Open Collections

UBC Undergraduate Research

Characterizing the influence of forest cover changes on streamflow variability at Fishtrap Creek, British.. Muenter, Luisa 2010

You don't seem to have a PDF reader installed, try download the pdf

Item Metadata

Download

Media
[if-you-see-this-DO-NOT-CLICK]
Muenter_Luisa_FRST_498_Graduating_Thesis_2010.pdf [ 1.07MB ]
[if-you-see-this-DO-NOT-CLICK]
Metadata
JSON: 1.0075622.json
JSON-LD: 1.0075622+ld.json
RDF/XML (Pretty): 1.0075622.xml
RDF/JSON: 1.0075622+rdf.json
Turtle: 1.0075622+rdf-turtle.txt
N-Triples: 1.0075622+rdf-ntriples.txt
Original Record: 1.0075622 +original-record.json
Full Text
1.0075622.txt
Citation
1.0075622.ris

Full Text

CHARACTERIZING THE INFLUENCE OF FOREST COVER CHANGES ON STREAMFLOW VARIABILITY AT FISHTRAP CREEK, BRITISH COLUMBIA by LUISA MUENTER  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF BACHELOR OF SCIENCE in THE FACULTY OF FORESTRY (Forest Sciences)  UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  December 16, 2010  ©Luisa Muenter, 2010  ABSTRACT Modelling hydrologic recovery following a forest disturbance can assist forest managers to practice forest management while taking into account the hydrologic response of certain forest activities and disturbances. Using Vegetation Resource Inventory (VRI) data that are available on a province-wide scale, a results-based approach of looking at hydrologic recovery following a major fire was carried out on the Fishtrap Creek Watershed. Stands were modelled using the Chapman Richards growth model, and the data were predicted backwards through time for the period for which stream discharge measurements are available. The stand forest parameters are used to calculate a measure of Equivalent Clearcut Area (ECA) at the watershed scale for each year of data available. Climate variables and equivalent clear cut area were used in a regression model to separate their effects from those of a wildfire on the following streamflow metrics: timing of the onset of freshet, total freshet runoff, and the timing and magnitude of the annual peak flow. The analysis identified the timing of the onset of freshet as the most sensitive metric to forest cover change. The challenges in using currently available forest inventory data for hydrologic applications are also discussed.  Keywords: ECA, Forest Cover Change, Hydrologic Response, Wildfire, Fishtrap Creek.  ii  Table of Contents Page  Abstract……………………..…………………………………………………………....…...…. ii List of Figures……………………………………………………………………......................  v  List of Tables……………………………………………………..........……………………..….  vi  Acknowledgments…………………………………………………..…........……………..……  vii  Introduction…………………………………………………………………………..…….........  1  Methods……………………………………………………………………..……………..……  4  Study Area..............................................................................................................……….  4  Data Collection...................................................................................................................  7  Backcasting forest polygons information..............................................................................  8  Calculation of ECA over time for each polygon...................................................................  10  Differences in ECA before and after burning .....................................................................  15  Results………………………………….................................................……………….…….…  16  Relations between stand height and age....................................................………….……  16  Harvest history and ECA……………………………………………………...................  18  Regression modelling of the relation between stream-flow parameters and climate variable ..  20  Regression analysis of start of Freshet runoff....................................................................  23  Regression analysis of amount (in mm) of Freshet runoff.................................................  25  Regression analysis of start on day of peakflow................................................................  27  Regression analysis of start of Peakflow discharge...........................................................  29  Height differences between first listed and second listed species in VRI ........................  31  Discussion……………………………………………………………………..…….................... 32 Assessing hydrologic response of Fishtrap creek related to Freshet period……..…....... 32 Assessing hydrologic response of Fishtrap creek related to Peakflow........................... iii  33  Practical implications from study related to post fire hydrology and effect of forest cover..............................................................................................  34  Effects of assumptions made on the models analyzed ................................................. 35 Recommendations for improving usability of VRI data in Result Based Analysis and analysis involving change over time….............….................................. Conclusion……………………………………………………………………..…....................  38 42  References……………………………………………………………………….......................... 43 Appendices………………………………………………………………………………………  iv  46  List of Figures Figure 1  Figure 2  Figure 3  Figure 4  Map showing Fishtrap Creek watershed and extent of the 2003 McClure fire. The watershed is the dark gray portion and the fire extent is the area shown in light gray.......................................................................................... Equivalent Clear cut Area-based stand height of dominant species (left) and Canopy closure (right). Created in R based on Model Values from Huggard and Lewis and Lewis (2008), with weighted averages 3x Winkler, 2x Hudson, and 1x Hardy Bristow........................................................................ Stand Height versus Age of polygons in Bonaparte Plateau from 2009 VRI. Example of spruce as leading species (n=1601)..................................  4  14 17  Chapman Richards models for spruce of measured height versus predicted stand height (n=1601) ………………………………………………………  18  Area harvested (in hectares) between 1950-2009 in Fishtrap Creek watershed, according to 2009 VRI records……………....………………...  19  Freshet start versus year (left) and freshet start versus April air temperature (TApr ºC) (right)................................................................................................  22  Assumptions for model 4, assumptions not met in case shown based on model for predicting start of freshet.................................................................  24  Figure 8  Freshet runoff (m3/s) versus year (left) and vs. winter precipitation (right)...  26  Figure 9  Start of peak flow (day of year) versus year (left) and versus winter precipitation (mm) and April air temperature (right) ………………………..  28  Peak discharge (m3/s) vs. year (left) and peak discharge (m3/s) versus winter precipitation (mm) (right). (n=1879) ……………………..……….  30  Height difference in (m) between dominant and co dominant species versus polygon unit in Fishtrap Creek (n=51)............................................................  32  Figure 5  Figure 6  Figure 7  Figure 10  Figure 11  v  List of Tables Table 1. Modification of Bonaparte VRI species definitions (SPEC_CD_1) to generalized leading species groupings used in SAS analysis using the Chapman-Richards equation............................................................................................................... 12 Table 2. Scaled ECA values for Fishtrap Creek watershed and the Bonaparte Plateau. Only every fifth year is shown ……………………………………………..  20  Table 3 Summary of the analysis comparing which climate variables are best predictors of hydrological variables using an F test……………………………................. 21  Table 4 Models for start of freshet to test effect of pre/post fire class variable and ECA on regression........................................................................................................... 23  Table5 Analysis of variance to compare models for start of freshet period (day of year). 25  Table 6 Models predicting amount of freshet runoff (m3/s) with climate variables, pre/post fire ………………………………………………..………………........... 27 Table 7 Summary of regression models to predict the day that peak flow occurs (day)…. 29  Table 8 Summary of regression models for Peak flow discharge (m3/s) based on winter precipitation (Pwin), Pre/post forest class variable (Prepost), and equivalent clearcut area (ECA)................................................................................................ 31  vi  Acknowledgments  I would like to thank first and foremost my advisors Dan Moore and Valerie Lemay for all their constant help and feedback throughout this course of this project. Their contributions and encouragements as I was challenged to become more proficient in R, and to puzzle through the unexpected kinks that work with inventory data always presents, made this an invaluable learning experience for me. I would also like to acknowledge Joel Trubilowicz, Georg Jost and Matthew Chuang for their help with R and ArcGis. I would like to express many thanks to Edward Fong from Forrex for his helpfulness in preparing the historical VRI data and answering my many questions regarding the idiosyncrasies of the data. Thank you to Dave Hutchinson for providing the climate data from the Kamloops Weather Station and Lynne Campo from the Water Survey of Canada for the discharge data. Thank you also to my lab mates, friends and family for their encouragements and good humour throughout my project.  E.L.M.  vii  Introduction For over a century, research has focused on how forest harvesting and natural disturbance influence hydrologic processes and especially the timing and magnitude of streamflow and stream water quality (Moore and Wondzell 2005). Understanding these connections between changes in forest cover associated with forest disturbance and the subsequent hydrologic response of a watershed is of fundamental importance in sustainable forest management to ensure the protection of community water supply and aquatic ecosystems (Foster et al. 2005, Maloney et al. 2008, Redding et al. 2003). Of particular importance in the assessment of cumulative effects of multiple disturbance is an ability to model hydrologic recovery associated with post-disturbance stand development (Redding et al. 2008, Varhola et al. 2010, Weiler et al. 2007, Woodsmith et al. 2004). These needs have become even more critical in British Columbia, given the widespread forest mortality associated with the current outbreak of Mountain Pine Beetle, and the projections of increased wildfire frequency associated with climatic warming. Much of what is known about the connections between forest disturbance and watershed hydrology draws upon paired-catchment experiments (Moore and Wondzell 2005). However, these studies have mainly focused on headwater catchments, which are typically less than 1 km2 in area, and it is difficult to extrapolate results from these studies to the catchments of interest in forest management, which are typically tens to thousands of square kilometres in area. Two alternative approaches for larger catchments are 1) retrospective analyses of stream flow variations that use statistical modelling to separate the effects of forest cover change from climatic variability (e.g., Lin and Wei 2009) and 2) computer simulation using a deterministic model to serve as a control for climatic variability (e.g., Seibert and McDonnell 2010). 1  The objective of this study was to gain insight into the effects of wildfire and post-fire salvage logging on watershed hydrology. The study focuses on Fishtrap Creek, a snowmeltdominated catchment located northwest of Kamloops, which has been gauged by Water Survey of Canada since 1971. In 2003, the McClure Fire disturbed over 60% of Fishtrap Creek's catchment area, and much of the burned area was subsequently salvage logged (Eaton et al. 2010). The availability of three decades of pre-fire streamflow data presents a unique opportunity to conduct a retrospective study. Research on Fishtrap Creek's response to the McLure fire has focused on suspended sediment transport, channel morphology and bed material transport, stream temperature and climate effects (Eaton et al. 2010a, Andrews et al. 2007, Leach and Moore 2008, Leach and Moore 2010, Schneider 2008, Eaton et al. 2010b). Using a statistically-based retrospective analysis, Eaton et al. (2010) found that the onset of the snowmelt freshet and the timing of peak annual flow occurred earlier after the fire, and freshet season (March-July) runoff appeared to increase, but there was no evidence for a change in the magnitude of peak flows. A limitation to the analysis presented by Eaton et al. (2010) is that it did not account for changes in forest conditions prior to the fire. Examination of air photographs and remote-sensing imagery indicates that there had been extensive forest harvesting in the catchment prior to the fire. The hydrologic effects of this harvesting could have reduced the ability of their analysis to detect the effects of the fire and post-fire salvage logging. The Equivalent Clearcut Area (ECA) is a commonly used index of the catchment-scale influence of forest disturbance. To compute ECA, each forest polygon must be assigned a value for hydrologic recovery (HR), which quantifies the extent to which the hydrologic function has 2  recovered to the pre-disturbance condition. A fresh clearcut is considered to have HR = 0%, and HR increases and ultimately reaches 100% as a forest stand develops. Once HR is assigned to each polygon, ECA is then computed for each polygon as 1 – HR/100. These polygon-scale values for ECA are then used to compute a catchment-scale ECA by computing the product of polygon area and polygon ECA, then summing these for all polygons. In British Columbia, HR and stand-level ECA are typically treated as a function of stand height (e.g., Huggard and Lewis 2008, Lin and Wei 2000), which ignores the effects of canopy cover and leaf area index. In this study, the statistical analysis involved fitting regression models using both climate variables and ECA as predictors, following the approach applied by Lin and Wei (2009) to the Willow River catchment, a large snowmelt-dominated catchment located north of Fishtrap Creek near Quesnel, British Columbia. Eaton et al. (2010) found a significant relationship for the prefire period between freshet runoff (April-July) and the precipitation totals for the preceding winter (a measure of snow accumulation) and the current freshet season. The observed runoff values for the post-fire period were consistently and, in two years, significantly higher than values predicted using the pre-fire relation. It is hypothesized that the addition of ECA as an additional predictor to those used by Eaton et al 2010) would represent the hydrologic effects of pre-fire forest harvesting, in addition to the effects of climatic variability. This additional indicator could improve the ability of the analysis to detect the effects of the fire-related disturbance. The effects of the fire and post-fire salvage were incorporated by including a categorical variable as a predictor. The study focused on the same four streamflow metrics as analysed by Eaton et al. (2010): (1) annual peak flow (m3/s); (2) the date of peak flow; (3) the date of the onset of spring freshet (defined by Eaton et al.. as the first date on which discharge 3  exceeded 2 m3/s); and (4) the runoff during the freshet period (March to July).  Methods and Materials Study Area Fishtrap Creek is located in the southern interior of British Columbia (Figure 1). It is a tributary of the North Thompson River, which flows into the Thompson River near Kamloops.  Figure 1 Map showing Fishtrap Creek watershed and extent of the 2003 McClure fire. The watershed is the dark gray portion and the fire extent is area shown in light gray. Source: (Eaton et al. 2010b) 4  The landscape of the area has uplands consisting of rolling hills and steep-sided, deeply incised valleys. The soils are thin and developed on glacial till throughout most of the Bonaparte plateau. Volcanic soils dominate the southern part of the watershed and sedimentary rocks consisting of mudstone, siltstone and fine clastic turbidites, and basalt volcanic rocks are common throughout the rest of the watershed (Scheider 2008). The elevation ranges between 700 m and 1,800 m, with 60% of the elevation above 1,200 m. The climate in the region is characterized by cold winters and hot, dry summers. The watershed is in the rain shadow of the Coast Mountains and receives a mean annual precipitation of 279 mm in the lower parts of the valley, increasing to 700 mm on the upper plateau (Andrews 2007). The mean annual air temperature is and 7.5 °C (at 370 m) and 2.1 °C (at 1,620 m) (Eaton et al. 2010b). Mean minimum temperature is -5.11°C (at 370 m) and -8.5 °C (at 1,620 m, whereas maximum mean temperatures are 18.9 °C (at 370 m) and 13.7 °C (at 1,620 m) according to ClimateBC (Eaton et al. 2010 ). Fishtrap Creek has a nival flow regime, in which stream flow is dominated by spring snowmelt (Redding et al. 2008). Spring freshet normally begins in April, with peak flows in May, followed by a prolonged flow recession through summer, autumn and winter. Post-freshet rain events occasionally interrupt the post-freshet recession trend. For the period between 1972 and 2009, the maximum discharge was 14 m3/s, the minimum discharge was 0.034 m3/s, and the mean discharge was 0.8 m3/s or, in terms of runoff, 180 mm/a. The Biogeoclimatic Ecological Classification (BEC) zones present in the watershed are Engelmann spruce subalpine fir (ESSF) (19.3% or 25.8 km2) at the highest elevations in the south-western part of the catchment, Montane Spruce (MS) (57.75% or 97.4 km2) in the high to 5  mid-elevation portions of the watershed, Interior Cedar Hemlock (ICH) (9.8 % or 17.82 km2 ) between the MS and Interior Douglas-Fir (IDF) zones, and ICH or IDF (10.9 % of the area; 27.615 km2) in the gullies and valley bottoms (see Figure A3). The natural disturbance regime in the area is characterized by stand-replacing fires occurring every 100 to 300 years. Insect infestations also occur frequently. There was an outbreak of Spruce Budworm in the early 1990s and, in the past 10 years, Mountain Pine Beetle infestations have had a strong influence in the western part of the watershed (Andrews 2007). Based on the Base Thematic Mapping (BTM) performed in the mid-1990s (Environment Canada, unpublished), 56% of the Fishtrap Creek catchment was young forest and 21% was old forest1 consisting of mainly white spruce (Picea glauca), subalpine fir (Abies lasiocarpa), and lodgepole pine (Pinus contorta) at higher elevations. At lower elevations, Douglas-fir (Pseudotsuga menziesii), ponderosa pine (Pinus ponderosa) and white spruce (Picea glauca) dominate the species composition, with some occurrence of trembling aspen (Populus tremuloides). According to the BTM, 23% was classified as recently logged (i.e. within the preceding 20 years). Salvage logging occurred in the westerly-most portion of the watershed as a result of Mountain Pine Beetle. The McClure Fire of 2003 burned about 62% of the catchment. Post-fire salvage logging took place in 2005 -2008.  1 Old Forest is defined according to the BTM terms (Province of British Columbia) as forest greater than or equal to 140 years old and greater than 6 m in height.  6  Data collection Stream flow is monitored at a weir by Water Survey of Canada (station number 08LB024). The station was established in 1971 and has provided nearly continuous data since. During the 2003 fire, the shed housing the data logger was damaged and was rebuilt in March 2004, resulting in a 6-month period without data collection. Climate stations are located both in McClure and in Kamloops. Though the McLure Station is just outside of the boundaries of Fishtrap Creek's catchment, it does not have a continuous record of data collection for this stream; as a result, data from the Kamloops station were used to compute metrics of interannual climate variability. To characterize forest cover, this study used data from the Vegetation Resources Inventory (VRI), a province-wide data base maintained by the Ministry of Forests and Range. The field measurements were taken by landowners and managers using standardized procedures and measurements. The VRI is updated annually, and the most current version is available for download from the Province of British Columbia Data Distribution Service website. Phase 1 of the VRI process includes photo-interpretation to separate the land into forest polygons, and projection of attributes in these polygons using a Growth and Yield Model called Variable Density Yield Prediction (VDYP). Phase 2 of VRI is ground sampling using a representative sample of polygons. For both phases, information is obtained via government, industry and other personnel. While there is a diverse nature of the contributors, all are required to adhere to a consistent set of standards for VRI. However, there is diversity in forest inventory data since VRI has not been implemented throughout the province (Moss et al. 2006). As a result, different versions of forest inventory with different standards over time can be found for a 7  location. Data were collected at different times, and then growth and yield modelling is applied to the attributes (e.g. height, age, volume, dead volume, crown closure) each year to project forward from the date of sampling.  Backcasting forest cover polygon information The calculation of ECA requires tree heights for each polygon in the VRI over the last 40 years. Using the VRI projected heights and ages that were obtained using VDYP, a meta-model was developed using nonlinear regression to backcast canopy height (model described later) in time. VDYP is based on permanent sample plots and has been validated for use in BC To obtain a broader range of polygon attributes, polygon data from the surrounding Bonaparte Plateau were used, as it has relatively homogeneous climate, topography and distribution of BEC zones throughout.  In order to backcast stand ages from the current year of VRI inventory, assumptions had to be made about when stands initiated following a disturbance. The main assumption was that a stand age of 100 years was used as the time at which a major disturbance (i.e., harvest or fire) occurred. Therefore, in back casting stands that were less than 40 years old at the time of the inventory, each stand was considered to be 100 years old in the year preceding establishment of the current stand. This assumption was based on the premise that stands would need to be 100 years to be considered harvestable. In addition, regardless of site index, the hydrologic recovery would be 100% by that age. Though there is no specific information concerning fire intensity or severity for the McClure Fire, the assumption was also made that stands were most likely to be 8  burned in a fire with high stand volume and canopy closure, as would be the case in mature stands of around 100 years of age. There were certain stands where no forest attribute data were available despite their classification as forested land under the British Columbia Land Classification Scheme (BCLS) standards of the VRI inventory data. Also, in some polygons the BCLCS status had been changed as a part of a fire mapping project after the McClure Fire to a non-forest type classification status. Based upon this fact, the conclusion cannot be drawn that those areas now classified as non-forest polygons were in fact not forested prior to the fire disturbance. To address the absence of meaningful VRI data in this regard, the Baseline Thematic Mapping (BTM) done in the mid 1990s can be used to infer the state of the forest cover pre-fire and premountain pine beetle infestation. The whole watershed according to the BTM land-use map was forested or recently logged, apart from a negligible area classified as small lakes in the western portion of the watershed. There were a significant number of VRI polygons in Fishtrap Creek which had no forest attribute data since they had been recently disturbed. As a result, there were no data to backcast stand height. These were areas that were either burned in the 2003 fire or were subjected to other recent forest disturbance. To be able to evaluate conditions prior to fire, a 2003 set of the VRI was obtained. Unfortunately, an area of 7,708 ha (45.7 %) of the 16,868 ha study area of the watershed was part of the Weyerhaeuser Tree Farm Licence #35. The licensee had only given permission to add this area to the VRI inventory in 2008. Thus, the area was not incorporated into the 2003 data set, and no data were available for this large section of the watershed under pre-fire conditions. 9  To backcast stands over the area of the watershed for which pre-fire data were not available, a method of imputation was applied to obtain likely polygon attributes prior to disturbance. For this imputation, polygons in the Bonaparte Plateau with forest attributes were used to substitute attribute data for polygons in Fishtrap Creek's catchment that lacked stand attribute data, using only BEC zone and BEC subzone attributes to find a suitable match. In particular, the Site Index BEC (SIBEC) database (Ministry of Forest 2010) was used to generate estimates of site index and species based on BEC zones. SIBEC is a result of scientists and foresters acknowledging that relating BEC site factors could expand the range of applications possible and would improve the ability to manage certain areas lacking some aspects of forest cover attribute data (Ministry of Forest and Range 2003). The SIBEC data based were used to assign site index and leading species to the forest polygons with missing forest cover attributes based on the BEC zone and subzone information available. In order to evaluate the significance of using the imputation, the equivalent clear cut area for the Bonaparte Plateau was also calculated to create a frame of reference as to what the prefire ECA in Fishtrap Creek could have likely been. The assumption was made that the Bonaparte Plateau had experienced a level of forest harvesting and forest management similar to that in the Fishtrap Creek catchment. The values imputed were based only on Bonaparte Plateau polygons with the BEC zones that occur in Fishtrap Creek's catchment, namely MS, ESSF, IDF, and ICH.  Calculation of ECA over time for each polygon The data analysis in the study involved the following steps: (1) the calculation of stand age and height through backcasting for each polygon for each year; (2) calculation of time series 10  of the ECA for each VRI polygon in the Fishtrap Creek catchment and scaling it to the entire catchment; (3) extraction of time series of climate parameters, including winter precipitation, precipitation during the freshet period, and air temperature during freshet; and (4) calculation of time series of relevant stream flow parameters, including peak flow, date of peak flow, freshet runoff, date of onset of freshet. Data analysis was mainly conducted using the R programming language. However, SAS Version 9.2 for Windows was used to fit nonlinear regressions rather than R, as this was better suited to deal with irregularities in the data set.  Calculation of stand age and height for each polygon by year For each polygon in the VRI data within the Fishtrap Creek catchment, stand age was calculated for the period 1970 to 2010 as follows: SA(i, t) = PA(i)  2009 + t  (1)  where SA(i,t) is the computed stand age for polygon i in year t, PA(i) is the projected age in the VRI data base for polygon i for the year 2009, and t is the year (1970 to 2010). For a stand that originated after 1970, Eq. (1) will generate negative ages. As noted earlier, these negative stand ages were replaced by an arbitrary age of 100 years, which corresponds to fully-recovered conditions prior to stand initiation. Nonlinear regression was used to generate models to predict stand height from tree age and stand attributes, using the projected values provided in the VRI. Separate models were fit for each species and residuals from the fitted models were plotted against several stand-level variables, including basal area, site moisture, and canopy closure to assess whether they should be incorporated into the model. Of the variables considered, only site  11  index improved model fit. Species from the VRI were grouped into more general combinations of species within each BEC zone for species that were not present or had too few occurrences to fit a regression. In these cases, the polygons were substituted into similar BEC zone-species combinations with enough incidences present to fit a regression (Table 1). Missing combinations of BEC zone and species were also assigned a grouping (Table A4. T1).  Table 1 Modification of Bonaparte VRI species definitions (SPEC_CD_1) to generalized leading species groupings (LEAD_SP) used fitting a Chapman-Richards model.  Category  LEAD_SP  SPEC_CD_1  Species definition  Spruce  SX  „S „ or „SX‟ or  SX= hybrid spruce (Picea sp.); S= spruce (Picea sp.);, SE=  „SE‟ or „SS‟ or  Engelmann Spruce (Picea engelmanii); SS= Sitka Spruce  „SW‟  (Picea sitchensis.); SW= White Spruce (Picea glauca)  „FD‟ or „FDI‟ or  FD= Douglas fir(Pseugtsuga menziesii), FDI=interior  „HW‟  Douglas fir, HW= Western Hemlock(Tsuga heterophylla)2  Douglas  FD  Fir and Hemlock True Firs  BL  „B‟ or „BL‟  B= Fir(Abies), B= Subalpine Fir (Abies lasiocarpa)  Poplars  AT  „AC‟ or „AT‟  AC= Black cottonwood(Populus balsimifera), AT= Trembling Aspen (Populus tremuloides)  Birches  2  EP  „EP‟ or „E‟  EP= Paper Birch (Betula papyerifera),E= Birch(Betula)  Western hemlock was added into the Douglas-fir category to minimize model complexity because there were only two incidences of occurrence .The growth and yield of the two species are similar in the area.  12  The coefficients from the Bonaparte source area were applied to the polygons in Fishtrap Creek's catchment, based on BEC zones and leading species to project height using the Chapman-Richards model using Equation 2:  h ( S , x , i) = (a 0  a 1  SI)  (exp((  1 . 0  ( a 2  a 3  SI )  x )))  1 . 9  a 4  a 5  SI ))  (  1 . 0 ))  (2) where h(SI, x) is the projected height of the stand of each polygon i, SI is site index, x is the stand age from equation 1 (SA), and a1 to a5 are the estimated coefficients. Four nonlinear regression were test to generate the coefficients used to predict height (equation 2). Therse four methods were (1) based on species only, (2) a nonlinear regression based on species, BEC Zone and BEC subzone, (3) a nonlinear regression using the Chapman Richards Model, and (4) a non linear regression using the Weibull model.  Calculation of time series of ECA The equivalent clearcut area was calculated by a model developed by Lewis and Huggard and Lewis (2008), based on a synthesis of three stand-level studies of hydrologic recovery (Winkler 2001, Hudson 2000, Hardy and Hansen-Bristow 1990). Huggard and Lewis (2008) fitted a smooth Supline Curve between stand-level ECA and tree height, by incorporating a relative weight to the data from each of the three studies. Huggard and Lewis (2008) (Figure 2).  13  The calculation of catchment-scale ECA was based on the following equation:  1 n ECA(t) =  Ai ECA(i, t) Ac i=1  (3)  where ECA(t) is the catchment-scale equivalent clear-cut area for year t, Ac is the catchment drainage area, Ai is the area of polygon i within the catchment, ECA(i,t) is the stand-level equivalent clearcut area for polygon i in year t, and n is the number of polygons in the catchment.  Figure 2: Stand-level models for ECA based on the height of the dominant species (left) and canopy closure (right) Huggard and Lewis (2008). 14  Calculation of time series of relevant stream flow and climate parameters Four streamflow metrics were selected for analysis. The maximum discharge in a given year was extracted as the peak discharge (Qpk, m3/s) and the day on which it occurred (tpeak) was expressed as day of year (Jan. 1 = day 1). The timing of the onset of freshet (tfreshet) was defined as the day of the year that discharge first exceeded 2 m3/s (Eaton et al. 2010). The total freshet runoff (mm) was computed for each year by integrating daily streamflow for the period from March to July, inclusive, and converting it to an equivalent depth of water over the catchment area. Climate variables used in the analysis included the following: winter precipitation (Pwin), defined as the total precipitation that fell between Nov. 1 and Mar. 31; freshet precipitation (Pfresh), defined as the total precipitation between Apr. 1 and Jul. 31; mean air temperature in April (TApr); and the mean air temperature in May (TMay).  Differences in ECA before and after burning In the regression modelling of streamflow metrics, the years up to 2002 were included in the pre-fire period, 2004 to 2010 was classified as post-fire. In the post-fire period, pre- and post-savage period were also analyzed separately to see if there was a significant change in the hydrologic parameters following salvage logging. Statistical models were fitted to relate each of the streamflow metrics to (a) climate variables, (b) ECA, and (c) a categorical variable distinguishing the pre- and post-fire periods. The predictors were tested in two different time periods, where the regression either used only the years preceding the fire (1971-2002) or the whole period for which records are available. 15  Goodness of fit parameters was used to evaluate which set of variables (pre/post fire category, ECA) and their possible interactions best predicted streamflow. F-tests and P values (probability of Type I error) compared against alpha of 0.05 were used to test for significance of the models. Models were tested using analysis of variance, and those that were significantly different were then compared using Akaike's Information Criterion (AIC), computed as follows: ESS AIC ( n , ESS , k , npar , L) =  k log(  )  ( k  npar ) (4) n  where ESS is the error sum of squares, n is the number of observations, npar is the number of parameters, and k = 2. A lower AIC indicates a better specification. The use of AIC allows models with different numbers of parameters to be compared. AIC tends to penalize models with more independent variables more so than adjusted R2, and often selects models with the fewest independent variables (Hofler 2010). A normal probability plot was examined to test whether the residuals were normally distributed. Independence and heteroscedasticity of residuals were tested by plotting them against the 1) climate variable(s) used, 2) year, and 3) ECA. A DurbinWatson test was used to assess the presence of autocorrelation in the residuals.  Results Relations between stand height and age Figure 3 provides an example of the relation between stand height and age for spruce (Picea glauca), illustrating the nonlinear relationship. The maximum height for is around 40 m, which is reached by an age of about 100 years. Much of the scatter in Figure 3 can be explained  16  by the inclusion of site index in the model.  Figure 3 Stand height versus age of polygons on the Bonaparte Plateau from 2009 VRI. Figure 4 shows that the fitted Chapman-Richards model provides a good fit for tree heights up to 30 to 35 m but diverges slightly for the species combination shown. However, that divergence did not influence the calculation of ECA, as the model developed by Huggard and Lewis (2008) predicts full recovery at a tree height of 25 m. Of the four non linear regessions tested to predict tree height, the regression using the Richard Chapman model was selected. It had low root mean square error and high coefficient of determination and was therefore used to generate the coefficients for the height prediction model  17  (equation 2).  Figure 4 Comparison of stand height as predicted by the fitted Chapman-Richards model versus measured height (n=1601).  Havervest history and ECA The harvesting history was reconstructed using the dates of harvesting provided for each polygon in the 2009 VRI data base (Figure 5). There has been ongoing harvesting since before the 1970s, with the most substantial harvesting in the past decade due to fire salvage and postMPB salvage (i.e. 2200 ha). The 1970s and 1900s also had large areas harvested both between 1300 and 1500 ha.  18  Figure 5 The area harvested in hectares between 1950-2009 in Fishtrap Creek watershed according to 2009 VRI records. Total area of watershed is 16,868 ha. (n = 1,879 polygons).  The ECA in Fishtrap Creek's catchment generally increased from 1971 to 2003, whereas for the Bonaparte Plateau it remained fairly steady (Table 2). The increase in ECA for Fishtrap Creek is partially due to the use imputed ages of burned polygons that increased steadily from an age of 62 in 1971 to 100 years in 2003. The ECAs in 1986 are similar for the Bonaparte Plateau and Fishtrap Creek. In 2006, the effect of the fire is clear in Fishtrap Creek, given that the ECA rose from 17.75 in 2001 to 69.1 in 2006. ECA on the Bonaparte Plateau did not increase between 2001 and 2006. The Fishtrap Creek area is not a significant portion of the area, and thus does not affect ECA for the Bonaparte Plateau. Based on these observations, the analysis for the 19  Bonaparte Plateau does provide an indication of the ECA conditions in Fishtrap Creek that might have been expected without the occurrence of fire in this area.  Table 2 Scaled ECA values for Fishtrap Creek watershed and the Bonaparte Plateau. Only every fifth year is shown . Year  Fishtrap Watershed ECA  Bonaparte Plateau scaled ECA  1971  12.61  19.738  1976  13.66  18.97  1981  16.77  18.79  1986  18.32  19.314  1991  19.2186  19.94  1996  20.11  20.59  2001  17.75  19.4936  2006  69.1  18.84  Regression modeling of the relation between streamflow parameters and climate variables Table 3 summarizes the results of the regression analyses tested for the pre-fire period in the case of the above analysis. For the freshet start date, only April air temperature was a significant predictor. For freshet runoff, both winter and freshet precipitation were significant  20  when used as predictors in a multiple regression model. The best model for predicting the annual peak flow included winter precipitation, which is a measure of snow accumulation. The date of the annual peak flow was best predicted by April air temperature.  Table 3 Summary of the analysis comparing which climate variables are best predictors of hydrological variables using an F test. Streamflow metric tfresh  ROfresh  tpeak  Qpeak  Climate predictor TApr  P value  Adjusted R2  4.975E-07  0.486  Pwin  3.723E+03  0.005826  Pwin  1.098E-06  0.5257  Pfresh  1.424E-02  0.2357  Pfresh+ Pwin  1.511E-06  0.5635  TMay, TApr, Pwin  3.776E-02  0.1768  Pwin  1.165E-01  0.04943  Tapr  5.011E-03  0.20873  TMay + Pwin  1.916E-01  0.04615  Pwin  1.772E-05  0.4462  The start of the freshet runoff period has a negative trend with time (Figure 6). The start of freshet is about two weeks earlier in the last decade than in 1970. Of the post-fire years, three have a start of freshet significantly earlier than average and the other four post-fire years have start of freshet occurring below the trend line (earlier than predicted). The post-fire years show a much higher range of variability than in previous (pre-fire) years (Figure 6, left hand side). The effect of the fire and post-fire salvage harvesting is evident in that the start of freshet over the  21  four years has a larger range than the pre-fire salvage years, but always occurs earlier than expected. The start of freshet has a negative relation with April air temperature (Figure 6, right panel). All the post-fire years lie below the trend line, indicating earlier start days than predicted for pre-fire conditions. Two of the post-fire data points fall outside the 95% prediction interval, supporting the significance of the apparent effect of the fire.  Figure 6 Date of the start of freshet versus April air temperature.  22  Regression analysis of start of freshet Results of the model fitting for freshet start date are summarized in Table 4. All the models are statistically significant. Models 3 and 6 have high coefficients of determination (R2= 0.582 and 0. 586). In terms of Akaike's Information Criterion, Models 1 (AIC= 195.47) and 3 (AIC= 197.092) have the lowest AIC.  Table 4 Models for start of freshet using predictor variables April air temperature (TApr), pre/post fire categorical variable, and ECA for the pre-fire period (1971-2002) or whole period (19712009). Model Predictors  Adjusted R2 P Value  AIC  RMSE  200.5169  4.757  0  Prefire period, TApr  0.422  2.48E-05  1  Prefire period, TApr , ECA  0.5175  6.789E-07  2  Whole period, Climate  0.3692  2.440E-05  258. 1852  6.3  3  Whole period TApr . ECA  0.5818  5.79E-08  243.0891  5.13  4  Whole Period, TApr , Pre/post  .4048  1.000E-07  202.3992  4.827  5  Whole period, TApr , ECA, prepost  0.5065  2.877E-05  197.0973  4.827  6  Whole period, TApr ,ECA , pre/post, Interactions (pre/post* Tapr, Pre/post* ECA)  0.5861  1.46E-6  245.2886  5.103  195.4715  4.34  Figure 7 provides an example of the diagnostic tests of the regression assumptions for 23  model 3. There is a significant lag-one correlation in the residuals, violating the assumption of independence in time (Figure 7). Model 2 also lacks independence because the residuals plotted against the year show a negative trend. In all of the nine models tested, four met all the assumptions (Models 1, 3, 5, 6). The model using only april air temperature for the pre fire period and the one using pre/post-fire period did not meet the required assumptions (models 0, 2 and 4). Of the models that met all assumptions, only models 1 and 3 had predictors which were all significant (e.g. for model 1, TApr t = -4.104, and ECA t = -2.672). Because these models do not include nested sets of predictors, ANOVA cannot be used to compare these models. AIC does not involve this restriction, and suggests that =model 1 provides a better fit than either models 3 or 5.  Figure 7 Diagnostic plots to test regression assumptions for model 4 for start of freshet runoff period.  24  Of the models based on the pre-fire period, models 1 is significantly different from model 0, indicating that the addition of ECA as a predictor is meaningful. As models 4, 5 and 6 have predictors that are not significant (based on t tests), only model 3 can considered as meaningful. Models 1 and 3 share TApr and ECA as predictors and are distinguished by the fact that model 1 is based only on the pre-fire period, while model 3 is based on the whole period.  Table 5 Analysis of variance to compare models for start of freshet period. For each model, those that provide statistically better fits are listed, along with the P value from the analysis of variance. Model  Models significantly different  P Values  0  1  1.207E-02  1  0  1.207E-02  2  3, 4, 5, 6  7.945E-05  3  2  7.945E-05  4  2,5  7.033E-04  4.111E-02  5  2,4  4.213E-04  4.111E-02  6  2  1.135E-03  7.033E-04  4.213E-04  1.135E-03  Regression analysis of freshet runoff  25  The amount of runoff during the freshet period displays no clear trend in time (Figure 8). However, it is positively related to winter precipitation (Figure 8). Runoff in the post-salvage years (2006-2010) consistently plots above the regression line, consistent with the expected increase in runoff associated with tree removal.  Figure 8 Freshet runoff (mm) versus year (left) and winter precipitation (right).  Freshet runoff had a best fit model that was a multiple regression with winter precipitation and freshet precipitation as predictor variables (Table 6). None of the models containing ECA or pre/post provided a better fit than models containing only climate variables as predictors. Therefore, no conclusion can be drawn regarding the effect of forest cover change on freshet runoff.  26  Table 6 Models predicting the amount of freshet runoff. Model  Predictors  Adjusted R2  0  Prefire period, Pwin, Ffresh,  1  P Value  AIC  RMSE  0.5635  1.511E-06  337.2319  37.23  Prefire period, Pwin, Ffresh, ECA  0.5525  7.190E-06  338.9348  37.7  2  Whole period, Pwin, Pfresh,  0.4861  2.363E-06  401. 447  39.07  3  Whole period Pwin, Pfresh, ECA  0.5116  3.095E-06  400.3589  38.08  4  Whole period, Pwin, Pfresh, , Pre/post  0.5561  6.422E-06  338.6705  37.54  5  Whole period, Pwin, Pfresh, ECA, prepost  0.5413  2.638E-05  340.5897  38.16  6  Whole period, Pwin, Pfresh, ECA , pre/post, Interactions: (pre/post* Tapr, Pre/post* ECA)  0.5413  2.638E-05  340.5897  38.16  Regression analysis of day of peakflow The date of peak flow exhibits a negative trend through time (Figure 9). It also has a negative relation with April air temperature. All of the post-fire years plot near or below the regression line, consistent with the finding from other studies that snowmelt peak flows occur earlier after removal of forest cover (Moore and Wondzell, 2005).  27  . Figure 9 Date of peak flow (day of year) versus year (left) and versus winter precipitation (mm) and April air temperature (right)  For the whole period analysis, models including ECA and/or pre/post (models 3, 4 and 5) had lower values of AIC than model 2, which only included TApr. Model 4, which included only pre/post as a measure of forest cover change, had the lowest AIC. This result supports the apparent advance in the date of peak flow following the fire, as seen in Figure 9.  28  Table 7 Summary of regression models to predict the day that peak flow occurs (day). Model  Predictors  Adjusted R2  P Value  0  Pre-fire period, TApr  0.2137  3.950E-03 255.5310  1  Pre-fire period, TApr , ECA  0.1877  1.679E-02 257.527  2  Whole period, TApr  0.2647  4.776E-04 296.7611  10.33  3  Whole period TApr . ECA  0.2752  1.152E-03 297.1307  10.26  4  Whole period, TApr , Pre/post  0.1968  1.420E-02 257.154  11.06  5  Whole period, TApr , ECA, prepost  0.1716  3.751E-02 259.0514  11.24  6  Whole period, TApr , ECA, Prepost + Interactions: (pre/post* Tapr, Pre/post* ECA)  0.1716  3.751E-02 259.0514  11.24  AIC  RMSE 10.95 11.13 on  Regression analysis of annual peakflow Annual peak flows in the post-fire period had much lower variability than for the pre-fire period and tended to be near the pre-fire mean. In the plot of annual peak flow versus winter precipitation, post-fire peak flows plotted near or below the regression line, suggesting that the fire and salvage harvesting did not produce an increase in peak flows, contrary to findings from previous studies focused on smaller catchments (Moore and Wondzell, 2005).  29  Figure 10 Annual peak flow vs. year (left) and peak flow versus winter precipitation (right).  None of the models including ECA or pre/post had lower AIC than the corresponding model that included only winter precipitation. This result indicates that there was no statistically detectable effect of forest cover change on annual peak flows.  30  Table 8 Regression models for annual peak flow based on winter precipitation (Pwin) and forest cover predictors. Model  Predictors  Adjusted R2  P Value  AIC  RMSE  0  Prefire period, Pwin  0.4465  1.287E-05  156.8976  2.456  1  Prefire period, Pwin, ECA  0.4397  6.387E-05  158.2185 2. 2.471 on (  2  Whole period, Pwin  0.4241  4.284E-06  181.0428  2.343  3  Whole period Pwin. ECA  0.4081  3.001E-06  183.0428  2.376 (36)  4  Whole period, Pwin, Pre/post  0.4466  5.306E-05  157.8105  2.456  6  Whole period, Pwin, ECA, prepost  0.4552  1.160E-04  158.1748  2.437  6  Whole period, Pwin , ECA, prepost + interactions: (pre/post* Pwin, Pre/post* ECA)  0.4552  1.160E-04  158.1748  2.437  Height difference between first listed species and second listed species in VRI The differences in the height between first-listed species in the VRI (referred to as dominant) and second-listed species (referred to as co-dominant species) were compared to assess whether using only the height of dominant species as the indicator for predictor variable of ECA is a significant variable (see Figure 11). These differences ranged from -3 m to 2 m, with the majority of height values falling within a 1 m height difference or less.  31  Figure 11: Height difference between dominant and co-dominant species versus polygon unit in Fishtrap Creek (n=51).  Discussion Assessing hydrologic response of Fishtrap creek related to freshet period Of the four hydrologic parameters examined here, only the “start of freshet” was significantly influenced by forest cover. The freshet runoff started earlier for all post-fire years and demonstrated some freshet years that were significant earlier, and did fall outside of the 95%  32  prediction interval for the pre-fire regression. The models incorporating ECA and April air temperature were the best regression models. The “start of freshet” is therefore the most sensitive streamflow metric to change in forest cover. Winkler (2009) and Huggard and Lewis (2008) generalized that a watershed is not highly sensitive to changes of forest cover when the area impacted is below a threshold of 20%. The indices of hydrologic recovery are a tool that can verify and support the significance of this threshold established in literature. The ECA prior to the fire was just below the threshold of 20%; after the fire, the ECA rose significantly and the effect is evident in the “start of freshet” indicator. Freshet runoff was higher than predicted based on the pre-fire regression in all years following the fire. Prediction of freshet runoff was not significantly improved by utilizing forest cover predictors and it is therefore insensitive to forest cover change. Based on plot-scale studies at sites disturbed in the McLure Fire, Winkler (in press) suggested that the effect of the forest canopy is largest in years where there is high winter precipitation. Unfortunately, there were no large snow pack years following the fire to test as to whether the freshet runoff would have responded more sensitively under large snow pack conditions.  Assessing hydrologic response of Fishtrap creek related to peakflow In terms of the “day of peak flow,” none of the post-fire years fell outside the prediction interval but the “day of peak flow” did tend to occur earlier than the average, consistent with previous studies (Moore and Wondzell, 2005). For peak flow discharge, there was a moderate relation with winter precipitation, and the inclusion of forest cover did not significantly improve model fit, suggesting that peak flows were not highly sensitive to forest cover change, including the fire and salvage harvesting. Lin and Wei (2010), however, found that incorporation of ECA 33  did contribute the prediction of peak flow, but in their case, peak flows had seen an increase following widespread harvesting. A possible contributing factor to the unusual results at Fishtrap Creek would be the rolling hill topography. Under conditions of uniform forest cover, the lack of strong aspect and elevation contrasts would mean that the snowpack would melt in a relatively uniform fashion. Following partial forest disturbance, the disturbed areas would melt earlier, resulting in a desynchronization between forested and disturbed areas. Particularly in lower snowpack years, this de-synchronization could cause a reduction in peak flows. The analysis performed is a results-based method of accounting for the hydrologic parameters based on available empirical data. This result-based analysis is a very commonlyused method of making management decisions, as such, the robustness of this analysis is highly important to consider. When considering the use of results-based management decisions in the context of the high variability of factors in watersheds, indicators must be carefully chosen and considered for their high sensitivity to changes in the watershed.  Practical implication from study related to post fire hydrology and effect of forest cover An important practical implication of the results of this study is related to recent research on the impact of climate warming on the timing of start of snow melt. For example, Stewart et al. (2005) found that the onset of snowmelt in the western United States advanced from one to four weeks over the last five decades. Such an advance is important in its own right, as the timing of stream flow is an important control on availability of water supply and aquatic ecology. Changing of the timing of snow melt is also important as it appears to be a sensitive 34  indicator of climatic warming, independent of the uncertainties associated with air temperature measurements. The results of this study highlight the need to differentiate the changes in snow melt timing caused by changes in forest cover from the changes due to altering climate patterns. In order to be able to differentiate these two factors, accurate forest records and frequent update of new changes in data records are necessary. This is a challenging task, given the demands of maintaining accurate forest cover records in a large forest area such as British Columbia, particularly given the current lack level of infrastructure and resources needed to maintain sophisticated and accurate levels of forest inventory data. Though stand height and canopy closure are the most significant indicators of ECA , the species distribution must also be taken into account. However, adding in additional species, though giving a more complete model, would not necessarily improve the estimation of the hydrological recovery and would confound the calculation to such a degree that the usefulness of the data could decrease considerably. Since the goal of ECA is to provide an indicator to guide forest managers, added complexity is to be avoided if it does not improve the model significantly.  Effects of assumptions made on the models analyzed Ideally a model to predict hydrologic recovery would be used that encompassed a wider range of stand complexity including, for example, canopy closure and a multi-species distribution. In the Fishtrap Creek inventory data, the primary species accounted for on average 71% of the species percent, whereas the second-listed species accounted for 21%. In many cases the second-listed species was not necessarily the co-dominant species (i.e. in some case it 35  was taller than first-listed species). If the choice had been made to perform the modeling on more than just the first species, the ECA would not have been substantially affected by an average height of 1.2 m height difference between the first and second species. In addition, only 4% of the polygons had a non-zero height difference. Restricting the use of a single species for the modelling is representative of stands in the ESSF and MS BEC zone in the dry climate of the southern interior, where single-species stands or stands heavily dominated by one species are common. Use of a two-factor species model would have added complexity and decreased the practicality for use as a management-decision tool. There is an apparent limitation to the number of factors included in the calculation of the ECA because it is based essentially only on the tree height and not combined with canopy closure as well. There is a high level of uncertainty projecting ECA when combining the many differing components of a watershed and forest stand. Only the statistical uncertainty can be quantified and addressed whereas there is an awareness that the other uncertainties inherent in this biologic process exist. Because canopy closure and height are correlated, including both in the model for stand-level ECA would have led to a high degree of confounding (Huggard 2008). There are discrepancies in the data available from Environment Canada concerning certain features of the watershed attribute and the VRI data. This brought to light challenges in using VRI to generate the forest cover projections in order to provide data which feed into the ECA calculations. For example, though the area of the watershed at the gauging station is 142 km2, the area of VRI polygons that were in the catchment was 168 km2. A further challenge inherent in the data itself emerged through use of the same shape file in ArcGIS, as in this application the external boundaries of each polygon were included in the extracted polygon, and 36  this resulted in an overestimation of the area. The factor was balanced by the fact that the watershed ECA is scaled upwards to the watershed scale resulting in a net cancellation the dimension of area and as such, the ECA term is therefore valid for interpretation. Estimating Equivalent Clearcut Area is based on forest polygons; however, the VRI definition of land class under the British Columbia Land Classification Scheme (BCLCS) complicates this data interpretation. Another important consideration is that data in the inventory do not always reflect changes that have occurred on the land base; for example, the inventory indicates that only 50% (i.e. 85.38 km2) of the Fishtrap Creek watershed was burned by the McClure fire, while in fact 70% of this area was burned. This kind of underestimation of changes in forest due to natural occurrences or harvesting activity is a reflection of the extent of contribution of licensees and landowners to updating the inventory information. Updating inventory records, especially for small-scale salvage or natural forest disturbances, is often given a high priority in terms of complete and timely reporting. . According to the BTM mapping, 23% the area of Fishtrap Creek was harvested in the 20 years preceding the 1990s data collection. The VRI records indicated that in the 1970s ad 1980s, 14% of the watershed was harvested , 8% less than the BTM estimation. This comparison indicates that there are likely forest harvesting activities and possibly other forest cover changes that are not reflected in the inventory data. In dealing with the VRI data, many complications and issues revealed problems that the inventory presents in attempting to use VRI for hydrologic and forest growth models. In a review of the effectiveness of the VRI program, the Inventory Review Panel (IRP) in 2007 addressed this with an assessment of the status of the inventory (Ministry of Forest 2007). 37  Different stakeholders connected with the inventory program all participated in the assessment and release a report entailed “The State of Forest Inventory in British Columbia.” Their observation was that the design of the program was in fact flexible as they had intended, but the implementation has less effective than hoped for. The Berlin, Easton, and Associates firm who worked on a main component of the mapping for the inventory commented that it was the most disproportionately complicated project to work on. The IRP identified that one of the problems in developing and maintaining the inventory is the fact that since its creation in 1992, the inventory projects had always been subject to very cyclic funding and that a consistent level of baseline funding would be necessary to steadily improve the inventory‟s short-comings (Ministry of Forests 2007).  Recommendations for improving the usability of VRI data in Result Based Analysis and studies involving change through These short-comings were underscored in the analysis of forest cover changes at Fishtrap Creek. Because participation in the VRI program is voluntary, there is an incomplete coverage of inventory throughout the whole province. The areas where there is active or recent harvest activity since the early 1990s tend to have the most complete and accurate set of inventory data. Areas with very low commercial activity and areas that are more remote, such as in the Northeastern section of the province, have especially low instances of VRI data entered (Lemay 2010). About 15 % of the area in the Kamloops Forest Region has been photo-interpreted and Phases I and II of ground sampling were completed in 2005. Many of the Forest Regions in the Province 38  have had more detailed inventory information added in recent years, but the discrepancy in data quality, quantity and consistency across the Province is a common problem. Polygons are being redefined as additions are made and changes to the management and changes to jurisdiction are made. This factor makes it difficult to compare the inventories through time. In the Fishtrap Creek watershed, Weyerhaeuser holds the Tree Farm Licence 38, which covers the majority of the south portion of the watershed and was only added in 2008. For this reason there is no baseline of comparison for the forest cover information for this very large portion of the watershed. The ability to be able to compare the current status of the vegetation cover to historical records is essential for a large number of applications for various stake holders. Forest managers, researchers, and policy makers need to have access to information concerning the change through time of the forest. All analysis involving change of growth and yield through time which are constantly altered by forest disturbance is of much greater relevance if a series of historic inventories are available. These analyses are much more difficulty currently, given that only the most recent years‟ VRI is available from the Ministry, and this data is replaced yearly with the most current version. Evaluating these data records over time would be significant to enable research to form a closer connection between the hydrologic response of our watersheds to the disturbances and changes occurring within in them. The protocols for the three parts of the VRI procedure -- aerial photograph interpretation, ground sampling and net volume adjustment factor calculations -- are very extensive. In fact, the protocol is so complicated that there is a considerable cost in both time and financial investment to implement it. The interpretation of the data frequently involves weighted averages, and such 39  interpretation takes expertise and resources to correctly complete (V. Lemay, 2010 personal communication). The protocol as it is today was created by a committee method, and the data resource fields that were accumulated from it are large, as each committee member‟s contribution was kept as an assurance. With input from such a wide range of fields, including fields with considerable overlap, there are redundancies in the protocol that are unwieldy. Also, many fields in the protocol tend to be only partially filled out, as each contributor must use their own judgement as to which types of data are relevant and feasible to collect and enter. Improving the quality of VRI inventory records could be achieved if a working group or task force could be mandated as responsible for the continual improvement of this project over a consistent long term time frame which would allow for more steady improvements and more systematic sampling methods and re -sampling methods. Shifting the cyclic nature of work on the VRI to a more continuous data collection method would result in a more accurate record of the actual state of the forest cover, and in turn, these data would be more useful and reliable for a diverse range of applications. In comparison, the federal inventory protocol is much more effective and has a systematic pattern of sampling to ensure complete coverage. The disadvantage with the federal inventory is that the scale is large and the data are not precise enough to be useful for numerous applications, such as the analysis in this study. (Lemay 2010). The provincial jurisdictions lack the adequate resources to be able to fund the inventory projects sufficiently. British Columbia‟s system of adding ground sampling to photo interpretation is ad-hoc and has a lack of emphasis for encouraging private land owners to contribute to the inventory – these are major weaknesses. In countries with massive forested area and lower population numbers such as Canada 40  and Russia, maintaining accurate records of the state of the forest cover is a huge challenge. Continuous data collection and the ability to interpret data through time would enable more meaningful and comprehensive types of analysis and research. Much of the current modelling focus in British Columbia has been directed towards forecasting shifting species range in the light of changing climatic conditions. Accurate forest cover records are the fundamental basis of accurate assessments of these types. A further recommendation would underscore the value of reassessing the complexity of the input fields in the inventory collection protocol, to achieve higher cohesiveness, while still maintaining flexibility of the data base. A third recommendation would be that a more systematic method of implementing the coverage of sampling and updating inventory information be developed. One possibility to address this recommendation might entail dividing the province into quadrants upon which rigorous data checking and re-sampling should occur. These four quadrants could be cycled through, to ensure more complete coverage and focus of emphasis in inventory data sampling and collection. (V. Lemay, 2010, personal communication). The availability of resources would be the determining factor upon whether this kind of more systematic coverage of sampling would be possible. The challenge of maintaining a high quality and complete inventory on such a large area requires the coordinated efforts of all stakeholders, and a need to frequently reassess the effectiveness of the program.  41  Conclusions The analysis of the effect of forest cover on hydrologic recover using the indicator of Equivalent Clear Cut Area (ECA) demonstrated that the start date of freshet was the most sensitive. The other hydrological parameters (freshet runoff, day of peakflow, and peakflow discharge) were less sensitive to the forest cover change, based upon the analysis in this study. The start of freshet may be a good indicator that could be used to monitor the hydrologic effect of forest cover change for results-based management. The peak flow parameters were insensitive indices for forest cover change related to the forest hydrology but this may be due to a real lack of increase in peakflow following the fire in Fishtrap Creek following the fire. The desynchronization of melt could be a significant driver of the behaviour of peak flow in plateau watersheds. This study also highlighted that there are challenges in using the current form of inventory on forest cover in British Columbia to do results-based analysis. It also tested the extent to which real changes in forest cover can be modelled with the inventory data alone. Addressing some of issues in using the inventory data would enable forest managers to perform results-based modeling of hydrologic cover more easily with the goal that it can serve as a practical decision making tool to asses impact of forest cover change. In particular, if the Ministry of Forest inventory branch could make historical records available to analysts, researchers, and forest managers, they could perform analyses involving change over time. For modelling, for example hydrologic recovery or forest growth and yield, historical records would result in more accurate results and ones representative of the specific parcel of land in question than the alternative of using imputation.  42  References Ager A, Clifton C. 2006. Software for calculating vegetation disturbance and recovery using equivalent clearcut area model. USFS Notes. PNW-GTR-637: 1-18. Alila Y, Beckers J. 2001. Using numerical models to address hydrological forest management issues in British Columbia. Hydrological Processes. 15: 3371-87 Beckers J, Smerdon B, Wilson M. 2009. Review of Hydrological Models of Forest Management and Climate Change Applications in BC and Alberta. FORREX Series 25: 1-179. Beeson P, Mertens S, Breshears D. 2001. Simulating overland flow following wildfire: mapping vulnerabilities to landscape disturbance. Hydrological Processes 15: 2917-30. Benda D, Tans P, Monson R. 2001. Partitioning net ecosystem carbon exchange with isotopic fluxes of CO2. Global Change Biology 7: 127-45. Burch G, Moore, I Burn J 1989., Soil hydrophobic effect on infiltration and catchment runoff . Hydrological Processes 3: 211-22. Cyzik K, Hogue T. 2009. Modeling post fire response and recovery using the Hydrological Engineering Center Hydrologic Modeling System HEC-HMS. Journal of American Water Resources Association 45: 702-14. Eaton BC, Andrews CAE, Giles TR, Phillips JC. 2010b. Wildfire, morphological change and bed material transport at Fishtrap Creek, BC. Geomorphology 118: 409-24. Eaton BC, Moore RD, Giles TR, Heise B, Owens P, Pettigrew E. 2010b. Fishtrap Creek Watershed Project. Streamline Watershed Management Bulletin 14: 12-13. Eaton BC, Moore RD, Giles TR. 2010a. Forest fire, bank strength, and channel instability: the Unusual response of Fishtrap Creek, British Columbia, Earth Surface Processes and Landforms: 35: 1167-83. Foster N, Beall F, Kreutzweiser D. 2005. The role of forests in regulating water: The Turkey Lakes Watershed Case Study. Forestry Chronicle 81: 143-148. Hardy J, Hansen Bristow K. 1990. Temporal accumulation and ablation patterns of the season snowpack in forests representing various stages of growth. In Proceedings of the 52nd Western Snow Conference, Sacramento, CA: 23-24. Hofler Richards. 2010. Akaike's Information Criterion and the Schwarz Criterion. Economics 6416 Class notes. University of Central Florida: 1-5. Huggard D, and Lewis D 2008. Effects of salvage options for beetle-killed pine stands on ECA. Febuary 2008 Update. Streamline Watershed Management Bulletin 31:1-10. Isaak DK, Luce C Rieman H, Bruce, E, Nagel D, Paterson E, Horen D, Parkes S, Changel G. 2010. Effect of climate change on wildfire on stream temperature and salmonid thermal 43  habitat in mountain river network. Ecological Applications. 20:1250-371. Laung-Zhang, L. 2003. Generalized Chapman-Richards Model and application to tree and stand modeling. Journal of Forestry Reseach 2: 23-25. Leach J, Moore RD. 2008. Stream temperature response to wildfire distrubance: Lessons from Fishtrap Creek. Streamline Watershed Management Bulletin. 12: 11-16 Leach JA, Moore RD. 2010. Above-stream microclimate and stream surface energy exchanges in a wildfire-disturbed riparian zone. Hydrological Processes 24: 2369-2381, DOI: 10.1002/hyp.7639. Lin Y, Wei X. 2009. The impact of large-scale forest harvesting on hydrology in the Willow watershed of Central British Columbia. Journal of Hydrology 359: 141– 149. Maloney D, Rex J, Redding T, Winkler R. 2008. Mountain Pine Beetle, Forest Practices, and Water Management. B.C. Min. For. Range, Res. Br., Victoria, B.C. Extention. Note 88: 1-11 Ministy of Forest 2010. Standards and Procedures for Net Volume Adjustment Factor. Version 4.4 Ministry of Forests and Range. 2007. Assessment of Status of Fires Inventory Report: 1-11. Ministry of Forests, Forest Science Program. 2009. Variable Density Yield Projection Volume 2.WinVDYP User Guide 2.2: 1- 58. Ministry of Forests. 2003. SIBEC Site Index Estimation is Support of Forest Management in BC. 04: 1-58. Ministry of Forests. 2008. Site Index Estimates by Site series: Report by biogeoclimatic Unit. SIBEC RDM: 1-145. Moody J, Martin D. 2001. Initial hydrologic and geomorphic response following wildfire in the Colorado Front Range. Earth Surface Processes and Landforms 26: 1049-70. Moss I, Marshall P, Lemay V. 2006 Assessment of the status of forest inventories in British Columbia. www.abcfp.ca/publications_forms/publications/committee_reports.asp Phillips JC Eaton B. 2009. Detecting the timing of morphologic change using stage-discharge regression: a case study at Fishtrap Creek, British Columbia, Canada. Canadian Water Resources Journal. 34: 1-16. Pierson F, Robichaud, Speath P. 2001. Spatial and temporal effects of wildfire on the hydrology of a steep rangeland watershed. Hydrological Processes 15: 2905-16. Pike R, Scherer R. 2003 Overview of the potential effects of forest management on low flow in snowmelt- dominated hydrologic regimes. BC Journal of Ecosystems and Management. 3: 2905-16. Redding T, Winkler R, Spittlehouse DS, Carlyle-Moses D. 2008. Mayson Lake Study examines hydrological processes. FORREX Forest Research and Extension Partnership 9-10-1. Redding T, Winkler R, Spittlehouse S, Moore, RD. Wei A, Teti 2007. Mountain Pine Beetle and Watershed Hydrology: A synthesis focused on the Okanagan Basin. 44  Schneider J. 2008. Impacts of climate change on catchment storage, stream flow recession and summer low flow. Diplom thesis :1-154 Seibert J, McDonnell JJ. 2010. Land-cover impacts on streamflow: a change-detection modelling approach that incorporates parameter uncertainty. Hydrological Sciences Journal 55: 316-332 Shakesby R, Doerr S. 2006. Wildfire as a hydrological and geomorphic agent. Earth–Science Review 74: 269-307. Silins U Stone M, Emelko M, Bladon K. 2009. Sediment production following severe wildfire and post fire salvage logging in the Rocky Mountain Headwaters of the Oldman Basin Alberta. Catena 70:189-97. Stewart IT, Cayan DR, Dettinger MJ. 2005. Changes toward earlier streamflow timing across western North America. Journal of Climate 18: 1136–1155. Varhola A , Coops, N, Weiler M, Moore RD. 2010. Forest canopy changes on snow accumulation and ablation: an integrative review of empirical results. Journal of Hydrology 392: 219-33. Weiler M, Coops N, Teti P, Boon, S. 2007. Equivalent clearcut area thresholds in large scale disturbed forests. Forest Investment Account-Forest Science Project. 81171 Progress Report.: 1-10. Wilson C 1999. Effect of logging on runoff and erosion on highly variable erodible granitic soils in Tasmania. Water Resources Research 35: 33-46. Winkler R, 2001. The effects of the forest structure on snow accumulation and melt in the southcentral portion of British Columbia. University of British Columbia: 1-178. Winkler R, Boon S. 2009. Summary of research into the effect of mountain pine beetle related stand mortality on snow accumulation and ablation in BC. Streamline Watershed Bulletin 13:1-7. Winkler R, Carlyle-Moses, Spittlehouse T, Giles T, Phillips, Owens P, Moore RD, Leach J, Einarson D, Heise H 2008. Watershed response to the McClure Forest Fire: Presentation Summaries from the Fishtrap Creek Workshop, March 2008. 12:1-11 Winkler R. 2008. Snow accumulation and ablation post wildfire on the Thompson Plateau. Forrex Form on Research and Extention in Natural Resources. Wildfire and Watershed Hydrology. Workshop Proceedings. June 2009: 1-59 Winkler Spittlehouse D, Golding D. 2005. Measured differences in snow accumulation and melt among clearcut, young, and mature forest in southern BC. Hydrological Processes 19: 51-62. Woodsmith RD, Vach K, MCDonnell J, Helvey 2004. Entiat Experimental Forest: Catchment scale runoff data before and after a 1970 wildfire. Water Resources Research 40: W11701.  45  APPENDICES APPENDIX I  Figure A1.F1 BTM land use zone for Fishtrap Creek (08LB024). Source: Environment Canada, Unpublished.  Table A1.T1 Baseline Thematic Mapping term descriptions (Province of British Columbia, 2001, Schneider 2008) Old Forest greater than or equal to 140 years old and greater than 6 meters in height. Areas defined as Recently Logged and Selectively Logged land uses are excluded Recently Burned Areas virtually devoid of trees due to fire within the past 20 years. Forest less than or equal to 15% cover.  Selectively Logged Areas where the practice of selective logging can be clearly interpreted on the Landsat TM image and TRIM area Young Forest Forest less than 140 years old and greater than 6 meters in height. Areas defined as Recently Logged and Selectively Logged land uses are excluded from this class.  APPENDIX II  Legend ESSF MS ICH IDF  Figure A2.F1 BEC zones present in Fishtrap Creek. Source: Environment Canada, Unpublished.  APPENDIX III BEC  ZONES ESSF MS ICH IDF  Figure A3.F1. Topography of Fishtrap Creek (08LB024). Source: Environment Canada, Unpublished.  APPENDIX IV  Table A4. T1. Substitution of missing combinations of species and BEC zones present in the Bonaparte VRI source data for use in the generation of growth model for Fishtrap Creek Area. Missing BEC zone/species Combination  Substitution  FD in ESSF  FD in MS  BL in SBPS  BL in SBS  HW in ESSF  FD in MS  LW  Use PL for the same zone  BL in IDF  AT in MS  Z, W, XH  AT from the same zone  APPENDIX V  Table A5. T1. For start of freshet period, comparison of modification of post-fire years with respect to ECA for snow accumulation (ECA1), Snow ablation (ECA1), and modification of ECA to include post-fire salvage logging (ECA2). Model  3 4 5  6  Independent Variables  Tapr+ ECA` Tapr+ ECA2 + pre/post Tapr +ECA2+ pre/post+ eca1 *pre/post + EC2 Tapr + ECA1+prepost*Tap r  R2  R2  Pvalue  R2  Pvalue  Delta 1, ECA1 0.603 0.61  (delta1, ECA2) 0.5760 0.5700  1.96E-07 1.04E-06  (delta2, ECA2) 0.579 0.580  5.56E-08 9.41E-07  0.63  0.5963  8.09E-08  0.6101  9.51E-07  0.61  0.6000  2.24E-06  0.5998  0.000001924  This table is an example of the analysis done to see if ECA better predicts the hydrologic parameters if ECA is modified to reflect snow accumulation (ECA1) and snow ablation (ECA2) according to Winkler (2008) description of the behaviour of these factors following fire. The delta 2 was modified from the delta 1 data (used in the regression analysis for this is study) to reflect change in snow ablation (ECA2) if the areas was salvaged. None of the modifications shown in Table A5. T1 significantly improved the results so it is not part of the result shown in the results section.  

Cite

Citation Scheme:

    

Usage Statistics

Country Views Downloads
Russia 1 0
China 1 0
United States 1 0
City Views Downloads
Saint Petersburg 1 0
Shenzhen 1 0
Ashburn 1 0

{[{ mDataHeader[type] }]} {[{ month[type] }]} {[{ tData[type] }]}
Download Stats

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.52966.1-0075622/manifest

Comment

Related Items