountain range orientation changes from north-south to northeast-southeast, and is also under the influence of the west-east oriented Strait of Juan de Fuca, making the determination of an ideal ageostrophic wind direction difficult. Also, the pressure slopes are for peak gradient only, and do not reflect the steady changes of orientation that occur as a low, or Chapter 2: Climatology of High-Wind Storms  68 surface front, moves inland. Near-peak gradients before or after the actual maximum may have had a more ideal orientation.  2.4.2 Cyclone Statistics  Mesquita et al. (2009) note that the background pressure field can have a significant influence on the associated pressure gradient magnitude of a deep low—e.g. an ETC developing in an already deep trough may have an extraordinary low pressure but without attendant steep gradients. There is a wide spread in the central pressures among the land-falling windstorm-generating ETCs from 1994 to 2012, and this might be taken to suggest that central pressure has only modest influence on wind speeds (Mass and Dotson 2010 has more discussion). However, the ETC landfalls occurred over a wide region (Figures 2.6 and 2.7, Table 2.8). When track is accounted for, such as focusing on the outcomes of NVI landfalls as is done here, the relationship between central pressure and surface wind magnitude becomes stronger. Linear regression on all east- to north-trending NVI lows from January 2008 to April 2013 (Figure 2.5) allows for some rough predictions. For a NVI low to match the three-station average peak gust magnitude of the Columbus Day Storm, an extremely rare (5.8 SD) 92.0 kPa ETC would be required. This allows for variance around the predicted value with ± ~4.4 m s-1 being 2 SD based on the standard residuals. Historically, cyclones with ~95.5 kPa central pressures have landed or passed close to the Cascadia region (Read 2008), and for NVI landfalls, this depth is a 3 SD event. The predicted average peak gust for a 95.5 kPa ETC is 23.3 m s-1, within the range of the top 25% of windstorms. However, given the variability around the trend-line, a potentially quite destructive 27.7 m s-1 is possible, an event outside of recent experience, as only the Columbus Day Storm has reached (and exceeded) this threshold. Lows with central pressures Chapter 2: Climatology of High-Wind Storms  69 down to 94.0 kPa (even lower) have been observed in the Northeast Pacific. Such a system tracking across NVI is a 4 SD event and is predicted to produce an average peak gust of 26.4 m s-1 at the study stations, higher than anything that occurred from 1994 to 2012. There is the possibility of ~31 m s-1 among the study stations, not quite the Columbus Day Storm, but almost certainly a historic storm. Note that this summary is for NVI tracks. SVI windstorms are expected to produce even higher wind speeds for a given central pressure, due to tracking closer to the study stations. So many ETCs undergo cyclolysis in Northeast Pacific, and even more so the Gulf of Alaska, that the region has been referred to as the cyclone graveyard (Mesquita et al. 2009). However, when high-wind generating ETCs are considered, it is clear that some cyclones do indeed land at or near peak intensity (Table 2.8). Though many of these intense lows begin filling rapidly upon landfall, they often reach the BC interior before becoming fully absorbed. A select few lows maintain central-pressure depth post-landfall, and can even strengthen slightly despite the interaction with mountainous country. Examination of available synoptic charts suggests that upper support moves inland with these lows, with enough strength to maintain low central pressures despite the rapid filling that can occur due to terrain-forced ageostrophic flow. The statistics on central pressure tendencies (Table 2.8) among 45 ocean ETCs suggest that studies focused on cyclogenic bombs miss many significant storms. Only 35% of high-wind generating ocean ETCs reached explosive cyclogenesis status based on a 21 hPa (24) h-1 cutoff. Indeed, the windstorms that struck southwest BC have exhibited a wide range of deepening and filling rates, from one to two hPa (3) h-1 upwards to 15 hPa (3) h-1. The most intense cyclogenic bomb in the 1994 to 2012 dataset, 12 Mar 2012, deepened at a rate similar to the intense 14 Nov Chapter 2: Climatology of High-Wind Storms  70 1981 ETC's 42 hPa (24) h-1 (Reed and Albright 1986), and above the Columbus Day Storm's 35 hPa (24) h-1 (Lynot and Cramer 1966), i.e. within the range of historical storms.  2.4.3 ETC Tracks for Southwest BC Windstorms  There are three general patterns that can be discerned amid the track complexity evident in Figure 2.6: i) many of these ETCs passed fairly close to the study region; ii) most of the ETCs moved north of, or in some cases right over, the three stations; and iii) when the ETCs near the Pacific Coast, there is a strong tendency to recurve toward the north. There are, of course, exceptions to these general patterns. The tendency for close tracks is not surprising given that the strongest pressure gradients tend to be near ETC cores. The propensity for the high-wind generating ETCs to track north of the study stations is likely related to the fact that many northeast Pacific cyclones tend to develop in the base of deep meridional troughs, or as secondary spinups in the base of a large GAK low, and this has the effect of reducing pressure gradients north of low centers (Read 2008). Therefore, those ETCs that track to the south of the region of interest tend not to produce high winds at the three study sites. Also, upper support around ETCs tends to favor winds from the southerly half of the compass rose. A low with a steep gradient on the north side tracking south of the station triad still might not produce high winds (Read 2008; e.g. 10 Nov 1975). The northward recurving of tracks is likely the result of ETC interaction with the large north-south trending mountain ranges of North America. Cyclonic vorticity can be enhanced in the northeast quadrant of the low due to airflow descending the west mountain slopes; the ETC then tends to migrate toward the area of vorticity spin-up (Carlson 1991). Chapter 2: Climatology of High-Wind Storms  71 On average, southeasters have markedly different tracks than westerly windstorms events (Figures 2.8 and 2.10). In other words, storm track direction appears to have a bearing on peak wind direction, especially for systems that follow paths close to the region of interest. When considering the vertical mixing of wind momentum, a southeast direction is better supported by upper-air flow from the south half of the compass rose; lows moving in from the south to southwest suggest upper-air flow from this direction (Chapter 3). In addition to this, as these cyclones near the coast, southwest BC ends up under the northeast quadrant of the low, a region with cyclonic curvature at the middle-atmospheric levels (e.g. 85 kPa) that tends to support SE airflow, in good alignment with surface winds from the same direction. As these storms track north of the study stations, mid-level winds tend to shift from SE to SW, still supporting SE winds, though not as strongly. As the storm moves inland, mid-level winds can become NW, better supporting westerly high winds, but typically the system is so degraded by terrain interaction at this time that the westerly winds do not reach the magnitude of the preceding southeasterlies. Storms with an easterly track—i.e. approaching from nearly due west—likely have good upper support for westerly winds. At the mid-levels, support for westerly high winds is best when the ETC tracks to the north, bringing the low's southwest quadrant over the region. For more distant tracks, especially those north of NVI, westerly windstorms tend to dominate (Table 2.9). This probably reflects the lower dependence on strong gradients that is characteristic of westerly high winds due to the long over-water fetch at CYVR. Turbulent drag is much lower over open water than over land (Stull 1988, Stull 2000), resulting in faster wind speeds for a given pressure gradient (Appendix B). A more distant cyclone, bringing a modest west-southwest pressure slope over the Georgia Strait, can trigger high westerly winds at CYVR whereas a system that brings an equal-magnitude east-southeast pressure slope over the same Chapter 2: Climatology of High-Wind Storms  72 region is less likely to have winds reach the same speed since SE winds are overland. Direct evidence for this in the presented data is somewhat weak as both westerly windstorms and southeasters have had a broad range of peak gradients (Table 2.3, Figure 2.3), reflecting the low correlation between pressure gradient and wind speed at these interior stations (Tables 2.5 and 2.6). However, southeasters do tend to have stronger gradients, indicated by an average 0.9 hPa (100) km-1 higher than westerly windstorms, though a fairly large SD, 1.6 (skewness 0.26) for SE and 0.8 (skewness 0.22) for W storms, suggest that this difference may not be that significant. For lows tracking over Vancouver Island, the total number generating westerly windstorms is reduced compared to those generating southeasters: seven verses 11 events, plus two storms in the southwest to south category. This does not mean that westerly high winds are uncommon with these landfall locations, just that the SE wind tends to dominate and yield the highest speeds during high-wind storm episodes, in part due to the fact that many cyclones degrade as they move inland, typically resulting in markedly reduced pressure gradients and other mechanisms of surface wind support, when the southwest pressure slope finally moves into the study region. A key storm formation region for southwest BC windstorms is encompassed by 40 to 50º N and 130 to 140º W (Figure 2.9). This area is close to a zone identified by Sanders and Gyakum (1980; their Figure 3) that had a smoothed frequency of one cyclogenic bomb per annum, and could be considered the classic windstorm formation zone. The second cyclogenesis region around Vancouver Island does not appear in the 1980 paper. This is almost certainly because lows developing so close to the rough costal terrain only rarely, if ever, reach cyclogenic bomb status, keeping them out of any analyses that uses the methods of Sanders and Gyakum. In the current study, the GAK region is under-represented relative to Sanders and Gyakum's findings—storms tracking this far north, even if they more frequently undergo explosive cyclogenesis than Chapter 2: Climatology of High-Wind Storms  73 in areas of the Pacific to the east and southeast, only rarely produce high winds at the three stations used in this analysis.  2.4.4 Seasonality  The occurrence of windstorms is strongly seasonal (Figure 2.12). A feature of global general circulation for coastal British Columbia is marked seasonality of precipitation, with the cool season, October to March, being wet due to the frequent visitation of ETCs (Hare and Thomas 1974). The warm season is relatively dry as the mean storm track shifts north and higher pressure associated with the subtropical desert belts also moves northward. This has the effect of reducing the frequency of ETCs that affect the region, making windstorms unlikely. That westerly windstorms dominate in the September to October period (Figure 2.13a) makes sense because these weather systems do not have to be as intense as southeasters to achieve high-wind status due to the over-water fetch at CYVR. Systems are likely to be weaker early in the storm season, a situation not usually conducive to a southeasterly event. However, historically significant windstorms have produced southeasterly gales, including an ETC on 21 Oct 1934 that brought damaging winds from both a SE and W direction to Vancouver (Vancouver Sun 1934, Read 2008), and the even more powerful 12 Oct 1962 Columbus Day Storm. A number of intense early-season storms appear to have had a tropical connection in the form of entraining the energy from degrading tropical cyclones, such as with Typhoon Freda in October 1962 (Nick Bond, WA State Climatologist, pers. comm. 02 Mar 2013). It is likely that with a longer period of record May would be added to the list of months that have high-wind events. This determination is based in part on the fact that two high-wind Chapter 2: Climatology of High-Wind Storms  74 storms occurred during the latter third of Apr and also because a westerly windstorm on 25 May 1982 brought peak winds of 17.0 m s-1 to CYVR, just below the cutoff in this study.  2.4.5 Return Intervals and Long-Term Climate Variability  Given a short 10 h time between the ETCs of 16 Nov 2009 (Table 2.3), it is possible to have as many as three individual windstorms sweep through the region in a single 24 h period, which has implications for automated schemes of isolating discrete events, such as the method of independent storms (e.g. Harris 1999). On the opposite end of the spectrum, as much as 3 y have passed without a windstorm. This high stochasticity limits the usefulness of mean return intervals (Keim and Muller 2007). The numbers also ignore seasonality, e.g. a return frequency of one event every three months in a region where most storm events happen from October to March will likely not have storms occur with such regularity. Windstorms arriving in closely timed series, sometimes called "storm trains," is a recurrent feature. Two significant examples are the three-storm series from 01-05 Dec 1967 and 11-15 Dec 2006, both of which contained at least one severe windstorm. When atmospheric conditions are suitable for the development of ETCs, often a series, or a family, of storms is spun up (Bjerknes and Solberg 1922). A storm-formation region that is strong enough and progressing eastward slowly may result in the development of several intense cyclones, each of which can bring high winds to the same region. The Pacific Decadal Oscillation (PDO) is a long-term spatiotemporal bi-modal change in the Pacific climate system (Mantua et al. 1997, Mantua and Hare 2002, Whitfield et al. 2010). A significant transition from cool-phase to warm-phase is recognized for the 1976-77 period (Hare and Mantua 2000). Mean storm tracks over the Pacific are known to shift in response to this Chapter 2: Climatology of High-Wind Storms  75 variability, with a move to the north during positive (warm-phase) conditions (Trenberth and Hurrell 1994, Moore and McKendry 1996), associated with a deepening of the mean Aleutian low and down-stream ridging over Western Canada.  The relationship between five-y windstorm frequency and both the NPI and PDO index (Figure 2.15) suggest that interdecadal variability has a marked influence on the occurrence of high-winds in the region. Tuller (2004) reports a moderately strong correlation between the NPI and PDO index; when one index shows a correlation with windstorm frequency, then the other is also likely to do so. The PDO analysis done here covers approximately one large cycle: part of the previous cool phase from ~1947 to 1975, to warm from ~1976 to 1998, then a stretch with shorter-period oscillations of roughly five years duration up to 2013 (NWFSC 2015). There is also a limited number of data points in the regressions. However, windstorm frequencies seem to follow even short-period PDO changes, with upward movement in the PDO index being followed by reduction in windstorm frequency such that peaks and troughs tend to line up among the five-y samples (Figure 2.16), lending some strength to the assessment. Using surface wind observations and pressure-triangle wind estimates, Tuller (2004) found reduced wind speeds at stations in southwest BC, including CYYJ and CYVR, during the onset of the recent warm phase of the PDO. Based on a wind speed cutoff of 40 km h-1, Abeysirigunawardena et al. (2009) also note a decrease in frequency of extreme events in the 1980s and show a reduction in expected speed for given return intervals during warm climate phases relative to neutral and cold phases. Given that the frequencies for various wind speed thresholds are for the period 1964 to 2012 (Table 2.11), these figures are essentially an average for the PDO phases. For example, the cold-phase decade of 1964 to1973 averaged 4.5 windstorms annually and the warm-phase decade 1984 to 1993 averaged 1.5 windstorms annually, yielding a mean of 3.0, which is close to Chapter 2: Climatology of High-Wind Storms  76 the 2.7 computed for 1964 to 2012 (Table 2.10). Thus, assuming the observed PDO response repeats over future cycles, one might expect a ~4.5/3.0, or ~50% increase in the frequency of events in Table 2.11 during PDO cold phases and a ~50% decrease during warm phases. However, this analysis does not account for possible non-linear relationships and differences in how the PDO may affect the frequency of events across wind-speed categories.  2.5 Conclusions and Recommendations for Further Research  High winds that cause widespread damage to trees and infrastructure, typically triggered by passing ETCs, are a recurring feature of the Lower Mainland and Greater Victoria, BC, climate. A comprehensive examination of high-windstorms that affect southwest BC including track typing based roughly on Keim and Muller (2007), the first of its kind for the region, reveals that strong variability characterizes windstorms. A range of synoptic structures, storm tracks and landfall locations have been responsible for damaging winds. Track directions are also quite variable, ranging from north to east to south, though many have tracks in the northeast quadrant. ETCs in all stages of development, from newly-forming open waves to mature and degrading stacked lows, have triggered high winds. Mean return frequencies have a marked degree of variability across intraannual, interannual and upwards to decadal scales, a result that is congruent with other wind climatology studies (e.g. Tuller 2004, Abeysirigunawardena et al. 2009) but also adds new information because it is the first to focus exclusively on independent high windstorms. Amid the erraticism in the ETCs paths, some patterns are present. Closer storm tracks appear associated with a greater probability of high-winds in the study area, with Vancouver Island landfalls being the most likely to produce a damaging windstorm. In most cases, high-Chapter 2: Climatology of High-Wind Storms  77 wind generating storms track north of, or right over, the region of interest with paths to the south only rarely bringing a damaging gale. The majority of ETCs follow a recurving track toward the northeast. A fairly large percentage of high-wind generating lows were at or near peak intensity during landfall. The preponderance of high-wind generating ETCs had central pressures below 100.0 kPa. These outcomes are similar to those found for windstorms just to the south of the study region (e.g. Mass and Dotson 2010). Southeasters seem more likely to move north-northeast to northeast and westerly windstorms east-northeast to east-southeast. This relationship does not appear to have been discussed explicitly before this study, probably because the current study appears to be the first to consider southeasterly and westerly windstorms separately. A moderately strong relationship between central pressure and peak winds has been found with northern Vancouver Island landfalls, with deeper lows more likely to generate damaging gusts. This kind of relationship has been considered before for other regions, especially for strong storms (e.g. Mass and Dotson 2010). However, no examination has employed the rigorous methodology used here where the distance between the ETC center and the region of interest is controlled for, and also with consideration of all low-pressure centers regardless of strength. As noted above, there appears to be a relationship between mean return frequencies and long-period climate variability associated with the PDO, with reduced windstorm numbers during the warm phase. Windstorm frequency appears to drop by a factor of roughly three between the strongest warm and cool phases of the PDO. This knowledge has utility. For example, during a multi-decadal warm phase, a power company may not need as many assets for storm repairs relative to a cold phase, thus conserving resources. Twenty to thirty years is essentially a human generation, enough time during a period of infrequent storms to allow Chapter 2: Climatology of High-Wind Storms  78 complacency to develop. Also, weaker tree material may accumulate over the years, creating greater hazard to infrastructure at the onset of a period of higher windstorm frequency. On average an event with peak winds in the range of 17.5 to 19.9 m s-1 can be expected nearly every year during the cool season; this is the type of storm that utility and transportation companies, foresters and city planners should be prepared for on a routine basis. Managers also need to be aware of the infrequent windstorm that can produce peak winds reaching or exceeding 25 m s-1 with gusts approaching 40 or more, events such as the 1962 Columbus Day Storm that caused severe destruction to trees and property and significant loss of life throughout the Western Cascadia region (Franklin n.d., Lynot and Cramer 1966). The most damaging windstorms (peak wind ≥22.5 m s-1) appear to have decade to multi-decade intervals depending on maximum wind speed. As noted above, these time intervals are long enough to invite complacency. Additional insight into coastal Cascadia windstorms could be gained by comparing the results from additional regions such as the Greater Seattle and Portland Metro Areas, an important aspect of track typing not done here due to the focus on one region. Determining the paths of the windstorm-generating ETCs that occurred from 1964 to 1993 could also provide useful information such as whether-or-not certain types of tracks are favored during specific PDO phases. Finally, isolating the variables that contribute to generally low correlations between maximum pressure gradient and peak wind/gust speed may shed further insight into the structure of windstorms as they interact with complex coastal terrain.  Chapter 3: Detailed Analysis of High-Wind Storms  79 3 Detailed Comparative Analysis of Major High-Wind Storms in Southwest British Columbia and Northwest Washington  3.1 Introduction   In the weather history of the Pacific Northwest (PNW)2, few windstorms are so well remembered as the destructive 1962 Columbus Day Storm (also called "Typhoon Freda"). For many locations in the region, this extratropical cyclone (ETC) produced the highest wind gusts on record (Lynot and Cramer 1966, Read 2008). The outcome of the tempest of 1962 conveys well the destructive potential for strong ETCs originating in the northeast Pacific Ocean. The damage in southwest British Columbia (BC) included the widespread loss of trees including thousands at Vancouver's Stanley Park, perhaps the largest disruption of electrical service on record with over 300,000 customers without power representing about 68% of those serviced by BC Hydro at the time, and the loss of at least seven human lives with at least four being direct deaths caused by wind-thrown trees (Bolwell 1962, CP 1962, Kheraj 2007). Roofs were torn from houses, siding stripped from buildings, windows were shattered, roads rendered impassable by debris, boats capsized, airport hangers destroyed along with the aircraft they sheltered and major bridges such as the Lions Gate were closed. This litany of damage occurred throughout the PNW, including at least 46 people killed, one of the largest windthrow events on record with over 25 million m3 of merchantable timber toppled, and nearly overwhelming damage to the power grid that included the destruction of gigantic steel transmission towers in some of the hardest-hit locations (Franklin n.d., CP 1962, Lynot and Cramer 1966).                                                 2	  For the purposes of this chapter, Pacific Northwest refers to "Cascadia", which includes northwest California, western Oregon and Washington and southwest British Columbia.	  	  Chapter 3: Detailed Analysis of High-Wind Storms  80 Following the Columbus Day Storm, Lynot and Cramer (1966) wrote, “In view of the economic havoc of severe windstorms in Oregon and Washington, it is regrettable that previous storms have not been more thoroughly analyzed and documented.” Since Lynot and Cramer's detailed analysis, many decades passed before comprehensive and rigorous accounts of windstorms in the PNW were undertaken. Starting in the early 2000s, I began a systematic catalogue of windstorms, including storm tracks and isotach maps, currently available on the Office of the Washington State Climatologist website (Read 2008). Mass and Dotson (2010) provide a brief historical review and a climatological and meteorological examination of windstorms, mainly focused on Washington (WA) and Oregon (OR), and their paper is in part informed by Read (2008). Neither of these efforts focuses on southwest BC, though there is overlap as some BC storms have also significantly impacted regions to the south. The analysis by Lynot and Cramer (1966) is among a small collection of studies that detail major windstorms (Smith 1950, Reed 1980, Reed and Albright 1986, Kuo and Reed 1988, Steenburgh and Mass 1996). These papers provide much insight into individual storms, and tend to be mainly focused on cyclone structure and/or the complexities involved during storm interaction with the PNW terrain, but generally do not look for patterns among different events. In contrast, Mass and Dotson (2010) present some storm tracks, look at synoptic patterns and also build a conceptual model of a PNW windstorm. However, Mass and Dotson (2010) do not examine mesoscale patterns (or the lack of) in surface wind, frontal movements and other storm responses between different ETCs, especially those following similar tracks. Much recent theory on the life cycle of ETCs is based on detailed examinations of individual and often major storms (e.g. Shapiro and Keyser 1990, Neiman and Shapiro 1993, Neiman et al. 1993, Clarke 2004). Broader examination of independent storms of various magnitudes, that follow similar paths to Chapter 3: Detailed Analysis of High-Wind Storms  81 provide some control for geographic influences, allows for exploration of whether particular features are present in all ETCs, the circumstance under which these features manifest, and how these features tend to affect a particular region. Pressure gradients are the known driver of wind (e.g. Stull 2000). However, in the rugged terrain of the western United States, Benjamin and Miller (1990) show a low correlation between pressure gradients (represented by geostrophic potential wind) and measured surface winds compared to other regions of North America that have less topographic relief. Other mechanisms in the generation of observed surface wind speeds during windstorms include the transfer of upper-air momentum to the surface. Frontal boundaries are sources of vertical mixing (Bergeron 1937, Lynot and Cramer 1966, Stull 2000, Clark et al. 2005; also see Neiman et al. 1993 for frontal transects of an ocean ETC showing vertical wind speeds), and provide a mechanism for bringing upper wind to the surface. Warm sectors, with low static stabilities and therefore potential for the creation of downdrafts, are also known to be prone to the occurrence of high surface winds (Lynot and Cramer 1966, Mass and Dotson 2010), making it important to track these features. Bent-back fronts have been recognized since the development of the occluded cyclone model, though they have had many different names throughout the decades, including non-frontal troughs, back-bent occlusions and bent-back warm fronts (Bjerknes 1930, Godske et al. 1957, Shapiro and Keyser 1990, Steenburgh and Mass 1996). A bent-back front develops along a baroclinic zone that wraps around the low-pressure center to the rear, or western half, of the low (Steenburgh and Mass 1996). On weather maps, this feature is often represented as an occluded front that trails the low and the leading frontal systems, or simply as a trough of warm air aloft (TROWAL) curling around the low, depending on the meteorological agency. In satellite photos, Chapter 3: Detailed Analysis of High-Wind Storms  82 the bent-back front often appears as an arc of clouds nearly encircling a dry slot of clear air, contributing to the classic comma-shape of intensifying ETCs. Marked pressure falls can occur along a bent-back front, supported in part by latent heat release from condensation, that enhance the surface pressure gradients, usually in the southwest quadrant of the ETC (Steenburgh and Mass 1996). Sometimes the pressure drop is strong enough that a new low develops southward of the old one. The bent-back front can take on the characteristics of a cold front, and, indeed may have been originally considered just this. Bjerknes and Solberg (1922) describe secondary cold fronts and note that if the secondary cold front is stronger than the leading one, all the air between the leading warm front and the secondary cold front, including the weak cold-sector behind the first cold front, can act as a large warm sector. It has been suggested that the strong winds along bent-back fronts in the PNW are primarily the result of the intense pressure gradients associated with this feature, as opposed to being primarily driven by any associated vertical mixing mechanisms (Mass and Dotson 2010). In contrast, analyses of Atlantic ETCs landing in the United Kingdom (UK) indicate that mixing may bring upper-momentum to the surface (Browning 2004, Clark et al. 2005, Baker 2009). Apparently along the tip of bent-back fronts, branches of the slantwise circulations, following a slope of constant wet-bulb temperature and characterized by evaporative cooling (i.e. evaporatively-cooled downdrafts), can descend to the surface boundary layer, where turbulent mixing can bring upper-momentum to the surface. These are sometimes termed "sting jets." Mass and Dotson (2010) note that little evidence exists for sting jet contributions to high surface winds during major PNW storms and suggest that the much steeper coastal terrain may be a limiting factor. Steenburgh and Mass (1996) also indicated that further Chapter 3: Detailed Analysis of High-Wind Storms  83 research is needed to better understand the development of bent-back fronts. Such work is currently underway (Cliff Mass, personal communication, 28 Feb 2015). Mass and Dotson (2010) point out that the highest winds for major ETCs tend to occur with the arrival of the bent-back front, which indirectly suggests that peak winds would occur right when or just after the low has passed a given location, as the bent-back front often trails close to the low-pressure center in the southwest quadrant. If this holds true for most ETCs, then forecasting the time of highest winds becomes a matter of accurately determining the sweep of the bent-back front within a given forecast window, especially the tip and associated strong pressure gradients. However, for a given region, it is not altogether clear that the highest winds are always associated with the passage of a bent-back front—this is one of the questions to be explored here. If peak winds can occur in association with other ETC features, such as in the warm sector as discussed above, then the location of the highest winds relative to the low-center position may vary widely, suggesting that the time of peak winds can also have wide variance, one that is dependent on the track of the low and what ETC structures are brought into the region of interest. This brings up a key question: Are there any clear patterns between track and storm structure that can help in the determination of the timing of peak winds for the region of interest?  There appears to be a relationship between ETC central pressure and peak wind speeds (Chapter 2). This suggests that tracking the depth of these storms is important, including the filling—if any takes place—of these lows post-landfall, as this is when the storm center often reaches its closest point to the region of interest. Among the individual storm studies, there has often been much interest in the conditions contributing to the deepening of ETCs, often with less emphasis on the decay (e.g. Reed and Albright 1986, Steenburgh and Mass 1996). Indeed some studies have been focused specifically on explosively deepening storms, or "cyclogenic bombs" Chapter 3: Detailed Analysis of High-Wind Storms  84 (Sanders and Gyakum 1980). Even though Steenburgh and Mass (1996) provided detailed mesoscale surface maps of the 1993 Inauguration Day storm, they do not mention the filling rate post-landfall and any implications for potential wind speed. In this case, the storm landed with a 97.2 kPa central pressure and began filling rapidly as the center tracked across the Olympic Mountains, reaching 98.4 kPa in three h despite a contribution from lee troughing east of the Olympics. Every PNW windstorm with a peer-reviewed analysis that includes central pressure post-landfall shows the commencement of filling by the time the low has reached the coastline. Interestingly, information pertaining to the increase in central pressure post-landfall is available for hurricanes, including empirical models that estimate the rates for different regions in the United States (Vickery 2005), but there appears no such examination for landfalling ETCs. Due to differences in storm formation mechanisms, post-landfall central pressure response of ETCs and tropical cyclones is not likely to be the same, save under specific circumstances. Maintenance of the low-pressure center of an ETC is largely due to 1) the balance of surface convergence that is largely influenced by the pressure gradient force, Coriolis force and friction; and 2) upper divergence that is supported by jet stream dynamics. Whereas in a hurricane, the central pressure is largely determined by the balance between surface convergence driven by convective strength that in turn also contributes to high pressure aloft that forces divergence (Godske et al. 1957, Simpson and Riehl 1981, Stull 2000). Therefore under the conditions of good upper support from an active jet stream, an ETC could in theory keep deepening post landfall as long at the upper divergence is strong enough to counterbalance enhanced surface convergence due to the increased turbulent drag. Given that ETCs can form in places such as the US Midwest, it is clear that upper divergence can be strong enough to support a deepening low over land. However, the ETCs of concern here form over the Pacific Ocean, a region with low Chapter 3: Detailed Analysis of High-Wind Storms  85 friction relative to terrain. In this situation, if upper support strength were to remain the same as a low approached the coast, weakening of the low post-landfall could occur, or at least a slowing or stalling of deepening rates, due to increased surface convergence. The mountainous terrain of the PNW would likely have a stronger filling effect compared to the open plains. Therefore it is no surprise that the detailed analyses of single storms done in the past have showed weakening as the ETCs neared and landed on the coast. However, there could be situations where the upper support enhances as the lows close in on the coast, at times becoming strong enough to counteract enhanced surface convergence even among the steep coastal terrain, resulting in continued deepening even post landfall. Such storms might be expected to have an increased potential for high winds in the region of interest due to the maintenance of a deep central pressure. It is unclear how often this happens with high-wind generating ETCs in the PNW. The conceptual windstorm model presented by Mass and Dotson (2010) has four stages: 1) pre-frontal, with the low well offshore, an emplaced cold surface layer over the region of interest, strong gap winds with light winds elsewhere and widespread precipitation; 2) post-frontal, of the leading front or fronts which often results in warming temperatures as the cold surface layer is scoured out, and also tends to shift the pressure gradient to a more favorable orientation for strong southerly surface flow with a resultant increase in wind speed in part due to the arrival of less stable, showery air where the transfer of upper-wind momentum to the surface can contribute to stronger, sometimes damaging wind gusts; 3) bent-back trough, when the area of strongest pressure gradients to the south and southwest of the low center moves through the region resulting in peak winds, 4) termination, with a wind direction shift to onshore and slowing speeds, save for those channels prone to W to NW winds. Though this model has general applicability to the region, it makes some key generalizations that may not apply to all high Chapter 3: Detailed Analysis of High-Wind Storms  86 windstorm events: Mainly that the ETC is tracking to the northeast, and, as mentioned above, that all ETCs bring a strong bent-back trough that tends to be the prime ingredient for peak winds. Also, when the focus is on a specific area, such as the Georgia Basin, a general conceptual model may miss some key details largely due to the complex geography of the region. Pressure slopes are a down-gradient measure of the pressure field orientation (Lange 1998); for example, a southeast pressure slope has a gradient roughly perpendicular to the axis of the Georgia Strait. Pressure slope, relative to local terrain, has a strong bearing on the surface wind response, both speed and direction, and is an important tool for understanding the dynamics during the interaction of an intense ETC with coastal terrain. In narrow channels such as the Georgia Strait, surface winds tend to be geostrophically balanced when the pressure gradient is approximately parallel to the basin orientation, also called a "cross-channel gradient" (Overland 1984). Winds occurring with a pressure field more orthogonally aligned, i.e. perpendicular to the terrain orientation, tend to be balanced via ageostrophic acceleration. The Georgia Strait falls within the 5 to 100 km width and 25 to 500 km length limitations mentioned in Overland (1984), and therefore these concepts likely apply. Building on the available knowledge, the goal of this study is to provide a detailed comparative analysis of Southwest BC windstorms and explore the interactions between the atmosphere and coastal terrain. Specific objectives include: i) for selected significant Lower Mainland and Greater Victoria windstorms that resulted from Vancouver Island landfalls, examine patterns and variability of synoptic conditions between storms at the 30 kPa, 50 kPa, 85 kPa and surface levels, ii) provide mesoscale surface maps to show the isobaric patterns and surface wind response around the peak wind phase of these storms, iii) determine the tracks and Chapter 3: Detailed Analysis of High-Wind Storms  87 central pressures of the selected windstorms in detail as they moved inland, iv) provide detailed frontal positions, especially for bent-back fronts, to help understand how air mass boundaries influence surface winds in the complex terrain of the region, v) examine surface pressure gradients and compare these to the surface wind response, and vi) build conceptual models for southeasterly and westerly windstorms that provide detail specific to the region of interest, including the variables that contribute to the timing of peak winds.  3.2 Methods  3.2.1 Study Region  The main region of interest is the Georgia Strait of BC and Northern Inland Waters of WA, which encompass the three long-term weather stations used to determine the most significant windstorms (see section 3.2.7). Synoptic analyses cover a large part of the Northeast Pacific Ocean and western North America (Figure 3.1). Mesoscale analyses look at a region that encompasses southwest BC including all of Vancouver Island and the Georgia Strait, and northwest WA including all of the Olympic Peninsula and Puget Sound region. As of ~2013, this region is home to approximately 8.1 million people (Statistics Canada 2014, US Census 2014).  Chapter 3: Detailed Analysis of High-Wind Storms  88  Figure 3.1 The study region. The larger map encompasses the area examined for the synoptic analyses and the inset shows the area considered for mesoscale analyses.  3.2.2 Synoptic Analyses  To determine storm tracks and other synoptic features of southwest British Columbia windstorms and to aid in the creation of mesoscale weather charts, surface and upper-air analyses Chapter 3: Detailed Analysis of High-Wind Storms  89 from several sources were used. The National Oceanic and Atmospheric Administration (NOAA) U.S. Daily Weather Maps Project of the Central Library has daily maps that are mostly focused on the U.S. but in some cases are at the North American scale, for the years 1871 to 2004 (NOAA 2014). The National Climatic Data Center produces 12-h surface and upper-air maps for the 1994 to present period (NCDC 2014a), and also has a nearly comprehensive collection of three-h, six-h, and sometimes 12-h charts covering scales from the U.S. to full hemispheres (NCDC 2014b). The Weather Prediction Center (WPC) produces detailed three-h surface maps at the United States and North American scales that include analyses combined with infrared satellite photos for Mar 2006 to present (WPC 2014). The Vancouver office of Environment Canada has surface and upper-air maps for the 1973 to present period in a microfilm repository. For many of the weather events in the 2000 to present timeframe (earlier in some cases), water vapor and infrared four-km resolution satellite imagery were collected in real-time from the National Weather Service (NWS) Seattle (SEW) and Portland (PQR) offices (NWS SEW 2014; NWS PQR 2014). Also, I collated forecast discussions and warning statements produced by the two offices as the storms progressed and these were useful aid in the interpretation of storm outcomes. Hourly and special surface observation data were used in the determination of storm tracks, the creation of mesoscale surface maps, and in the determination of peak wind/gust magnitude and pressure gradients. Data for land-based stations and offshore buoys were obtained from Environment Canada, the NOAA National Data Buoy Center (NDBC 2011), the University of Washington and the Plymouth State Weather Center (PSWC 2011). In some cases raw METAR reports were collected from the NOAA Aviation Weather Center (AWC 2011), and the Chapter 3: Detailed Analysis of High-Wind Storms  90 Emergency Weather Network (EWN 2012). Data were compiled and analyzed using Microsoft Excel, SAS and StatPlus, and weather maps were constructed manually using Adobe Illustrator. The examination primarily focused on events from 1994 to present. During this era, an increasing number of weather stations became automated. In the case of wind measurement, automation provides much more consistent and less biased readings than manual observations (e.g. Tuller 1980). The 19-y study interval also includes a number of significant windstorms and also a wide variety of scenarios that aid in understanding the synoptic features that contribute to high winds in southwest BC.  3.2.3 Mesoscale Analyses of Low Tracks  Mesoscale surface observation maps using 50 to 60 stations in the region were analyzed by hand (Figure 3.2). The exact number of sites for a given event depended on factors including station inceptions and decommissions, communication loss due to power interruption and data loss due to instrument failure. An additional 20 off-map stations were also considered in the development of the charts. Official surface analysis maps were also consulted, thus the mesoscale maps were drawn using a nested approach. On the meso-maps, wind speed is shown in m s-1, with a full barb representing five m s-1 and a pennant 25 m s-1—this creates a result very close to using 10 knots for a full barb. Isobars are drawn at four hPa intervals. To help control for biases between different station networks, when determining the position of isobars, fronts and troughs, first-order stations were given priority. In the context of this study, a first-order station is one that has a nearly continuous record in excess of five decades, and often appears at a key airport--e.g. Olympia (KOLM), Sea-Tac (KSEA) and  Vancouver (CYVR). First-order stations always report sea-level pressure (SLP) on the hour and  Chapter 3: Detailed Analysis of High-Wind Storms  91  Figure 3.2 Weather stations used for the development of mesoscale surface analysis maps. Some stations not shown, including Qualicum Beach (VOQ), North Cowichan (VOO) and Victoria Heartland (WVV). The key stations Vancouver, Victoria and Abbotsford are highlighted in Orange. Weather station codes are actually four digits, with Canadian observation site designations being lead with a "C" (e.g. YVR becomes CYVR). Stations in the contiguous 48 United States are lead with a "K" (e.g. SEA becomes KSEA). also record an entire suite of meteorological variables including sky cover, precipitation type and other weather conditions, temperature, dew point, relative humidity, wind speed, direction and gust, altimeter and visibility. Buoy and Coastal-Marine Automated Network (C-MAN) station data were given second priority, with the rest of the stations put at a tertiary level. One consistent disparity appeared during the analysis for the "island stations" in the Georgia Strait: Saturna (CWEZ), Entrance (CWEL), Ballenas (CWGB) and Sisters (CWGT) Chapter 3: Detailed Analysis of High-Wind Storms  92 islands. The sea-level pressure reports from these four reporting sites seem to be consistently low by about 0.5 to 1.5 hPa, sometimes higher, relative to the buoys and stations on major land areas in the vicinity. Accordingly, pressure records from the island stations were given lowest priority. Regular observation times can be off by as much as 30 minutes between stations, a difference sometimes compensated for through simple linear interpolation. Canadian buoys frequently report observations far off the hour, and indeed most have observations taken closer to the half hour. During periods of rapid pressure changes, differences in reporting of 15 or 30 minutes can result in significant error relative to surrounding stations that share near synchronous observation times. For example, if the pressure is falling approximately four hPa h-1 in a particular region, a station that reports 30 minutes before the rest can show a two hPa disparity in the pressure field relative to the other stations. For the mesoscale analysis, the landfall points and analyzed positions of low pressure centers are determined from the surface maps created during this study, and then supplemented with existing official weather maps and satellite photos. Generally, the low center was placed at geometric center of the smallest closed isobar. Since extratropical cyclones are not necessarily symmetrical, and isobaric positions are not always precise, this approach probably does not pick up the exact center of circulation. However, the method has the benefit of being easy to perform and replicate. The southern quadrants of ETCs tend to be the most important regions for high-wind generation, and many lows tend to have significantly weaker pressure gradients in the north quadrants. Due to these factors, for cases where lows have a strong north-south elongation, the center is considered to be the point between the southernmost point of the minimum-pressure isobar and a line drawn between the locations where the isobaric lines become roughly parallel. In the case of a low developing in at the intersection of the warm, cold and occluded fronts, the Chapter 3: Detailed Analysis of High-Wind Storms  93 low center is considered to be at the point-of-occlusion--sometimes these are called a triple-point, or point-of-occlusion lows.  Minimum central pressure is determined via the lowest available pressure during the particular observation hour, rounded down to the nearest full hPa. Due to limited observations over the ocean, this method likely has considerable error. Estimated central pressures were also verified against available surface maps from professional meteorological organizations, though these are not without error. In rare instances pressure trends at coastal stations are used to determine a central pressure for a low that is still offshore, but close to landfall. If the barometer at a particular station located near the landfall point of the cyclone indicates a low pressure of 97.5 kPa one or two hours after a given surface map's observation time, and the lowest pressure indicated by any station at the observation time is 97.7 kPa, then the low is assumed to have a central pressure of at least 97.5 kPa. Of course, the central pressure is a changing feature over time, and lows typically begin to weaken as they move into the rough terrain of southwest BC, so the estimated central pressure could even be lower (or higher in the rare case of a deepening cyclone) than this method allows.  3.2.4 Plotting Frontal Systems  Frontal positions were determined using traditional methods (e.g. Stull 2000, Vasquez 2002 and 2008). Due to the moderating effects of the Pacific Ocean, frontal boundaries can be hard to detect and at times may only be discerned through observations of wind direction and/or dew point changes. Therefore frontal positions and types are not always exact. Satellite photos proved a useful aid for the placement of some fronts.  Chapter 3: Detailed Analysis of High-Wind Storms  94 3.2.5 The Special Case of Bent-Back Fronts  The location of bent-back fronts around their parent cyclones and their surface progress was carefully determined. At times, this yields surface analysis maps that may look somewhat different from traditional professional meteorological approaches, largely due to the presence of two or more significant fronts, placed to each side of the low, among other frontal features. The "fractured-front" model of ETC evolution presented by Shapiro and Keyser (1990) is useful for interpreting the mesoscale maps presented here. One key difference from traditional mapping of these fronts is that the barbs denoting the front type are placed in the direction that the front is moving, which is the original intent of the frontal symbols and is similar to the approaches used in detailed examinations of ocean ETCs (e.g. Neiman and Shapiro 1993, Neiman et al. 1993, Browning 2004). Often in official weather maps the barbs are pointing away from the direction of movement. In this paper, bent-back fronts may be designated as cold fronts or occlusions, largely based on surface temperature changes that occur when the boundary moves through. If strongly cold air is associated with the bent-back front, it is given a cold-front designation in a fashion similar to Neiman and Shapiro (1993) and Browning (2004) and is here referred to as a secondary cold front as Bjerknes and Solberg (1922) might have done. Note that for ETCs that have a strong secondary cold front, it can be useful to think of the leading cold front as a warm front (Bjerknes and Solberg 1922). A number of surface weather indicators are used to determine the arrival time of bent-back fronts in the study area, and descriptions of the passage of these features can be found in Steenburgh and Mass (1996), Read (2008), Mass and Dotson (2010) and Read and Reed (2013). With strong bent-back fronts, most if not all of the characteristics listed below will be present. Chapter 3: Detailed Analysis of High-Wind Storms  95 For weak manifestations, only some features may be evident, or many of the variables will simply exhibit a minor signal. The pattern change must occur after it is clear that the leading front—be it an occlusion or a cold front—and sometimes the low center, has passed. A strong wind direction shift in the S-SW winds that occur behind typical leading cold and occluded fronts tends to occur with the bent-back feature, with the veering occurring quite suddenly. This direction change can be 45 to 90º. A common outcome is that if SW winds arrive with the leading front, then the bent-back front brings W-NW winds. This wind direction shift is most pronounced over open water and at coastal stations. Inland stations may experience only small wind direction changes, apparently due to terrain constrictions. There is an escalation in wind speed with the approach and passage of a bent-back front, often suddenly, and sometimes to damaging levels. Precipitation, sometimes heavy in stronger storms, is likely to occur. The temperature typically drops, sometimes quite sharply, and if the temperature gets low enough, snow can result. A rapid jump in pressure (i.e. pressure surge) often accompanies the bent-back front. In strong storms, this surge can easily be three to five hPa h-1 and in extreme cases more than 10 hPa h-1. For coastal PNW reporting sites, nearly every all-time-maximum pressure surge has occurred in association with the bent-back fronts of land-falling cyclones (e.g. Read 2008). Bent-back fronts tend to move at roughly the same speed as the low center. Sometimes the wrap-around in the southwest quadrant may run slightly ahead of the low and move into the southeast quadrant. At other times, the bent-back front at the surface may be slowed by terrain features and lag somewhat behind the low. The close association of frontal forward progress with that of the cyclone's center distinguishes the bent-back occlusion from the leading warm, cold and occluded fronts which can move faster than and therefore race far ahead of the parent low. Chapter 3: Detailed Analysis of High-Wind Storms  96 Well-defined bent back-fronts were also tracked via satellite imagery. When available, satellite photos were used extensively to identify and follow these fronts. A curl of clouds that wraps westward over the pole-ward side of the low-pressure center and then equator-ward along the west side indicates the location of the bent-back front (Figure 3.3). In mature lows, the band of clouds may wrap around the low center more than once.  Figure 3.3 Some examples of easily detectable bent-back fronts using water-vapor satellite imagery. Shown is the 11 Dec 2006 windstorm (a), 02 Apr 2010 (b) and 12 Mar 2012 (c) just before landfall in all cases. The major 15 Dec 2006 windstorm is also shown (d), which has strong wrap-around moisture, but the exact position of the bent-back front is not as clear as with the other ETCs in this figure. Images courtesy of the US. National Weather Service and are in the public domain.   Chapter 3: Detailed Analysis of High-Wind Storms  97 3.2.6 Wind and Gust  The wind averaging period varies between station types and networks. For NWS Automated Surface Observation System (ASOS) stations (e.g. Sea-Tac), National Data Buoy Center (NDBC) Coast Meteorological Automated Network (C-MAN) stations (e.g. Smith Island) and Canadian staffed stations (e.g. Vancouver), wind is a two-min average speed (Giselle Bramwell personal communication 12 Jul 2011, NDBC 2014, NCDC 2013b). NDBC buoys (e.g. Cape Elizabeth) report an eight-min average (NDBC 2012). Canadian fully automated stations (e.g. Pitt Meadows) report a 10-min average wind (Giselle Bramwell, personal communication 12 Jul 2011). The parameters for wind gust vary between different station networks and time periods, and it is important to keep this reality in mind when comparing the values between storms. In the US, ASOS became implemented in the 1993 to 1996 timeframe, depending on station. Before ASOS, the highest instantaneous reading, often recorded on strip-charts, indicated the maximum gust. Instantaneous gust is considered to be a one to three second (s) average due to the inertial response of the Belfort cup-based anemometer systems used by the NWS (Krayer and Marshall 1992). The initial ASOS stations continued to utilize cup-based anemometers, but gust became a five-s block average of one-s samples (NIST 2012). This completely changed the character of gust. In tests conducted by the author, and based on one datum offered by NWS Portland during a windstorm on 12-13 Dec 1995, peak instantaneous gust appears significantly higher than peak five-s gust, perhaps by an average of about 1.19 (Read 2008). The National Institute of Standards and Technology (NIST) suggest a lower conversion factor of 1.065 based on theoretical estimations (NIST 2012). From 2004 to 2007, depending on station, the ASOS cup-based anemometers were switched to sonic sensors. With this change, the NWS changed the parameter Chapter 3: Detailed Analysis of High-Wind Storms  98 for gust again, this time to the international engineering standard of three-s, and also using a moving average (NIST 2012). The NDBC stations used in this study report five-s gust (NDBC 2014). This averaging period appears to have been consistent throughout the period of record. Canada modernized many of its first-order weather stations in the 1990s, incorporating the 78D anemometer, a digital cup-based system (Richards and Abuamer 2007, Nav Canada 2013). For these systems, gust is the highest instantaneous report (MANOBS 2013). However, the 78D displays a five-s vector-average wind speed; therefore gust is a five-s average. Not all stations in the Canadian networks used in this paper have equipment of the same make. This results in wind data that could have significant bias relative to the information from the first-order stations. For example, Campbell Scientific stations are used in some instances, mainly for a network of purely automated stations, such as at White Rock (CWWK). Some of these reporting sites may not meet World Meteorological Organization standards as closely as the first-order airport stations. Using the CWWK example, this station is not located at an airport but on a modest slope at a city facility with a two story building immediately to the east, a parking lot to the south and a cluster of tall trees to the west. The anemometer sensor is placed on a mast that is taller than the building, perhaps close to the standard 10 m, but there is still much more clutter in the vicinity than compared to the open, flat expanse of an airport. During the course of the study, CWWK winds have proven to be consistently lower than those from nearby airport stations. Some of this is also due to a longer averaging period for these stations: 10-minutes for mean wind. No attempt is made to correct for these variations, but it is important to recognize the differences between stations as they explain, at least in part, some of the patterns seen (e.g. some locations always having lower wind speeds) in certain data presentations, such as peak gust maps. For considerations of peak wind magnitude, first-order stations are given priority. Chapter 3: Detailed Analysis of High-Wind Storms  99 Peak gusts used for analysis in tables and maps were taken from hourly and special METAR observations. Many stations report a peak gust for the given standard observation hour with the code PK WND; this is the best available record for peak gust, since the report also includes direction and time to the min. First-order US stations (e.g. Bellingham, Sea-Tac, Quillayute) and many of the automated Canadian stations (e.g. Sand Heads, Entrance Island, Ballenas Island) report the PK WND code. Canadian staffed stations (e.g. Comox) sometimes report maximum wind and gust for the past hour. For the NDBC buoys and coastal C-MAN stations, peak gust for each ten-min observation window is found in the continuous winds data. For those stations that do not include a PK WND observation in their METAR, including many of the long-term Canadian airport stations (e.g. Victoria, Vancouver and Abbotsford), automated AWOS stations in the US (e.g. Bremerton, Arlington and Mt. Vernon) and Canadian buoys, peak gust is determined from the highest gust reported in the regular and special observations (if available). Gust direction and time is based on the direction of the wind report and the time of the observation, and may be off from the actual values. For Canadian stations, peak gust can also be found in the daily data available from Environment Canada—this is used to double-check peak gusts gleaned from the regular and special observations. Gust magnitude and direction are available in the daily data, but not time. For those stations that do not report gust but do report wind (few in recent years, with more in the earlier years of the study), peak gust is estimated via a 1.3 gust factor, a conservative value that reflects the tendency for steadier wind at higher speeds (Davis and Newstein 1968, Peterka and Shahid 1998). Gusts estimated in this manner are shown in italics. Sometimes, especially during major events, stations can stop sending observations, either due to storm-related damage, power loss and/or communications failure. If a large set of observations is missing at a critical time, then the peak gust is considered Chapter 3: Detailed Analysis of High-Wind Storms  100 lost. This is either indicated by a blank in the tables and the absence of a value on the maps. If only a few observations are missing, or the lost reports are not at a critical time, the indicated peak gust on the maps is shown with a tilde (~) ahead of the value to indicate that the actual value could have been higher.  3.2.7 Windstorm Definition  Three first-order stations were selected as the basis for determining the presence of a high windstorm in the south coast study area: Vancouver International (CYVR), Victoria International (CYYJ) and the Abbotsford Airport (CYXX). The NWS defines high winds as when either the one-min average wind is ≥18 m s-1 (35 knots) and or gusts are ≥25.7 m s-1 (50 kt) (NWS 2012). In this study, the high wind cutoff is lowered by one knot for wind, to ≥17.5 m s-1 (34 kt), and two knots for gust, ≥24.7 m s-1 (48 kt). This allowed for a three percent error in wind measurement known for the conical three-cup sensor used in Canada (Devine 2008). Lowering the cutoff also brought additional storms into the analysis. Any wind meeting one or both of the above thresholds at just one of the three stations counted as a windstorm. There were 58 independent events during the 1994 to 2012 period (Chapter 2). These windstorms were then put into subcategories largely based on peak wind direction at CYVR. Categorization by wind direction is important due to the complex local geography and its potential influence on wind speeds at the three sites. Windstorms largely fell into westerly (n=27) and southeasterly categories (n=21). One storm, 15 Dec 2006, produced high S winds at CYXX ahead of the incoming low, followed by high W winds at CYVR as the ETC tracked inland (Chapter 2). In essence, this ETC could be put in two categories; for the purpose of this analysis, the storm is put in the westerly windstorm category due to producing the Chapter 3: Detailed Analysis of High-Wind Storms  101 strongest winds from this direction among the three stations. Twenty of the 58 total windstorms were the result of ETCs crossing Vancouver Island. Of these, 18 had either southeasterly or westerly peak wind direction. This study focuses primarily on six southeasterly and two westerly windstorms that resulted from Vancouver Island landfalls, although additional storms are considered in some examinations. Four of the selected storms landed in northern Vancouver Island with the remaining four crossing southern Vancouver Island. The eight storms were chosen in-part based on the availability of certain types of data like satellite photos and surface maps which became more widely archived for dates later in the 1994 to 2012 study period: This eliminated events from before 1999. Of the windstorms from 1999 to 2012, the strongest events were selected based on the average peak gust for CYYJ, CYVR and CYXX. One exception is the 15 Dec 2000 westerly windstorm. The 03 Mar 1999 southeasterly windstorm was used instead because of a decision to have at least two southeasters among the four selected storms that crossed southern Vancouver Island. Also, the 14 Dec 2001 westerly windstorm was favored over the 15 Dec 2000 event due to the former being later in the period and therefore having more information available—the average peak gusts between the two storms differed by just ~0.5 m s-1 (one kt).  3.2.8 Geostrophic Wind and Three-Point (Two-Dimensional) Pressure Gradients  To create a wind-related dataset independent of wind measurement, and also to explore local pressure gradients, geostrophic potential wind magnitude is calculated using hourly sea-level pressure data from a number of first-order stations in Washington and British Columbia. The calculation of the U and V wind components requires a right triangle, sometimes called a pressure-wind or geostrophic wind triangle. Eight such triangles were established in the region Chapter 3: Detailed Analysis of High-Wind Storms  102 covered by the mesoscale analysis (Figure 3.4). Since stations are rarely aligned perfectly, five of the triangles required two pressure interpolation points, with two triangles requiring a single interpolation locus. The remaining one did not require any pressure extrapolation. Interpolation assumed a simple linear relationship between the stations. This is almost certainly a source of error, but examination of results obtained by altering pressure data across a range of values (0.5 to 5.0 hPa) revealed that the triangles yield fairly robust output for gradient measures, especially for strong differentials, and perhaps more sensitivity with pressure slope calculations. Larger triangles are less sensitive to error in pressure measurement because the output values are effectively "averaged" over longer distances. The method for estimating geostrophic wind magnitude, Mg, follows that of Stull (2000). Assumptions used in the calculation of Mg are an atmospheric density (ρ) of 1.225 g/m3 (Stull 2000) and a latitude (ϕ) that is the average of the three stations used in the calculation. The value of Mg best represents the center of the triangle. Note that the geostrophic wind is a theoretical wind that assumes non-accelerating conditions. In practice, upper-level winds, largely free of turbulent drag, are best approximated by the geostrophic wind. Surface winds are not expected to approach geostrophic values (e.g. Myers 1954). For example, it has been estimated that a one-min average wind at six m height would likely reach about 50% of Mg (Lynot and Cramer 1966). The geostrophic wind, being a theoretical construct, provides insight into the nature of observed wind conditions in the boundary layer: Mg is a key component of methods for modeling surface winds, such as the gradient wind and boundary-layer gradient-wind. Calculating Mg also provides a measure of the pressure-gradient magnitude and the orientation of the isobars, information useful for the interpretation of synoptic conditions during  Chapter 3: Detailed Analysis of High-Wind Storms  103  Figure 3.4 Stations used in the construction of the geostrophic wind triangles. White-filled circles with black outlines are station locations. Station identifiers are listed at the end of each triangle's designation. Gray lines denote the station triads. Black lines and orange shading delineate the right triangles used in the calculations. Triangle endpoints not touching a station are interpolation points for sea-level atmospheric pressure.  storms. Unlike traditional one-dimensional two-station gradients, this two-dimensional method provides a means of more closely assessing the actual strength of the pressure differential since the measure is always perpendicular to the isobars. All pressure gradients are reported in hPa (100 km)-1, to make them easily comparable between station triads. To help determine the effective pressure gradient in cases of ageostrophic acceleration, the calculated Mg is adjusted downward by taking the cosine of the difference between the pressure slope and the orientation of the major terrain features and then multiplying the value with the Chapter 3: Detailed Analysis of High-Wind Storms  104 original Mg. In the case of the Georgia Strait geostrophic wind triangle, the terrain barriers are considered to roughly trend along the bearing from Entrance Island to Comox, 305º. In other words, a pressure slope of 125º is near the ideal for southeasterly ageostrophic winds in this location.  3.2.9 Conceptual Modeling  To help develop conceptual models of the synoptic conditions and surface wind response of southeasterly and westerly windstorms as they affect the study region, 6-h observations of wind, wind direction, gust, sea-level pressure, temperature, dew point, pressure gradient and pressure slope were gathered and charted for CYYJ, CYVR, CYXX and Comox (CYQQ) for the duration of a given ETC's life cycle. The pressure gradient and slope is derived from the Georgia Strait geostrophic wind triangle (Figure 3.4). All this information is then compared to the bearing and distance of the low center from Vancouver. Patterns identified from the synoptic and mesoscale analyses are also used in the development of these conceptual models.   3.3 Results  3.3.1 General Windstorm Statistics   Commonalities among the tracks of the eight windstorms selected for detailed analysis include (Figure 3.5): 1) all ETCs had some tendency to recurve as they neared North America; 2) all lows moved to the north, or right over the study region, and; 3) at closest approach to the  Chapter 3: Detailed Analysis of High-Wind Storms  105  Figure 3.5 Tracks of eight ETCs that produced significant high-winds in southwest BC during the 1994 to 2012 period. In this study, these storms are given detailed analyses. Paths with a blue shade denote those ETCs that landed on south Vancouver Island, and tracks in red and orange are from those storms that moved onto the north end of the island. The gray tracks trace the paths of two major windstorms that occurred outside of the study period: the 12 Oct 1962 "Columbus Day Storm" (AKA "Typhoon Freda") and the 13-14 Nov 1981 "Friday-the-13th" windstorm. They are included for reference. research area, track direction remained in the northeast quadrant, though with much variation within this arc, ranging from nearly due north to east-northeast. Differences among the paths include; 1) some ETCs started with a path direction toward the east-southeast only to later turn to the northeast while others had a more southern origination and maintained a track direction Chapter 3: Detailed Analysis of High-Wind Storms  106 roughly to the northeast, and; 2) cyclogenesis for some of the selected ETCs began well off the map, while happening very close to the West Coast for other windstorms, resulting in a wide variety of track lengths. Table 3.1 Peak wind with timing and direction and peak gust at CYVR, CYYJ and CYXX for the selected eight windstorms. Storm Event (LST) Peak Wind Time (LST) Peak Wind Direc-tion Desig-nation YVR Dir-ection (º) YVR Peak Wind (m s-1) YYJ Dir-ection (º) YYJ Peak Wind (m s-1) YXX Dir-ection (º) YXX Peak Wind (m s-1) YVR Peak Gust (m s-1) YYJ Peak Gust (m s-1) YXX Peak Gust  (m s-1) 03Mar99 0600 SE 130 17.5 130 15.4 170 14.4 22.1 24.7 21.6 14Dec01 0400 W 300 22.6 210 13.9 270 14.4 26.8 15.4 27.8 15Nov06 1300 SE 160 15.9 260 11.8 180 20.6 21.1 19.0 27.3 11Dec06 1400 SE 150 18.0 140 12.9 170 24.2 23.7 21.1 30.4 15Dec06 0234 W 290 23.1 270 15.4 180 18.5 26.2 21.6 25.2 12Nov07 0939 SE 130 17.0 140 15.9 180 22.6 21.1 24.7 30.9 02Apr10 1400 SE 140 19.0 140 13.9 180 18.5 24.7 23.1 22.1 12Mar12 0900 SE 140 18.0 130 18.5 180 11.3 24.7 26.2 17.5  At the three study stations, peak wind exceeded 20 m s-1 during five out of the eight storms, and peak gust exceeded 25 m s-1 during six out of the eight storms, with gusts in excess of 30 m s-1 in two instances both at CYXX (Table 3.1). During the six southeasters, peak winds tended to be SE at CYVR and CYYJ and S at CYXX. During the two westerly windstorms, peak winds were WNW at CYVR, and varied at the other two stations, from SW to W at CYYJ and S to W at CYXX. Peak winds for all storms occurred in the 12 h between 0200 and 1400 local standard time (LST) on the respective day for each event. A broader look at peak gusts indicates much higher wind speeds than those at the relatively sheltered key study stations (Tables 3.2a and b). All of the windstorms brought winds above 30 m s-1 at one or more stations, with the highest values often occurring at the more exposed stations along the coast or near interior waterways. In half of the windstorms, gusts exceeded 40 m s-1, with one exceeding 50 m s-1. Looking at the regional average peak gust, there is a fairly  Chapter 3: Detailed Analysis of High-Wind Storms  107 Table 3.2a Peak gust direction, magnitude and timing for key areas in and around the region of interest for the first four (arranged by date) of the eight windstorms selected for detailed analysis. Values led by a tilde (~) are the highest available from incomplete datasets. Times followed by an "E" are estimated. Storm: 02-03 Mar 1999 13-14 Dec 2001 15-16 Nov 2006 11-12 Dec 2006 Loca-tion Dir º M m s-1 t UTC Dir º M m s-1 t UTC Dir º M m s-1 t UTC Dir º M m s-1 t UTC Coast                         KHQM 240 26.8 1542 290 24.7 0945 150 ~30.4 2101 170 25.7 2002 46041       291 25.9 1000 163 ~27.7 1827       DESW1 156 35.6 0935 290 34.1 1001 160 36.6 1750 157 37.4 1953 KUIL 220 24.7 1454 270 22.1 0910 150 ~25.2 1731 140 ~24.7 2017 TTIW1 258 39.6 1531 288 33.3 0843 180 41.6 2049 192 43.3 2222 46087             173 31.0 1849 194 30.4 2330 46206 88 22.8 0845       176 ~26.6 1626 138 ~26.7 2026 CWEB       330 31.4 0558 80 29.8 0903  30.9 0100 Average   29.9     28.6     31.1     31.3                             Juan de Fuca                       CWSP 260 25.7 1922 280 30.9 1001 280 26.2 0226 260 20.9 0100 CWQK 240 32.4 1853 260 33.4 0942 270 28.8 0108 110 24.7 1907 KCLM 240 12.3 1653 280 20.6 1044 10 22.1 1913 230 15.9 2243 CWLM 240 29.3 1907 250 27.3 1222 250 28.3 0242 230 25.2 0127 46088             258 24.6 0303 123 23.9 1842 Average   24.9     28.1     26.0     22.1                             Puget Lowlands                     KSHN 220 23.1 1741 240 20.1 0931 220 19.5 0139 210 19.0 2333 KOLM 150 21.1 1102 270 18.0 1425 190 19.0 2028 190 16.5 2230 KTCM 180 24.7 1143 240 21.6 1218 160 23.7 2158 200 17.5 2206 KSEA 180 26.8 1208 220 18.0 0552 190 18.5 2238 220 17.5 2317 WPOW1 177 30.3 1250 157 22.5 2340 180 20.3 0209 193 24.1 2251 KPAE 180 25.7 1157 150 17.5 0150 180 20.1 0219 190 20.1 2345 Average   25.3     19.6     20.2     19.1                             Northwest Interior                     KNUW 140 28.3 1157 260 26.2 1141 160 29.3 2108 160 27.3 2240 SISW1 148 30.9 1015      151 34.1 1900 143 29.9 2137 KFHR 150 24.2 1120 220 17.0 0520 140 ~26.2 1941 140 22.1 2207 KBLI 150 28.3 1725 270 18.5 1549 180 28.3 2059 150 28.3 2028 Average   27.9     20.6     29.5     26.9                             Chapter 3: Detailed Analysis of High-Wind Storms  108 Storm 02-03 Mar 1999 13-14 Dec 2001 15-16 Nov 2006 11-12 Dec 2006 Loca-tion Dir º M m s-1 t UTC Dir º M m s-1 t UTC Dir º M m s-1 t UTC Dir º M m s-1 t UTC Georgia Strait Interior (Lower Mainland)               CYXX 180 21.6 1900 270 27.8 1300 180 28.3 2100 170 30.4 2238 CYVR 130 22.1 1400 290 26.8 1600 130 21.1 1825 150 23.7 2221 CWSB 100 26.2 1700 300 30.4 1300 90 29.3 1700 90 25.2 2100 CWMM 150 19.5 1820 260 19.5 1014 170 19.5 2231 140 18.5 2246 CYYJ 130 24.7 1145 230 15.4 0500E 260 18.5 0500 140 21.1 2051 CYQQ 110 18.0 1500 130 19.0 2200 130 25.7 1700 130 27.8 2200 Average   22.0     23.2     23.7     24.5                             Georgia Strait                       CWEZ 140 34.0 1150 150 22.3 0000 140 30.4 2025  30.9 2133 CWVF 140 30.9 1736 300 28.8 1732 140 25.2 1827 200 26.2 0129 CWEL  28.3 1206 290 32.9 1459 130 22.1 1655 130 24.2 2313 CWGB 110 27.8 0731 310 30.9 1357 120 27.8 1729 140 25.2 2114 CWGT 130 26.8 1115 300 32.9 1345 120 33.4 1701 150 29.8 2326 46146 109 21.7 1543 292 27.2 1141 110 ~21.3 1641 102 ~16.8 1441 Average   28.3     29.2     26.7     25.5                             Grand Average 26.3     25.1     26.3     25.1    high variability in terms of which location receives the strongest winds. The Puget Lowlands and Georgia Strait Interior tend to have the slowest winds on average, whereas the Coast and Georgia Strait tend to have the fastest, with the Northwest Interior being a close runner-up. Some storms broadly affect the entire area with strong winds (e.g. 15 Dec 2006), while others are much more focused (e.g. 12 Mar 2012). Looking at the "grand average" of all stations for each storm, there is some variability among these selected windstorms with a range of 24.3 to 29.0 m s-1. The strongest storm, 15 Dec 2006, had an average some 2.0 m s-1 faster than the next most powerful event, 12 Nov 2007.  The timing of peak gusts among the stations tends to be highly variable during individual storms, with peaks generally occurring over a period of 10 to 20 h as the ETC progresses through the region (Tables 3.2a and b). This variance tends to be less within specific regions, but can still  Chapter 3: Detailed Analysis of High-Wind Storms  109 Table 3.2b Peak gust direction, magnitude and timing for key areas in and around the region of interest for the remaining four windstorms selected for detailed analysis. Values led by a tilde (~) are the highest available from incomplete datasets. Times followed by an "E" are estimated. Storm: 14-15 Dec 2006 12-13 Nov 2007 02-03 Apr 2010 11-12 Mar 2012 Loca-tion Dir º M m s-1 t UTC Dir º M m s-1 t UTC Dir º M m s-1 t UTC Dir º M m s-1 t UTC Coast                         KHQM 180 28.3 2303 160 31.9 1536 200 25.7 2007 170 17.0 0403 46041       173 27.1 1343       138 30.5 1200 DESW1 195 35.9 0250 165 39.3 1436 217 36.9 2040 140 29.9 1336 KUIL 290 26.2 0815 150 ~25.2 1302 210 24.2 2320 140 22.6 1413 TTIW1 274 35.1 0950 164 39.1 1402 211 42.1 2312 173 28.8 1510 46087 268 29.1 1007 169 29.0 1455       153 27.3 1543 46206 295 28.6 0726 152 ~24.6 1126 110 27.1 1338 148 28.7 1238 CWEB  51.4 0916 260 29.3 1926 140 35.5 2030 140 36.5 1514 Average   33.5     30.7     31.9     27.7                             Juan de Fuca                       CWSP 280 44.7 1100 270 22.6 2059 90 20.1 1500 210 23.8 1800 CWQK 240 43.7 1054       120 36.0 1800E 90 24.7 1322 KCLM 240 ~24.2 1016 250 25.7 1103 160 17.0 18:06 160 18.5 1413 CWLM 260 34.5 1055 130 22.6 1444 220 25.2 0332 100 21.1 1247 46088 247 29.6 1006 132 29.3 1441 131 25.9 2020 126 24.5 1126 Average   35.3     25.1     24.8     22.5                             Puget Lowlands                     KSHN 190 ~24.7 0528 240 15.9 2053 220 18.5 0215 140 13.4 1110 KOLM 190 ~23.7 0753 180 20.6 1844 200 18.0 0054 170 20.1 1415 KTCM 200 30.9 0636 170 23.7 1748 180 19.0 2009 180 18.5 1839 KSEA 220 30.9 0844 190 21.1 1951 180 21.1 2057 170 18.0 1704 WPOW1 196 31.6 0921 198 24.9 1237 179 24.8 2018 175 20.0 1038 KPAE 180 29.3 0646 200 17.5 1251 190 27.8 2223 160 21.1 1032 Average   28.5     20.6     21.5     18.5                             Northwest Interior                     KNUW 250 30.9 1017 160 27.8 1343 150 27.8 2053 140 21.1 0947 SISW1 251 33.8 1102 135 33.3 1544 153 29.0 2027 145 25.3 0523 KFHR 220 26.8 1024 140 27.3 1405 130 23.7 2026 140 23.7 1219 KBLI 160 24.7 0749 150 32.9 1609 150 27.3 2104 140 22.6 1533 Average   29.1     30.3     27.0     23.2                             Chapter 3: Detailed Analysis of High-Wind Storms  110 Storm: 14-15 Dec 2006 12-13 Nov 2007 02-03 Apr 2010 11-12 Mar 2012 Loca-tion Dir º M m s-1 t UTC Dir º Loca-tion Dir º M m s-1 t UTC Dir º Loca-tion Dir º M m s-1 Georgia Strait Interior (Lower Mainland)               CYXX 170 25.2 0600 180 29.8 1739 180 22.1 0332 180 17.5 2000 CYVR 290 26.2 1100 130 21.1 1400 120 24.7 2016 130 24.7 1519 CWSB 290 32.9 1200 100 28.8 1600 100 27.3 2200E 90 26.8 0700E CWMM 180 15.9 1025 150 19.0 1321 150 17.0 2210 140 15.9 1753 CYYJ 270 21.6 1150 140 24.7 1500 120 21.1 1900 130 26.2 1400 CYQQ 330 14.4 0920 130 31.9 1700 140 28.3 0000 130 32.4 1700 Average   22.7     25.9     23.4     23.9                             Georgia Strait                       CWEZ  25.7 1004 160 39.6 1436 150 30.4 1906 160 28.8 1206 CWVF 280 27.8 1100 140 27.3 1538 170 27.3 0227 160 27.3 1502 CWEL 280 27.3 0939 130 24.2 1609 120 27.8 2200E 140 26.8 1344 CWGB 320 23.1 1006 140 26.8 1359 130 30.9 1523 120 27.8 1430 CWGT 330 24.2 0939 120 31.9 1556 130 32.9 2239 120 32.9 1430 46146 307 23.2 0941 135 21.8 1441 120 26.0 2041 111 24.8 1241 Average   25.2     28.6     29.2     28.1                             Grand Average 29.0     27.0     26.3     24.3    be quite extreme, sometimes in excess of 10 h. This is especially true for large regions (e.g. Coast). However, even in regions where the stations are closely spaced (e.g. Northwest Interior), peak gust timing often varies by two to four h, though sometimes it can be quite uniform, <one h. Looking at some characteristics of the high-wind-generating ETCs (Table 3.3), at closest approach to CYVR, most of the storms were moving roughly in a northeast direction, with one each in the north, north-northeast, east-northeast and east categories. All had minimum central pressures below 98.0 kPa, and seven had sub-98.0 kPa central pressures at the time of landfall on Vancouver Island. All of the ETCs exhibited fast deepening rates at some stage in their evolution, generally in excess of 5 hPa (3) h-1. Seven of the eight ETCs persisted long enough for a 24 h rate of deepening to be calculated: Of the seven, four deepened at >20 hPa (24) h-1, or one Bergeron at the latitude of 50º N. A description of Bergerons is in Sanders and Gyakum (1980). Chapter 3: Detailed Analysis of High-Wind Storms  111 Table 3.3 General track and central pressure statistics for eight windstorms with either north Vancouver Island (NVI) or south Vancouver Island (SVI) tracks. Storm Event (LST) Storm Track Desig-nation General Track Direction at Closest Approach to CYVR ETC Min Central Pres (kPa) ETC Central Pres at Landfall (kPa) ETC Max Deepen-ing Rate (hPa [3] h-1) ETC Max Filling Rate (hPa [3] h-1) ETC Max Deepen-ing Rate (hPa [24] h-1) ETC Absolute Central Pressure Change (hPa) 03Mar99 SVI NE 96.8 97.6 15 9 38 45 14Dec01 SVI ENE 97.9 97.9 6 1 8 8 15Nov06 NVI E 97.6 98.2 7 7 17 20 11Dec06 NVI NE 97.8 97.8 8 9 9 18 15Dec06 SVI NE 97.4 97.6 7 6 20 32 12Nov07 NVI NNE 96.7 96.7 9 4 31 39 02Apr10 SVI NE 97.6 97.7 10 6 33 36 12Mar12 NVI N 96.2 96.2 8 10 41 49  3.3.2 Synoptic Analysis   From a synoptic standpoint, all eight windstorms appear to have followed, roughly, typical development cycles (see Shapiro and Keyser 1990, Carlson 1991, Neiman and Shapiro 1993, Neiman et al. 1993, Stull 2000, and Lackmann 2013 for discussions of cyclogenesis), with cyclogenesis taking place in a region of marked baroclinicity, under a strong jet stream and in the diffluent region just downstream of the 50 kPa trough axis. The 03 Mar 1999 storm appears to have had the most classic pattern of cyclogenesis and cyclolysis, with the low developing on the left side of a jet max on the lower eastern side of the 50 kPa trough, and with strong warm advection (seen in the 50 kPa temperatures) moving up the Pacific Coast ahead of the low being closely followed by marked cold advection behind and into the base of the low (Figures 3.6 to 3.8). The ETC occluded before reaching Vancouver Island, and the surface low center migrated under the 50 kPa trough, becoming vertically stacked with the low isolated from the jet support. The mature ETC fell apart just after landfall. Chapter 3: Detailed Analysis of High-Wind Storms  112  Figure 3.6 Synoptic analysis 1200 UTC 02 Mar 1999. Orange shading depicts the 30 kPa jet stream, with orange and brown numbers labeling isotachs in m s-1. For the 50 kPa level, black lines denote the heights in dm; upper lows and highs are also marked in black. Isotherms in ºC are indicated with white dashed lines.  At the surface: Lows and central pressures in hPa are indicated with dark blue, and tracks in light blue. Highs are indicated in red. Troughs are delineated with an orange dashed line. Key fronts are also indicated. Not all surface features are shown.     Chapter 3: Detailed Analysis of High-Wind Storms  113   Figure 3.7 Synoptic analysis 0000 UTC 03 Mar 1999. See Figure 3.6 caption for the key.    Chapter 3: Detailed Analysis of High-Wind Storms  114   Figure 3.8 Synoptic analysis 1200 UTC 03 Mar 1999. See Figure 3.6 caption for the key.  The 50 kPa troughs associated with the eight storms had a wide range of form. Most troughs assumed U-shapes, some of them quite broad and even nearly meridional in the most extreme cases while others were sharply-defined and almost V-shaped. The trough axes were generally located around 135 to 145º W at the time of cyclogenesis, with a position around 140º W being the most common. The upper trough appears to move eastward more progressively with the storms that have more meridional tracks. For example, the 14 Dec 2001 trough shows the Chapter 3: Detailed Analysis of High-Wind Storms  115 most rapid movement, covering about 20º longitude in 24 h (Figures 3.9 to 3.11). Conversely, the 50 kPa trough associated with the ETC that had the most latitudinal track, 12 Mar 2012, hardly moved at all (Figures 3.12 to 3.14). This trough also had the most amplified form at the time of ETC closest approach, almost assuming a V-shape. Lows with the upper trough following closely behind tended to maintain integrity better upon interaction with the steep coastal terrain, such as 14 Dec 2001 and 15 Nov 2006, than those ETCs that ran ahead of the trough—i.e. ETCs tended to fall apart more easily at landfall when either the trough flattened out or when the system became vertically stacked, or cut-off from the jet support. The 30 kPa jet streams present during the eight storms largely mirror the patterns seen with the 50 kPa troughs, as there is a close association between jet streams and regions where the 50 kPa height lines are tightly packed. ETCs that remained immediately on the eastern side of a trough typically had the best jet support, as the left exit region of the jet maximum is often located in the vicinity. Surface lows that moved away from the jet support became vertically stacked and weakened slowly when over the ocean and then more rapidly upon landfall. Unique cases include the 15 Nov 2006 ETC. This storm continued to deepen post-landfall while being supported by the right-entrance of a jet maximum, a feature that moved inland with the surface low, though the left exit of a second jet streak may have also played a role in the maintenance of the system later on in its evolution (Appendix C Figures C.1 to C.3). The 12 Nov 2007 ETC spun up in the vicinity of the right entrance region of a jet streak, then subsequently moved under the jet axis to the support of the left exit region just before landfall (Appendix C Figures C.4 to C.6). The 14 Dec 2001 storm is another case where the upper support moved ashore almost in lock-step with the surface low (Figures 3.9 to 3.11). This ETC also maintained depth despite landfall (section 3.3.3 has further details on central pressure).  Chapter 3: Detailed Analysis of High-Wind Storms  116  Figure 3.9 Synoptic analysis 1200 UTC 13 Dec 2001. See Figure 3.6 caption for key.    Chapter 3: Detailed Analysis of High-Wind Storms  117  Figure 3.10 Synoptic analysis 0000 UTC 14 Dec 2001. See Figure 3.6 caption for key.    Chapter 3: Detailed Analysis of High-Wind Storms  118  Figure 3.11 Synoptic analysis 1200 UTC 14 Dec 2001. See Figure 3.6 caption for key.    Chapter 3: Detailed Analysis of High-Wind Storms  119  Figure 3.12 Synoptic analysis 0000 UTC 12 Mar 2012. See Figure 3.6 caption for key.    Chapter 3: Detailed Analysis of High-Wind Storms  120  Figure 3.13 Synoptic analysis 1200 UTC 12 Mar 2012. See Figure 3.6 caption for key.    Chapter 3: Detailed Analysis of High-Wind Storms  121  Figure 3.14 Synoptic analysis 0000 UTC 13 Mar 2012. See Figure 3.6 caption for key. At the 85 kPa level, some of the ETCs had a discrete structure with many closed height-lines (Figure 3.15). Most, however, were in the base of deep troughs, with weak height gradients to the north, but strong to the south, mirroring surface conditions to some extent (Figures 3.16 to 3.17). In all cases, strong gradients near the low center were brought over the study region (Figures 3.15 to 3.17 and Appendix C Figures C.13 to C.18). For southeasters, there appears to be good support for surface winds with a southerly component, with 85 kPa airflow inferred Chapter 3: Detailed Analysis of High-Wind Storms  122 from the height lines ahead of the lows being from the S to SW for north Vancouver Island landfalls and from the SE to S for south Vancouver Island landfalls. In the case of the two analyzed westerly windstorms, surface winds from the NW, right along the axis of the Georgia Strait, were well supported at the 85 kPa level behind the lows.   Figure 3.15 Twelve-hourly analysis at the 85 kPa height covering 1200 UTC 02 Mar to 1200 UTC 03 Mar 1999. Light gray lines denote the lowest several height lines during the initial time-step, with black marking the second and dark gray the third. Dashed lines denote the -5ºC isotherm.   Chapter 3: Detailed Analysis of High-Wind Storms  123  Figure 3.16 Twelve-hourly analysis at the 85 kPa height covering 1200 UTC 13 Dec to 1200 UTC 14 Dec 2001. See Figure 3.15 caption for the key.   Chapter 3: Detailed Analysis of High-Wind Storms  124  Figure 3.17 Twelve-hourly analysis at the 85 kPa height covering 0000 UTC 12 Mar to 1200 UTC 13 Mar 2012. See Figure 3.15 caption for the key.  3.3.3 Mesoscale Analysis   The mesoscale surface maps created for the eight storms are focused on conditions around the time of peak winds (Figures 3.18 to 3.25 and Appendix D Figures D.1 to D.13). Usually the second map in the three-chart sequence marks the time of maximum wind at one or more of the Chapter 3: Detailed Analysis of High-Wind Storms  125 study stations. Interestingly, all the sequences for southeasterly windstorms depict a landfalling low. This suggests a tendency to have the strongest winds occur just before or around the time of low-center landfall on Vancouver Island. The key exceptions are the two westerly windstorms where the strongest winds happened after the ETC passed to the east. The general trend of major terrain features is from northwest to southeast, emphasized by the long axis of Vancouver Island and the Georgia Strait. The orientation of key frontal boundaries in all eight scenarios often had roughly the same orientation as the geography, sometimes termed "negative tilt" in weather forecasting circles (think in terms of a trend-line on an x-y graph: positive tilt is southwest to northeast). These fronts also tend to move inland ahead of the low, with the obvious exception of any bent-back or secondary boundaries. This tends to put at least one of the fronts, be it warm, cold or occluded, very close to if not right on top of the study stations at about the same time that the low center is making landfall.  Tracking warm fronts through the complex coastal terrain proved difficult. Warm fronts tended to move slowly and seemed strongly impeded by the steep terrain in the region. In at least one case, the warm front stalled upon reaching the coast, allowing the cold front to catch up, and the resulting occlusion then swept inland (Appendix D Figures D.10 to D.12). Some storms had large and progressive warm sectors move into the study region (Figures 3.21 to 3.22 and Appendix D Figures D.1 to D.3). In other windstorms, apparent warm sectors managed to migrate into the interior valleys and then stalled against the Coast and Cascade Mountain ranges, allowing the cold front to catch up and create a seclusion of warm air (Figures 3.18 to 3.20 and Appendix D Figures D.4 to D.6). It is possible that the warm air indicated to the northeast of the 03 Mar 1999 low is the result of the warm seclusion process described by Shapiro and Keyser  Chapter 3: Detailed Analysis of High-Wind Storms  126  Figure 3.18 Mesoscale surface analysis for 1400 UTC 03 Mar 1999. Wind speed is 2.5 m s-1 per half-barb; 25 m s-1 for a pennant. A full circle represents calm conditions. Solid isobars are drawn in four hPa increments; dashed isobars showing further detail are at two hPa resolution.    Chapter 3: Detailed Analysis of High-Wind Storms  127  Figure 3.19 Mesoscale surface analysis for 1600 UTC 03 Mar 1999. See Figure 3.18 caption for the key.   Chapter 3: Detailed Analysis of High-Wind Storms  128  Figure 3.20 Mesoscale surface analysis for 1800 UTC 03 Mar 1999. See Figure 3.18 caption for the key.   Chapter 3: Detailed Analysis of High-Wind Storms  129  Figure 3.21 Mesoscale surface analysis for 0100 UTC 14 Dec 2001. See Figure 3.18 caption for the key.   Chapter 3: Detailed Analysis of High-Wind Storms  130  Figure 3.22 Mesoscale surface analysis for 0500 UTC 14 Dec 2001. See Figure 3.18 caption for the key.   Chapter 3: Detailed Analysis of High-Wind Storms  131  Figure 3.23 Mesoscale surface analysis for 1000 UTC 14 Dec 2001. See Figure 3.18 caption for the key.   Chapter 3: Detailed Analysis of High-Wind Storms  132  Figure 3.24 Mesoscale surface analysis for 1500 UTC 12 Mar 2012. See Figure 3.18 caption for the key.   Chapter 3: Detailed Analysis of High-Wind Storms  133  Figure 3.25 Mesoscale surface analysis for 1700 UTC 12 Mar 2012. See Figure 3.18 caption for the key.   Chapter 3: Detailed Analysis of High-Wind Storms  134  Figure 3.26 Mesoscale surface analysis for 1900 UTC 12 Mar 2012. See Figure 3.18 caption for the key.  (1990), and not the original warm sector. One storm arrived fully occluded, with no indication of warm sector air entering the region (Appendix D Figures D.13 to D.15). During the detailed analysis of the 50 to 60 stations used in creating the mesoscale maps, it became clear that in many instances the warm-sector air simply lifts over a commonly-entrenched cold surface layer that occupies the Salish Sea region. This cold surface layer may be supplied via gap winds, such as through the Fraser Canyon, or is the cool, stable marine air emplaced after the passage of an earlier storm. Some warm fronts appear to erode the denser, shallow layer more quickly than others. Sometimes a surface warm front did not show up in Chapter 3: Detailed Analysis of High-Wind Storms  135 sheltered areas such as on the east side of the Olympics where the cold air can get dammed against the mountains in the offshore-flow environment ahead of an incoming ETC, but will surface further inland, such as on the east side of the Puget Sound. There is evidence of warm air surfacing in isolated pockets that may be temporary in nature, especially in the south Puget Sound region, perhaps a reflection of down-sloping winds off of the Cascade Mountains. Primary cold fronts appear to progress through the region more steadily and uniformly than warm fronts. However, at times cold fronts seem to be slowed by the higher terrain, with subsequent deformation of the shape of the boundary as the front sweeps around obstacles (e.g. Appendix D Figures D.4 to D.6). In some cases, the cold front occluded as it moved inland (e.g. Figures 3.18 to 3.20 and 3.24 to 3.26), apparently in part due to the slowing of the leading warm front as it encountered the rugged coastal terrain. The resulting occlusions tend to progress in the manner of cold fronts. Sometimes the leading cold fronts and occlusions mimicked warm fronts when they scoured out any cold surface layer that was in place—during the wet season, moist, cool air off the pacific is often warmer than trapped continental polar air. These two types of boundaries appear to be more effective at eroding the entrenched air than warm fronts. This is reflected in the less-kinked shape of these fronts compared to warm fronts. Identifiable bent-back fronts were associated with at least seven out of the eight windstorms. The one exception had a triple-point low structure and deepened as it moved inland, indicating a storm still approaching peak intensity (Appendix D Figures D.1 to D.3). The other triple-point low, also still deepening at landfall appears to have had a bent-back front, though this one trailed further behind the low than many of the others (Figures 3.21 to 3.23). The boundary moved quite slowly, produced a very weak surface temperature change if at all and turned out to be easier to track via sharp drops in dew point. Maximum winds tended to occur with this Chapter 3: Detailed Analysis of High-Wind Storms  136 feature, though with a slower build-up than with some other bent-back fronts. This particular bent-back front may have been more a surface reflection of the upper-level trough that moved in behind the low than an actual front, perhaps a combination of weak surface front and upper-trough reflection. This also points to a possible relationship between the position of the bent-back front and upper (50 kPa) trough, as seen in the synoptic analysis, with the front being located near the base of the trough during some storms (Figures 3.7, 3.10 and Appendix C Figures C.7, C.11 and C.14). At least one bent-back front had a weak to moderate signature at many locations (Appendix D Figures D.13 to D.15). This bent-back front faded fast when it reached the coast. Almost no temperature change occurred as the front passed coastal stations, such as at Estevan Point (Figure 3.27). Maximum winds from this storm tended to occur well ahead of the bent-back front, and appear more related to the closest approach of the low before it began rapidly filling post-landfall, or perhaps to the leading occlusion. Important exceptions occurred at some north WA coastal stations including Tatoosh Island, which had an intense peak wind (32.9 m s-1) associated with the bent-back front. Interestingly, even this station also did not report a marked change in temperature with the arrival of the front. Other bent-back fronts were strong, such as with 11 Dec 2006 (Appendix D Figures D.4 to D.6). This front brought sharp temperature drops along the coast, as exemplified by Solander Island as opposed to Estevan Point to control for track differences between the 11 Dec 2006 and 02 Apr 2010 storms (Figure 3.27a). The cold air was in a narrow band, as temperatures rebounded to more typical pre-storm offshore-flow levels, seen before t -5 h, within several hours. Intense winds were associated with this bent-back front, including the maximum gusts at many of the affected stations. At Estevan Point, peak winds associated with the bent-back front  Chapter 3: Detailed Analysis of High-Wind Storms  137  Figure 3.27 Comparison of bent-back front surface responses for the 02 Apr 2010, 11 Dec 2006 and 15 Dec 2006 windstorms, using data from Estevan Point, Solander Island and Buoy 46206. Wind direction (WDIR, thin dashed lines) and temperature (T, heavy solid lines) are shown. Available Buoy 46206 data (reporting is spotty at times) are included mainly due to the absence of wind direction reports from Estevan Point during the 15 Dec 2006 windstorm. Chapter 3: Detailed Analysis of High-Wind Storms  138 were 19.5 m s-1 during the 02 Apr 2010 storm and 22.6 m s-1 during 11 Dec 2006. At Solander Island, maximum winds reached 26.8 m s-1 during the 11 Dec 2006 storm; during the 02 Apr 2010 storm, the bent-back front appears to have skirted south of this station, and a 19.5 m s-1 peak was apparently associated with the upper trough tracking immediately behind the low. In terms of wind, the 2006 Hanukkah Eve Storm brought with it perhaps the strongest bent-back front among the eight storms: peak gusts at many locations occurred with the arrival with this feature (Table 3.2b). The intensity of this front is reflected in the presence of marked preferential pressure falls associated with the boundary (Appendix D Figures D.7 to D.9). Strong temperature falls also occurred with this front (Figure 3.27b), and the cooling persisted unlike during 11 Dec 2006, suggesting a much larger field of cold air moved in behind this storm. Estevan Point, just clipped by the northern arc of the bent-back front, experienced a maximum wind of 26.8 m s-1. Stations a bit further south, more equivalent to Estevan Point during the 02 Apr 2010 storm relative to storm track position, had similar peaks, with Tatoosh Island reporting 27.1 m s-1 for example. Many of the bent-back fronts, whether weak or strong, appear to follow immediately behind the ETC center and wrap into the base of the low in a long, gentle arc, sometimes taking on the form of an additional front with similar tilt to the leading boundaries. This puts the low center between the leading occlusion (or warm/cold fronts) and the bent-back front, like a pearl inside two halves of an oyster shell. In one case the front wrapped around the low in a tight spiral, with the effect of bringing multiple frontal passages to parts of northern Vancouver Island (Figures 3.24 to 3.26). In contrast, another intense ETC at near peak intensity near landfall, almost a twin of the previously mentioned storm, did not have evidence for such a tight spiral and instead had a bent-back front with more typical form (Appendix D Figures D.10 to D.12). Chapter 3: Detailed Analysis of High-Wind Storms  139 Looking at the isobaric patterns, the strongest gradients tend to be on the south sides of ETCs, with shallower gradients to the north. Sometimes the difference is very dramatic (e.g. Figure 3.20), and at other times the low has a more symmetrical form (e.g. Figure 3.24). Preferential pressure falls can occur along any of the frontal boundaries and sometimes locally enhance gradients, providing a potential boost to surface winds. During the passage of several of the analyzed ETCs, there is the clear tendency for lee troughing to the north and northeast of the Olympic mountains. With four hPa isobaric increments, this sometimes shows up as a closed surface low (e.g. Appendix D Figure D.11), or simply a widening of the isobaric separation in region (e.g. Appendix D Figure D.1). The pressure gradients around the incoming ETCs can change fairly dramatically over short time periods and tend to be at their strongest on the coast and/or along frontal systems. By simply measuring from the maps, maximum pressure gradient magnitudes at the most focused location around the low centers have ranged from 6.9 hPa (100) km-1 in one case to 14.8 hPa (100) km-1 during three events. Surface winds for north Vancouver Island landfalls, all southeasters in this examination, in the Georgia Strait and in the Northern Inland Waters tend to be SE to ESE as these lows move inland (i.e. at the time of peak wind), with the strongest speeds in the vicinity of the Georgia Strait and along the coast. Winds further south, such as in the Puget Lowlands, tend to be more S. Indeed, the airflow seems to follow an arc that reflects the terrain, with a S direction over the Puget Sound becoming SE in the Northern Inland Waters and the Georgia Strait to nearly due E over the Queen Charlotte Sound. These winds are usually supported by an east to southeast pressure slope. Terrain may strongly influence these general directions in places: for example, when E winds are blowing at Atkinson Point (CWSB), they may be coming from the S at Pam Rocks (CWAS) (Figure 3.26). Wrapping around the low, in the area of northwest pressure slope, Chapter 3: Detailed Analysis of High-Wind Storms  140 N to NW winds tend to sweep across north Vancouver Island (Appendix D Figures D.13 to D.15), and gradually migrate south- and east-ward as weather systems move inland. Sometimes, a marked cold frontal boundary may carry W to NW winds into the southern Georgia Strait (e.g. the 15 Nov 2006 windstorm, not shown), but this tends to be after the time of peak winds. Also, if the low eventually tracks due north of the study region, winds will tend to shift more to the S and even SSW over the Georgia Strait, again well after peak winds. For south Vancouver Island landfalls that produced southeast windstorms, surface winds tend to have a pattern similar to north Vancouver Island lows. However, the wind direction shifts dramatically north of about the latitude of the ETC center, with a tendency for NW winds even in the northern Georgia Strait and a region of surface convergence in the northeast quadrant of the low (e.g. Figure 3.19). For westerly events, there is a dramatic shift from E to SE wind to W to NW as the lows track inland. Intense W to NW winds surge down the Georgia Strait and through the Strait of Juan de Fuca when a southwest to west pressure slope is carried into the region. With a nearly uniform air mass in place post-front, the strong W flow appears to impede the S winds in the Puget Lowlands, and results in an area of convergence in the lee of the Olympic Mountains (e.g. Appendix D Figure D.9). Mass (1981) has further discussion on the Puget Sound convergence zone. The distribution of high-wind gusts is fairly similar among the four ETCs that landed on northern Vancouver Island (Figures 3.28 to 3.31). The entire coast is prone to extreme gusts, as are the regions including the Queen Charlotte Strait, Georgia Strait and Northern Inland Waters. Sometimes high wind gusts occur in the Puget Lowlands and the Strait of Juan de Fuca. For south Vancouver Island tracks (Figures 3.32 to 3.35), the distribution of high-wind gusts is less uniform, with the two southeasters tending to mainly affect the region south of their respective  Chapter 3: Detailed Analysis of High-Wind Storms  141  Figure 3.28 Low center track and peak gusts for the 15 Nov 2006 north Vancouver Island ETC. Peak gusts are shown to the nearest m s-1. High-wind criteria speeds (~ ≥25 m s-1) are indicated with red-filled circles. The time indicated for the storm center positions (large circle with "X") is in UTC, with the starting day being the windstorm designation. Central pressure is indicated in kPa.   Chapter 3: Detailed Analysis of High-Wind Storms  142  Figure 3.29 Low center track and peak gusts for the 11 Dec 2006 north Vancouver Island ETC. See Figure 3.28 caption for the key.   Chapter 3: Detailed Analysis of High-Wind Storms  143  Figure 3.30 Low center track and peak gusts for the 12 Nov 2007 north Vancouver Island ETC. See Figure 3.28 caption for the key.   Chapter 3: Detailed Analysis of High-Wind Storms  144  Figure 3.31 Low center track and peak gusts for the 12 Mar 2012 north Vancouver Island ETC. See Figure 3.28 caption for the key.   Chapter 3: Detailed Analysis of High-Wind Storms  145  Figure 3.32 Low center track and peak gusts for the 03 Mar 1999 south Vancouver Island ETC. See Figure 3.28 caption for the key.   Chapter 3: Detailed Analysis of High-Wind Storms  146  Figure 3.33 Low center track and peak gusts for the 14 Dec 2001 south Vancouver Island ETC. See Figure 3.28 caption for the key.   Chapter 3: Detailed Analysis of High-Wind Storms  147  Figure 3.34 Low center track and peak gusts for the 15 Dec 2006 south Vancouver Island ETC. See Figure 3.28 caption for the key.   Chapter 3: Detailed Analysis of High-Wind Storms  148  Figure 3.35 Low center track and peak gusts for the 02 Apr 2010 south Vancouver Island ETC. See Figure 3.28 caption for the key. storm tracks, while the westerly windstorms affected much of the coast, the Strait of Juan de Fuca, Northern Inland Waters, Georgia Strait and Queen Charlotte Strait. The storm tracks show some tendency for ETCs to slow down just before landfall, and in some cases almost stall (Figures 3.29, 3.31, 3.32 and 3.35). Once ashore, the speed of these ETCs tends to increase and can resume values equivalent to what the storm exhibited over-ocean. The two westerly windstorms do not show as much slowing at landfall as the six southeasters (Figures 3.33 and 3.34). In terms of track direction, both westerly windstorms had east-northeast tracks as they reached closest approach to the study region. The six southeasters generally had northeast to north directions at closest approach. The one exception tracked further Chapter 3: Detailed Analysis of High-Wind Storms  149 away from the study region than the two westerly storms, across northern Vancouver Island, a situation more conducive to SE winds (Figure 3.28). However, even this low brought a modest westerly wind in its wake, with CYYJ reporting peak winds from the W. With regards to central pressure tendencies, ETCs fall into two groups: Those that began weakening around the time of landfall, and those that did not (Figure 3.36). Six of the eight analyzed windstorms are in the weakening category. Four of these showed rising central pressures within one to three h of landfall, with the other two starting to fill earlier. Of the two storms that did not immediately weaken upon reaching the coast, one maintained depth, though just barely since the 98.0 kPa closed isobar slowly shrank over time (Figures 3.21 to 3.23), while the other deepened an additional six hPa overland. The average for all eight storms shows peak depth occurring around an hour before landfall, and filling at one to two hPa h-1 afterward.  3.3.4 Comparative Analysis of Surface Response  Aside from the eight windstorms examined in detail, an additional ten southeasterly and westerly windstorms occurred during the study period (Figure 3.37). Information from these ETCs is brought into this and following sections. For the comparative analysis of surface response, 14 of these storms are used, including the eight selected for detailed analysis. There is good correspondence between the pressure slope over the Georgia Strait and the bearing between CYVR and the ETC center among the 14 strongest windstorms that landed on Vancouver Island (Figure 3.38). Using observations from the 14 storms, a linear regression run on the two variables, when the jump between zero and 359º is accounted for, has an R2 of 0.59 (n=75, slope=0.780). The pressure slope for southeasterly windstorms that track across northern Vancouver Island tends to begin around 90 to 120º when the low is offshore, shifting to roughly  Chapter 3: Detailed Analysis of High-Wind Storms  150  Figure 3.36 ETC central pressure trends before and after landfall. Blue lines highlight two storms that did not exhibit filling for several hours after landfall. Dashed lines depict those storms with northern Vancouver Island tracks and solid lines those that moved into southern Vancouver Island. The orange line is an average of all eight storms. 180 to 220º as the ETC tracks inland. Westerly windstorms appear to have similar outcomes. For ETCs with south Vancouver Island tracks that produce southeasters, the pressure slope tends to be more strongly easterly, around 30 to 90º as the low approaches, shifting to around 140 to 180º as the low tracks inland. For westerly windstorms, the pressure slope again starts out around 30 to 90º, but swings more strongly to the west, around 200 to 270º, even higher, as the low tracks to the east and northeast of the study region. For northern Vancouver Island tracks, peak winds for southeasterly windstorms tend to occur at the time the ETC center has reached closest approach (Figure 3.39). Winds gradually climb over 12 to 24 h as the low approaches, and sometimes escalate more dramatically in  Chapter 3: Detailed Analysis of High-Wind Storms  151  Figure 3.37 The tracks of an additional ten ETCs that produced either southeast or westerly windstorms during 1994 to 2012. These storms were not analyzed with the detail of the initial eight presented in the preceding figures. Paths with a blue shade denote those ETCs that landed on south Vancouver Island, and tracks in red and orange are from those storms that moved onto the north end of the island. The gray track traces the path of two major windstorms that occurred outside of the study period: the 12 Oct 1962 "Columbus Day Storm" (AKA "Typhoon Freda") and the 13-14 Nov 1981 ("Friday-the-13th" windstorm) and are provided for reference. the final ~six hours at some stations. Wind direction tends to be SE, sometimes E to NE, ahead of the ETC when the low center is to the southwest to west-southwest, shifting to S and then SW as the low center moves inland to the north and northeast (Figure 3.40). Wind direction has Chapter 3: Detailed Analysis of High-Wind Storms  152 marked variability at times. For the two westerly windstorms available in the data, maximum wind speed tends to occur after the low has passed inland (Figure 3.41), though perhaps this generalization is better applied to those stations that are prone to strong westerly winds, such as CYVR. CYQQ, for instance, showed a tendency to have peak winds before the low landed. Wind direction tends to be offshore ahead of the low, ranging from NE to SE depending on the station with the low center to the west-southwest; and then shifting abruptly to the SW to NW as the low tracks inland to the north and then northeast (Figure 3.42). Looking at southern Vancouver Island tracks, southeasterly windstorm peak winds tend to occur around the time of closest approach, quite similarly to storms crossing the north half of the Island (Figure 3.43). Winds seem to steadily rise toward peak speeds over about a 12 to 24 h period. Wind direction tends to start out NE at many of the stations when the low is tracking toward the region, shifting to SE as the low center reaches closest approach and then becoming S to SW as the low moves inland (Figure 3.44). For the westerly windstorms, maximum winds occurred after the low center had passed through the region in nearly all cases (Figure 3.45). Wind direction tended to start out similarly to the southeasters though there appears to be more variability, and then shifts abruptly to the WSW to NW after the low tracks through (Figure 3.46). In general, surface winds in the region of interest begin to be influenced by incoming ETCs usually when the low travels inside 1500 km, and perhaps more typically inside 1000 km, with marked variation.    Chapter 3: Detailed Analysis of High-Wind Storms  153  Figure 3.38 Georgia Strait pressure slope (thick lines) compared to the bearing between CYVR and the low-pressure center (thin lines) for 14 of the 18 Vancouver Island-landfalling ETCs that produced either a southeaster or westerly windstorm from 1994 to 2012. NVI is north Vancouver Island and SVI is south Vancouver Island. The approximate time of landfall is indicated by gray shading.   Chapter 3: Detailed Analysis of High-Wind Storms  154  Figure 3.39 Distance of the low-pressure center from CYVR and two-min wind speed in six-h increments. For southeasterly windstorms with ETC landfalls on northern Vancouver Island.   Chapter 3: Detailed Analysis of High-Wind Storms  155  Figure 3.40 Bearing of the low-pressure center from CYVR and two-min wind direction in six-h increments. For southeasterly windstorms with ETC landfalls on northern Vancouver Island.   Chapter 3: Detailed Analysis of High-Wind Storms  156  Figure 3.41 Distance of the low-pressure center from CYVR and two-min wind speed in six-h increments. For westerly windstorms with ETC landfalls on northern Vancouver Island. During the study period, there were only two events in this category.  Figure 3.42 Bearing of the low-pressure center from CYVR and two-min wind direction in six-h increments. For westerly windstorms with ETC landfalls on northern Vancouver Island. During the study period, there were only two events in this category.   Chapter 3: Detailed Analysis of High-Wind Storms  157  Figure 3.43 Distance of the low-pressure center from CYVR and two-min wind speed in six-h increments. For southeasterly windstorms with ETC landfalls on southern Vancouver Island.   Chapter 3: Detailed Analysis of High-Wind Storms  158  Figure 3.44 Bearing of the low-pressure center from CYVR and two-min wind direction in six-h increments. For southeasterly windstorms with ETC landfalls on southern Vancouver Island.   Chapter 3: Detailed Analysis of High-Wind Storms  159  Figure 3.45 Distance of the low-pressure center from CYVR and two-min wind speed in six-h increments. For westerly windstorms with ETC landfalls on southern Vancouver Island.   Chapter 3: Detailed Analysis of High-Wind Storms  160  Figure 3.46 Bearing of the low-pressure center from CYVR and two-min wind direction in six-h increments. For westerly windstorms with ETC landfalls on southern Vancouver Island.  3.3.5 Pressure Gradients and Surface Wind Speed   Maximum pressure gradients for the eight detailed ETCs varied considerably between storms and regions (Table 3.4). Nevertheless, there are some patterns. For north Vancouver Chapter 3: Detailed Analysis of High-Wind Storms  161 Island landfalls, the northernmost station triads tended to experience the strongest gradients, followed by the north WA coast. For south Vancouver Island landfalls, the north WA coast tended to experience the strongest gradients. Overall, the most intense gradients tend to occur on the coast. Looking at the pressure slopes at the time of peak gradient, for north Vancouver Island landfalls, there is a general tendency for the orientation to shift from southeast in the southern regions to east and northeast in the northern regions and even due north at the extreme north end. For southern Vancouver Island landfalls, the pressure slope has a similar pattern, though along the south and central Vancouver Island coast the orientation tends to be shifted more to the northeast than east to east-southeast as with the more northern storms. Table 3.4 Maximum two-dimensional pressure gradients, with associated pressure slopes, for the eight detailed windstorms. Pressure slope averages in italics were determined from the vector components. Geostrophic wind triangle locations are in Figure 3.4.  Maximum Pressure Gradient (hPa [100] km-1), Pressure Slope (º) Windstorm KAST-KHQM-KOLM KHQM-46087-CWSP 46087-CWEB-CYQQ CWEB-CWRU-CYZT CWRU-CYZT-46207 KOLM-CYVR-CYHE CYYJ-CYQQ-CWSK North Vancouver Island Landfalls 15-Nov-06 4.4, 130 12.3, 111 6.2, 98 15.2, 83 10.0, 20 5.8, 131 8.2, 102 11-Dec-06 3.6, 120 11.6, 110 8.6, 121 9.9, 65 10.1, 358 5.7, 117 8.7, 118 12-Nov-07 3.5, 160 10.7, 114 7.3, 87 15.0, 80 11.8, 347 4.8, 116 8.5, 99 12-Mar-12 3.7, 113 5.8, 113 8.0, 101 18.3, 76 15.2, 3 4.3, 99 7.2, 90 Average 3.8, 131 10.1, 112 7.5, 102 14.6, 76 11.8, 2 5.2, 116 8.2, 102 South Vancouver Island Landfalls 03-Mar-99 6.6, 162 14.7, 119 8.0, 41 5.2, 300 5.5, 6 8.1, 136 5.4, 34 14-Dec-01 5.1, 137 7.3, 95 5.6, 267 6.5, 83 5.2, 312 4.6, 150 5.7, 229 15-Dec-06 11.1, 167 10.5, 219 9.7, 60 6.0, 30 7.2, 358 7.6, 156 7.3, 221 02-Apr-10 5.4, 102 15.2, 108 6.6, 62 7.1, 73 6.3, 11 5.1, 145 6.0, 116 Average 7.1, 142 11.9, 127 7.5, 45 6.2, 48 6.1, 354 6.4, 147 6.1, 185  Of the eight ETCs, some landed on the coast with extreme pressure gradients near the core (Figure 3.47). In the case of the depicted 12 Mar 2012 windstorm, going from south to north, peak pressure gradients among the station triads nearly doubled with each step. An exponential Chapter 3: Detailed Analysis of High-Wind Storms  162 curve fitted to the pressure gradients from four of the five coastal geostrophic wind triangles, South Washington to North Vancouver Island (i.e. from the edge to near the center of the cyclone), has an R2 of 0.92 just before landfall. For the 15 Nov 2006 and 12 Nov 2007 windstorms (not shown), these values are 0.82 and 0.73 respectively. These ETCs can be said to have funnel-shaped pressure profiles. Other ETCs land with a more even distribution of pressure gradients (Figure 3.48). For the depicted 11 Dec 2006 windstorm, an exponential curve fitted to the coastal pressure gradients near the time of landfall has an R2 of 0.44, and in fact the last three data-points have similar magnitudes—in essence a roughly even distribution instead of increasing toward the center. The 14 Dec 2001 windstorm (not shown) had the same kind of profile, but even more strongly, with an R2 of 0.15 for an exponential fit. Pressure profiles for this type of storm might be called V-shaped. Some lows appear to end up with bowl, or U-shaped profiles, i.e. with broad centers containing shallow pressure gradients surrounded by a region of stronger gradient far from the center (Figure 3.49). For the depicted 02 Apr 2010 windstorm, an area of strong pressure gradient developed along the south side of the low, with a broad center passing right through the South Vancouver Island Coast geostrophic wind triangle. Along the north side of the low, the pressure gradient remained weak: in this instance the "bowl" had a lopsided shape. The 03 Mar 1999 windstorm had similar form (Figure 3.50), but the area of strongest gradient occurred immediately south of the low center, supporting a funnel-shaped profile, with the north side being a shallow bowl. Chapter 3: Detailed Analysis of High-Wind Storms  163  Figure 3.47 Coastal pressure gradients in hPa (100) km-1 for the 12 Mar 2012 windstorm based on hourly observations. Geostrophic wind triangles are arranged south to north with the warmer colors representing the southern sites.   Chapter 3: Detailed Analysis of High-Wind Storms  164  Figure 3.48 Coastal pressure gradients in hPa (100) km-1 for the 11 Dec 2006 windstorm based on hourly observations.    Chapter 3: Detailed Analysis of High-Wind Storms  165  Figure 3.49 Coastal pressure gradients in hPa (100) km-1 for the 02 Apr 2010 windstorm based on hourly observations.    Chapter 3: Detailed Analysis of High-Wind Storms  166  Figure 3.50 Coastal pressure gradients in hPa (100) km-1 for the 03 Mar 1999 windstorm based on hourly observations.   Chapter 3: Detailed Analysis of High-Wind Storms  167  Figure 3.51 Coastal pressure gradients in hPa (100) km-1 for the 15 Dec 2006 windstorm based on hourly observations. The 15 Dec 2006 windstorm produced a unique result that does not quite fit in any of the above categories (Figure 3.51). The pressure gradient profile at the time of landfall appears somewhat inverted with the strongest gradient in the southern-most region, with decreasing magnitude toward the low center. Strong gradients also existed to the north side of the low, unlike a number of the examined ETCs. Maximum pressure gradients did not tend to coincide between the regions as with some storms (e.g. 12 Mar 2012): For the South Coast Vancouver Chapter 3: Detailed Analysis of High-Wind Storms  168 Island geostrophic wind triangle, the gradient peaked well ahead of landfall at 2100 UTC 14 Dec 2006, even before the leading occlusion had arrived. The South WA coast experienced its peak at 0700 UTC 15 Dec 2006, apparently after the passage of the leading cold front and before the arrival of the bent-back front. The North WA coast had its peak around 1200 UTC 15 Dec 2006, this after the bent-back front had swept through.  Figure 3.52 The CYYJ-CYQQ-CWSK, or Georgia Strait, geostrophic wind triangle and key stations in and around the region encompassed by the polygon. Stations marked with red-filled circles are those primarily used in the analysis of wind speed and pressure gradients during analyzed windstorms. Looking specifically at the Georgia Strait pressure-wind triangle (Figure 3.52), the surface wind response relative to the local pressure field in and around the region of interest can be examined. Among the four detailed ETCs with northern Vancouver Island landfalls, 12 Mar 2012 appears to have had the most predictable results (Figure 3.53). Wind speeds generally increased with elevating pressure gradients as the low approached and then decreased as the ETC  Chapter 3: Detailed Analysis of High-Wind Storms  169  Figure 3.53 Surface pressure and wind response for the 12 Mar 2012 windstorm using the Georgia Strait geostrophic wind triangle. Solid black and dark blue lines demark 0.25 and 0.40 times the calculated geostrophic wind speed (Mg) in km h-1. The dotted black line is the pressure slope orientation in degrees. The solid gray and light blue lines indicate the geostrophic potential wind speed adjusted, generally downward, based on the relationship between pressure slope and the ideal ~125º orientation in the Georgia Strait for supporting ageostrophic winds. Dashed gray lines plot the sea-level pressure in kPa for the three stations used in the calculations. Orange diamonds indicate the wind speed at CYQQ, with the dashed orange line showing wind direction. Black circles denote the average wind speed for each hourly observation among these six stations: CYYJ, CYQQ, Entrance Island, Ballenas Island, Sisters Island and Buoy 46146; blue circles show the highest wind speed among the same six stations. landed between 1500 and 1800 UTC 12 Mar 2012, began filling and moved away to the north (Figures 3.24 to 3.26). The pressure slope remained generally east to southeast during this windstorm, not always in line with the ideal orientation for ageostrophic winds, which would be perpendicular to the general terrain orientation (Overland 1984). Surface wind speeds and pressure gradients had a strong relationship for all the stations save Squamish for this windstorm (Table 3.5).  Chapter 3: Detailed Analysis of High-Wind Storms  170 Table 3.5 Coefficient-of-determinations for linear regression between hourly pressure gradient and two-min wind speed for windstorms that landed on northern Vancouver Island (n=36 in all cases, save 46146 where it varied from 23 to 36). Six-station average compares the average wind speed among the six stations for each hour to the hourly pressure gradient. Six-station peak compares the highest wind speed among the six stations for each hour to the hourly pressure gradient. Windstorm CYYJ CYQQ CWSK CWEL CWGB CWGT 46146 6-Station Average 6-Station Peak North Vancouver Island Landfalls 15-Nov-06 0.37 0.16 0.01 0.64 0.51 0.36 0.30 0.49 0.38 11-Dec-06 0.29 0.65 0.23 0.63 0.22 0.36 0.60 0.55 0.42 12-Nov-07 0.40 0.63 0.16 0.52 0.58 0.62 0.79 0.73 0.66 12-Mar-12 0.72 0.73 0.11 0.76 0.75 0.76 0.65 0.81 0.79 Average 0.45 0.54 0.13 0.64 0.52 0.53 0.59 0.65 0.56 South Vancouver Island Landfalls 03-Mar-99 0.47 0.01 <0.01 0.62 0.45 0.39 0.56 0.53 0.61 14-Dec-01 0.07 0.25 0.08 0.49 0.61 0.57 <0.01 0.42 0.37 15-Dec-06 0.28 0.02 0.03 0.08 0.14 -0.01 0.41 0.15 0.11 02-Apr-10 0.09 0.14 0.01 0.30 0.16 0.15 0.27 0.21 0.20 Average 0.45 0.54 0.13 0.64 0.52 0.53 0.59 0.65 0.56  The 11 Dec 2006 windstorm did not produce such a predictable outcome (Figure 3.54). In this case, wind speeds gradually escalated well ahead of the peak in pressure gradient. Indeed, during the time of maximum support for ageostrophic SE winds, 2300 to 0100 UTC 11-12 Dec 2006, the measured wind speeds began a marked decline. Regression between wind speed and pressure gradient showed lower fits among the study stations, perhaps best indicated by an R2 of 0.55 for the six-station average, compared to 0.81 during the 12 Mar 2012 windstorm. For the ETCs with southern Vancouver Island tracks, the 03 Mar 1999 southeasterly windstorm had the strongest relationship between surface pressure gradient and wind speed, with an R2 of 0.53 for the six-station average wind (Table 3.5). This storm had the lowest pressure gradient among the eight storms (Table 3.4), yet had some of the strongest wind gusts in the Georgia Strait (Tables 3.2a and b). Winds escalated with a steepening of the gradient around  Chapter 3: Detailed Analysis of High-Wind Storms  171  Figure 3.54 Surface pressure and wind response for the 11 Dec 2006 windstorm using the Georgia Strait geostrophic wind triangle. See Figure 3.53 caption for the key.  Figure 3.55 Surface pressure and wind response for the 03 Mar 1999 windstorm using the Georgia Strait geostrophic wind triangle. See Figure 3.53 caption for the key. Chapter 3: Detailed Analysis of High-Wind Storms  172  Figure 3.56 Surface pressure and wind response for the 15 Dec 2006 windstorm using the Georgia Strait geostrophic wind triangle. See Figure 3.53 caption for the key. 0600 UTC 03 Mar 1999 and continued strong during a northeast pressure slope better supportive of geostrophic as opposed to ageostrophic winds in the Georgia Strait (Figure 3.55). Winds remained high for many hours before trailing off after 1700 UTC along with decreasing pressure gradients, this despite the pressure slope swinging to the southeast to an orientation that is supportive of ageostrophic winds in the Strait. Winds associated with the 15 Dec 2006 south-Vancouver-Island crossing westerly windstorm had the lowest relationship between pressure gradient and wind speed among the eight cases, exemplified by an R2 of 0.15 for the six-station average wind speed (Table 3.5). Coefficient-of-determinations proved particularly low among the Gulf Islands sites—Entrance, Ballenas and Sisters Islands—locations that tended to have a rather high relationship between surface pressure gradient and wind speed during most of the storms. Wind speeds escalated around 0900 UTC 15 Dec 2006, during a time of low pressure gradient, and then began to slowly Chapter 3: Detailed Analysis of High-Wind Storms  173 taper off as the gradient climbed to its peak at 1200 UTC (Figure 3.56). The pressure slope hovered around southwesterly at this time. After dropping off, the wind and pressure gradient went through a second round of escalation, peaking around 1800 UTC.  3.4 Discussion  3.4.1 General Windstorm Statistics  Storm track, both position and direction, clearly has a bearing on the region in which the strongest winds occur (Tables 3.2a, 3.2b and 3.3, and Figure 3.5). In the Strait of Juan de Fuca, a narrow east-west channel in the coastal mountains, the strongest wind events occurred with lows tracking relatively close (south Vancouver Island) and/or moving with an east-northeast to east direction (e.g. 14 Dec 2001 and 15 Dec 2006). The Georgia Strait, and to a somewhat lesser extent parts of the Lower Mainland and Greater Victoria, is also prone to a W to NW wind during the same track types. These two regions are also prone to intense SE winds apparently from lows that track through either half of Vancouver Island. For the south Vancouver Island tracks, it seems that nearly geostrophic winds will occur ahead of the low in the northeast quadrant of the ETC, and for the north Vancouver Island tracks strong SE ageostrophic winds will develop in the southeast quadrant. The Northwest Interior of WA has a similar wind response to the Georgia Strait, though some stations are not prone to W to NW winds (e.g. KBLI). The Puget Lowlands is most prone to strong winds during south Vancouver Island strikes, but even in this scenario not all storms produce high winds. All but one of the eight selected storms had a central pressure at landfall below 98.0 kPa (Table 3.3), and the single exception was very close with a reading of 98.2 kPa. In all cases Chapter 3: Detailed Analysis of High-Wind Storms  174 minimum central pressure—occurring offshore, but usually close to Vancouver Island—did go below 98.0 kPa. Given that the storms were selected based on being among the strongest of the 20 events that landed on Vancouver Island from 1994 to 2012, the deep low pressures among the storms can be considered another reflection of the relationship between central pressure depth and peak wind speeds (Mass and Dotson 2010, Chapter 2). The storm track direction designation (Table 3.3), which is generally based on how the low tracked during its closest approach to the study region, is a rough measure at best. Track direction has strong variance (Figures 3.5 and 3.28 to 3.35). Therefore track direction designations are rough approximations. However, these track directions appear useful in a broad sense as they appear to reflect the upper steering currents, and therefore upper-air conditions over the region, as the storms were at or near their peak wind phase. Using the approximate value for 50ºN to determine an adjusted cutoff of 21 hPa (24) h-1 for 1 Bergeron for explosive cyclogenesis status (see Sanders and Gyakum 1980), four of the eight selected windstorms were cyclogenic bombs (Table 3.3). Interestingly, the storm of 15 Dec 2006 does not quite make this cutoff despite producing the highest average peak gust (Tables 3.2a and b). If however, one were to use the average latitude that the ETC resided during its deepening stage, about 43.8ºN, the cutoff is lowered to 19.2 hPa (24) h-1, just bringing the storm into explosive cyclogenesis status. This illustrates the challenges of creating discrete categories. In any event, three of the storms were not cyclogenic bombs. One, 14 Dec 2001, did not persist long enough for a 24 h central pressure change to be calculated. The other two endured longer than a day, but simply did not deepen at a very fast rate for a long period of time. However, all three exhibited rates of intensification in excess of one hPa h-1 for periods shorter than 24 h. Perhaps a more meaningful measure of bomb cyclogenesis is one that is made over shorter spans Chapter 3: Detailed Analysis of High-Wind Storms  175 of time, perhaps three or six h. There may be some relationship between rapid deepening near the time of landfall and high wind potential for PNW storms south of the study region (e.g. 03 Nov 1958 and 07 Feb 2002 from Read 2008). Among the eight sample storms, only two showed rapid and persistent deepening within ~three h of landfall (Figure 3.36), so this does not appear to be a necessary condition for a strong windstorm.  3.4.2 Synoptic Analysis  Upper-air dynamics appear to be critical to the maintenance of low-pressure centers as they progress inland. Of the eight storms, two became strongly vertically stacked during the 24-h sequences (Figures 3.6 to 3.8 and Appendix C Figures C.13 to C.15). These two ETCs disintegrated rapidly upon landfall, reflected in part by strong and persistent central pressure rises (Figures 3.18 to 3.20, 3.36 and Appendix D Figures D.13 to D.15). Four other storms were at various levels of maturity as they reached the coast and these also began to fill around the time of landfall (Figures 3.12 to 3.14, 3.24 to 3.26, 3.36, Appendix C Figures C.4 to C.12 and Appendix D Figures D.4 to D.12). All six of these ETCs showed a tendency to begin filling before landing on the coast. In contrast, storms that had not yet reached peak development, with strong upper support moving inland with the surface lows, tended to maintain integrity, this despite the presence of steep geography and terrain-forced ageostrophic flow (Figures 3.9 to 3.11, 3.21 to 3.23, 3.36, Appendix C Figures C.1 to C.3 and Appendix D Figures D.1 to D.3). Both the 14 Dec 2001 and 15 Nov 2006 storms had triple-point low structure and had good upper support following immediately behind the surface low-pressure centers: this is probably not coincidental, and may be diagnostic of a low with the capability of maintaining depth or even deepening post-landfall. Chapter 3: Detailed Analysis of High-Wind Storms  176 Historically, some of the most intense Pacific Northwest windstorms had track directions nearly due north up the coast, with the major windstorms of 12 Oct 1962 and 14 Nov 1981 being two key examples. These historical events had steep upper-air troughs strongly supporting southerly airflow over the region (Lynot and Cramer 1966, Reed and Albright 1986). Of the eight windstorms, the two that exhibited nearly due north tracks around the time of closest approach to the study region also had sharp U-shaped 50 kPa troughs with height lines supporting upper-air flow from the SSW to SW over the region as the lows tracked inland (Figures 3.12 to 3.14 and Appendix C Figures C.13 to C.15). Five of the remaining six ETCs had broader troughs, in some cases very weakly defined, with nearly zonal flow in the upper-levels (Figures 3.9 to 3.11 and Appendix C Figures C.1 to C.12). These had east to northeast tracks. The one exception had a steep trough like the storms with track directions more to the north (Figures 3.6 to 3.8), but tracked to the northeast, confounding the apparent pattern—though there is some evidence of this track becoming more northerly post-landfall at the time the surface low became absorbed (Figure 3.28). There is clearly some variability in any relationship between the amplitude and wavelength of the 50 kPa trough and a given storm's track direction, but a rough pattern appears to be present.  3.4.3 Mesoscale Analysis  3.4.3.1 Broad ETC Characteristics, Lee Lows and Surface Wind Response  Shapiro and Keyser (1990) identify four stages in their conceptual model of ETC development: Stage I, a continuous broad frontal region that is where the incipient cyclone begins to develop; II, frontal fracture with obvious cold and warm fronts extending from near the Chapter 3: Detailed Analysis of High-Wind Storms  177 center of the low (i.e. approximating a triple-point structure, or a point-of-occlusion low); III, frontal T-bone shape with an obvious bent-back warm front; and IV, warm-core seclusion with the low center now trailing the point-of-occlusion and often characterized by repeated wrapping of the bent-back front around the low. Out of the eight analyzed ETCs, only one had a multiply-wound bent-back front: 12 Mar 2012 (Figures 3.24 to 3.26). This suggests that only one of the eight windstorms landed in Stage IV. It appears that most landed while at Stage III (Figures 3.18 to 3.20 and Appendix D Figures D.4 to D.15) and the two ETCs with triple-point-low structure landed while in Stage II (Figures 3.21 to 3.23 and Appendix D Figures D.1 to D.3). However, both the 03 Mar 1999 and 02 Apr 2010 storms could be considered at a later state of maturity than 12 Mar 2012, as these two ETCs had moved far from their upper support well before landfall. In fact, the 03 Mar 1999 ETC exhibited a multiply-wound spiral long before reaching the coast (Read 2008, Mass and Dotson 2010 their Figure 22c). The frontal structures on the maps (Figures 3.18 to 3.20) could be indicative of warm seclusion near the center typical of Stage IV, with the old wound frontal structure so weakened that it was undetectable in the surface observations at landfall. The indicated bent-back front may have developed as the ETC neared the coast and entrained air out of the BC interior down the rear flank of the low. This explanation does not entirely work for 02 Apr 2010 (Appendix D Figures D.13 to D.15), as there is no obvious area of warm seclusion, at least at the surface. Many of the ETCs exhibited slowdowns in their forward speed as they neared landfall (Figures 3.29 to 3.32 and 3.34 to 3.35). However, the two westerly windstorms did not slow as much at landfall as the six southeasters (Figures 3.28 and 3.33). This may be due to the fact that both storms had an upper trough rapidly migrating eastward behind the surface lows, almost as if the surface low and upper trough were moving in lock-step. In essence, the upper steering Chapter 3: Detailed Analysis of High-Wind Storms  178 currents were strong in an easterly direction, a situation that is conducive to the steady progress of storm systems, especially compared to those that have weaker upper support. This upper-wind direction is also supportive of westerly surface winds. Mesoscale lows in the lee of the Olympic mountains appeared in at least three of the eight analyzed windstorms. These tend to occur with Froude numbers ≥~1 (Reed 1980, Overland 1984). The development of lee lows is thought to be highly sensitive to the mixed-layer depth, the height of which is generally determined by the point at which the atmosphere becomes statically stable. For example, lee lows might tend to form in a relatively shallow statically stable layer due to a tendency for the air to be deflected around the mountains instead of a situation with a deeper planetary boundary layer that simply moves over the barrier. However, a strong mesolow that developed during a major windstorm on 13 Feb 1979 apparently occurred in a weakly stable air layer extending from the surface to far in excess of the height of the Olympic Mountains (Reed 1980). The obvious mesolows identified in the current study were present in two types of air-masses: in the warm sector during the 15 Nov 2006 windstorm (Appendix D Figure D.2), and in the previously emplaced air-mass ahead of the warm sector during the 12 Nov 2007 windstorm (Appendix D Figure D.11). The former case is interesting because warm sectors tend to be the least stable air mass associated with PNW windstorms. The latter case is likely a shallow boundary-layer situation—stable marine air, perhaps, emplaced behind an earlier storm system, or continentally modified air from the interior being moved northward ahead of the incoming low and interacting with the Olympic Mountains. The formation of a mesolow downstream of the Olympics affects sea-level pressure measurements at surrounding stations (see Ferber and Mass 1990 and Mass and Ferber 1990). Significant modification of the isobars in the vicinity of CYYJ occurred during the 15 Nov 2006 Chapter 3: Detailed Analysis of High-Wind Storms  179 and 12 Nov 2007 windstorms (Appendix D Figures D.2 and D.11). This would have an influence on the gradients and pressure slope determined from geostrophic wind triangles that use CYYJ, including the one for the Georgia Strait (Figure 3.4). With somewhat lower pressure favored on the south leg of the triangle, the tendency would be for a reduction in calculated pressure gradients for storms passing to the north. Also ahead of an incoming low, there would be a bias in favor of more offshore-oriented pressure slopes (e.g. instead of southeast, perhaps east-southeast). After the low tracks inland, if the lee low persists the bias would shift more in favor of an onshore pressure slope (e.g. instead of southwest, perhaps west-southwest). The influence on the smaller Northwest Interior geostrophic wind triangle (used in Chapter 2) would likely be in the same direction but even stronger. There may be other local influences on surface pressure in the region that also affect the two-dimensional pressure gradient calculations. With northern Vancouver Island landfalls, high-wind criteria peak gusts tend not to occur in the Strait of Juan de Fuca and the Puget Lowlands (Tables 3.2a and b and Figures 3.28 to 3.35). This appears due in part to the distance that the low is from these locations—the strongest surface pressure gradients tend to be carried further north. For the Strait of Juan de Fuca this is also because the pressure slope does not tend to reach an angle that is conducive to a strong westerly wind down the narrow east-west channel, again due to the more distant track. A strong pressure gradient from an angle that supports westerly winds generally does not occur until well after the low has moved inland and weakened, or the low simply falls apart as it moves inland. Of interest, one case of high winds in the Strait of Juan de Fuca appears to be the result of a foehn that occurred ahead of an incoming ETC (the peak gust for Port Angeles in Figure 3.30), and another is due to better support for westerly winds as the ETC continued to deepen even while moving inland (Figure 3.28). Chapter 3: Detailed Analysis of High-Wind Storms  180 The southern Vancouver Island landfalls produced a more diverse set of high-wind gust distributions. This is in part due to the presence of two westerly windstorms among the four cases. Also, the 02 Apr 2010 windstorm tracked further north than the other three and, with a recurvature to a northerly direction as it neared the northern Georgia Strait, is perhaps better placed with the northern Vancouver Island storms. The 15 Dec 2006 windstorm (Figure 3.34), strongest of the eight, produced a gust distribution that seems like a combination of the 03 Mar 1999 (Figure 3.32) southeaster and 14 Dec 2001 (Figure 3.33) westerly windstorm. With a few exceptions, the 15 Dec 2006 ETC produced the strongest SE to S winds among the eight storms in the southeast-wind-prone regions south of the track as the low approached the coast, and then produced the most intense W wind in the Strait of Juan de Fuca among the eight storms as the low tracked inland (Tables 3.2a and b), in essence acting like two separate major storms in combination. Peak winds during southeasterly windstorms tend to occur just before or during landfall of the ETC centers (Figures 3.18 to 3.20, 3.24 to 3.26, 3.39, 3.43 and Appendix D Figures D.1 to D.6 and D.10 to D.15). This may be the result of a sudden "blunting" of low-pressure systems upon interaction with the steep coastal terrain, causing the storms to have more of a bowl-shaped profile instead of a funnel and therefore reducing pressure gradients near the center (Figure 3.57). Careful examination of the wind traces and mesoscale maps in Browning (2004) for the 16 Oct 1987 windstorm suggest a similar pattern amid the lower relief of the United Kingdom (e.g. EGJJ), though the presence of and focus on a strong bent-back front confounds peak wind/gust timing. Interestingly, the low center filled from ~95.2 kPa to ~96.0 kPa over a three h period as it made landfall, and pressure gradients immediately around the low center appear to weaken, save on the south side, thus showing trends similar to many landfalling PNW windstorms. Chapter 3: Detailed Analysis of High-Wind Storms  181  Figure 3.57 Conceptual model of the "blunting" of strong ocean ETCs as they interact with the rough terrain of coastal BC. Panel a depicts a peak intensity low that is leaving upper support as it moves inland. Panel b shows a low nearing peak intensity that moves inland along with good upper support. Pressure gradients (in blue) are for 100 km. Orange dashed lines show how the low may have deepened had it remained over the open ocean. Time intervals are uneven to reflect the tendency for ETCs to slow down as they make landfall. Vertical scale is exaggerated. Frontal boundaries are often moving through the study region at around the same time that the low center is making landfall. Given that strong vertical mixing capable of bringing upper-level wind momentum to the surface tends to occur along fronts (Bergeron 1937, Lynot and Cramer 1966, Stull 2000, Clark et al. 2005), it is unclear if the low center position relative to the Chapter 3: Detailed Analysis of High-Wind Storms  182 region of interest, the pressure gradient profile of the low (e.g. funnel vs. bowl), or fronts are the primary cause of the peak wind timing. It could be a combination of them all, perhaps with one dominating over the others during different windstorms. By looking at the 2D surface pressure gradients compared to the surface wind response and the timing of frontal passages, association between frontal passage and peak wind speed can be examined (sections 3.4.3.2 and 3.4.4.2).  3.4.3.2 Identification and Movement of Frontal Systems  Surface warm fronts were often difficult to track, especially inland where complex terrain appears to contribute to differential slowing of forward progress. Cold fronts, and perhaps leading occlusions, seemed less impeded by terrain. Both these tendencies were recognized early in the development of the frontal ETC model (Bergeron 1937). In the PNW, east to northeast pressure slopes and their associated offshore winds can trap a cold, dense surface layer against the coastal mountains like the Olympics, shielding inland basins such as the Puget Lowlands and Georgia Strait from warm air incursion. Warm fronts tend to move right over the cold surface layer, in effect creating a "hidden front." Typically it takes the stronger trailing cold front to mix out this cold air, sometimes with the effect of a cold front raising surface temperatures due to the cool marine air being a bit warmer than the inland cold air mass. The trapped cold air layer may not be of uniform height, and downslope winds can cause some local mixing and warming. This has the result of an apparent warm front surfacing in places—sometimes just isolated pockets that can be transient features as parts of the cold surface layer advect back in—leading to a very sinewy and sometimes broken surface warm-air boundary. Godske et al. (1957) showed how downsloping winds can make the determination of warm front arrival difficult. At other times, when warm air intrusion is strong enough to overwhelm the cold surface layer, the warm front Chapter 3: Detailed Analysis of High-Wind Storms  183 can get trapped against the Cascade Mountains. This can result in the seclusion of the warm air in places as the trailing cold front sweeps inland, eventually resulting in the development of an occlusion. During southeasters and looking specifically at the study region, the strongest winds tend to occur within the warm sector (e.g. Figures 3.18 to 3.20 and 3.55), when one is present. In many cases, the warm front occludes before moving inland (e.g. Figures 3.24 to 3.26), and warm sector air does not reach the Georgia Strait. Thus, maximum winds tend to occur either in air that had been previously emplaced ahead of the storm, usually a shallow, stable cold layer, or after an occluded front has moved through. This may have the effect of mitigating the maximum wind speeds during many windstorms, as these air masses are likely to be relatively more stable, though in the case of a warm occlusion that sweeps out the cold surface layer, there will likely be a decrease in stability (Lynot and Cramer 1966, Mass and Dotson 2010). It is noteworthy that the 03 Mar 1999 windstorm produced the lowest pressure gradient in the Georgia Strait among the eight storms (Table 3.4), but some of the strongest SE winds—this with an apparent warm sector (or warm-core seclusion) in place. The tendency for higher winds in the warm sector appears to be due to lower stabilities with attendant vertical mixing bringing upper-wind momentum to the surface (Lynot and Cramer 1966, Mass and Dotson 2010), and the fact that the airflow is not strongly impeded by air dams (e.g. the cold wraparound to the north) because the more buoyant air simply lifts over any barriers. Cold fronts, and at least some occlusions, appear to be more effective than warm fronts at eroding away any entrenched continental polar air that may be in place ahead of the incoming ETC. This is perhaps in part due to a more supportive environment for onshore flow behind the fronts. As high-wind-generating ETCs progress inland, winds tend to be out of the easterly half Chapter 3: Detailed Analysis of High-Wind Storms  184 of the compass rose ahead of and in the warm sector (Figure 3.40 and e.g. Appendix D Figures D.1 to D.6), a direction that can keep the cold surface layer dammed up against the east slopes of coastal mountain ranges, such as the Olympics. Winds tend to switch to the W sector as the leading cold front or occlusion moves through, and even more so as the low center moves to the east, allowing the surface layer to be carried away. The bent-back front and trough is known to be associated with damaging winds, sometimes the strongest winds of a given storm (Steenburgh and Mass 1996, Mass and Dotson 2010). Interestingly, there is variability in the strength of bent-back fronts even among a selection of the strongest windstorms from 1994 to 2012. Not every landfalling ETC that has produced high winds has a bent-back front present, certainly not one that is strong or easily discernable. The strength of the winds associated with bent-back fronts appears to be related to temperature gradients (Cliff Mass, personal communication, 29 Oct 2014). For the eight analyzed storms, the strongest bent-back fronts—the ones often associated with the fastest winds—do seem to be associated with a sharp surface-temperature drop (Figure 3.27). The temperature falls are similar to what occurred with the passage of the wraparound band during the intense 16 Oct 1987 windstorm in the UK (Browning 2004), further supporting a relationship between the intensity of the front and magnitude of cold-air entrainment. The source of the cold air appears to be from the BC interior. When Vancouver Island lows track in from the southwest to west, the pressure slope in the northeast quadrant is supportive of offshore winds. This is a situation where very cold continental polar air from the BC interior, when it is in place, can be carried through coastal gaps (Lange 1998, Lange 2003). The cold air, being moderated first by the descent from higher elevations and then from the heat flux from the relatively warm waters of the Queen Charlotte Sound, tends to wrap around the north side of Chapter 3: Detailed Analysis of High-Wind Storms  185 lows that land on Vancouver Island, moving down the back-side (Figure 3.58). This cold air seems to be transported most strongly along the bent-back (secondary cold) front as temperatures appear to be the most depressed very near the boundary. Overall, the arrangement of fronts in Figure 3.58 bears a resemblance to the fractured-front model of ETCs as described by Shapiro and Keyser (1990) and also described by others based on analysis of Atlantic ETCs (Neiman and Shapiro 1993, Neiman et al. 1993, Browning 2004, Clark et al. 2005), though it appears that in the study region, the fronts become squeezed relatively close together perhaps due to the slowing of leading boundaries against the high coastal terrain, creating structures that are strongly elongated along a northwest to southeast axis. Indeed, many of these high-wind storms exhibited a frontal arrangement typical of a storm at or near peak intensity, with the bent-back front just curling into the base of the low. The entrainment of a fresh pool of cold air that is significantly cooler than the original cold sector has some implications. As described by Bjerknes and Solberg (1922), when a secondary cold front is stronger than the leading one, the entire "cold sector" behind the leading cold front can act as a warm sector. This suggests the possibility that significant convection can develop along the bent-back front. Heavy rain does occur with these features, including during the 15 Dec 2006 storm (Read 2008, Mass and Dotson 2010, Read and Reed 2013), supporting the idea of convection. Such convection can lead to preferential pressure falls, perhaps strong enough in some cases to lead to a migration of the surface low toward the bent-back frontal boundary in a storm-relative sense, enhancing the local pressure gradient along the southern side of the ETC. Preferential pressure falls along the wraparound band are indicated in Steenburgh and Mass (1996), and seem to have been present in the 15 Dec 2006 windstorm. With the warm sector expanded to include everything between the leading warm front and the secondary cold front,  Chapter 3: Detailed Analysis of High-Wind Storms  186  Figure 3.58 Conceptual model of the entrainment of a fresh pool of cold air as an ETC lands on Vancouver Island, in this case using a mesoscale map from the 11 Dec 2006 windstorm. Large arrows show the general motion of surface airmasses, with color indicating relative temperature (blue cold, orange warm). Cold air drawn from the BC interior is carried around the low and down the rear flank, creating a secondary cold front (a). This has the effect of turning the initial cold advection field into a de-facto warm sector (b). Remnants of the original warm sector are secluded against the Cascade Mountains (c). A small wedge of air emplaced ahead of the storm remains (d). There is an implied quasi-stationry frontal boundary (e) between the colder, drier interior air and warmer, moister marine air, with the coastal mountains acting as a barrier between the two airmasses. then arguably any leading occluded front acts as a de-facto warm front—perhaps a weak one but with warmer air than the new cold advection. A quasi-stationary front (or perhaps a pseudo-front) dividing marine modified air along the coast from the continentally modified air in the interior is implied by this setup—this feature wraps into the low center along the boundary demarking the new cold air advection and can act as a local cold or warm front depending on Chapter 3: Detailed Analysis of High-Wind Storms  187 how the low tracks. Given the entire arrangement, warm-sector-like conditions, with lower static stabilities (relative to the even colder air moving in behind the secondary cold front) may be present over a larger-than-expected area. As discussed earlier, one of the key reasons that high winds are favored in warm sectors is that the warm air is not impeded by the cooler air wrapping around the low in the northeast and northwest quadrants: The more buoyant air simply rides over these potential "air dams." Thus surface airflow in the warm sector can move more freely compared to the surface air in occluded parts of the storm. With a new warm sector created by the presence of a secondary cold front, the potentially large air mass behind the leading cold front can, in theory, also move more freely as it is more buoyant than the invading, denser air wrapping around the low. There are some exceptions to the idea that the strongest bent-back fronts are associated with the entrainment of a fresh batch of cold air from the BC interior. This includes the 14 Dec 2001 windstorm (Figures 3.21 to 3.23). At many reporting sites, little temperature change occurred with the passage of the bent-back front, yet a period of intense W to NW winds, often with a long duration, occurred at many places prone to strong winds from this direction, but also in some unusual locations. In the Lower Mainland, high winds penetrated well inland such that CYXX reported a higher gust than CYVR, an extremely unusual happenstance due to CYVR having a long overwater fetch in the western quadrant and therefore being favored for high readings during westerly winds. The dew point tended to fall sharply with the arrival of the boundary, suggesting strong subsidence, possibly a reflection of entrainment of drier air from the BC interior, air that moderated to sea-level background temperature levels as it moved downslope. There are a unique combination of characteristics with this ETC, including the maintenance of central pressure depth post-landfall that is related to the presence of good upper Chapter 3: Detailed Analysis of High-Wind Storms  188 support right over the region, a triple-point low structure (Stage II of Shapiro and Keyser 1990) and a very long bent-back front that appears to have resided right along the axis of the 50.0 kPa trough (Figure 3.10). The bent-back front may in fact have been directly linked to this feature, thus explaining the slow progression and protracted episode of strong winds. This storm warrants further analysis, as it may hold some important clues about the nature of bent-back fronts and strong westerly windstorms. Another exception to cold-air-entrainment and bent-back-front-strength idea can be found with the front of the 02 Apr 2010 windstorm (Appendix D Figures D.13 to D.15). Intense winds occurred in a narrow region right along the tip of the bent back front, striking Tatoosh Island directly. In the hourly observations, little surface temperature change occurred with the passage of this feature (Figure 3.27), even at Tatoosh. Interestingly, given the late-season timing, there was no source of strongly cold air available to this storm, unlike events that occur in the winter. Interior temperatures were 2 to 4ºC as the low tracked inland—these would warm even further after descent to coastal elevations. The focus here has been on surface temperatures. With bent-back fronts, barcolinicity appears to increase with height into the middle troposphere (Neiman and Shapiro 1993). I did not examine vertical temperature profiles, an endeavor that could provide more understanding about the nature of the relationship between baroclinicity and the strength of bent-back fronts. Considering just the primary study region, the Georgia Basin, preferential pressure falls along any attendant bent-back front may contribute to peak winds during southeasters, but this appears to be primarily a concern with south Vancouver Island systems. North Vancouver Island landfalls are often far enough from the study region, due to the length and northwest-southeast trend of the island, that the bent-back front may not pass close enough to the Georgia Strait to Chapter 3: Detailed Analysis of High-Wind Storms  189 have a strong influence (e.g. Figures 3.24 to 3.26, Appendix D Figures D.10 to D.12). South Vancouver Island ETCs tracking inland are more likely to carry the tip of the bent-back front right through the study region (e.g. Figures 3.18 to 3.20 and Appendix D Figures D.7 to D.9). Also, the ETC first encounters terrain when it is much closer to the Georgia Strait, meaning there is less time, and terrain, to disrupt both the ETC and the front before they reach the study area.  3.4.4 Pressure Gradients, Frontal Passages and Surface Wind Speed  Ultimately all lows have stronger gradients near their cores due to the simple fact that along the furthest reaches of the depression there is a point where pressure gradients may fall to near zero, depending in part on scale. This discussion focuses on the region within approximately 300 km of the low center (i.e. at the mesoscale)—in other words, what is happening immediately around the low where much of the high wind action typically takes place. The narrative is also focused around the time of landfall, when the strongest winds tend to occur in the study region.  3.4.4.1 Pressure Profiles and Slopes  Central pressure tendencies are largely the result of the balance between surface convergence and upper-air divergence (Godske et al. 1957), making it tempting to relate upper support to ETC surface pressure profiles. However, there does not appear to be a perfect relationship between upper support and the type of profile that an ETC assumes as it makes landfall. Determinations are made more difficult by being limited to just eight samples. Nevertheless, there are some hints at a few possible tendencies. Chapter 3: Detailed Analysis of High-Wind Storms  190 Focusing on coastal pressure gradients, which provide the best profile of the ETC since terrain modification of the pressure gradient field around the low may be strong as the low moves inland, storms that landed with relatively good upper support (e.g. 15 Nov 2006, 12 Nov 2007, 12 Mar 2012) appear to be the most likely to have a funnel-shaped surface pressure profile, resulting in gradients with increasing magnitude toward the low center (Figure 3.47). Interestingly, two of these ETCs had strong secondary (bent-back) cold fronts with the tips near the low center—in the case of the 12 Mar 2012 storm, the front had wrapped around the low at least 1.5 times. Steenburgh and Mass (1996) and Mass and Dotson (2010) suggest that the strongest pressure gradients tend to occur near the tips of bent-back fronts, apparently the result preferential pressure falls along the band due in part to latent heat release (i.e. convection). It is possible that the close proximity of the end of the bent-back front to the low center is a key contributor to the intense pressure gradients in the cores of these storms, confounding to some extent a focus just on the jet support (of course, the jet support could be contributing to any convective activity). Perhaps in the case of the 11 Nov 2007 storm, pressure falls near the tip of the bent-back front may have been strong enough that the low center had migrated toward the tip in a storm-relative sense. Those lows that recently moved away from upper support by the time of landfall (e.g. 11 Dec 2006) may have a tendency for more of a V-shaped profile, with pressure gradients approximately the same across a broad area around the low center (Figure 3.48). And lows that are well away from good upper support (e.g. 02 Apr 2010) may have a propensity to develop a U-shaped profile with shallow gradients near the center (Figure 3.49), likely due to rapid filling in an environment of terrain-induced ageostrophic flow. Chapter 3: Detailed Analysis of High-Wind Storms  191 There are exceptions to these rough generalizations. The 03 Mar 1999 windstorm appears to have had a funnel-shaped profile (Figure 3.50), at least on the south side, right at landfall. In fact, both the 03 Mar 1999 and 02 Apr 2010 windstorms show very weak gradients on the north side with focused areas of intense gradient on the south side, right near the center in the case of 03 Mar 1999 and away from the center in 02 Apr 2010. Both of these storms had weak upper support at landfall. They also had relatively weak bent-back fronts, but it appears that the strongest pressure gradients were nevertheless associated with the tips of these features. These two cases appear to support the idea that preferential pressure falls can be associated with the ends of bent-back fronts even in the case of very mature ETCs.  There is, in fact, a marked tendency for the steepest pressure gradients to occur immediately on the south side of most landfalling ETCs (Table 3.4), with some exceptions. This could be the signature of preferential pressure falls along associated bent-back fronts (Steenburgh and Mass 1996, Mass and Dotson 2010). Other dynamics outside of the bent-back front are also at play. ETCs tend to develop in the base of deep troughs, or quasi-stationary Gulf-of-Alaska lows in what is sometimes termed secondary-spinup. There may be a surface high-pressure cell to the south often off the CA coast or over the United States desert Southwest region. ETCs developing in this situation have a lower pressure field to the north relative to the south, a situation that favors steeper gradients on the south side of incoming lows. This is a reminder that ETCs have asymmetric form, a fact that can confound symmetric conceptual models. Even with the tendency for higher pressure to the south of incoming lows, it is nevertheless interesting that the maximum gradients seem to generally occur on the south side of the low even in cases when the bent-back front has obviously wrapped around the low center Chapter 3: Detailed Analysis of High-Wind Storms  192 multiple times based on examination of satellite imagery (e.g. Figures 3.47 and 3.50). This suggests that the contribution from the tip of bent-back fronts may be the lesser. Returning to the generalization about upper support and pressure profiles, another exception is the 14 Dec 2001 windstorm which had a V-shaped profile at landfall and had good upper support moving inland with the ETC (not shown). The low developed just off the Vancouver Island coast and had a relatively short existence over the Pacific before making landfall—perhaps it had not had the time to develop the funnel form seen with landfalling storms that had decent upper support. In other words, a funnel-shaped profile may be more likely at Stage III than Stage I or II. A more striking exception to the above generalization is the 15 Dec 2006 windstorm (Figure 3.51). This storm had a somewhat inverted profile, with the highest gradient occurring on the south Washington coast, gradually decreasing towards the low center. The ETC had good upper support as it neared the shore (Appendix C Figures C.10 to C.12), and then moved away from the left exit region of the jet streak around the time of landfall, in a manner somewhat similar to 11 Dec 2006 though remaining closer to the jet axis. The inverted profile may be indicative of preferential-pressure falls along the bent-back front overtaking the original low center in terms of supporting intense pressure gradients—this happening right as the low tracked inland (see Steenburgh and Mass 1996, Mass and Dotson 2010). This source of pressure gradient intensification may have been a major factor in the measured extreme wind speeds. Most of the other ETCs did not have a bent-back front with the magnitude of the 15 Dec 2006 storm. Pressure slopes at the time of maximum gradient (Table 3.4) seem to be largely dictated by the storm track, especially in the case of the northern Vancouver Island landfalls—their orientations delineate a low just off the north coast. South Vancouver Island landfalls seem to Chapter 3: Detailed Analysis of High-Wind Storms  193 result in more variability, though some of the same tendency seen with the northern tracks is evident, just shifted south by about one region, and with a more northerly slope component over the southern 2/3 of Vancouver Island, likely a reflection of being in the northeast quadrant of incoming lows. There is much variability of pressure slope orientation in the Georgia Strait region. South Vancouver Island ETCs often track right across the Georgia Strait thus explaining the highly varied result. Sometimes the northeast side of an incoming low has the sharpest gradient, likely due to the presence of strong high pressure over the BC interior, and at others it occurs on the southeast side, with the low just skirting through the north end of the geostrophic wind triangle, putting the region in the steepest gradients to the south of the low center. Some ETCs tracking northeast go right through the center of the triangle, bringing inland a sharp southwest pressure slope in their wake. Those systems with well-developed bent-back fronts may have the strongest pressure gradients in the southwest quadrant of the low.   3.4.4.2 Surface Pressure Gradients, Winds and Frontal Passages  Looking at the surface pressure gradient and wind response of the eight detailed windstorms, some stations tended to have a strong relationship between the two variables while others did not (Table 3.5). It appears that the more land-locked sites tended to have lower R2 values, probably due to terrain interference of surface airflow. In this regard, Squamish stands out the most strongly. However, this station is located in a narrow north-south valley that is not prone to strong winds during most of the southeasterly and westerly windstorms that strike Greater Victoria and the Lower Mainland. Synoptic setups favorable to N or S winds may show a stronger fit between pressure-gradient and wind speed at Squamish. The four ETCs with south Vancouver Island landfalls tended to have weaker fits than the north Vancouver Island storms. Chapter 3: Detailed Analysis of High-Wind Storms  194 One likely contributing factor to the lower coefficient-of-determinations is the fact that ETCs with more southern tracks tend to pass right through the geostrophic wind triangle. When a low moves almost exactly between the stations, then all sites can show similar sea-level pressure readings, which results in the calculation of a weak pressure gradient despite the fact that there may be a steep one close to the low center (further discussion below).  The moderately strong to strong relationship between wind speed and pressure gradient at most Georgia Strait stations during the north Vancouver Island southeasterly windstorm on 12 Mar 2012 (Figure 3.53 and Table 3.5) may indicate a situation where the surface pressure field, as opposed to other features such as frontal bands with associated vertical mixing bringing down upper-level momentum, primarily drove the wind response. The ETC did carry an occluded front through the region (Figures 3.24 to 3.26). However, both the strongest pressure gradient and peak wind occurred well ahead of this feature, by approximately five h, with the wind slowing down considerably before the boundary moved through. This contrasts with 11 Dec 2006, another north Vancouver Island southeasterly windstorm, where the relation between measured wind speeds and pressure gradient was not as strong. Wind speeds decreased even as the pressure gradient continued to climb and the pressure slope shifted to 125º, nearly ideal for ageostrophic SE winds in the Georgia Strait (Figure 3.54 and Table 3.5). The occluding cold front associated with this ETC moved into the Georgia Strait around 0000 UTC 12 Dec 2006 right as the low landed and began filling (Appendix D Figures D.4 to D.6). Immediately afterward, the wind speed began to diminish, this despite a continued increase in wind support from a still-escalating pressure gradient with good orientation for ageostrophic acceleration. It appears that the arrival of cooler, more stable air behind the front closed off a good vertical mixing source. Indeed, ahead of the cold front, a warm advection field became Chapter 3: Detailed Analysis of High-Wind Storms  195 secluded over the Georgia Strait (Appendix D Figure D.5). As noted earlier, lower static stabilities in warm sectors can support the vertical mixing of upper wind momentum to the surface (Mass and Dotson 2010).  For ETCs that landed in south Vancouver Island, the 03 Mar 1999 windstorm showed the best fit between pressure gradient and wind speed (Figure 3.55 and Table 3.5). The pressure gradient and wind escalated between 0600 and 0700 UTC 03 Mar 1999, around the time that the leading warm front moved into the region and then stalled against the Coast Ranges. Note that this "warm front" may, in fact, be the arrival of the region of warm seclusion near the low center that is characteristic of mature Stage IV cyclones, and therefore may not be part of the original warm sector (see 3.4.3.1). While this de-facto warm sector remained in place, the winds continued strongly, reaching their maximum around 1300 UTC—this is the approximate time of landfall for the low center. Winds appear to slow a little after landfall, when the primary cold front arrived around 1400 UTC ±1 h depending on the station (Figures 3.18 to 3.20). For some stations peak winds appear to have occurred right with the frontal passage (e.g. CYQQ and CYVR). A second wind maximum, not quite as strong as the first and not showing up at all stations, occurred in this cooler air just ahead of the bent-back front. After this boundary moved through around 1800 UTC, wind speeds tapered off more strongly. During the phase of strong winds, the pressure slope is northeasterly. Surface winds, out of the SE, were therefore quasi-geostrophic. This condition can occur even in very narrow channels, with ageostrophic winds tending to dominate when the pressure gradient is orthogonal to the terrain orientation (Overland 1984). Given that the onset of winds occurred with the arrival of the warm front and then began to slow down after cold advection arrived, and also given that this storm produced the lowest pressure gradients among the eight while producing some of the highest gusts (Tables 3.2a and Chapter 3: Detailed Analysis of High-Wind Storms  196 b), it appears that the surface winds may have been significantly influenced by vertical mixing of upper momentum in the relatively unstable de-facto warm sector. Interestingly, peak wind appears to have occurred around the time of landfall (± ~2 h), and this coincided with the passage of the leading cold front. Note that the 11 Dec 2006 case described above had a similar response, with the strongest winds occurring inside the warm sector, then slowing down markedly behind the primary cold front, with this happening around the time the ETC center made landfall. Westerly winds tend to occur in the colder, more stable air behind the low. Therefore, surface warm sectors do not appear to be an important factor. Indeed, with westerly windstorms, perhaps a strong case can be made that the timing of peak winds is typically controlled by a bent-back or leading cold front, with the main winds occurring just as the boundary passes. However, the strongest pressure gradients are often near the low center, and this alone could produce a sudden wind strike as the pressure slope quickly realigns after the ETC moves to the north and east. In the case of 15 Dec 2006, association with the secondary cold front is strong (Appendix D Figures D.7 to D.9). For 14 Dec 2001 an initial surge of westerly winds appears related to the passage of the cold/occluded fronts that are aligned right through the triple-point low center (Figures 3.21 to 3.23), with a slow climb toward peak when a weakly-defined bent-back occlusion (and/or upper trough) arrived. The 15 Dec 2006 westerly windstorm had the weakest relationship, in general, between pressure gradient and wind among the eight detailed windstorms (Figure 3.56 and Table 3.5). The ETC tracked right through the middle of the geostrophic wind triangle (Figures 3.4 and 3.34), with the center moving to the northeast of the stations by 0900 UTC 15 Dec 2006. This likely resulted in a low estimation of pressure gradient, probably during the time when the Chapter 3: Detailed Analysis of High-Wind Storms  197 storm's steepest pressure differential moved through the region. The value of 7.3 hPa (100) km-1 at 1200 UTC appears to just catch the steep gradient on the southwest side of the low. The highest winds occurred from 0900 to 1000 UTC, then slowly trailed off. This appears to be the time that the bent-back front swept through much of the north and central Georgia Strait, reaching the south end (e.g. CYVR) by 1100 UTC. In this regard, peak winds can be considered frontally-supported, but perhaps mainly due to the intense pressure gradient associated with the boundary (Appendix D Figure D.8) more than due to vertical mixing mechanisms. Given that the Georgia Strait stations used in this part of the analysis were on the north side of the ETC track, and that wind speeds in the region tended to be the lowest among the eight storms (Tables 3.2a and b), plus the fact that the tip of the bent-back front tracked well to the south (Appendix D Figures D.7 to D.9), a sting-jet is ruled out as a factor in the observed wind speeds (see Browning 2004 and Clark et al. 2005). Incidentally, the fastest winds after the low tracked inland occurred with a southwest pressure slope, another case of strong quasi-geostrophic winds as with 03 Mar 1999, but from the opposite direction. Also during the 15 Dec 2006 windstorm, there is a secondary wind and pressure gradient maximum well behind the storm, peaking around 1800 UTC 15 Dec 2006. This is not an uncommon feature with westerly wind events and occurred among a number of storms during the 1994 to 2012 study period. The secondary peak appears to be related to the passage of the trailing upper trough. Indeed, the highest winds during the 14 Dec 2001 windstorm also may have been influenced by the passage of the upper trough, which appears to have been in close association with the bent-back front (Figures 3.9 to 3.11 and 3.21 to 3.23).   Chapter 3: Detailed Analysis of High-Wind Storms  198 3.4.5 Comparative Analysis of Surface Response  Data are plotted in six-h increments largely due to the availability of surface maps (Figures 3.38 to 3.46). It is important to recognize that the actual timing of key events may have occurred between the charted observations, and that only broad conclusions can be drawn from this analysis. Also, the discussion is largely focused on the Georgia Strait region. And keep in mind the relationship between the low center bearing from CYVR and the regional pressure slope (Figure 3.38). Wind direction had marked variability, especially at times of lower wind speed (Figures 3.39 to 3.46). This is almost certainly due to the airflow interacting with local terrain under specific pressure-slope conditions (Lange 1998 and Lange 2003), and makes determining general trends more challenging. Using conceptual models for various pressure slopes helps interpret observations and also in the development of a conceptual model for windstorms triggered by Vancouver Island land-falling ETCs. Roughly speaking, in the Georgia Strait, pressure slopes of 90 to 120º as evidenced by incoming southeasterly windstorms with northern Vancouver Island landfalls (Figure 3.38 and 3.40), are generally supportive of SE winds. For the south Vancouver Island ETCs that produce southeasterly windstorms, pressure slopes in the range of 30 to 60º ahead of the incoming system are more supportive of E to NE winds, with 60 to 90º favoring SE (Figure 3.44). When the pressure slope shifts to above 180º, the potential for a strong W to NW wind is present, especially when the low center is relatively close as this is generally where the strongest surface pressure gradients are found (Figure 3.46). Values around 220 to 250º are seemingly the most supportive, or at least the most common during westerly windstorms, though certainly pressure slopes with a more westerly orientation have been present.  Chapter 3: Detailed Analysis of High-Wind Storms  199 Given that the pressure slope response is roughly similar between southeasterly and westerly windstorms as they track inland, this suggests that upper support for W winds is a contributing factor in the Georgia Strait. Westerly windstorms appear to generally result from ETCs tracking east-northeast to east-southeast, as opposed to northeast to north, though there is some overlap (Table 3.3 and Figures 3.5 to 3.14, 3.37 and Appendix C Figures C.1 to C.15). Given the tendency for windstorms to result from lows moving right over or north of the study region, tracks that are close to due east tend to bring the southwest quadrant of the low over the Georgia Strait. This region tends to have good support for NW winds at the 85.0 kPa level (Figures 3.16 and Appendix C Figure C.19). For northern Vancouver Island-crossing westerly windstorms, the tendency for wind maximums ahead of the low at some stations is likely the result of exposure to winds from a specific direction. CYQQ, for example, is well exposed to strong SE winds, and therefore has the tendency for maximum winds to occur during the southeast wind phase, which usually happens as the low approaches the region. For those stations prone to westerly winds, such as CYVR, the peak arrives behind the low when the pressure slope is most favorable for W to NW winds. During southeasters, especially those crossing the north half of Vancouver island, CYXX has a tendency for a rapid onset of extreme winds. This appears to be associated with the arrival of the warm sector—in nearly all the cases with a sudden occurrence of high winds, the temperature is at or near its maximum for the day when the strongest gusts arrive. Sometimes dew-point depressions are very high during these strong wind episodes, including 19ºC in the case of the 12 Nov 2007 storm, indicating subsidence that could be related to either descending air in the unstable warm sector, or downsloping off of the Cascade Mountains. Also during the 12 Nov 2007 event, Port Townsend experienced a Foehn off of the Olympic Mountains that Chapter 3: Detailed Analysis of High-Wind Storms  200 reached high-wind criteria speeds and was also accompanied by a sharp increase in temperature and drop in relative humidity. There may be a link here. Apparently a shallow, stable surface layer being overrun by warm-sector air that, in places such as CYXX and Port Townsend, appears to have reached the surface due to terrain forcing. The 14 Dec 2001 ETC center tracked right over CYVR, explaining the abrupt break in the line indicating bearing (Figure 3.46). In response to the storm center moving right through the study region, the wind direction at all four stations had a fairly sharp and consistent change to the WSW to NW. Unlike many westerly windstorms, this event had a more gradual onset of winds, starting off fairly strong and then gradually climbing to maximum speed over roughly six h.  3.4.6 Windstorm Conceptual Models  An integration of the key points discussed in the previous sections informs the construction of conceptual models for strong windstorms triggered by ETCs that land on north and south Vancouver Island. Keep in mind that the discussion primarily focuses on the results of the eight strongest windstorms, and sometimes includes information from an additional six weaker storms. The range of scenarios that triggered all 58 windstorms in southwest BC from 1994 to 2012 is far broader (Chapter 2). The focus is on the primary study region, the Salish Sea Basin. The following conceptual models can be considered a nested subset of the general windstorm model developed by Mass and Dotson (2010). There are a number of relevant details that the general model does not address. Windstorm track direction can vary considerably, easily ranging from north to east to even southeast at the time of closest approach to the study region (Chapter 2). A tendency for high-wind-generating ETCs to fall into two broad classes based on peak wind direction, southeasters and westerly windstorms, appears to be linked in part to track Chapter 3: Detailed Analysis of High-Wind Storms  201 direction (Chapter 2 and Section 3.4.4 this chapter), and the surface wind response from these two types of storms may not fit neatly within the Mass and Dotson conceptual model. For instance, looking at southeasters, peak winds appear to regularly occur during stage one (pre-frontal) as opposed to stage three (bent-back trough), and for some storms wind maxima occur during stage 2 (post-frontal) especially when a warm sector is emplaced. The pressure slope orientation around bent-back fronts may not be equally supportive of SE verses W winds, and in the case of the study region it appears that bent-back fronts may have a more dominant role in determining the peak wind phase of westerly windstorms. Indeed, when the focus is on a specific region, especially one that is in the interior of the coastal mountain ranges, often the bent-back front may not play a significant role simply due to weakening as the storm system moves inland. Also, the centers of some windstorms can track so far to the north that the most extreme pressure gradients associated with their bent-back fronts do not reach the study area.  It is here proposed that, for the study region, the Mass and Dotson (2010) conceptual model be modified, but still retaining four stages: 1) pre-landfall, when the low center is far offshore such that terrain interaction is not strongly affecting the central pressure tendencies and characterized by gradually escalating wind speeds in the study region, usually E to SE, 2) landfall, when the low nears the coast and moves inland, often characterized by rising central pressure and weakening pressure gradients around the core and for southeasters is the typical time of peak winds, 3) bent-back trough (or post-landfall), when the low and bent-back front reach their closest approach and then move away from the study site, typically triggering the highest winds during westerly windstorms, and is often characterized by continued weakening of the low due to terrain interaction, 4) termination, characterized by the low tracking well inland with continued filling or being absorbed, and with fading onshore winds at the study site. Chapter 3: Detailed Analysis of High-Wind Storms  202  3.4.6.1 Conceptual Model of Synoptic Evolution  Regardless of the landfall location on Vancouver Island, high-wind-generating ETCs tend to form near the base of a 50.0 kPa trough with an axis approximately around 140º W, varying from ~135 to 145º. These lows will have good 30.0 kPa jet support, often forming in the left-exit region of a jet streak, though some may spin up under the right entrance region. A few lows may spin up under the right-entrance region and then eventually move under the left-exit region and receive continued support, or a new jet streak may move into the vicinity of the low and also contribute support. Some ETCs may develop fairly far north, between 45 to 50º N and then track east-southeast to southeast for a time before recurving to the northeast as they near the coast. Others will develop further south and then track east for a time before recurving, while a few will steadily track northeast to Vancouver Island. A rare few will have nearly due north track directions at the time of closest approach; most will fall between north-northeast and east-northeast. Southeasters generally occur with track directions from the northeast to north, and westerly windstorms east-northeast to east-southeast. There is some overlap, and those intermediate storms have the possibility of bringing both high SE and W winds. Most ETCs exhibit a slowdown just before landfall, and some even temporarily stall. Once ashore, forward movement tends to increase again, sometimes back to over-ocean speeds. Those lows that have strong upper support moving inland with them appear to slow the least at landfall. The ETCs tend to deepen rapidly over a 6 to 24 h period as they approach the coast, the duration often depending on how far offshore the storm began, with some reaching one or more Bergerons (e.g. "explosive cyclogenesis" status). For these windstorms, the central pressure Chapter 3: Detailed Analysis of High-Wind Storms  203 tends to fall below 98.0 kPa for a time, with some storms going below 96.5 kPa, and usually remains close to minimum up to the point of landfall. Many of these ETCs move away from robust upper support as they approach the coast, and tend to fill rapidly upon landfall with a corresponding reduction in surface pressure gradients and winds. Filling typically begins one to three h before landfall. Other ETCs move inland with conditions of strong upper support and may maintain central pressure depth, or even deepen, post-landfall. However, pressure gradients around the low center may still weaken to some extent as the low is "blunted" from interaction with the coastal mountains and resulting terrain-forced ageostrophic flow (Figure 3.57). Incoming ETCs begin to influence the surface winds in the study region at a distance of around 1000 to 1500 km, depending on the synoptic setup (e.g. how close the ETC is to any storms that may be downstream).  3.4.6.2 Conceptual Model of Surface Wind Response for Southeasterly Windstorms  For southeasters landing on northern Vancouver Island, the pressure slope tends to shift to 90 to 120º when the incoming low is well offshore and persists up to landfall (Table 3.6). This generally results in SE ageostrophic winds in the Georgia Strait. Sometimes winds are E to NE, especially for stations that are close to gaps in the coastal mountains (e.g. CYXX). As the low tracks inland to the north and northeast of the study region, the pressure slope tends to shift to around 180 to 220º, with winds becoming S to SW, usually at lower speeds than the pre-landfall phase. For southern Vancouver Island landfalls, the pressure slope in the pre-landfall phase tends to be around 30 to 90º, generally supporting NE winds in the region (Table 3.6). However, SE quasi-geostrophic winds may develop, especially at locations with a long overwater fetch, and Chapter 3: Detailed Analysis of High-Wind Storms  204 perhaps more likely with an emplaced warm sector. As the lows track inland, the pressure slope tends to shift to approximately 140 to 180º. Surface winds tend to broadly shift to the SE when the low center is close, and then to the S and SW as the low moves well inland. For both north and south Vancouver Island landfalls, winds escalate over a period of 12 to 24 h as the ETC nears the coast, with the maximum speed happening around the time of landfall—essentially the moment of closest approach before the low is strongly affected by the coastal terrain. Peak wind speeds in strong to severe storms (return interval ≥~3 y, Chapter 2) generally reach ~18 to 23 m s-1 with gusts of 25 to 30 at one to two of the three key stations CYVR, CYYJ and CYXX, with well-exposed Georgia Strait sites sometimes experiencing ~25 to 30 m s-1 winds with gusts in the range of 35 to 50. The rare catastrophic storm can produce gusts of 30 to 40 m s-1 at one or more of the three key stations (e.g. 12 Oct 1962). These wind speed maxima are approximately the same for all classes of windstorm considered herein. In addition to the peak winds occurring near landfall, the forward movement of many ETCs is reduced as they near the coast, a characteristic that may protract the high-wind phase. This is also the time when any associated leading warm, cold or occluded fronts may sweep through the Georgia Strait, presenting the possibility of local wind speed escalations due to upper-wind momentum transfer to the surface via vertical mixing along the air-mass boundaries. The arrival of a warm front may result in the sudden onset of high winds at some stations, especially CYXX, as airflow shifts from NE to E then to SE to S. Bent-back (occlusions or secondary cold) fronts generally have a lesser role in strong SE winds during northern Vancouver Island landfalls, mainly because the tips of these fronts often pass to the west and north of the Georgia Strait with these more distant landfalls. For the southern tracks, the situation is different, as the tip of the bent-back front can be carried right through the study region, Chapter 3: Detailed Analysis of High-Wind Storms  205 Table 3.6: Summary of wind response for the Lower Mainland and Greater Victoria windstorm conceptual model. NVI refers to north Vancouver Island landfalls, and SVI south Vancouver Island landfalls. These acronyms are followed by the peak wind direction designation for the two landfall classes (e.g. SE or W). All figures are generalizations.  Pressure Slope (±~30º) Stage NVI SE SVI SE NVI W SVI W 1) Pre-landfall 90 60 60 60 2) Landfall 120 90 120 90 3) Post-landfall / Bent-back trough 180 150 210 210 4) Termination 210 180 240? 240      General Wind Direction Stage NVI SE SVI SE NVI W SVI W 1) Pre-landfall SE; E-NE coastal gaps E-NE; SE at favored sites SE; E-NE coastal gaps E-NE 2) Landfall SE SE SE SE 3) Post-landfall / Bent-back trough S-SW SE-S W-NW W-NW 4) Termination S-SW; WSW-W coastal gaps S-SW; WSW-W coastal gaps W-NW W-NW      Wind Speed Character Stage NVI SE SVI SE NVI W SVI W 1) Pre-landfall Gradually accelerating Gradually accelerating Gradually accelerating Gradually accelerating 2) Landfall Peak Peak Peak for SE-prone sites Peak for SE-prone sites 3) Post-landfall / Bent-back trough Slowing Slowing; Peak if strong bent-back trough reaches region Peak for W-prone sites Peak for W-prone sites 4) Termination Slowing Slowing Slowing Slowing   Chapter 3: Detailed Analysis of High-Wind Storms  206 potentially bringing the strongest pressure gradients of the ETC into the Georgia Strait. In this case, peak winds, still generally out of the SE, may occur just ahead of the bent-back front.  3.4.6.3 Conceptual Model of Surface Wind Response for Westerly Windstorms  For westerly windstorms triggered by ETCs that land on northern Vancouver Island, the pressure slope response is similar to southeasters (Table 3.6). The main difference is that wind direction shifts more westerly, generally SW to NW as the low tracks inland. Winds moving down the Georgia Strait under these circumstances may be quasi-geostrophic (e.g. a southwest pressure slope with NW winds). Storm track direction seems to play a role in determining whether-or-not a westerly gale occurs, with those ETCs moving east-northeast to east-southeast being the most ideal, perhaps in part due to upper support (e.g. 85.0 kPa winds) being favorable over the Georgia Strait after the low has moved inland. For southern Vancouver Island tracks, pressure slopes begin in similar fashion to southeasters, but swing more sharply to the west, ~200 to 270º, as the low tracks inland (Table 3.6). This generally supports WSW to NW winds, quasi-geostrophic to ageostrophic depending on the situation. Westerly windstorms differ markedly from southeasters in that they tend to have a rather abrupt onset. Winds ahead of the low may be light to moderate out of the NE to E, sometimes SE, and may show some acceleration as the low nears the coast, with stations favored for high readings during offshore-wind situations peaking just before landfall in a fashion typical of a southeaster. As the low tracks to the north and east of (sometimes over) the study region, the pressure slope goes through an abrupt shift to the southwest and even west. At this time, WSW to NW winds quickly sweep down the Georgia Strait, often reaching peak values within minutes Chapter 3: Detailed Analysis of High-Wind Storms  207 to a few hours of the initial wind shift, then slowly trailing off. In this regard, associated primary or secondary (bent-back) cold fronts may play a role in determining the timing of W wind arrival. The former may have a