UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Structure and morphology of the convective boundary layer in mountainous terrain De Wekker, Stephanus Franz Joseph 2002

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

Item Metadata

Download

Media
831-ubc_2003-792129.pdf [ 31.25MB ]
Metadata
JSON: 831-1.0052567.json
JSON-LD: 831-1.0052567-ld.json
RDF/XML (Pretty): 831-1.0052567-rdf.xml
RDF/JSON: 831-1.0052567-rdf.json
Turtle: 831-1.0052567-turtle.txt
N-Triples: 831-1.0052567-rdf-ntriples.txt
Original Record: 831-1.0052567-source.json
Full Text
831-1.0052567-fulltext.txt
Citation
831-1.0052567.ris

Full Text

STRUCTURE AND MORPHOLOGY OF THE CONVECTIVE BOUNDARY LAYER IN MOUNTAINOUS TERRAIN By STEPHANUS F R A N Z JOSEPH DE W E K K E R M.Sc. (Soil, Water, Atmosphere; focus meteorology), Wageningen University, 1996 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE F A C U L T Y OF G R A D U A T E STUDIES ATMOSPHERIC SCIENCE Department of Earth and Ocean Sciences We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH C O L U M B I A November 2002 © Stephanus Franz Joseph De Wekker, 2002 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of (^AMTH AK,n (Or/=ArJ fc/<?K>c£S The University of British Columbia Vancouver, Canada Date A/QUFM/^/I AT. Zr9<Q?_ DE-6 (2/88) A B S T R A C T Mountainous terrain can significantly modify convective boundary layer (CBL) structure and morphology compared to the situation over flat terrain and consequently can affect the transport and mixing of air pollutants. It is uncertain i f and how concepts of the C B L that have been developed in flat terrain are applicable in mountainous terrain. This dissertation attempts to address this issue by investigating the morphology and structure of the C B L in a deep, narrow mountain valley, near a mountain base, and over a mountain range. An integrative approach was taken by making use of three data sets, a numerical modeling system, and a Lagrangian particle dispersion model. The data set obtained in a deep, narrow valley was used to evaluate the performance of a mesoscale modeling system in complex terrain. C B L structure was captured surprisingly well given that surface layer and turbulence parameterization schemes are strictly valid only for flat and homogeneous terrain. In the deep, narrow valley, a distinction could be made between a 'conventional C B L ' with characteristics similar to those over flat terrain and a 'valley C B L ' which extends beyond the height of the conventional C B L and is not well-mixed in its upper part. Upvalley flows prevailed throughout the valley C B L but the horizontal wind structure was inhomogeneous, with larger wind speeds on the eastern than on the western side of the valley. Aerosols were found up to the height of the valley C B L . Near a mountain base, enhanced heating aloft was related to a reduced boundary-layer growth, resulting in a depression of the C B L height. Vertical and horizontal advection of warm air associated with the thermally driven circulation along the mountain slope play a role in the enhanced heating aloft. Over a mountain range, it was found that aerosol layer heights exceed C B L heights without the formation of separate elevated aerosol layers as found in previous studies. Particles are carried above the C B L up to the 'mountain C B L ' height by venting mechanisms associated with the presence of thermally driven mesoscale flows. The 'valley C B L ' and 'mountain C B L ' heights are more important parameters for air-pollution studies than the conventional C B L height since they indicate the height up to which air pollutants can be transported and/or mixed. 11 T A B L E O F C O N T E N T S A B S T R A C T i i T A B L E OF CONTENTS i i i LIST OF T A B L E S vi LIST OF FIGURES vii LIST OF A C R O N Y M S xvi A C K N O W L E D G E M E N T S xvii 1. I N T R O D U C T I O N 1 1.1. REVIEW OF PREVIOUS W O R K 1 1.2. R E S E A R C H OBJECTIVES A N D M E T H O D O L O G Y 11 2. C B L M O R P H O L O G Y IN A V A L L E Y : T H E M A P - R I V I E R A F I E L D S T U D Y 17 2.1. INTRODUCTION 17 2.2. D A T A 17 2.3. N U M E R I C A L M O D E L SETUP 23 2.4. DIURNAL R A N G E OF SURFACE V A R I A B L E S A T BOSCO D l SOTTO 27 2.4.1. Observations 27 2.4.2. Model evaluation 29 2.5. SPATIAL SURFACE WEND FIELD IN THE RIVIERA V A L L E Y 30 2.6. TURBULENT SENSIBLE H E A T F L U X FN THE RIVIERA V A L L E Y 33 2.7. T E M P O R A L EVOLUTION OF THE V A L L E Y ATMOSPHERE AT BOSCO D l SOTTO 39 2.7.1. Observations 39 2.7.2. Model evaluation 41 2.7.3. Heating of the valley atmosphere 43 2.8. SPATIAL STRUCTURE OF A L O N G - V A L L E Y WEND 45 2.9. SPATIAL STRUCTURE OF POTENTIAL T E M P E R A T U R E 49 2.10. SPATIAL STRUCTURE OF C B L HEIGHTS 53 2.11. SPATIAL STRUCTURE OF T U R B U L E N T KINETIC E N E R G Y 54 2.12. IMPLICATION OF C B L M O R P H O L O G Y FOR PARTICLE DISPERSION 56 2.13. CONCLUSIONS 58 3. C B L M O R P H O L O G Y N E A R A M O U N T A I N B A S E : T H E P A C I F I C ' 9 3 F I E L D S T U D Y 61 3.1. INTRODUCTION 61 3.2. OBSERVATIONS OF A ' C B L HEIGHT DEPRESSION' 61 3.3. M O D E L I N G OF A ' C B L HEIGHT DEPRESSION' 67 in 3.3.1. Model setup 67 3.3.2. C B L heights from a numerical model 68 3.4. COMPARISON B E T W E E N OBSERVATIONS A N D N U M E R I C A L M O D E L 71 3.5. CONCLUSIONS 77 4 . CBL MORPHOLOGY AND AEROSOL LAYER STRUCTURE OVER A MOUNTAIN RANGE 78 4.1. INTRODUCTION 78 4.2. D A T A 79 4.3. OBSERVATIONS OF THE AEROSOL L A Y E R HEIGHT 83 4.4. M O D E L SETUP A N D E V A L U A T I O N 89 4.4.1. Model setup 89 4.4.2. Evaluation of model output with observations 91 4.5. L P D M SIMULATION A N D COMPARISON WITH OBSERVATIONS 93 4.6. C B L HEIGHTS A N D A L HEIGHTS 99 4.6.1. C B L heights from model simulations 99 4.6.2. C B L height and topography 100 4.6.3. Comparison of C B L heights and L P D M results 102 4.6.4. Comparison of modeled C B L height and observed al height 105 4.7. CONCLUSIONS 107 5. DISCUSSION 109 5.1. C B L M O R P H O L O G Y IN A V A L L E Y 109 5.2. C B L M O R P H O L O G Y N E A R A M O U N T A I N B A S E 112 5.3. C B L M O R P H O L O G Y OVER A M O U N T A I N R A N G E 113 5.4. TOWARDS A N INTEGRATIVE A P P R O A C H TO C B L M O R P H O L O G Y IN MOUNTAINOUS TERRAIN 116 5.5. A REVISIT OF THE C B L HEIGHT DEFINITION FOR MOUNTAINOUS TERRAIN A N D SOME FINAL R E M A R K S 121 6. GENERAL CONCLUSIONS AND RECOMMENDATIONS FOR FUTURE RESEARCH 124 6.1. G E N E R A L CONCLUSIONS 124 6.2. RECOMMENDATIONS FOR FUTURE R E S E A R C H 127 7. BIBLIOGRAPHY 129 APPENDICES A l . M E S O S C A L E M O D E L I N G 139 i v A l . l . R A M S D E S C R I P T I O N 139 A 1 . 2 . H Y P A C T D E S C R I P T I O N 147 A 1 . 3 . S C A L E S O F I N T E R E S T A N D G R I D S P A C I N G C O N S I D E R A T I O N 148 A 1 . 4 . S H O R T W A V E R A D I A T I O N C O R R E C T I O N I N R A M S 155 A 2 . S U R F A C E B O U N D A R Y C O N D I T I O N S F O R M A P - R I V I E R A C A S E S T U D Y 158 A 3 . S E N S I B L E H E A T F L U X E S I N T H E R I V I E R A V A L L E Y 164 A 4 . A I R C R A F T M E A S U R E M E N T S D U R I N G M A P - R I V I E R A 166 A 4 . 1 . G R I D D I N G M E T H O D 166 A 4 . 2 . A I R C R A F T C R O S S S E C T I O N S 168 A 5 . C B L H E I G H T D E T E R M I N A T I O N W I T H T H E ^ - M E T H O D 178 A 6 . L A N D U S E I N T H E S T A A A R T E ' 9 7 I N V E S T I G A T I O N A R E A 179 A 7 . C O M P A R I S O N O F E C M W F C B L H E I G H T S D U R I N G S T A A A R T E W I T H O B S E R V E D A L H E I G H T S A N D C B L H E I G H T S F R O M R A M S 180 A 8 . G L A C I E R A N D C L O U D E F F E C T S O N C B L H E I G H T D U R I N G S T A A A R T E 182 A 9 . L I D A R M E A S U R E M E N T S D U R I N G S T A A A R T E 183 A 9 . 1 . L I D A R B A S I C S 183 A 9 . 2 . L I D A R C R O S S S E C T I O N S F R O M S T A A A R T E 184 v LIST OF TABLES 1.1. Characteristics of the investigations in chapters 2, 3, and 4. Lx, Ly, and H are defined in Fig. 1.3 15 2.1. Site identification, location, measurement height, and some surface characteristics for the ten surface stations measuring turbulence 23 2.2. Characteristics of the four grids used in the M A P case study. N X and N Y are the number of grid points in the west-east, and north-south direction, respectively 24 4.1. Details of the flight tracks shown in Fig. 4.1. The A L heights (which are leg-averaged values) will be discussed in section 4.5 83 4.2. Characteristics of the four grids used in S T A A A R T E modeling case study. N X and N Y are the number of grid points in the west-east, and north-south direction, respectively 90 A l . l . LEAF-2 biophysical parameters by land use class number used in R A M S . The parameters include albedo (a), surface emissivity (e), leaf area index (LAI), fractional vegetation coverage (vfrac), aerodynamic roughness length (zo), aerodynamic displacement height (Zd), and root depth (root) 143 A2.1. Meaning of the polygon colors in Fig. A2.2 159 A2.2. Details of the cross-referencing between land use classes from the Swiss data set and land use classes according to LEAF-2 159 A2.3. Land use classes according to BFS (1993) 163 A3.1. Characteristics of the ten surface stations measuring turbulence 164 A4.1. Specifics of the time and the orientation of the interpolated cross sections, and the figure number where each cross section can be found 168 A9.1. Details of flight legs 1-17 (morning flights). Orientation is either parallel (P) or perpendicular (O) to the mountain divide. The time indicates the approximate starting and ending times of each flight leg. A L heights are leg-averaged values. These could not be determined for some flight legs 185 A9.2. As Table A9.1 but for flight legs 18-35 (afternoon flights) 185 vi L I S T O F F I G U R E S 1.1. Structure of the atmosphere above a mountain range (after Ekhart, 1948) 6 1.2. Conceptual model of the boundary layer in mountainous terrain on a fair weather day in the early morning just before sunrise (a), in the morning just after sunrise (b), at noon (c), and in the late afternoon (d). Light grey shading is the topography. Dark grey shading is a strongly stable nocturnal inversion. Arrows denote horizontal and vertical flows. The wavy lines depict the upper limit of surface generated convection. Vertical profiles of potential temperature are shown at several locations in the topography (after Fiedler et al., 1987) 11 1.3. Schematic of the different topographic settings in which C B L morphology is studied in this dissertation. Lx, Ly, and Ff depict horizontal and vertical scales of the topography. Ly denotes the horizontal scale in the along-valley direction. Typical values for the topographies studied are given in Table 1.1 12 2.1. Topography of the Riviera Valley and surroundings. Contour lines are drawn every 400 m. Terrain above 2000 m asl is shaded. The dashed lines depict the cross-valley and along-valley cross sections. The main valley site, Bosco di Sotto, (46.26° N , 9.01° E, 250 m asl) is located near Claro in the valley center 18 2.2. Flight pattern in the cross-valley (top) and along-valley (bottom) directions 19 2.3. Surface (a) and 500 mb (b) weather maps from the Swiss Meteorological Institute for 25 August 1999 1200 UTC. (c) A V H R R satellite image at 1419 UTC (channel 2 image, 0.725-1.10 pm). Satellite image courtesy of the University of Dundee (http://www.sat.dundee.ac.uk/) 21 2.4. Location of the surface turbulence stations. See table 2.1 for details. Contour lines are drawn every 100 m 22 2.5. Temperature (a), mixing ratio (b), wind speed (c) and wind direction (d) at the surface at site A l (Bosco di Sotto) for all valley wind days during M A P -Riviera. The closed circles are the data for 25 August 1999. Up-valley wind direction is indicated the horizontal black line in (d) 28 2.6. Observed (filled circles) and modeled (open circles) net radiation (a) and potential temperature (b) at Bosco di Sotto on 25 August 1999 29 2.7. Horizontal wind vectors plotted as a function of time at Bosco di Sotto, as observed at 12 m agl (top panel) and modeled at 10 m agl (bottom panel) for 25 August 1999 30 2.8. Modeled and observed surface wind field in the Riviera Valley at 0900, 1200, and 1500 UTC. Observations at various surface stations are indicated by the bold arrows. Topography is shown with a contour interval of 400 m and the shading corresponds to modeled wind speed. Dashed arrows depict some major flow features in the valley. The axes scale corresponds to that on Figs. 2.1 and 2.4 31 2.9. Observed versus modeled half-hourly-averaged surface sensible heat flux for all the surface stations listed in Table 2.1 on 25 August 1999 between 0800 and 1600 UTC 33 vn 2.10. Observed (squares) and modeled (dashed line) surface sensible heat flux at site A l (a), site B (b), and site F2 (c) on 25 August 1999. Site locations are shown in Fig. 2. The shaded area indicates the range of modeled sensible heat fluxes at the nine model grid points surrounding the observation sites 34 2.11. Contour lines of topography (solid lines) and slope azimuth angle (dashed lines) around surface station F2. The axes scale corresponds to that on Figs. 2.1 and 2.4. The squares illustrate the model grid with a horizontal resolution of 333 m 35 2.12. Theoretical incoming solar radiation for a slope of 30° having azimuth angles of 200° (solid line) and 240° (dashed line) 36 2.13. Spatial distribution of surface sensible heat flux at 0900 UTC (a), 1200 UTC (b), and 1500 UTC (c). Topography is indicated by contour lines drawn every 400 m. The axes scale corresponds to that on Figs. 2.1 and 2.4 38 2.14. Vertical profiles of observed potential temperature (a), specific humidity (b), and horizontal wind vectors (c) at the valley center (Bosco di Sotto) at 0739 UTC, 0915 UTC, 1208 UTC, and 1508 UTC on 25 August 1999. The surface potential temperature measured at 1.5 m at the different times is indicated in (a) with symbols. The approximate height of the ridge is indicated with a grey rectangle 39 2.15. Vertical profiles of observed and modeled potential temperature and specific humidity at Bosco di Sotto on 25 August 1999 at 0739 (a,b), 0915 (c,d), 1208 (e,f), and 1508 UTC (g,h). Model output is for 0700 (a,b), 0900 (c,d), 1200 (e,f), and 1500 UTC (g,h). The approximate height of the ridge is indicated with a grey rectangle 42 2.16. Vertical profiles of observed and modeled horizontal wind vectors for the same times as in Fig. 2.15. The approximate height of the ridge is indicated with a grey rectangle 42 2.17. Vertical profiles of the various modeled terms in the heat budget equation (see text, eq. 2.1) for Bosco di Sotto, averaged between 1200 and 1500 UTC. The approximate height of the ridge is indicated with a grey rectangle 44 2.18. Interpolated cross sections of the along-valley wind component (m s"1; up-valley is positive) from aircraft data between 0706 and 0742 (a), 0908 and 0919 (c), 1118 and 1218 (e), 1330 and 1342 (g) and from model output at 0700 (b), 0900 (d), 1200 (f) and 1400 UTC (h). The location of the west-east cross section is depicted in Fig. 2.1. The horizontal distance is relative to Claro(see Fig. 2.1) 46 2.19. Interpolated cross sections of the along-valley wind component (m s"1; up-valley is positive) from aircraft data between 1222 and 1330 UTC on the western (a) and eastern (b) sides of the valley and from model output at 1300 UTC (c). The location of the cross section in (a) and (b) is depicted in Fig. 2.20. The horizontal distance is relative to Claro (see Fig. 2.1) 48 2.20. Topography of the Riviera Valley as in Fig. 2.1. The aircraft flew part of the time between 1222 and 1330 UTC on the western side and part of the time on the eastern side of the valley. A l l the measurements taken in the western rectangle are shown in Fig. 2.19a and all the measurements taken in the eastern rectangle are shown in Fig. 2.19b 48 vin 2.21. As in Fig. 2.18 but for potential temperature (Kelvins). The dotted line indicates the C B L height determined from the Pi-method 50 2.22. Sketch of isentrope deformations near a valley sidewall associated with the presence of an upslope flow. The thick solid line represents the valley and sidewall surfaces. The thin solid lines are isentropes. The dashed vector indicates the presence of an upslope flow 51 2.23. As in Fig. 2.19 but for potential temperature (Kelvins). The dotted line indicates the C B L height determined by the Pz'-method 52 2.24. Cross section of modeled T K E (m2 s"2) at 1300 UTC. The location of the west-east cross section is depicted in Fig. 2.1. The asterisks denote the height of the along-valley flight legs. The 0.03 m 2 s"2 isoline and the C B L height calculated from the P/-method are shown by the solid and dashed line, respectively 55 2.25. Vertical profiles of T K E in the western (a) central (b) and eastern (c) part of the valley around 1300 UTC. Dashed lines connected with open circles are from along-valley aircraft legs; plusses and triangles are the model values averaged over 10 km and 5 km lines in the along-valley direction, respectively 55 2.26. Particle distribution across (a) and along (b) the valley at 1200 UTC. The dashed line denotes the C B L height determined by the Pz'-method from model output. The location of the west-east cross section is depicted in Fig. 2.1 57 2.27. Observed aerosol concentrations (number per cm 3 for aerosols > 0.5 pm) on a valley cross section at 1200 UTC. The location of the west-east cross section is depicted in Fig. 2.1 58 3.1. Map of the Lower Fraser Valley of British Columbia. The region of interest in this study is indicated in (a) by the rectangle, which is magnified in (b). The lidar paths are indicated by L7 and T l while H and L denote the Harris Road tethered balloon station and the Langley radiosonde station, respectively. The horizontal dotted line near H denotes the approximate location where the aerosol layer starts to decrease in height in Fig. 3.2a. Contour intervals in (a) and (b) are 200 m and 100 m, respectively. Darker shades of grey represent higher elevations in (a). Elevation higher than 700 m is shaded in (b) 62 3.2. Flight tracks L7 (a) and T l (b) on 4 August 1993. The gray scale is proportional to lidar backscatter with light colors indicating high backscatter intensities. The depression in the A L height is indicated by the dotted line in (a). The arrow in (b) indicates a region with relatively shallow aerosol layer heights, corresponding to the location where flight tracks L7 and T l intersect 63 3.3. Vertical profiles of potential temperature at Harris Road and Langley on 4 August 1993 at 1452 and 1600 PST. C B L height is indicated with a horizontal dashed line. The black circle near the surface is the surface temperature measured at Harris Road at 1500 PST 65 ix 3.4. Vertical profiles of potential temperature on a cross section of the Rhine Valley at 1300 CEST on 22 September 1992 during the TRACT field study (adapted from De Wekker (1995)) 66 3.5. Variation of C B L heights with horizontal distance from the mountain base as a function of time for a 11.5° slope (a) and as a function of slope steepness (b). C B L heights are determined by the T K E method 69 3.6. C B L heights on a cross valley section as determined by different methods using model output from 1700 PST 70 3.7. Particle distribution at 1500 PST. The C B L height determined by the T K E method from model output is indicated with the heavy black line. Particles were released continuously from the surface 70 3.8. Differences in heating rates as a function of height between Harris Road and Langley, averaged between roughly 1000 and 1300 (a, 'morning'), and 1300 and 1600 PST (b, 'afternoon') for 4 August 1993. Arrows indicate the minimum and maximum differences in heating rates. Error bars indicate a measurement error of 1 K 72 3.9. Idealized topography used in the model simulations showing the columns A and B whose heat budget terms were compared. A slope steepness of 11.5° is used in the simulation 72 3.10. Differences in heating rates as a function of height between the atmospheric columns B and A (depicted in Fig. 3.9), as averaged between 1000 and 1300 (a, 'morning'), and 1300 and 1600 PST (b, 'afternoon'). Arrows indicate the minimum and maximum differences in heating rates 73 3.11. Observed differences in heating rates as a function of height between Harris Road and Langley, averaged between roughly 1300 and 1600 PST for 4 August 1993 (a). The corresponding modeled difference in heating rate as a function of height is shown by the bold line in (b). The contribution of the individual terms to the total heating rate is also shown. Arrows indicate the corresponding locations of minimum and maximum differences in heating rates. Error bars in (a) indicate a measurement error of 1 K 74 3.12. Cross section of potential temperature and wind field at 1500 PST. The C B L height determined by T K E method from model output is indicated with the heavy black line 75 3.13. As Fig. 3.11 but for 1 August 1993 76 4.1. Topographic map of the experimental area. Contour lines are drawn every 1000 m, the darkest shade representing terrain over 3000 m asl. The Jungfraujoch station is depicted by the white asterisk. The approximate locations of the flight tracks perpendicular (07, 016, 024, 034) and parallel (P6, P23, P30) to the mountain divide are depicted by the dashed lines. The inner square represents the innermost grid in the mesoscale model run described in section 4.4 80 4.2. Surface (a) and 500 mb (b) weather maps from the Swiss Meteorological Institute for 1200 UTC 30 July 1997. (c) A V H R R satellite image at 1419 UTC (channel 2 image, 0.725-1.lOum). Satellite image courtesy of the University of Dundee (http://www.sat.dundee.ac.uk/) 81 x 4.3. Global radiation at various surface meteorological stations in and around the investigation area. The stations are: Payerne (46.82°N, 6.95°E); Ulrichen (46.50°N, 8.32°E); Adelboden (46.50°N, 7.57°E); Grimsel-Hospiz (46.57°N, 8.33°E); and Jungfraujoch (46.55°N, 7.98°E) 82 4.4. Cross sections of lidar backscatter ratio for flight track P6 (a), P23(b), and P30(c) at 0653, 1308, and 1409 UTC, respectively. The color scale is proportional to lidar backscatter ratio 84 4.5. Cross sections of lidar backscatter ratio for flight tracks 07 (a), 016 (b), 024 (c), and 034 (d) at 0702, 0913, 1317, and 1520 UTC, respectively. The color scale is proportional to lidar backscatter ratio. The Rhone Valley, Aletsch Glacier, and Jungfraujoch station (JFJ) are indicated in (a). As an example of the results from the semi-objective method of determining A L height, see the black dots in (d) 85 4.6. Illustration of the determination of the A L height with the semi-objective method. See text for details 86 4.7. Flight-leg-averaged A L heights for legs parallel to the mountain divide (open circles) and perpendicular to the mountain divide (closed circles) 89 4.8. Map of Europe with the four R A M S grids represented by the trapezoids. The innermost grid is the area within the inner square in Fig. 4.1 and is also shown in Fig. 4.11 90 4.9. Observed (squares) and modeled (solid lines) vertical profiles of potential temperature (a), relative humidity (b), wind speed (c), and wind direction (d) for Payerne at 1200 UTC 30 July 1997. The vertical grid spacing in the model is indicated in (a) with the small horizontal lines 92 4.10. Observed (squares) and modeled (solid lines) vertical profiles of potential temperature for Milan at 1200 UTC (a), above the Jungfraujoch at 0900 UTC (b) and above the Jungfraujoch at 1500 UTC (c) on 30 July 1997 93 4.11. Topographic map of the innermost R A M S grid. Contour lines are drawn every 1000 m, the darkest shade representing terrain over 3000 m asl. The solid rectangles indicate the release areas of the particles. The solid line indicates the location of the cross section in Fig. 4.12. For the meaning of the dashed lines and the dotted rectangle, see the text 94 4.12. Cross section of the particle distribution at 0900 UTC (a), 1200 UTC (b), and 1500 UTC (c) in the area between the dashed lines in Fig. 4.11. The dashed line, solid line, and light dotted line in (a) are topography cross sections on the western side, center, and eastern side of this area, respectively (locations of cross sections are shown in Fig. 4.11). The heavy dotted lines in (a), (b), and (c) represent the A L height at 0913, 1317, and 1520 UTC, respectively (from Fig. 4.5) 95 4.13. Horizontal distribution of particles at 1530 UTC. Red dots denote particles released in the rectangle north of JFJ, yellow dots denote particles released in the rectangle south of JFJ. For the location of the release areas, see Fig. 4.11. For the meaning of dashed and solid arrows, see the text 97 2 3 4.14. Diurnal course of aerosol surface area per unit volume S a (pm cm") at the Jungfraujoch for 500 mbar synoptic flows from the west (circles) and northwest (squares), averaged over the seven year period 1991 to 1997 xi (Lugauer, 1998). The diamonds depict the diurnal range of particle numbers arriving inside the rectangle shown in Fig. 4.11 98 4.15. C B L height from the parcel method plotted against C B L height from the Ri-method for all grid points in the innermost R A M S domain at 1200 UTC (a). C B L height from the T K E method plotted against C B L height from the Ri-method for all grid points in the innermost R A M S domain at 1200 UTC (b) 100 4.16. C B L heights from the i?/-method in the innermost R A M S grid plotted against elevation for 0800 (a) and 1300 UTC (b). The values of the parameters T ((&lopo-ah}/atopo) and r (correlation coefficient) are also indicated 101 4.17. Diurnal course of the parameters T and r as a function of time of day for all grid points in the innermost R A M S domain 102 4.18. Cross section of the particle distribution at 0900 (a), 1200 (b), and 1500 UTC (c) in the area between the dashed lines in Fig. 4.10. C B L height from the i?z'-method is indicated by a dashed line. Values of the parameters T and r are also indicated 103 4.19. Cross section of potential temperature and windfield with superimposed C B L heights as determined using the Ri-method. The cross section is the same as in Fig. 4.18 104 4.20. Cross section of the particle distribution at 1200 UTC at the same location as in Fig. 4.16 but for a simulation with no topography. C B L height as determined using the i?/-method is indicated by a dashed line 105 4.21. Observed A L heights (closed circles) and C B L heights from model output. Squares are C B L heights from model output averaged over a 15 km wide band perpendicular to the mountain divide. Plusses are C B L heights from model output averaged over a 15 km wide band parallel to the mountain divide 105 5.1. Sketch of the C B L morphology on the afternoon of 25 August 1999. Dotted ellipses are isotachs of the along-valley wind. The dashed lines denote C B L heights. The lower dashed line is the conventional C B L height and the upper dashed line is the 'valley C B L height'. Arrows denote upslope flows 109 5.2. Idealized vertical profiles of potential temperature for an inversion that breaks up by subsidence (a), and a schematic of potential temperature structure evolution in the Riviera Valley on 25 August 1999 (b). The approximate height of the ridge is indicated with a grey rectangle 110 5.3. Sketch of the depressed C B L height near a mountain base. The arrows denote vertical winds. The dashed line denotes the C B L height 112 5.4. Conceptual picture of the situation on the afternoon of 30 July 1997. h is the C B L height and h a the A L height. The depicted mechanisms are (1) mountain venting, (2) cloud venting, (3) advective venting, and (4) advection of aerosols from airmasses elsewhere. h a is referred to as the 'mountain C B L ' 113 5.5. Schematic of vertical potential temperature profiles in the S T A A A R T E investigation area south of JFJ (solid line) and north of JFJ (dashed line). The numbers correspond with the mechanisms indicated in Fig. 5.4. The subscripts's' and 'n ' refer to south and north of JFJ, respectively. The thin xii arrows denote turbulent eddies. The diagonal hatching denotes the valley surface. h a n And h a s are referred to as the 'mountain C B L ' (see text) 115 5.6. (a) Cross section of lidar backscatter ratio around 1335 UTC 28 August 2001 during CHAPOP. Lidar wavelength is 532 nm and the covered region is from 8.5° to 9.7°E. (b) Same as (a) but for a lidar wavelength of 1064 nm and between 8.8° and 9.07°E. (c) Vertical profiles of potential temperature (solid line) and mixing ratio (dotted line) at 1400 UTC 28 August 2001, in the Leventina Valley (the valley on the left hand side of Fig. 5.6b) 118 5.7. Schematic of vertical potential temperature profile in the Leventina Valley at 1400 UTC during CHAPOP. The numbers correspond with the mechanisms indicated in Fig. 5.4. The thin arrows denote turbulent eddies, h Is the C B L height, h a ] is the A L height in the valley, and h a 2 is the A L height over the mountains. The diagonal hatching denotes the valley surface. hai and h a 2 could also be referred to as a 'valley CBL ' height and a 'mountain C B L ' height, respectively (see text) 119 A l . l . Arakawa type C grid stagger used in R A M S . T represents the location of thermodynamic variables; U represents west-east velocity; V represents north-south velocity. W, the vertical velocity, is not shown, but is located half a grid distance above and below T 141 A1.2. Idealized topography of two superimposed sine waves (a) and the corresponding wavelet spectrum (b). Also shown are the 90% significance line (dotted line) and the position of the maxima at 4 and 20 km (dashed line) 151 A1.3. STAAARTE'97 investigation area. Contour interval is 1000 m. The darkest color represents terrain over 3000 m. The locations of the cross-sections used in the wavelet analysis are indicated by black lines 152 A1.4. Topography of four N-S cross sections (a) and four E-W cross sections (b) with the corresponding wavelet spectra (c and d, respectively). The 90% significance lines corresponding to each spectrum are indicated in c, and d 153 A1.5. Incoming shortwave radiation R short as a function of time for Bosco di Sotto (MAP case study). Open squares are observed values, closed squares are modeled values without a reduction factor, and plusses are modeled values with a reduction factor 156 A1.6. Incoming shortwave radiation R S h 0 r t as a function of time for the Jungfraujoch station (STAAARTE case study). Open squares are observed values, plusses are modeled values with reduction factor 157 A2.1. Distribution of land use classes in the Riviera Valley based on a combination of the Swiss land use data set and LEAF-2 land use classes (according to Table A2.2). Height contours are drawn every 400 m 160 A2.2. Differences between the USGS data set and the land use classes used in the R A M S simulations (see Fig. A2.1) for a) short grass and b) crop/mixed farming. The black filled polygons are the locations where the two data sets correspond. For meanings of the other shading types, see Table A2.1. Height contours are drawn every 400 m 161 X l l l A2.3. Distribution of soil moisture in the Riviera Valley on 25 August 1999. Height contours are drawn every 400 m 162 A3.1. Comparison between observed and modeled surface sensible heat fluxes at ten different sites in the Riviera Valley (listed in Table A3.1) on 25 August 1999. Squares are observations, dashed lines are the model output at the grid point closest to the observation. The shaded area represents the range of sensible heat fluxes measured at the nine model grid points surrounding the observation location 165 A4.1. Interpolated cross section of potential temperature from (a) model output and (b) model output sampled at locations along a flight track (shown by the dotted line) and interpolated according to the Delauney triangulation method. Contour lines are in K and drawn every I K 167 A4.2. As Fig. A4.1 but for wind speed. Contour lines are in m s"1 and drawn every l m s " 1 167 A4.3. Interpolated cross sections of potential temperature (K), specific humidity (g kg"1), cross-valley wind component and along-valley wind component (m s"1) for 0706-0742 UTC 169 A4.4. Same as Fig. A4.1 but for 0851-0900 UTC 169 A4.5. Same as Fig. A4.3 but for 0908-0919 UTC 170 A4.6. Location of the data points of the cross sections in Fig. A4.3 (a), Fig. A4.4 (b), and Fig. A4.5 (c) 170 A4.7. Same as Fig. A4.3 but for 0742-0848 UTC. Only data points on the western side of the valley are considered 171 A4.8. Same as Fig. A4.3 but for 0742-0848 UTC. Only data points on the eastern side of the valley are considered 171 A4.9. Location of the data points of the cross sections in Fig. A4.7 (a), and Fig. A4.8(b) 172 A4.10. Same as Fig. A4.3 but for 1118 -1218 UTC 173 A 4 . l l . Same as Fig. A4.3 but for 1330-1342 UTC 173 A4.12. Same as Fig. A4.3 but for 1506 -1548 UTC 174 A4.13. Location of the data points of the cross sections in Fig. A4.10 (a), Fig. A 4 . l l (b), and Fig. A4.12 (c) 174 A4.14. Same as Fig. A4.3 but for 1222-1330 UTC. Only data points on the western side of the valley are considered 175 A4.15. Same as Fig. A4.3 but for 1222-1330 UTC. Only data points on the eastern side of the valley are considered 175 A4.16. Same as Fig. A4.3 but for 1344-1510. Only data points on the western side of the valley are considered 176 A4.17. Same as Fig. A4.3 but for 1344-1510 UTC. Only data points on the eastern side of the valley are considered 176 A4.18. Location of the data points of the cross sections in Fig. A4.14 (a), Fig. A4.15 (b), Fig. A4.16 (c), and A4.17 (d) 177 A6.1. Distribution of land use classes in the STAAARTE'97 investigation area 179 A7.1. Development of the C B L in a 0.5° x 0.5° area around the Jungfraujoch from R A M S (squares) and E C M W F (diamonds). The stars depict the C B L height xiv determined subjectively from E C M W F potential temperature profiles for 1200 and 1600 UTC (Fig. A7.2) 180 A7.2. Vertical profiles of potential temperature from E C M W F at 8°E, 46.5°N at 1200 UTC (filled squares) and 1600 UTC (open squares). The C B L height from E C M W F is indicated by a dashed line for 1200 U T C and by a solid line at 1600 UTC 181 A9.1. Cross sections of lidar backscatter ratio for flight legs 1 to 17. These legs were flown in the morning of 30 July 1997. The times of each leg can be found in Table A9.1. The exact location of each leg is depicted in Fig. A9.2...186-188 A9.2. Topography of the S T A A A R T E investigation area along with the location of flight legs 1-10 (a) and flight legs 11-17 (b) 188 A9.3. Cross sections of lidar backscatter ratio for flight legs 18 to 34. These legs were flown in the afternoon of 30 July 1997. The times of each leg can be found in Table A9.2. The exact location of each leg is depicted in Fig. A9.4...189-191 A9.4. Topography of the S T A A A R T E investigation area along with the location of flight legs 18-27 (a) and flight legs 28-35 (b) 191 xv LIST OF A C R O N Y M S A B L Atmospheric Boundary Layer agl Above Ground Level A L Aerosol layer asl Above Sea Level A V H R R Advanced Very High Resolution Radiometer BATS Bio sphere-Atmosphere-Transfer S cheme BFS Bundesamt fur Statistik C B L Convective boundary layer CHAPOP Characterization of High Alpine Pollution Plumes COST European cooperation in the field of scientific and technical research D L R German Aerospace Research Establishment E C M W F European Centre for Medium-Range Weather Forecasts ETH-Ziirich Eidgenossische Technische Hochschule Zurich GDAS Global Data Assimilation System H Y P A C T HYbrid Particle And Concentration Transport model JFJ Jungfraujoch L D A S Land Data Assimilation System L E A F - 2 Land Ecosystem Atmosphere Feedback Model, version 2 lidar Light detection and ranging L P D M Lagrangian Particle Dispersion Model M A P Mesoscale Alpine Programme NCEP National Centers for Environmental Prediction N O A A National Oceanic and Atmospheric Administration PNNL Pacific Northwest National Laboratory PSI Paul Scherrer Institut PST Pacific Standard Time (= UTC-8) R A M S Regional Atmospheric Modeling System sodar sound detection and ranging S T A A A R T E Scientific Training and Access to Aircraft for Atmospheric Research Throughout Europe T K E Turbulent Kinetic Energy TRACT Transport of Air Pollutants over Complex Terrain U B C The Univerity of British Columbia USGS United States Geological Survey WaSiM Water balance Simulation Model xvi A C K N O W L E D G E M E N T S Many people helped to make this study possible. First of all, I very much would like to thank my supervisor Douw Steyn, for being supportive of my research in terms of funding, encouragement, intellectual support and advice. He also gave me the opportunity to participate in the MAP-Riviera field study in Switzerland. Dave Whiteman encouraged me to pursue a Ph.D. degree in the first place and I would like to thank him for the support and advice he has offered me in the past years. I also would like to thank the following people: the members of my Ph.D. committee Ian McKendry, Roland Stull, and Dave Whiteman, the university examiners Phil Austin and Mike Novak, and the external examiner Bob Banta, for providing useful suggestions for improvement; Mathias Rotach for hosting two visits at ETH-Zuerich and for many stimulating discussions; Marco Andretta, Eva van Gorsel, Andreas Weigel, Alexandra Weiss, and Massimiliano Zappa for providing MAP data; Karsten Jasper for providing soil moisture output from the hydrological model WaSiM-ETH; Jerome Fast and Shiyuan Zhong for hosting several visits to PNNL, providing the numerical model RAMS version 4.3 and HYP ACT, and helping me with the use of the model; Stephan Nyeki, Urs Baltensperger, and Markus Furger of PSI for involving me in the STAAARTE project, providing STAAARTE data and hosting a visit at PSI; Meinolf Kossmann, Alberto Martilli and Ming Zhao for useful suggestions and stimulating discussions; Mathias Muller for helping me to learn programming in EDL and his assistance in the analysis of STAAARTE data; Paul Bovis for his help of using various windows-based software programs; Rolf Hertenstein and Greg Poulos from Colorado Research Associates, Boulder for their assistance in extracting tendency terms from RAMS; and all other people whose name I may accidentally have forgotten to mention here. A special note of acknowledgment goes to Prof. Lee Gass. After I hit an obstacle on the road to this dissertation, he played a crucial role in getting me back on the right track and contributed hugely to my personal and academic development in these years. All numerical simulations in this study were performed on a linux cluster in the department of physics at UBC. I would like to thank Matt Choptuik for kindly allowing me to use the cluster and his assistance in getting RAMS to run on it. xvii I am grateful to N O A A Climate Diagnostics Center, Boulder, Colorado, USA, for providing NCEP Reanalysis data from their Web site at http://www.cdc.noaa.gov/, and to the British Atmospheric Data Centre for providing access to the European radiosonde dataset. D. Luethi from ETH provided the E C M W F data that were used in the M A P -Riviera case study. The land use data for the Riviera valley were made available through a collaboration with the Institute for Atmospheric and Climate Science ETH, permission # X X X X X X by 'Bundesamt fuer Landestopographie'. I wish to thank the following organizations who funded this work: The University of British Columbia (UBC) provided two years of support via University Graduate Fellowships. The Isaac Walton Killam Memorial Trust provided two years of support via its pre-doctoral fellowship. The U B C geography department provided personal funding through several teaching assistantships. Grants from The Natural Sciences and Engineering Research Council of Canada to Douw Steyn supported several research assistantships, covered the costs of involvement in the MAP-Riviera field study, provided day-to-day funding for general research costs, and also funded several conference visits. Lastly, I am very grateful to my family whose continuous support and encouragement I needed so much to complete this dissertation. Stephan de Wekker November 2 0 0 2 xviii 1. INTRODUCTION Transport and mixing processes throughout the lower atmosphere are generally well understood in flat and homogenous terrain. These processes can be very different in mountainous terrain than in flat terrain, caused by interactions between the terrain and the overlying atmosphere. Many of the worst air pollution episodes occur in regions of complex terrain, motivating research towards better understanding of the processes leading to these episodes. Despite the realization of the differences in transport and mixing processes in complex terrain, concepts that have been developed for flat, homogeneous terrain, are simply applied to complex, mountainous terrain. The evaluation of the applicability of conventional concepts in mountainous terrain and the development of new concepts that specifically address the effects of mountainous terrain may be necessary to advance our knowledge on air pollution transport and mixing in mountainous terrain. This can only be achieved if our understanding of boundary layer structure and morphology in mountainous terrain is improved. This dissertation attempts to address this issue. To this end, a variety of data sets is examined and idealized and realistic numerical simulations are performed. A detailed description of the research objectives/questions and the overall strategy is presented in section 1.2. Next, a short review of previous work on the research topics in this dissertation is given. 1.1. REVIEW OF PREVIOUS WORK Hilly and mountainous terrain are more the rule than the exception on the Earth's surface. Estimates of the fraction of the Earth's surface covered by hilly and mountainous terrain range from 20% (Louis, 1975) to 70% (Strobach, 1991), depending on the definition. These terrain types affect many atmospheric processes, and orographically induced phenomena are common in the Earth's atmosphere (Atkinson, 1981; Blumen, 1990; Barry, 1992; Whiteman, 2000). For example, on fair weather days, cumuli develop frequently over mountain ridges. On weather maps and satellite images, the presence of orography is often revealed by disturbances that form on the upwind or downwind side of extensive mountain ranges such as the Rocky Mountains, the Andes, or the Alps. 1 Orographically induced phenomena are present on a large range of space- and time scales (Blumen, 1990; Barry, 1992) and they often occur simultaneously, making the investigation of an individual phenomenon a challenging task. In the past century and a half or so, our understanding of atmospheric processes over orographically complex terrain has increased greatly. More and better observations, laboratory experiments, and theoretical and numerical modeling approaches have helped us answer many questions but, of course, many new questions arose. An exciting aspect of mountain meteorological research in its current stage of development is the unknown number of phenomena that are yet to be discovered and explained. It sometimes seems as i f every new field experiment provides evidence for the existence of some new phenomenon whose explanation challenges the mountain meteorologist. The research presented in this dissertation, which focuses on boundary-layer phenomena in mountainous terrain, is a good example of this. The planetary or atmospheric boundary layer (ABL) is the atmospheric layer that directly interacts with the Earth's surface on a time scale of a few hours or less. In flat, homogeneous terrain, boundary-layer dynamics under daytime, convective conditions have been investigated intensively (e.g., Stall, 1988; Plate et al., 1998). The A B L that exists during these conditions is also called the convective boundary layer (CBL). The C B L grows by sensible heat input from the surface and downward sensible heat flux at the top of the CBL. The convergence of the sensible heat flux causes the C B L to warm and grow in depth. The growth rate of C B L depth is affected primarily by the characteristics of the fully developed inversion at sunrise and by the upward sensible heat flux at the surface (Driedonks, 1982a). Subsidence at the top of the C B L wil l slow its growth rate. Ball (1960), Lil ly (1968), Deardorff (1972), Stull (1973), Tennekes (1973), and Carson and Smith (1974) were among the first who developed successful models for C B L growth over flat, homogeneous terrain (e.g., Tennekes and van Ulden, 1974, Driedonks, 1982b). Other terms that are synonyms for C B L are convective mixed layer, mixed layer, or mixing layer (AMS, 2000). As these terms imply, convectively generated turbulence 2 dominates over shear generated turbulence to produce vigorous mixing in these ABLs . Because of the vigorous mixing, potential temperature, moisture, tracer concentrations and momentum are well-mixed in the vertical and are thus usually independent of height within the CBL. At the top of the CBL, there is usually an increase in potential temperature and wind speed and a sharp decrease in humidity and pollutant concentration. A layer in which potential temperature increases with height or in which the temperature lapse rate is less than the dry adiabatic lapse rate is considered a potential temperature inversion. The base of the inversion is often referred to as the C B L depth but other definitions of the C B L depth also exist as will be shown shortly. Note that there is a distinction between the depth and the height of the C B L : the CBL height is the height of the C B L above sea level (asl) while the CBL depth is the height of the C B L above ground level (agl). Especially for orographically complex terrain, this is an important note to remember; the terms are often used incorrectly. Of course, as long as one knows the elevation of a location, C B L height can be calculated easily at that location from C B L depth and vice versa. The C B L depth is an important scaling parameter in air-pollution and boundary-layer meteorology. Within the COST project of the European Union, a group of scientists agreed upon the following definition of C B L depth which they refer to as mixing height (Seibert et al., 2000): "The mixing height is the height of the layer adjacent to the ground over which pollutants or any constituents emitted within this layer or entrained into it become vertically dispersed by convection or mechanical turbulence within a time scale of about an hour". They note that this definition corresponds to the top of the entrainment zone; a zone at the top of the C B L that is not well-mixed and where turbulence intensity declines. This means that the C B L height is located somewhere within the capping inversion, not necessarily at its base. Many other definitions have been given in the literature for C B L depth, including the following: • height above the surface through which relatively vigorous vertical mixing occurs (Holzworth, 1972). • height above the ground of the layer of the atmosphere adjacent to the surface where vigorous mixing occurs as a result of thermal and mechanical turbulence (Norton and Hoidale, 1976). 3 • level of a potential barrier to the dispersion of pollutants at the interface between stable and less stable air (Maughan et al., 1982). • height at which a ground-based unstable to neutral vertical temperature profile becomes stable (Baxter, 1991). • height up to which pollutants would mix over a relatively short period of time of 1-2 hours (Baxter, 1991). • height that defines the vertical extent of vigorous thermal turbulence during daytime heating and thus sets a limit to upward mixing of pollutants (Myrick et al., 1994). In theoretical studies, where the C B L depth is used for scaling purposes, it is also defined as the height above the surface where the turbulent sensible heat flux is most negative. In this dissertation the term C B L height will be used more frequently than C B L depth. Also, the above definition by Seibert et al. (2000) will be assumed initially. Later in the dissertation, it will prove useful for mountainous terrain to distinguish between the C B L height and the height to which pollutants (released from the surface) can reach. Over flat terrain, these heights are identical. To determine the C B L height, Seibert et al. (2000) recommend the use of a modified parcel method and a Richardson number (Ri-) method using a surface excess temperature. This latter method, used throughout this dissertation, is described in Appendix 5. Since mesoscale numerical models can resolve the three dimensional field of turbulent kinetic energy (TKE), C B L heights are sometimes diagnosed from T K E as well. T K E is negligibly small above the boundary layer compared to T K E within the boundary layer so that the C B L height is determined as the height at which T K E falls below a certain cut-off value (e.g. Cai and Steyn, 1993). Conventionally, C B L heights have been determined from vertical temperature profiles. In recent years, the development of remote sensors such as sodars (sound detection and ranging) and lidars (light detection and ranging) has provided alternatives to this conventional method. Particularly, downward looking airborne lidars aboard an aircraft have proven useful for the investigation of the spatial variability of C B L heights 4 (e.g., Melfi et al., 1985; Kiemle et al., 1995; Hayden et al., 1997; Hageli et al., 2000). Such data will also be used in this dissertation. A lidar emits short laser pulses into the atmosphere and detects the backscattered radiation, which is proportional to aerosol concentration and aerosol size. The C B L is characterized by high aerosol loading and relative humidity relative to those in the overlying free atmosphere. The high relative humidity causes swelling of the aerosols which results in an enhanced backscattered radiation. For these reasons, a sharp change in the backscatter profile at the top of the aerosol layer (AL) indicates the location of the C B L top. Several studies over flat terrain have shown a good correspondence between estimates of the C B L height from lidar and temperature profiles (van Pul et al., 1994; Hayden et al., 1997; Marsik et al., 1995) while others have found the C B L height from lidar measurements to be slightly higher (Coulter, 1979). Overall however, it has been assumed that A L heights and C B L heights are equal. This is also supported by the fact that lidar backscatter images frequently show small scale A L top structure that resembles the C B L top structure found in laboratory studies (Deardorff et al., 1980) and large eddy simulations (Sullivan et al., 1998). Thermal eddies penetrating an inversion layer are the most prominent example. In this dissertation, both the structure (thermal and dynamical) and the height of the C B L are studied. These characteristics and their spatial variability are denoted by the term 'morphology' here. Knowledge of C B L morphology is required in many boundary-layer- and air-quality studies, and should be considered in the meteorology of ozone episodes in regions of complex terrain (Steyn et al., 1997). Also, since the C B L height affects, for example, the winds inside the C B L (and therefore the dynamical structure), knowledge of it will be important for aerial spraying- and fire weather operations, regional climatology investigations, studies of air-pollution dispersion, and weather forecasting. This knowledge will also aid in assessing the effects of air pollutant emissions and emission control strategies and assist in the forecasting of pollution episodes. The research in this dissertation aims towards gaining a better understanding of C B L height and structure in mountainous terrain. Next, a brief review of C B L morphology in mountainous terrain will be presented. 5 One of the first attempts at describing the thermal structure of the A B L over mountainous terrain was made by Ekhart (1948). Ekhart analysed radiosonde data and found that the diurnal amplitude of temperature in mountainous regions not only increases near the surface, but also at higher elevations. He observed a secondary maximum in the diurnal temperature range above the mountain ridges which is not observed over flat terrain, and concluded that the atmosphere above the mountain ridges is very different from and independent from the valley atmosphere. Based on analyses of the diurnal temperature range and the winds at various locations, he makes a distinction between four different atmospheres: the slope atmosphere of at most a couple of hundred meters, a valley atmosphere, determined by along valley flows, a large-scale mountain atmosphere which is produced by the thermal and dynamical effects of the entire mountain range, and the free atmosphere which is the unperturbed atmosphere at great horizontal and vertical distance from the mountain chain. These different atmospheres are depicted in Fig. 1.1. free atmosphere slope atmosphere | - v a l l e y atmosphere Fig. 1.1. Structure of the atmosphere above a mountain range (after Ekhart, 1948) Ekhart (1948) concludes his paper with the following statement: "It will be an extremely interesting and profitable task for Alpine meteorological research to expand on those notions on the structure of the mountain atmosphere , and in particular to find mathematical bases for a clearer separation of the different regions and to analyze their thermal and dynamic structure in detail. To achieve this goal, it would be necessary to make more use of modern aerological resources ..." (Ekhart, 1948, translated by Whiteman and Dreiseitl, 1984). 6 The research in this dissertation is a small step towards achieving this goal. By the time of Ekhart's (1948) publication, a considerable amount of research in mountain meteorology had already been dedicated to the determination of C B L structure in valleys. Early investigations were mainly focussed on the evolution of diurnal winds. Observations in numerous valleys have shown diurnal winds in valleys with winds generally blowing up the valley during daytime and down the valley during nighttime, termed upvalley- and downvalley flows, respectively (Whiteman, 1990). Similarly, flows blowing up the slope during daytime and down the slope during nighttime are termed upslope- and downslope flows, respectively. These thermally driven valley and slope winds have been researched theoretically and from observations since the 19th century starting with a description by Fournet (1840). The historical development of the understanding of these flows is an intriguing one (see Hawkes, 1947) which has led to the currently recognized theory of Defant (1949, 1951) and his well-known schematic of the diurnal development of slope- and valley flows. The first numerical simuations of these flows were made by Thyer (1966). It is now well-known that valley winds are produced by horizontal pressure gradients that develop as a result of temperature differences that form along the valley axis, or temperature differences between the air in the valley and the air at the same height over the adjacent plains (Whiteman, 1990). Similarly, slope flows are driven by temperature differences between the air near the slope surface and the atmosphere at the same height at a certain distance from the slope. The duration, intensity, vertical extent and other characteristics of slope- and valley flows depend on many factors such as valley width, ridge height, valley orientation, slope steepness, vegetative cover, and latitude. Furthermore, slope and valley winds are usually not the only winds that determine the wind characteristics in valley atmospheres. Regional and synoptic-scale flows are almost always present and the interaction between these flows and the more local valley flows often results in complex flow patterns. The presence of cross-valley flows (blowing from one sidewall to the other) can also make the wind pattern more complicated and variable. 7 Simultaneous observations of the wind- and temperature structure in a valley came much later and led to a conceptual model of the temporal evolution of the convective boundary layer in deep valleys (Whiteman, 1982). The role of slope flows in redistributing energy gained at the surface over the entire valley atmosphere plays an important role in this conceptual model. Also the compensatory sinking motions, that are produced over the valley center related to the withdrawal of mass by upslope flows, are a key aspect in this conceptual model. The spatial structure of wind and temperature including the C B L height along and across a valley has not been given much attention in previous research. In many studies, such as those where mass budgets are calculated, it is assumed that cross-valley temperature- and wind structure is homogeneous and that the along-valley structure is simple, with monotonically increasing/decreasing or constant flows along the valley (Freytag, 1987). Hewson and Gi l l (1944) presented temperature data on a cross section of the Columbia Valley near Trail, BC. Their observations (from aircraft data) showed that cross-valley differences in temperature of about 3-4 °C can occur on a constant height surface due to unequal heating of the slopes. Cross-valley circulations were also found by Hennemuth (1985) in a deep Swiss valley. While considerable past research has focussed on C B L structure in valleys, C B L structure over mountain ridges and slopes has received relatively little attention. Holzworth (1964) and Raymond and Wilkening (1980) noted the frequent presence of deep afternoon CBLs over mountainous terrain in western North America, with C B L heights up to 2-3 km over the mountain ridges. Cramer and Lynott (1961) and Cramer (1972) show potential temperature analyses of the daytime C B L over mountainous terrain from which similar C B L heights can be derived. Braham and Draginis (1960) made aircraft observations over a mountain range and investigated the nature of thermals. Their cross sections of potential temperature in the afternoon C B L reveal the presence of thermals over the highest peaks and ridges. Banta and Cotton (1981) studied the C B L structure in the Colorado mountains and found that during daytime, strong westerly gradient winds in a neutral layer aloft were mixed downward to the surface and 8 overpowered upvalley and upslope flows. Banta (1982, 1984) found that while the C B L in the early morning is very inhomogeneous, the deep afternoon C B L tends to be horizontally homogeneous. He also found that terrain features seem to have a smaller effect in localizing thermals than was expected from previous research. Relatively homogeneous afternoon CBLs have also been found over slightly undulating terrain by Lenschow et al. (1979) and over more complex terrain in Germany (Fiedler, 1983). An extensive analysis of over 300 radiosonde ascents over Israel was performed by Dayan et al. (1988). They found that orography was the major factor determining C B L height variability, rather than differences in synoptic conditions or land use. Both terrain following and non-terrain following C B L heights were observed. Based on lidar observations, Nyeki et al. (2000) concluded that C B L heights did not follow topography. Radiosonde and aircraft data from a field study in the German Black Forest region (Fiedler, 1992) showed a variety of behaviors, from CBLs that follow the underlying terrain to CBLs that seem unaffected by terrain irregularities (De Wekker et al., 1997; Kalthoff et a l , 1998; KoBmann et al., 1998). Investigators found it difficult to arrive at general conclusions regarding circumstances under which different behaviours occur, but factors such as atmospheric stability and synoptic wind speed seem to play an important role. In general, relatively small C B L depths were found over the mountain ridges and C B L growth models applied to these sites overestimated the growth of the C B L . De Wekker (1995) and Binder (1997) proposed the use of an effective sensible heat flux for use in growth rate models that implicitly takes into account the effects of advection, orography shape and other effects that influence C B L growth over mountain ridges. Unfortunately, relative contributions of individual effects are not yet known. Stull (1998) derived an equation for the tendency of the C B L top to become more horizontal in the course of the day by applying the mass conservation equation for a C B L over hilly terrain. He found that advection, entrainment and friction contribute to making the C B L follow the terrain while gravitational forces tend to make the top of the C B L more level. The derived equation applies for free convection conditions. If there is more than just a light wind, the advection term was found to be dominant in making the top of the C B L follow the orography. This was also found by Dayan et al. (1988). A major shortcoming of Stull's theory is that it does not take into account the effect of thermally-driven flows 9 over mountainous terrain. Those flows are known to affect C B L heights to a large extent (KoBmann et al., 1998; De Wekker, 1995). Stall's theory shows further that deeper CBLs are less terrain following than shallower ones and that the mixed layer top is less level over orographical features that have a larger horizontal extent. These latter implications of the theory are generally corroborated by the observations mentioned above. A number of studies have appeared in the last decade or so in which large-eddy simulations were performed to investigate the effect of orography on the CBL. In all cases idealized topography was used. Walko et al. (1992) investigated the impact of moderate hills on the C B L and did not find significant differences in the mean profiles of atmospheric variables between flat and hilly terrain. Gopalakrishnan et al. (2000) found that for horizontal length scales of less than about 5 km, topography has very little impact on the mean properties of the C B L , even with hills as high as 30% of the C B L depth. At larger horizontal scales, topographical features as small as about 10%> of the depth of the C B L have some effect on the mean characteristics of the C B L . The conceptual model that arises from previous research on the diurnal development of the C B L over mountainous terrain is depicted in Fig. 1.2 (Fiedler et al., 1987; Whiteman, 2000). In the morning hours cold air collects in the valleys to produce an atmospheric stability there that is larger than over the mountain ridges. After sunrise, this leads to a slower development of the C B L in the valleys than over the mountain ridges. After the inversion in the valleys has broken up, the development of the C B L is faster there than over the mountain ridges. This leads to a C B L in the early afternoon that more or less follows the terrain. If the C B L in the valleys keeps growing faster there than over the mountain ridges, the top of the C B L eventually becomes level. With this conceptual model, there is no clear separation in the C B L between a slope atmosphere, a valley atmosphere, and a mountain atmosphere as called for in Ekhart's (1948) model. The broad and general intent of this dissertation is to extend the current concepts of the daytime C B L morphology over mountainous terrain as depicted in Figs. 1.1 and 10 1.2. An overview of research objectives and questions and the approach that is taken in this dissertation is presented in the next section. horizontal distance (km) horizontal distance (km) elz) Fig. 1.2. Conceptual model of the boundary layer in mountainous terrain on a fair weather day in the early morning just before sunrise (a), in the morning just after sunrise (b), at noon (c), and in the late afternoon (d). Light grey shading is the topography. Dark grey shading is a strongly stable nocturnal inversion. Arrows denote horizontal and vertical flows. The wavy lines depict the upper limit of surface generated convection. Vertical profiles of potential temperature are shown at several locations in the topography (after Fiedler et al., 1987). 1.2. RESEARCH OBJECTIVES AND METHODOLOGY In this dissertation, aspects of C B L morphology are examined in different topographic settings; in a deep, narrow valley; near a mountain base; and over a mountain range. These three topographic settings are sketched in Fig. 1.3. A mountain range can be considered as consisting of a multitude of archetypal topographic units having a multitude of horizontal and vertical scales. Valleys and mountain bases are two examples of such archetypal topographical units. Other examples include mountain ridges, plateaus, and basins. Aspects of C B L morphology that this dissertation focusses on are stability, wind structure, and C B L height. 11 VALLEY MOUNTAIN BASE MOUNTAIN RANGE Fig. 1.3. Schematic of the different topographic settings in which C B L morphology is studied in this dissertation. Lx, Ly, and H depict horizontal and vertical scales of the topography. Ly denotes the horizontal scale in the along-valley direction. Typical values for the topographies studied are given in Table 1.1 Previous studies have mainly been concerned with examining atmospheric processes in individual topographic units, primarily in valleys (see previous section), without considering how these processes may affect or interact with the CBL morphology of an entire mountain range. Speaking in terms of Ekhart's (1948) concept outlined in Fig. 1.1, investigators have focussed on examining the valley atmosphere and, to a much lesser extent, the slope atmosphere, while relatively little attention has been given to a more integrative approach in which the atmospheric structure of the large scale mountain atmosphere is also considered. In this dissertation, an integrative approach is taken by making use of a variety of data sets, a sophisticated numerical modeling system, and a Lagrangian particle dispersion model. First, CBL morphology in a valley and near a mountain base are studied in chapters 2 and 3, respectively, followed by an investigation of CBL morphology over an entire mountain range in chapter 4. 12 Three data sets are used in this dissertation, two of which contain downlooking aerosol lidar data. Downlooking aerosol lidars have given a new stimulus to boundary-layer research in recent years by providing a tool to investigate C B L structure over large horizontal scales. Lidar data, however, cannot provide needed information about the concurrent thermodynamic structure of the atmosphere. In this dissertation, the interpretation of lidar data is assisted by a mesoscale numerical model. This combined use of lidar data and a mesoscale numerical model is a novel, and, as will be shown in chapters 3 and 4, a useful approach to gain additional understanding of C B L morphology over mountainous terrain. The third data set includes aircraft data which provide an excellent means to describe and investigate C B L morphology in a valley and to evaluate the performance of a mesoscale modeling system. This data set also contains extensive turbulent surface sensible heat flux measurements. Coupling between the surface and the overlying atmosphere is established through the fluxes of heat, mass, and momentum. An investigation of sensible heat fluxes is therefore expected to help understand C B L morphology in complex terrain. The analysis of the data sets is supplemented by mesoscale numerical simulations and simulations of particle dispersion. The objectives of these simulations are to evaluate the performance of a mesoscale modeling system in very complex terrain, to assist in the interpretation of the observational data, to investigate the mechanisms producing the observed morphology, and to investigate the net effect of modeled C B L morphology on the dispersion of a passive tracer. Idealized simulations and two case studies are performed. Based on the findings and on knowledge of previous research, concepts will be formulated concerning the dynamics of C B L morphology over mountainous terrain. The improved understanding of C B L morphology over mountainous terrain from these observational and numerical investigations leads to a conceptual picture describing some important aspects of C B L morphology over mountainous terrain. As noted before, an improved understanding of C B L morphology in mountainous terrain has many applications in air-pollution meteorology, aviation, weather forecasting, and regional climate modeling. For example, surface concentrations of air pollutants have 13 been known to increase rather than decrease as the C B L height increases i f elevated pollutant layers are entrained (McKendry and Lundgren, 2000). It is also of importance to aerosol scientists who try to explain the diurnal course of aerosol concentration at high-alpine stations. The diurnal course of aerosol concentration is often attributed to the diurnal development of the C B L in mountainous terrain (Lugauer, 1998). The organization of the dissertation is outlined below. This section ends with a set of research questions, which clarify the main focus of this dissertation. In chapter 2, C B L morphology is investigated in a deep, narrow valley using the MAP-Riviera data set. This data set comes from a field study carried out in the Riviera Valley in Switzerland in the summer and fall of 1999 (Rotach et al., 2002). The Atmospheric Science group at U B C participated in this field study with many universities and research institutes from across the world. The initial plan was to perform simultaneous measurements in the slope atmosphere, valley atmosphere, and mountain atmosphere (following terminology of Fig. 1.1). Such a data set would be ideal for the investigation of multi-scale phenomena of the C B L in mountainous terrain and their interactions. Unfortunately, airborne measurements in the slope and mountain atmosphere were not accomplished as planned. Despite this set-back, a very extensive and unprecedentedly detailed data set was obtained. The MAP-Riviera field study is unique in its successful operation of a network of turbulence sensors on the valley floor and sidewalls and of an aircraft that collected data across and along the valley on a number of days under a variety of weather conditions. Data from one undisturbed fair weather day are used in this dissertation. A mesoscale numerical model is used to further investigate C B L morphology in the Riviera Valley and to evaluate the model performance in an environment of very complex terrain. The data set provides the unique opportunity to evaluate the turbulent sensible heat fluxes at the surface and the mean and turbulence structure in the valley atmosphere. Such a model evaluation has never been performed in an environment this complex. In chapter 3, C B L morphology is investigated near a mountain base using the PACIFIC'93 data set. This field study was carried out in the Lower Fraser Valley in 14 British Columbia, Canada in the summer of 1993 (Steyn et al., 1997). The data set includes downlooking lidar data, tethersonde data and rawinsonde data. Data from one intensive observational period is used. Observations and idealized mesoscale numerical simulations show that the mountain range exerts a profound influence on the C B L by reducing the C B L height in its vicinity. The objective of this study is to examine the processes underlying this phenomenon. For this purpose, the individual terms in the heat budget equation are examined using idealized two-dimensional numerical simulations. Chapter 4 describes aspects of C B L morphology over a mountain range using the STAAARTE'97 data set. This field study was carried out in the Bernese Oberland, a mountain range in the Swiss part of the European Alps, in the summer of 1997 (Nyeki et al., 2000). The data set includes data from a downlooking lidar aboard an aircraft that flew over the Alps during one day. Preliminary results from this field study were presented by Nyeki et al. (2000). In that study, it was concluded that C B L heights did not follow topography. The objective of this chapter is to determine the processes underlying this behaviour. For this purpose, realistic simulations are made with a mesoscale numerical model and a Lagrangian particle dispersion model. The results from the individual chapters are brought together and discussed in chapter 5. Conclusions and an outlook for future research are given in chapter 6. An extensive appendix concludes this dissertation. Some characteristics of the investigations in chapters 2, 3, and 4, including the typical horizontal and vertical scales, are summarized in Table 1.1. Table 1.1: Characteristics of the investigations in chapters 2, 3, and 4. Lx, Ly, and H are defined in Fig. 1.3. Topographic setting chapter Dimensionality spatial scales (km) Lx Ly H main numerical simulations idealized realistic Valley 2 3-D 2 15 2 X Mountain base 3 2-D 10 1 X Mountain range 4 3-D 4, 50 10, 50 1,3 X 15 One consequence of the integrative approach used in this dissertation is that an in-depth investigation of each of the three data sets was not possible. This is particularly so for the MAP-Riviera and Pacific'93 data sets where only one day was selected for investigation. An investigation of multiple days could easily lead to several PhD dissertations. The selected data sets enable an integrative approach and facilitate the investigation of the following research questions: • What characteristics of C B L morphology are observed in a deep, narrow valley, near a mountain base, and over a mountain range and how can these characteristics be explained? • Can a mesoscale numerical model simulate the C B L characteristics in a deep, narrow valley, near a mountain base, and over a mountain range? • What differences and similarities can be found between the C B L morphology in a deep, narrow valley, near a mountain base, and over a mountain range? • How do simulated and observed turbulent sensible heat fluxes compare in a deep, narrow valley? What is the spatial variability of observed and modeled turbulent sensible heat fluxes and how can this affect C B L morphology? • The assumption that the A L height is equal to the C B L height has been justified by many studies over flat terrain (see chapter 1.1). Can this assumption also be made over mountainous terrain so that lidar data can be used to determine C B L heights over mountainous terrain? If not, how do the behaviours of C B L and A L heights differ over mountainous terrain? • What can be learned from airborne aerosol lidar data regarding C B L morphology in mountainous terrain? 16 2. C B L M O R P H O L O G Y IN A V A L L E Y : T H E M A P - R I V I E R A F I E L D S T U D Y 2.1. I N T R O D U C T I O N In this chapter, data from the MAP-Riviera field study (Rotach et al., 2002) and a mesoscale numerical model are used to study various aspects of C B L morphology in a deep and narrow valley. Surface and airborne observations will be presented and discussed and the mesoscale numerical model will be evaluated with these observations. The model is then further used to assist in the interpretation of the data. Simulation results are often interpreted without proper evaluation with observational data. In some previous field studies in other valleys, only single vertical soundings have been used for model evaluation. In other cases, no data were available at all. The MAP-Riviera field study provides a unique and extensive data set for model evaluation including high resolution upper air data and turbulent surface flux data. This data set therefore provides an excellent opportunity to evaluate a mesoscale numerical model in very steep and complex terrain, while investigating C B L morphology. This chapter is organized as follows. First the data set and mesoscale model setup will be described. This includes a discussion of the boundary and initial conditions for the simulations. Then, the observations and simulation results are presented. The diurnal course of selected surface variables is then discussed including turbulent surface sensible heat fluxes on the valley floor and sidewalls. This is followed by a discussion of rawinsonde and aircraft data focused on the temporal evolution and the spatial structure of temperature and along-valley winds. Simulations are evaluated with the observations and are used to interpret the observed phenomena. Finally, the validated model output is used to drive an L P D M simulation to investigate the influence of C B L morphology on the dispersion of a passive tracer released at the surface. 2.2. D A T A A comprehensive boundary-layer field study, described by Rotach et al. (2002), was carried out from August to October 1999 in the Riviera Valley of southern Switzerland. This field study was part of the larger scale Mesoscale Alpine Programme (MAP) field project and is referred to as the MAP-Riviera field study. M A P is a major 17 international research effort to improve knowledge on meteorological and hydrological processes over the Alps. A general description of M A P can be found in Bougeault et al. (2001). Eight scientific projects were defined within M A P (Binder and Schar, 1996). Two of these projects, dealing with boundary-layer and hydrological processes over steep orography, were performed in the MAP-Riviera field study. A map of the topography in the area is shown in Fig. 2.1. The Riviera Valley is located in the southern Alps, about 100 km north of Milan, between the small towns of Bellinzona in the south (240 m asl) and Biasca in the north (300 m asl). The Riviera Valley is a U-shaped valley with several tributary valleys. Its orientation is from southeast to northwest (155° - 335°). The valley is narrow and steep with a valley floor width of 2 km, a depth between 2 and 2.5 km, and slope angles of roughly 30° on the eastern slope and 35° on the western slope. The valley floor has a length of about 15 km, 0 5 10 15 20 x (km) Fig. 2.1. Topography of the Riviera Valley and surroundings. Contour lines are drawn every 400 m. Terrain above 2000 m asl is shaded. The dashed lines depict the cross-valley and along-valley cross sections. The main valley site, Bosco di Sotto, (46.26° N , 9.01° E, 250 m asl) is located near Claro in the valley center. 18 and an average slope angle of less than 0.5°. The highest peak is at 2727 m asl (Pizzo di Claro). The topography of this valley is typical of the southern Alps. The valley floor consists of agricultural land and a number of small villages. A highway and railroad run along the valley. The slopes are covered mainly with deciduous trees up to 1000 m asl, with conifers above. The treeline is at about 1800 m asl with areas of bare ground and short grass at higher elevations. A detailed map of the land use in the area is shown in Appendix 2, Fig. A2.1. During eight 'flight days', an instrumented light aircraft operated by MetAir (Neininger et al., 2001) flew specified patterns inside the valley. Two different flight patterns were flown on these days, one with cross-valley flight legs at different heights and another with along-valley flight legs on both sides of the valley at different heights. A schematic of this flight pattern is depicted in Fig. 2.2. During the flight days, rawinsondes were released at the valley center approximately once every three hours. Fig. 2.2. Flight pattern in the cross-valley (top) and along-valley (bottom) directions. 19 The flight days cover a variety of weather situations, from overcast days where mechanically driven turbulence is expected to play a dominant role to dry, cloudless, convective days. The present study uses data from the 25 August 1999 (day 237) flight day, which had dry, convective weather. Theoretical sunrise was at 0435 UTC and sunset was at 1817 UTC. On this day, aircraft measurements were taken in the morning and afternoon and there was a good background coverage and quality of surface data. The surface and 500 mb weather maps for 25 August 1999 are shown in Fig. 2.3a and b, respectively. A high-pressure ridge extending from North Africa to Scandinavia influenced the weather in the investigation area. Synoptic flows were weak to moderate (15 m s"1 at 500 mb) from northwesterly directions (-300°). The combination of northerly synoptic flows with southerly valley flows is known as inverna in this region and is a common wind pattern in the southern Alps (Furger et al., 2000). The satellite image for 1419 UTC on this day, shown in Fig. 2.3c, shows cloudless conditions over the Alps and a major part of central Europe. Radiosondes were launched on 25 August 1999 at 0739, 0915, 1208, 1508, and 1800 UTC. The aircraft operated between 0649 and 0942 UTC in the morning and between 1112 and 1541 UTC in the afternoon. In the morning, three across- and one along-valley flights were made. In the afternoon, three across- and two along-valley flights were made. Appendix 4 describes the exact times and presents data for all the flight legs. Data from selected flight legs will be used in the current chapter. Accuracy of rawinsonde and aircraft data is on the order of 0.1-0.5 K for temperature and 0.5 m s"1 for wind speed. A detailed description of the observed parameters and their specifications can be found in Rotach et al. (2002). 20 Fig. 2.3. Surface (a) and 500 mb (b) weather maps from the Swiss Meteorological Institute for 25 August 1999 1200 UTC. (c) A V H R R satellite image at 1419 U T C (channel 2 image, 0.725-1.10 urn). Satellite image courtesy of the University of Dundee (http://www.sat.dundee.ac.ulc/). 21 In addition to the airborne observations, the data set for this case study also includes a unique set of surface turbulent flux data obtained at ten different measurement sites on the valley floor and sidewalls. The measurement sites exhibit a large heterogeneity in slope and surface characteristics. Details of the measurement sites are provided in Table 2.1; the locations are shown in Fig. 2.4. A l l slope sites, except for site C, were located on the west-facing slope. 10 12 1 4 1 6 18 x (km) Fig. 2.4. Location of the surface turbulence stations. See table 2.1 for details. Contour lines are drawn every 100 m. 22 Table 2.1. Site identification, location, measurement height, and some surface characteristics for the ten surface stations measuring turbulence. Site Location lat(°N), lon(°E), elev (m asl) Measurement height (m agl) Surface characteristics A l 46.2572, 9.0131, 250 3.56 Valley floor, mixed agriculture A2 46.2500, 9.0153, 250 1.15 Valley floor, mixed agriculture B 46.2647, 9.0311, 760 23.78 Slope; forest (mean height of trees ~ 15m) C 46.2494, 9.0056, 340 6.33 Slope; vineyard D 46.2467, 9.0269, 256 2.62 Valley floor; mixed agriculture E l 46.2667, 9.0372, 1060 12.70 Slope; meadow E2 46.2706, 9.0364, 1030 22.68 Slope; forest (mean height of trees ~ 13m) F l 46.2700, 9.0553, 1750 6.30 Slope; sparse vegetation F2 46.2728, 9.0608, 2110 1.30 Slope; shrub (-75 m below ridgeline) G 46.2742, 9.0317, 870 5.25 Slope; forest (bridge over small tributary valley) Small-aperture scintillometers were installed at A2 and D measuring over a path length of about 100 m. Sonic anemometers were installed at the other sites. Specific information about the instrumentation can be found in Rotach et al. (2002). Measurements of turbulent sensible heat and momentum fluxes were made at all sites. At a few sites, latent heat fluxes were also measured. In the present study only sensible heat flux measurements will be presented. In the next section, the mesoscale numerical model, which wil l be evaluated with the data described above and which will be used to develop a more comprehensive understanding of the C B L morphology than is possible from the data alone, will be described. 2.3. N U M E R I C A L M O D E L S E T U P The mesoscale numerical model used is the Regional Atmospheric Modeling System (RAMS, Pielke et al., 1992), in which land-surface processes are represented by 23 the Land Ecosystem Atmosphere Feedback Model, version 2 (LEAF-2, Walko et al., 2000). Additional details on the mesoscale modeling system can be found in Appendix 1.1. The simulations use two-way interactive, nested grids. The model domain consists of four nested grids with horizontal grid spacing of 9, 3, 1, and 0.333 km, respectively. The outermost grid covers central Europe including the Alps while the innermost grid is the area shown in Fig. 2.1. A l l four grids have 38 vertical levels, with a grid spacing from 70 m near the surface to 1000 m near the model top at about 16 km. The vertical grid spacing varies with height by a factor of 1.09 until it reaches its maximum value of 1000 m. Some issues that were considered to determine appropriate horizontal and vertical grid spacings are presented in Appendix 1.3. Thirteen soil nodes were used to a depth of 0.9 m below the surface. Details of the four grids used in the simulations are given in Table 2.2. Table 2.2. Characteristics of the four grids used in the M A P case study. N X and N Y are the number of grid points in the west-east, and north-south direction, respectively. Grid N X N Y Grid spacing W-E Size N-S Size AT(s) (km) (km) (km) 1 89 89 9 801 801 30 2 53 53 3 159 159 15 3 65 65 1 65 65 7.5 4 71 92 0.333 23.667 30.667 3.75 The simulations cover 36 hours from 1200 UTC 24 August to 0000 UTC 26 August 1999. The five outermost lateral boundary points in the largest domain were nudged using an implementation of the Davies (1976) scheme toward E C M W F objective analysis fields and rawinsonde data to allow changes in large-scale conditions to influence the model simulations. Top boundary nudging towards analysis fields is used to suppress gravity wave reflections at the model top. The magnitude of the boundary forcings was assumed to vary linearly in time between the 6-hour intervals for which E C M W F analysis fields were available. Nudging towards objective analysis fields was only applied to the outermost grid; no interior nudging was applied. 24 The land use and topography in grids 1, 2, and 3 were derived from 30 arcsecond (~ 1 km) resolution data from the USGS (United States Geological Survey) data set. The USGS land use data set is based on 1 km resolution Advanced Very High Resolution Radiometer (AVHRR) data spanning April 1992 through March 1993 (see Appendix 2). For the innermost grid (with 333 m grid spacing), topography and land use were obtained from a 100 m resolution Swiss topography and land use data set (BFS, 1993). Land use types in the Riviera Valley from the 1 km USGS data set were found to give a poor representation of reality, as is demonstrated in Appendix 2. Thus, use of the high resolution Swiss BFS data set was deemed necessary. Most land use classes in the BFS data set did not match those in LEAF-2. Also, the BFS data set contains many more classes (69) than LEAF-2 (31). Therefore, land use classes from the BFS data set had to be cross-referenced to land use classes of LEAF-2 , using an approach described in Appendix 2. The BFS data set did not distinguish between needleleaf and broadleaf trees. By inspecting photographs and satellite-derived vegetation patterns in the Riviera Valley (Rotach, van Gorsel, personal communication), the boundary between broadleaf and needleleaf trees was put at 1000 m asl with only needleleaf above and broadleaf below that boundary. The resulting land use in the Riviera Valley as it was used in the mesoscale simulations is shown in Appendix 2, Fig. A2.1. Modeling studies have noted the importance of a correct soil moisture initialization for simulating atmosperic processes in a numerical model (Grasso, 2000; Jacobson, 1998). Soil moisture is often used inappropriately as a tuning parameter to obtain good agreement with atmospheric observations. Also, despite the fact that spatial variability of soil moisture in complex mountainous terrain is expected to be very important for boundary-layer processes, it is usually neglected or inadequately initialized because of a lack of data. Soil measurements were taken at a few sites in the Riviera Valley but there is insufficient information about its spatial variability. To improve the soil moisture initialization for the present case study, soil moisture distribution for the innermost grid was obtained from a WaSiM simulation (Water balance Simulation 25 Model, a hydrological model developed at ETH-Zurich (Jasper, 2002)). The simulation was made in hourly time steps from 1 January 1999 through 23 August 1999 to obtain the soil moisture initialization for 00 UTC 24 August 1999. The spatial resolution is 500 x 500 m and the simulation was done for the two catchment areas covering the major part of the innermost grid (i.e., the Ticino and Verzasca catchment areas). Figure A2.3 in Appendix 2 depicts the volumetric soil moisture (ratio of the volume of water in a soil sample and the volume of the total soil sample) for the area. It can be seen that volumetric soil moisture in the Riviera Valley and adjacent sidewalls was rather inhomogeneous, with typical values around 0.31. These soil moisture values compared well with observations of soil moisture at sites A l and B in the valley (for details about soil moisture measurements, see Zappa et al., 2000). At high elevations where bare soil is present, the soil moisture is considerably reduced. For the other grids, a constant volumetric soil moisture of 0.28 was assumed. This value is somewhat lower and is believed to be more representative of the inner Alps. The southern part of Switzerland covering the Riviera Valley has a more humid climate than the inner Alps (e.g., Frei and Schar, 1998). Sensitivity studies showed that the boundary-layer structure inside the Riviera Valley is not very sensitive to the specific value of the soil moisture in the outer grids. Thus, the lack of information about soil moisture for these grids is not crucial to this study. Soil type in the simulations was set to loamy sand, which is the predominant soil type in the Riviera Valley (Jasper, 2002). The volumetric soil moisture for this soil type at saturation is 0.41. The initialization of the soil temperature offset (i.e., the difference between air and soil temperature) at the 13 soil levels was based on observations of the soil temperature at site A l . Soil temperature offsets at 50 cm, 20 cm, and 5 cm depth at 1200 UTC were found to be approximately 7 K, 6 K, and 5 K respectively (the soil being cooler than the air temperature). The modeled incoming shortwave radiation was compared to the observed incoming shortwave radiation at site A l (Table 2.1) and at two meteorological stations north and south of the Riviera Valley. It was found that the model overestimated the incoming shortwave radiation somewhat. The procedure that was followed to reduce this 26 error is outlined in Appendix 1.4. It is interesting to note that Zhong and Doran (1994) also found a discrepancy between the observed shortwave radiation and the parameterization in R A M S . An investigation of the causes of this discrepancy is outside the scope of this dissertation. In the following section, observations in the Riviera Valley will be discussed and used to evaluate the mesoscale model. First, surface data will be shown, followed by airborne data. 2.4. DIURNAL RANGE OF SURFACE VARIABLES AT BOSCO DI SOTTO 2.4.1. OBSERVATIONS Andretta et al. (2001) selected valley wind days for the Riviera Valley in the M A P field study period between 21 August and 16 October 2000 to facilitate their investigation of the near surface turbulent characteristics of the valley boundary layer. Valley wind days are characterized by a regular pattern of the diurnal evolution of the wind field (Barry, 1992; Whiteman, 2000). The criteria for the selection of valley wind days included a strong diurnal range of pressure gradient between two sites north and south of the Riviera Valley, weak synoptic flows and a large diurnal range of global radiation (fair weather days). The days that were selected are: 21, 25, 31 August; 2, 7, 8, 9, 11, 12, 14, 15, 24 September; and 1, 7, and 16 October. These days cover about 20% of the entire measuring period. Figure 2.5 shows the diurnal course of temperature, mixing ratio, wind speed, and wind direction on valley wind days for Bosco di Sotto in the center of the valley (site A l , Fig. 2.4). The diurnal range on 25 August 1999, the case study in this chapter, is highlighted in the figure with filled circles. It can be seen that 25 August 1999 was the warmest (up to 28 °C) and most humid (up to 16 g kg"1) of the valley wind days. Global radiation values also were the highest of the valley wind days (not shown). Compared to all the other valley wind days in the field study, the thermal forcing of the boundary layer on this day is expected to be relatively large, making it a suitable day for the investigation of C B L morphology in this valley. Wind direction shows an onset of upvalley flows at about 0800 UTC, about 3.5 hours after sunrise (0435 UTC). Upvalley 27 flows cease shortly after sunset (1817 UTC). Wind speed increases after the onset of the upvalley flow to about 4 m s"1 at around 1300 UTC and decreases (and becomes more variable) afterwards. The daytime wind is approximately aligned with the upvalley direction indicated by the horizontal solid black line at 155°. The wind speed and direction on 25 August do not deviate significantly from the average behaviour. 12 16 Time (UTC) 24 8 12 16 20 24 Time (UTC) ,(b). 360-(d) 270 J Q T3 § 90 8 12 16 Time (UTC) 20 24 ~ o ° 0 ° a,° 8 12 16 20 24 Time (UTC) Fig. 2.5. Temperature (a), mixing ratio (b), wind speed (c) and wind direction (d) at the surface at site A l (Bosco di Sotto) for all valley wind days during MAP-Riviera. The closed circles are the data for 25 August 1999. Up-valley wind direction is indicated by the horizontal black line in (d). 28 2.4.2. M O D E L E V A L U A T I O N Figure 2.6 shows the diurnal course of observed and modeled net radiation and potential temperature at site A l . (a) 400 A observations 12 16 Time (UTC) 20 24 (b) 304 „ 300 e 2 296-I 292 288 J V 8 12 16 Time (UTC) 20 24 Fig. 2 . 6 . Observed (filled circles) and modeled (open circles) net radiation (a) and potential temperature (b) at Bosco di Sotto on 25 August 1999. The data are half-hourly averaged values while the model output represents instantaneous hourly values. The temperature at the first model level at -35 m was extrapolated downward to observation height (1.5 m) using Monin-Obukhov similarity functions (e.g., Stull, 1988). Net radiation is somewhat overestimated by the model in the early morning and late evening because the model does not take into account the shadowing effect of the adjacent sidewalls. Colette and Street (2002) recently modified the radiation code in a numerical model to account for the shadowing effect and found that inversion-layer breakup in steep valleys was slightly retarded by this. The self-shading effect on the radiation budget in sloping terrain is more significant and is taken into account by R A M S . Some clouds were present in the afternoon, which explains the short-term reduction of the radiation values at around 1300 UTC. The agreement in the diurnal range of potential temperature is good, as the model underestimates the minimum and maximum temperature by only 1-2 K. The decrease in temperature in the early evening is somewhat underestimated by the model as well. The rate of increase in potential temperature after sunrise is particularly well simulated. This is important for 29 this study since the boundary-layer structure is studied between approximately 0700 UTC and 1500 UTC. Figure 2.7 shows the observed and modeled horizontal wind vectors at site A l as a function of time at Bosco di Sotto. The 10-m wind speed was obtained by extrapolating the wind speed downward from the first model level (~35 m) .using Monin-Obukhov similarity functions. The model clearly shows upvalley flows with speeds that are generally somewhat weaker in the early afternoon and stronger in the late afternoon and evening than the observed wind speeds. Upvalley flows start about two hours later in the model than in the observations at this particular location. In the next section, the modeled spatial surface wind field in the Riviera Valley will be discussed and evaluated with available observations. - observation 12 m agl 2 ms"1 . . \ t \ \ ^ \ \ V \ \ . i \ \ \ / _ , model 10 m agl , v \ \ \ w w \\\ I ' 1 ' 1 ' 1 ' 1 4 8 12 16 20 Time (UTC) Fig. 2.7. Horizontal wind vectors plotted as a function of time at Bosco di Sotto, as observed at 12 m agl (top panel) and modeled at 10 m agl (bottom panel) for 25 August 1999. 2.5. SPATIAL SURFACE WIND FIELD IN THE RIVIERA V A L L E Y Figure 2.8 shows the modeled surface wind field (10 m agl) at 0900, 1200 and 1500 UTC in the Riviera Valley. Also shown is the wind observed at the various surface stations listed in Table 2.1 and some major flow patterns. At 0900 UTC, the upvalley wind at the valley entrance makes a sharp turn towards the western sidewalk 30 6 8 10 12 14 16 18 6 8 10 12 14 16 18 x (km) x (km) 6 8 10 12 14 16 18 x (km) Fig. 2.8. Modeled and observed surface wind field in the Riviera Valley at 0900, 1200, and 1500 U T C . Observations at various surface stations are indicated by the bold arrows. Topography is shown with a contour interval of 400 m and the shading corresponds to modeled wind speed. Dashed arrows depict some major flow features in the valley. The axes scale corresponds to that on Figs. 2.1 and 2.4. 31 This sidewall, in contrast to the opposed sidewall, is lit by the sun at this time, and a cross-valley wind component is induced towards the heated slope. Wind fields observed by aircraft and modeled wind fields (shown later) also show this cross-valley wind component at higher elevations around this time. Wind speeds on the eastern sidewall are very weak at this time. On the valley floor around and just north of site A l , modeled winds are rather weak and indeterminate. This is not only the case at 0900 UTC but also at later times and explains some of the discrepancies between observed and modeled winds at site A l in Fig. 2.7. By 1200 UTC, upslope flows have started at the eastern sidewall and the winds near the entrance are more aligned with the valley axis. Upslope flows still prevail at the surface at 1500 U T C on the eastern sidewall while, on the western sidewall, a downslope flow component is observed. It can be seen in Fig. 2.8 that in the course of the day, upvalley flows become most intense on the eastern side of the valley and resemble a jet-like feature. Simulations also suggest the existence of recirculation patterns on the valley floor (indicated by curved arrows in Fig. 2.8) which are probably induced by wind speed differences across the valley. Winds near the exit of the valley near Biasca where there is another bifurcation (see Fig. 2.1) are also relatively high and generally directed upvalley. Noticeable is the chaotic and disorganized behaviour of the wind in the central part of the valley during the day. The model predicts flows that are directed downvalley at times and there is no evidence of a horizontally homogeneous upvalley flow. Only at the eastern side of the valley does the flow seem well-organized. Unfortunately, the observational data at the surface is not well-suited to verify some aspects of the modeled surface flow patterns. Comparison between the different observed wind speeds is also complicated by the different measurement heights. In general, however, the valley and slope flows that are seen in the observations are captured by the simulations. Aircraft data shown later also indicate disorganized flows at upper levels, and also show downvalley directed flows during daytime on the western side of the valley, although not to such an extent as in the model. The complex and disorganized wind structure is thought to be due to the topographic setting of the Riviera Valley, which does not open into a plain. The Riviera Valley flow is not entirely induced by the valley itself but is part of a larger-scale 32 upvalley flow that splits off near the mouth of the valley and flows partly towards the Calanca and Mesolcina valleys to the north (see Fig. 2.1). Furthermore, tributary valleys may also have contributed to the complex flow pattern in the Riviera Valley. 2.6. TURBULENT SENSIBLE HEAT FLUX IN THE RIVIERA V A L L E Y Turbulent sensible heat input from the valley surface provides the energy needed to drive the evolution of wind systems and boundary layers in valleys. The MAP-Riviera data set provides an unique opportunity to study the turbulent surface heat exchange in the valley and over the sidewalls and to evaluate the performance of a mesoscale model in its simulation of the sensible heat flux. A l l the sites listed in Table 2.1 measured sensible heat flux. A few sites also measured latent heat flux. Inspection of the energy balance at Site A l showed that most of the available energy on the valley floor is used for evapo(transpi)ration, with Bowen ratios generally smaller than 0.5. A direct comparison between modeled and observed sensible heat fluxes for all 10 surface sites is shown in Fig. 2.9. 0 5 0 .0 0.0 0.1 0.2 0.3 0.4 0.5 Modeled w T (Kms1) Fig. 2.9. Observed versus modeled half-hourly-averaged sur-face sensible heat flux for all the surface stations listed in Table 2.1 on 25 August 1999 between 0800 and 1600 U T C . 33 The modeled values are taken at the grid point closest to the observation site. It is obvious that there is a large scatter but also that there is no clear systematic under- or overestimation. To investigate reasons for the large scatter in more detail, consider the observed and simulated turbulent sensible heat fluxes at three locations on the valley floor and west-facing slope shown in Fig. 2.10 (graphs for all the sites are shown in Appendix 3). 0.30 site A1 S. JE3 • B " 1 i i i i i i 1 i i I • • ' I • — ! |-4 6 8 10 12 11 16 18 20 time (UTC) Fig. 2.10. Observed (squares) and modeled (dashed line) surface sensible heat flux at site A l (a), site B (b), and site F2 (c) on 25 August 1999. Site locations are shown in Fig. 2. The shaded area indicates the range of modeled sensible heat fluxes at the nine model grid points surrounding the observation sites. 3 4 Observations are shown with the squares, modeled sensible heat are shown with the dashed line. The shaded area indicates the range of modeled sensible heat fluxes at nine grid points surrounding the observation site. It is clear that the range is large, which indicates that spatial variability of the modeled sensible heat flux is large. The variability in topographic parameters such as slope steepness and azimuth angle (or similarly, slope orientation) can explain this. Figure 2.11 shows the topography and azimuth angle around site F2 along with nine grid points surrounding the site. 16.0 16.4 16.8 x (km) Fig. 2.11. Contour lines of topography (solid lines) and slope azimuth angle (dashed lines) around surface station F2. The axes scale corresponds to that on Figs. 2.1 and 2.4. The squares illustrate the model grid with a horizontal resolution of 333 m. It can be seen that the difference in azimuth angle between the observation point and the closest grid point can be substantial. The same applies for the slope steepness, resulting in a very different timing and amount of radiation received at those two points. This, of course, can result in substantial differences in sensible heat flux. Differences in 35 surface characteristics such as roughness length and soil moisture could also cause some variability but are expected to be less significant than the effects due to differences in slope and azimuth angles. As an example, a calculation was made for the incoming shortwave radiation at two locations with a slope angle of 30° and an azimuth angle of 200° and 240°, respectively. The diurnal course of the shortwave radiation at the top of the atmosphere at those two sites is plotted in Fig. 2.12. 1400 1200 4 1000 4 800 600 400-200-- • — 200 azimuth x 240° azimuth /7^y f x • x I • > i x i >< i >< / X / x X X X X / x / x / x 8 12 \ X \ X \ x - r -16 20 24 Local Solar Time Fig. 2.12. Theoretical incoming solar radiation for a slope of 30° having azimuth angles of 200° (solid line) and 240° (dashed line). The figure shows that this difference in azimuth angle can cause a time shift of the onset of the radiation and maximum radiation of about one hour. The sun sets in the west (287°) which explains the larger radiation values just before sunset at the slope with the 240° azimuth angle than at the slope with the 200° azimuth angle. Figure 2.11 also implies that measurements on slopes in complex topography are often not representative of a larger area represented by a grid cell in a mesoscale numerical model, even with the high resolution used in the current simulation. Also, it is known that measurements of turbulent heat fluxes in complex terrain are subject to errors. This is revealed, for example, by the fact that the sum of the sensible and latent heat flux 36 from eddy-correlation measurements in complex terrain often differs from the available energy at the surface, i.e., the energy balance is often not closed (Bernhofer, 1992; Lee and Black, 1993; Panin et al., 1998, Twine et al., 2000). These errors in the measurements should also be taken into account when modeled surface heat fluxes are evaluated. Although these considerations make it difficult to assess the performance of the model, Fig. 2.9 and 2.10 show that the observed values of sensible heat flux lie within the range of values produced by the model and that there is no clear under- or overestimation of the sensible heat flux. The spatial distribution of the modeled surface sensible heat flux in the Riviera Valley is shown in Fig. 2.13 for three different times. The spatial variability is large and follows the spatial variability of the incoming solar radiation in complex topography (Whiteman, 1990). Sunlit slopes show larger sensible heat fluxes than shaded slopes so that there is a shift from relatively high values on the western sidewall in the morning to relatively high values on the eastern sidewall in the afternoon. Sensible heat fluxes at lower elevations can even be negative at noon. High soil moisture contents at lower elevations give rise to small Bowen ratios. It can also be seen in Fig. 2.13 that there is an increase of sensible heat flux with height, with the lowest values on the valley floor where the net radiation is also small (Matzinger et al., 2002). Fig. 2.10 shows that there is a general tendency for sensible heat fluxes to increase with height in the observations as well. Maximum observed and simulated sensible heat fluxes are of the order of 0.1, 0.2, and 0.3 K m s"1 for site A l (250 m asl), B (760 m asl), and F2 (2110 m asl), respectively. Notice that the investigation of sensible heat fluxes was done in terms of kinematic heat fluxes, rather than dynamic heat fluxes. This way, a possible additional error due to differences in pressure and temperature between observation and model is removed. 37 6 8 10 12 14 16 18 6 8 10 12 14 16 18 x (km) x (km) 6 8 10 12 14 16 18 x (km) Fig. 2.13. Spatial distribution of surface sensible heat flux at 0900 U T C (a), 1200 U T C (b), and 1500 U T C (c). Topography is indicated by contour lines drawn every 400 m. The axes scale corresponds to that on Figs. 2.1 and 2.4. 38 2.7. T E M P O R A L E V O L U T I O N OF T H E V A L L E Y A T M O S P H E R E A T B O S C O D I S O T T O 2.7.1. OBSERVATIONS Figure 2.14 shows observed vertical profiles of potential temperature, specific humidity, and horizontal wind vectors on 25 August 1999 at four selected times at Bosco di Sotto in the valley center. 295 300 305 310 Potential Temperature (K) 2 4 6 8 10 12 Specific Humidity (gkg1) Fig. 2.14. Vertical profiles of observed potential temperature (a), specific humidity (b), and horizontal wind vectors (c) at the valley center (Bosco di Sotto) at 0739 UTC, 0915 UTC, 1208 U T C , and 1508 U T C on 25 August 1999. The surface potential temperature measured at 1.5 m at the different times is indicated in (a) with symbols. The approximate height of the ridge is indicated with a grey rectangle. 39 The shaded rectangle indicates approximate ridge level height. At 0739 UTC, an inversion from the previous night is present and light and variable winds are observed. As was shown in Fig. 2.5 and 2.7, surface observations show a change from down-valley to up-valley winds at about 0800 UTC. At 0915 UTC, a C B L is observed up to 650 m asl capped by a weak inversion. Also shown in the figure are the potential temperatures observed at the surface. These indicate that the layer near the surface is unstably stratified during daytime even though the radiosonde data may imply a stable layer close to the surface at 0915, 1208, and 1508 UTC. The C B L grows only a few hundred meters between 0915 and 1208 UTC and reaches a height of around 800 m asl after that. The warming during the day is clearly not confined within the C B L . During the day, significant heating takes place up to about 1800 m asl, which is well below the average ridge height (-2300 m asl). There is some evidence of a shallow, more stable layer around 2 km. Above, the atmosphere is close to neutral. More heating occurs in the morning than in the afternoon and the heating is uniform in a major part of the valley atmosphere. Thus, roughly three layers can be identified in the lower troposphere: a rather well-mixed lower layer, a middle stable layer up to about ridge height which becomes more stable near its top, and an almost neutral layer aloft. In the conventional definition, the C B L would be the lowest well-mixed layer since it exhibits characteristics that are similar to those over flat terrain, i.e., well mixed with an inversion at its top. However, above the 'conventional C B L ' in a valley, the atmosphere is affected considerably by mesoscale flows and surface processes on the valley sidewalls on a timescale of a few hours or less. It may therefore be appropriate to include the effect of these flows and surface processes in a definition of the 'valley C B L ' . As will be shown later in this chapter, this new definition agrees better with the height up to which aerosols and pollutants are transported. In chapters 4 and 5, it is demonstrated that this also applies for the mountain C B L in general. In the remainder of this chapter, the term C B L will be used in its conventional context unless stated otherwise. Notice that the specific humidity profiles show a somewhat different structure than the potential temperature profiles. There is often no significant change in specific humidity at the top of the C B L as determined from the potential temperature profiles. The atmosphere seems to become more moist through the morning almost uniformly up 40 to heights above 1000 m, which is well above the top of the C B L . This behaviour is different from observations over flat terrain where, at the top of the C B L , the specific humidity usually shows a sudden decrease. Only in the layer between 1500 and 2000 m asl is there a decrease in specific humidity. There are also various sublayers with increased specific humidity in this layer. It is possible that these sublayers are caused by mountain venting processes as will be demonstrated later in this chapter with an L P D M simulation. It could be argued that the effect of these processes should be included in the definition of a 'valley C B L ' as was also implied above. A transition from upvalley flows to large-scale northwesterly flows takes place in a layer between 1500 and 2000 m asl during the day corresponding to the height of the 'valley C B L ' . Maximum upvalley flow speeds are on the order of 5 m s"1. At 1800 UTC (not shown), a ground-based inversion of about 100 m depth starts to form and temperatures in the valley atmosphere drop somewhat. The valley flow is still directed upvalley at this time and now reaches speeds of a little more than 5 ms"1. 2.7.2. M O D E L E V A L U A T I O N Figures 2.15 and 2.16 show a comparison between modeled and observed potential temperature, specific humidity and horizontal wind in the lower atmosphere. The potential temperature profiles show that the three-layer structure with the shallow mixed layers in the afternoon is well captured by the model, although it was unable to simulate the capping inversion just below ridge height. This may be due to the rather coarse vertical grid spacing (-200 m) at that elevation. However, it is noted also that the entire layer between roughly 2 and 3 km asl is warmer in the observations than in the model. A comparison of modeled and observed vertical temperature profiles at Milan and Payerne also show this, implying that this may be a feature unrelated to processes in the valley but is rather associated with synoptic scale processes. The E C M W F initialization may not have sufficiently captured this. 41 0 I I ol I 0 I I ol I 295 300 305 310 315 320 0 2 4 6 8 10 12 14 295 300 305 310 315 320 0 2 4 6 8 10 12 14 Potential Temperature (K) Specific Humidity (gkg'') Potential Temperature (K) Specific Humidity (gkg'') Fig. 2.15. Vertical profiles of observed and modeled potential temperature and specific humidity at Bosco di Sotto on 25 August 1999 at 0739 (a,b), 0915 (c,d), 1208 (e,f), and 1508 U T C (g,h). Model output is for 0700 (a,b), 0900 (c,d), 1200 (e,f), and 1500 U T C (g,h). The approximate height of the ridge is indicated with a grey rectangle. 3 -(a) (b) i £ ; " r ^ : 5 ms" 1 (d) 2 -to e 2: 1 1-0 -i > _ • ->, N co i \ i ! 1 " o -^ 1 0-> v » !=• \ \ v> 0 -1 \ Fig. 2.16. Vertical profiles of observed and modeled horizontal wind vectors for the same times as in Fig. 2.15. The approximate height of the ridge is indicated with a grey rectangle. 4 2 The surface based inversion at 0700 UTC is not well captured in the model. Even at earlier times, there was no pronounced surface-based inversion present. However, when temperatures at the lowest model level are extrapolated to the surface (2 m agl) using M - 0 similarity functions, a strong inversion is the result. Extrapolated surface temperatures are even about 2 K colder than those observed (see Fig. 2.6b). The more intense heating of the valley atmosphere between 0900 and 1200 UTC than between 1200 and 1500 UTC is well simulated. The modeled specific humidity (Fig. 2.15d) starts to decrease at lower heights in the valley atmosphere than the observed specific humidity. The tendency of the valley atmosphere to moisten with time is well captured, though. Modeled wind (Fig. 2.16) corresponds well at this location and the decrease in wind speed below ridge height and the increase aloft is well captured. Wind direction changes from upvalley to northwesterly at heights just below the average ridge height in both the model and the observations. 2.7.3. HEATING OF THE V A L L E Y ATMOSPHERE As the numerical model produces a thermodynamically balanced data set, the possibility exists to examine the relative contribution of the different terms in the temperature tendency equation. This equation can be written as: de de de ee e „de n — = -u v w— +—K — + R , dt dx dy dz dx( dx( / ' Ti " ~ni ly ' v which states that the total heating rate (I) is due to horizontal temperature advection (II), vertical temperature advection (III), turbulent diffusion (IV), and radiation (V). The tendency terms were extracted from the model for a grid point near Bosco di Sotto and averaged between 1200 and 1500 UTC. The magnitudes of these averaged tendency terms are shown as a function of height in Fig. 2.17. It can be seen that the behaviour of the various terms is rather complicated. Generally, however, horizontal and vertical advection are the dominant terms. In the layer between 1000 and 2000 m, the figure indicates that horizontal advection rather than subsidence contributes significantly to heating above the C B L . It is very well possible that the heating that occurs in the C B L 43 over the slope not only affects the region near the slope but also the central part of the valley. The heat that is produced along the slopes is transported horizontally by the mean flow in the valley to regions in the center of the valley above the C B L . As wil l be shown later, along-valley temperature gradients were significant and the wind field in the Riviera Valley is very complex, with disorganized winds in the center of the valley. This results in an inconsistent behaviour of the heat budget terms and difficulties in explaining this behaviour. It is clear though that subsidence is absent in a major part of the valley atmosphere, in contrast to many idealized simulations of the valley atmosphere and estimations of heat budget terms from observations. Hennemuth (1985), however, did a mass budget calculation in the Swiss Dischma valley and concluded there were rising motions in this deep valley, rather than sinking motions during daytime. In that case, she implied that heating came from turbulent heat flux divergence and not from horizontal advection. As can be seen in Fig. 2.17, there are regions within the valley atmosphere with significant heating due to turbulent diffusion but the term is not as large as the advection terms. Total —A— Radiation —A—Turbulent Diffusion —a— Horizontal Advection \—m— Vertical Advection Fig. 2.17. Vertical profiles of the various modeled terms in the heat budget equation (see text, eq. 2.1) for Bosco di Sotto, averaged between 1200 and 1500 UTC. The approximate height of the ridge is indicated with a grey rectangle. 44 In the following sections, aircraft data and model output will be presented to investigate the spatial structure of the valley atmosphere. First, the along-valley wind component will be studied, followed by an investigation of the temperature and turbulent kinetic energy structure. 2.8. S P A T I A L S T R U C T U R E OF A L O N G - V A L L E Y WIND Figure 2.18 shows aircraft observations and simulations of the along-valley wind component in a valley cross section at four selected time intervals. The location of the west-east cross section is indicated in Fig. 2.1. It can be seen that at the location of this particular cross section, ridge height is around 2000 m asl. It should be kept in mind though, that the average height of the ridges enclosing the Riviera Valley is around 2300 m asl. The interpolated cross sections from observations were created from 1 Hz aircraft data. Before contouring, the irregular data were interpolated to a regular grid using triangulation with linear interpolation. Some background information on the method can be found in Appendix 4.1. The density of the datapoints from which these figures were created is shown in Appendix 4.2. The observations in the early morning show that upvalley winds occur first at the surface, and are strongest at the eastern sidewall. During the day, the upvalley flow layer grows to a height of about 2 km in the early afternoon, while the wind speed maximum on the eastern side becomes more pronounced. Clearly, the upvalley flow structure is horizontally inhomogeneous, with generally stronger winds in the eastern part of the valley than in the western part. In the vertical direction, there are regions in the valley where a double maximum in the along-valley wind component can be seen, a maximum near the surface and one higher up around 1 km asl. Maximum wind speeds increase during the day and reach up to about 7 m s"1 in the afternoon. On the western side of the valley, there is a small region with downvalley directed flows in the observations. These may be related to the existence of a recirculation pattern as was implied from the surface 45 wind field in Fig. 2.8. The layer with upvalley flow corresponds in depth to the layer that was heated in the valley atmosphere, which was called the 'valley C B L ' (see Fig. 2.14). west (a) 0706-0742 UTC east west east (b) 0700 UTC 3 h '' -4000 (c) 0908-0919 UTC 2000 0 2000 horizontal distance (m) -2000 0 2000 horizontal distance (m) 4000 S _ ' r- \ -c- <=—T -4000 e) 1118-1218 UTC 2000 0 2000 horizontal distance (m) -4000 (f) 1200 UTC -2000 0 2000 horizontal distance (m) -4000 -2000 0 2000 horizontal distance (m) 4000 -4000 -2000 0 2000 horizontal distance (m) 4000 (g) 1330-1342 UTC (h) 1400 UTC -2000 0 2000 horizontal distance (m) -4000 -2000 0 2000 horizontal distance (m) Fig. 2.18. Interpolated cross sections of the along-valley wind component (m s*1; up-valley is positive) from aircraft data between 0706 and 0742 (a), 0908 and 0919 (c), 1118 and 1218 (e), 1330 and 1342 (g) and from model output at 0700 (b), 0900 (d), 1200 (f) and 1400 U T C (h). The location of the west-east cross section is depicted in Fig. 2.1. The horizontal distance is relative to Claro (see Fig. 2.1). 46 The onset of upvalley flows at higher elevations occurs one to two hours later in the model than in the observations. A delay in the onset of upvalley flows in the model was also present at Bosco di Sotto (section 2.4). The modeled cross sections fail to show a wind maximum on the eastern side of the valley in the morning hours. In the afternoon, however, the modeled wind maximum moved towards the eastern side of the valley and a good agreement is found with the observed wind maxima (cf. Fig. 2.18e and 2.18h). The modeled wind maximum is located closer to the surface, though. Also, modeled upvalley flows on the eastern side of the valley are about 2 m s"1 larger than observed. Downvalley flows on the western side of the valley in the afternoon are also seen in the modeled wind field. Figures 2.19a and 2.19b show observed upvalley wind cross sections in the along valley direction for the eastern and western parts of the valley between 1220 and 1330 UTC. The along-valley flight pattern of the aircraft (Fig. 2.2) shows that data are for the most part obtained over the sidewalls and may not be representative for the valley center. The data in Fig. 2.19a and 2.19b are representative for the area in the western and eastern rectangles in Fig. 2.20, respectively. This should be taken into account when comparing the observations with the simulations shown in Fig. 2.19c for 1300 UTC. The flow pattern is rather complicated and disorganized, both in the model and in the observations. Most consistent and also present at other times (not shown) are the relatively high wind speeds (up to 7 m s"1) close to the bifurcation zone at the valley entrance at the southern end of the cross section. Cross-valley wind components (not shown) are also largest in that area. Figs 2.19a and 2.19b indicate that wind speeds are generally larger on the eastern side of the valley than on the western side. This was not only the case at the time shown but also at other times. This implies that the inhomogeneous wind structure seen in Fig. 2.19 is not limited to this particular cross section but is also found at other locations along the valley. At some regions on the western side, flows are even directed downvalley as was also seen in the cross-valley sections before. 47 south (a) 1222-1330 UTC north 0 5000 horizontal distance (m) Fig. 2.19. Interpolated cross sections of the along-valley wind component (m s"1; up-valley is positive) from aircraft data between 1222 and 1330 UTC on the western (a) and eastern (b) sides of the valley and from model output at 1300 U T C (c). The location of the cross section in (a) and (b) is depicted in Fig. 2.20. The horizontal distance is relative to Claro (see Fig. 2.1). 0 5 10 15 20 x (km) Fig. 2.20. Topography of the Riviera Valley as in Fig. 2.1. The aircraft flew part of the time between 1222 and 1330 UTC on the western side and part of the time on the eastern side of the valley. A l l the measurements taken in the western rectangle are shown in Fig. 2.19a and all the measurements taken in the eastern rectangle are shown in Fig. 2.19b. 48 Reiter et al. (1983) observed a similar inhomogeneous behaviour of the wind structure in German and Austrian valleys with higher wind speeds on one side of the valley than on the other side. This behaviour was explained as a result of differential heating between the two sidewalls. It is questionable whether such an explanation applies to this case since the larger wind speeds were observed to be consistently present on the eastern side of the valley during the day and did not shift from one side of the valley to the other as in Reiter et al.'s (1983) observations. The shift of the modeled wind maximum, however, is consistent with Reiter's observations and can be explained by cross-valley differences in the radiation budgets during the day. In the morning hours the western valley sidewall is lit by the sun, resulting in a cross-valley component that accelerates the flow towards the western sidewall. The absence of this feature in the observations may indicate that the intensity of the cross-valley flows were somewhat overestimated by the model. From the observed and simulated wind fields, one can conclude that the wind field is inhomogeneous in the along- and cross-valley directions. It can be argued that these inhomogeneities induce divergence/convergence patterns and thus vertical motion fields that enhance mixing inside the valley atmosphere. Furthermore, the wind field differences across the valley may induce a recirculation pattern on the scale of the valley-width near the surface, resulting in stagnant flows at locations on the valley floor or even downvalley directed flows. These horizontal eddies (but on a larger horizontal scale) are known to exist in mountainous terrain (Whiteman, 2000). 2.9. SPATIAL STRUCTURE OF POTENTIAL TEMPERATURE Figure 2.21 shows aircraft observations and simulations of potential temperature in a valley cross section at four selected time intervals. The dotted lines indicate diagnosed C B L heights and will be discussed later. The observations show a cross-valley temperature structure that is rather homogeneous. The relatively stable layer below ridge height is clearly visible and is persistent during the day as was also seen in the vertical temperature profiles at site A l . In the observations, the region near the slopes is not entirely captured by the aircraft observations since the aircraft did not fly closer than 49 west (a) 0706-0742 UTC east west (b)0700 UTC -4000 (c) 0908-0919 UTC 2000 0 2000 horizontal distance (m) -2000 0 2000 horizontal distance (m) (d) 0900 UTC -4000 -2000 0 2000 horizontal distance (m) -2000 0 2000 horizontal distance (m) (e) 1118-1218 UTC (f) 1200 UTC (g) 1330-1342 UTC 1R 2 \ ^ ^ ^ , ———' -4000 -2000 0 i . i i i i 2000 i ' 4000 horizontal distance (m) (h) 1400 UTC -2000 0 2000 horizontal distance (m) east 4000 -2000 0 2000 4000 -4000 -2000 0 2000 4000 horizontal distance (rm) horizontal distance (m) Fig. 2.21. As in Fig. 2.18 but for potential temperature (Kelvins). The dotted line indicates the C B L height determined from the /^'-method. about 100 m to the slopes. In the afternoon cross sections, downcurving isentropes can be seen near the slopes, especially on the eastern slope which was lit by the sun at that time. Downcurving isentropes indicate warmer temperatures near the slope which are the driving mechanism for upslope flows as illustrated in Fig. 2.22 (e.g. Atkinson, 1981). Slope flows were observed at the surface stations on the valley sidewalls with speeds around 2 m s"1 as was shown in section 2.5. 50 6 + 2A9 9 +A9 9 4 Fig. 2.22. Sketch of isentrope deformations near a valley sidewall associated with the presence of an upslope flow. The thick solid line represents the valley and sidewall surfaces. The thin solid lines are isentropes. The dashed vector indicates the presence of an upslope flow. It is interesting to note that several studies have found a rather inhomogeneous temperature structure across a valley from which the existence of cross-valley circulations could be explained (Hewson and Gill, 1944; Hennemuth, 1986). Given the steepness of the Riviera Valley, it may be surprising that horizontal temperature gradients and cross-valley circulations were not clearly seen in the observations. The horizontal temperature structure shows more irregularities in the model than in the observations. This is partly due to the fact that the modeled fields include the temperature field directly along the slope. The irregularities in the modeled temperature field are also partly caused by numerical noise. This is particularly visible in the western part of the valley above ridge height in the morning. The numerical noise is caused by the treatment of horizontal diffusion in the model. Zangl (2002) recently pointed out these numerical artifacts in steep terrain and presented a method to reduce the numerical errors. Besides these irregularities, the horizontally homogeneous temperature structure and the stable layer below ridge height are fairly well modeled. The modeled isentropes are downcurved near the slopes which is indicative of the presence of an upslope flow as was illustrated above (Fig. 2.22). This deformation of the isentropes is not as clearly present in the observations. The stability in the layer below about 1500 m is less in the model than in the observations, especially in the morning hours. The model was not very successful in 51 simulating a stable ground-based inversion in the early morning hours as discussed before. Figures 2.23a and 2.23b show observed along-valley cross sections of potential temperature for the eastern and western parts of the valley between 1220 and 1330 UTC. Fig. 2.23c shows the modeled potential temperature cross section along the center of the valley at 1300 UTC. south north (a) 1222-1330 UTC horizontal distance (m) Fig. 2.23. As in Fig. 2.19 but for potential temperature (Kelvins). The dotted line indicates the C B L height determined by the fli-method. The along-valley temperature structure shows more irregularities in the isentropes than the cross-valley structure. The inhomogeneity is even more accentuated in the modeled cross section along the valley center, shown in Fig. 2.23c. Consequently, horizontal temperature gradients on a scale of a few kilometers can be significant. Notice that both in the model and in the observations, isentropes are curved down as well as up and do not show a consistent behaviour. The differences between the eastern and western side are not very large indicating a rather homogeneous behaviour of the temperature field in the cross-valley direction. Modeled fields show more spatial irregularities than 5 2 the observations and it is difficult to assess whether these structures are real, or are artifacts of the model. As explained before, some of the irregularities may be caused by the treatment of numerical diffusion in the numerical model. 2.10. S P A T I A L S T R U C T U R E OF C B L H E I G H T S The dotted line in Figs. 2.21 and 2.23 depicts the CBL height, determined with a i?/-method following Vogelezang and Holtslag (1996). This approach is described in Appendix 5 and has been recommended as one of the preferred methods for the determination of CBL heights (Seibert et al., 2000). The i?z'-method gives somewhat larger values for the CBL height than does a simple parcel method (Holzworth, 1964), primarily because it includes an 'excess' temperature at the surface and wind shear generated turbulence. A comparison of several methods for determining CBL heights from model output will be shown in chapter 4. For the determination of CBL heights from aircraft data with the Z?z'-method, surface observations of potential temperature, turbulent sensible heat flux, and friction velocity are also needed (see equation A2.1). These surface observations were taken from the tower at site A l . It can be seen that the CBL height does not show significant variability over the valley floor, either in the cross-valley or in the along-valley directions. In the course of the day, the CBL height increases from about 700 to 1300 m asl. This is higher than the CBL heights determined from visible inspection of radiosonde profiles that were shown in Fig. 2.14. It should be mentioned though that the determination of CBL height is rather sensitive to the surface variables that are needed as input, in particular the surface potential temperature. Especially in cases where no pronounced inversion is present to cap the CBL as in this case, there may be differences of a few hundred meters. The CBL height from the simulation shows somewhat more variability. This is particularly so in the along-valley direction where the model simulates a complex flow pattern, including a deceleration near the middle of the valley and at some locations even down-valley directed flow (as was shown in section 2.8). This produces a region with convergence and resulting vertical motions which have an impact on the temperature 53 structure and therefore the CBL height. CBL depths generally become smaller higher up on the slopes. This has also been observed in other studies (e.g. KoBmann et al., 1998; De Wekker, 1995), but some studies (e.g., Moll, 1935) have shown an increase of CBL depth up the slope. As surface variables vary in an unknown way along the sloping sidewalls, CBL heights from aircraft data were not determined there. It can be expected though that CBLs are deeper over the sunlit slopes than over the shaded slopes, as predicted by the model. CBL heights cannot, by definition, be determined from the 7?/-method if the surface sensible heat flux becomes negative. This explains why on the eastern sidewall in the early morning (Fig. 2.21b), and on the western sidewall in the afternoon (Fig. 2.2lh), CBL heights are not shown. 2.11. SPATIAL STRUCTURE OF TURBULENT KINETIC ENERGY The CBL is characterized by relatively high values of turbulent kinetic energy (TKE) and an attempt was made to evaluate observed and modeled TKE. TKE was calculated from aircraft data by analysing individual flight legs along the valley. Over each individual flight leg of about 10 km along the valley, 10 Hz data were detrended and high-pass-filtered (cut-off of 40 s, data provided by Andreas Wei gel, ETH). In this way TKE values were obtained at set heights for the western, central, and eastern parts of the valley. These locations are shown with asterisks in Fig. 2.24 along with the modeled TKE field at this cross section for 1300 UTC. The observed TKE values are plotted against height in Fig. 2.25 for the western, central, and eastern parts of the valley. Also shown in this figure are the vertical profiles of TKE from the mesoscale model. For the modeled TKE, values were sampled at the location of the asterisks in Fig. 2.24 and averaged in the along-valley direction over a distance of 10 km. This was done because the observed TKE values are for a flight path of about 10 km. It can be seen that observed TKE values are generally higher than modeled TKE values. The tendency of the observed TKE to be relatively large on the eastern side of the valley at elevations above 1000 m and relatively large on the western side of the valley at 54 elevations below 1000 m is well captured by the model. These relative maxima are related to enhanced levels of wind shear in those regions, as could be seen in Fig. 2.18. To assess the sensitivity of modeled TKE values to averaging distance, 5-km-averaged values in the along-valley direction are also shown in Fig. 2.25. It can be seen that the averaging distance does not change the profile shape. Fig. 2.24. Cross section of modeled T K E (m 2 s"2) at 1300 UTC. The location of the west-east cross section is depicted in Fig. 2.1. The asterisks denote the height of the along-valley flight legs. The 0.03 m 2 s"2 isoline and the C B L height calculated from the /^'-method are shown by the solid and dashed line, respectively. Fig. 2.25. Vertical profiles of T K E in the western (a) central (b) and eastern (c) part of the valley around 1300 UTC. Dashed lines connected with open circles are from along-valley aircraft legs; plusses and triangles are the model values averaged over 10 km and 5 km lines in the along-valley direction, respectively. 55 It should be noted that the magnitude of the observed TKE depends on the details of the high-pass filtering and detrending and that one should be careful when comparing TKE quantitatively. Furthermore, as Stull (1998) points out, the issue of whether TKE (and other turbulence statistics) can even be defined in complex terrain is unresolved and needs further research. With this in mind, the discussion given in this section should be considered preliminary. Even though there is considerable spatial variability in the TKE field, the CBL height as determined by the ifo'-mefhod follows the terrain rather uniformly as indicated by the dashed line in Fig. 2.24. The solid line is the TKE=0.03 m 2 s"2 contour line which is sometimes used as a cut-off value for CBL height (e.g. Cai and Steyn, 1993). It can be seen that these two methods for diagnosing CBL heights correspond fairly well. 2.12. IMPLICATION OF CBL MORPHOLOGY FOR PARTICLE DISPERSION It was previously shown that the potential temperature structure is rather homogenous across the valley. Consequently, CBL heights do not show large spatial variability over the valley floor, even though the wind field is complex and spatially inhomogeneous. To investigate the effect of the CBL morphology on dispersion characteristics in the valley, an LPDM simulation was made with HYP ACT whereby particles were released continuously at the entrance of the valley in a line across the valley floor at the surface (for a description of HYP ACT, see Appendix 1.2). The release of particles started at 0800 UTC. The resulting particle distribution across and along the valley at 1200 UTC is shown in Fig. 2.26 along with the observed and modeled CBL heights. It can be seen that there is a tendency for the particles to accumulate on the western sidewall. The surface wind field shown in Fig. 2.8 indicated that in the morning hours, when the eastern sidewall is still shaded, flow entering the Riviera Valley makes a bend towards the western sidewall. This cross-valley wind component causes particles to be transported to this sidewall, carried up the slope and subsequently transported out of the CBL. Further aloft, the synoptic northwesterly flow carries the particles in a down-valley direction. This process can also be seen on the along-valley cross section in Fig. 2.25. On the eastern side of the valley, particle concentrations stay relatively low since high wind speeds there cause particles to disperse very quickly. It is difficult to evaluate 56 these results with available observations. A cross section of aerosol distribution in the area is shown in Fig. 2.27. It can be seen that somewhat higher concentrations are found at elevations around 2 km on the western sidewall. This may indicate that particles indeed tend to accumulate on that side of the valley. The figure furthermore shows that significant aerosol concentrations are found above the C B L height. wes t e a s t s 1 -4000 -2000 0 2000 horizontal distance (m) 4000 a 1 -5000 0 5000 horizontal distance (m) 10000 Fig. 2.26. Particle distribution across (a) and along (b) the valley at 1200 U T C . The dashed line denotes the C B L height determined by the ^'-method from model output. The location of the west-east cross section is depicted in Fig. 2.1. 57 west east Fig. 2.27. Observed aerosol concentrations (number per cm3 for aerosols > 0.5 um) on a valley cross section at 1200 UTC. The location of the west-east cross section is depicted in Fig. 2.1. 2.13. C O N C L U S I O N S Aspects of CBL morphology in a deep, narrow valley were investigated with the MAP-Riviera data set and a mesocale numerical model during one fair weather day. The vertical temperature structure is characterized by three layers. The first layer from the surface is considered the 'conventional CBL' , the first and second layer together can be considered the 'valley CBL' . The third layer represents a transition zone between the valley atmosphere and the free atmosphere. Daytime heating occurs in both the first and second layers and is more intense in the morning than in the afternoon. Potential temperature structure was shown to be rather homogeneous in the cross-valley direction while wind structure showed a complex behaviour. A wind maximum was present on the eastern side of the valley during the entire day. This was also evident from cross sections taken in the along-valley direction. The wind and temperature structure in the along-valley direction show a disorganized behaviour. Conditions in the Riviera Valley were not conducive to an homogeneous upvalley flow. 58 Aircraft data, radiosonde data, and mean and turbulent surface data allowed a thorough evaluation of a mesoscale model in very steep and complex terrain. Among other things, the mesoscale model was initialized with a spatially heterogeneous soil moisture field, which was obtained off-line from a hydrological model. Simulations agree well with observed temperature and wind at the surface, and net radiation data. The comparison of the surface turbulent sensible heat fluxes exhibits large scatter. Differences in slope steepness and orientation between measurement sites and locations of grid points, and the inherent uncertainty of taking turbulence measurements in complex terrain can explain this scatter. Given these problems in evaluating turbulent flux measurements in complex terrain, it is encouraging that observed sensible heat fluxes lie within the range of modeled sensible heat fluxes that result from taking values from the nine grid points surrounding the measurement site. The three-layer structure in the vertical profile of potential temperature was well simulated although the model failed to reproduce an increased stability at the top of the second layer. An investigation of modeled temperature tendency terms at a grid point in the center of the valley revealed that heating of the atmosphere above the CBL in the afternoon was primarily due to horizontal advection. It is concluded that heat is transported horizontally from the CBL over the slopes by the mean flow to regions in the center of the valley. The spatially inhomogeneous wind field was also well reproduced by the model. The larger upvalley flows on the eastern slope than on the western slope are visible in the modeled wind field as well. However, in the model, the wind maximum shifts from the western to the eastern side of the valley during daytime, consistent with the shift in the incoming radiation on the corresponding sidewalls, while in the observations, the wind maximum remains on the eastern side of the valley. CBL heights diagnosed from simulations and observations correspond well over the valley floor. The maximum depth of the CBL is around 1000 m and does not show large spatial variability over the valley floor. There is good qualitative agreement in the vertical and horizontal structure of turbulent kinetic energy. An LPDM simulation shows that particles that are released on the valley floor are carried aloft by upslope flows to regions above the CBL but stay below average ridge 59 height. There is more transport of particles up the slope on the western sidewall because upslope flows on the eastern sidewall are overpowered by the strong upvalley flows there. Aerosol concentrations from aircraft measurements also show higher values along the western sidewall. Overall, the MAP-Riviera field study provided a data set which is particularly well-suited for the evaluation of a mesoscale model. This chapter shows that the mesoscale model used in this dissertation performed well over complex terrain. 60 3. CBL MORPHOLOGY NEAR A MOUNTAIN BASE: THE PACIFIC'93 FIELD STUDY 3.1. INTRODUCTION In comparison to the C B L structure in a valley, C B L morphology near a mountain base (or similarly, near the base of extensive mountain ridges) has received relatively little research attention. Some similarities may be expected since slope flows, for example, occur on both the slopes of individual mountains and on valley sidewalls. While it is generally accepted that processes occurring on the slopes have a profound influence on atmospheric processes in a valley (also see previous chapter), the effects of slope processes near a mountain base are relatively unknown and poorly documented. In this chapter, C B L morphology near a mountain base is investigated by focussing on a specific phenomenon, the so-called C B L height depression, using observations from Pacific'93 and a mesocale numerical model. Observational evidence for the phenomenon is presented and it is shown that this phenomenon also occurs in numerical simulations over idealized topography. Observations and simulations are then combined to investigate the mechanisms underlying the phenomenon. This is done by analyzing heating rates near the mountain base and away from it. The correspondence between the observed and modeled heating rates is encouraging, and allows an in-depth numerical investigation of the processes causing the phenomenon. This way, a better insight into atmospheric processes near a mountain base is obtained. 3.2. OBSERVATIONS OF A 'CBL HEIGHT DEPRESSION' In this section, characteristics of the C B L height depression will be presented using data from the Pacific'93 field campaign that was carried out in July and August 1993 in the Lower Fraser Valley of British Columbia, Canada. The general objective of Pacific'93 was to develop a better understanding of the mechanisms leading to high ground-level ozone concentrations in the Lower Fraser Valley (Steyn et al., 1997). Figure 3.1a shows a map of the Lower Fraser Valley, which is bounded by the Coast Mountains to the north and the Cascade Mountains to the southeast. The maximum height of the Coast Mountains is 1000 to 1400 m in this region and the terrain is characterized by steep 61 slopes (> 15 °). A complex coastline is present to the west and south. The densely populated valley floor extends more than 60 km inland, having a wide range of land use: urban areas including the city of Vancouver, agricultural and horticultural fields, parks and forests, bogs and swamps, rivers and lakes. The region of interest in this study is indicated by the rectangle in Fig. 3.1a, which is magnified in Fig. 3.1b. Surface based and airborne measurements were conducted under convective weather conditions with light to moderate synoptic winds, generally from the west (Pottier et al., 1997). A detailed description of the Pacific'93 field study and the various measurement systems was presented by Steyn et al. (1997) and Pottier et al. (1997). a) -123.00 -122.90 -122.80 -122.70 -122.60 -122.50 Longitude (°E) Fig. 3.1. Map of the Lower Fraser Valley of British Columbia. The region of interest in this study is indicated in (a) by the rectangle, which is magnified in (b). The lidar paths are indicated by L7 and T l while H and L denote the Harris Road tethered balloon station and the Langley radiosonde station, respectively. The horizontal dotted line near H denotes the approximate location where the aerosol layer starts to decrease in height in Fig. 3.2a. Contour intervals in (a) and (b) are 200 m and 100 m, respectively. Darker shades of grey represent higher elevations in (a). Elevation higher than 700 m is shaded in (b). 62 As noted in chapter 1, airborne lidar data have proven very useful in field campaigns to examine the spatial variability of the C B L height. As part of Pacific'93, a 1.064 pm downward looking Nd-Yag lidar was flown on an aircraft (Convair 580) at approximately 4200 m asl. The horizontal and vertical resolution of the lidar measurements was approximately 200 m, and 10 m, respectively. Lidar measurements are described in detail by Hoff et al. (1997). Background information about lidar measurements in general can be found in Appendix 9.1. Figure 3.2a and b show airborne lidar data in the flight tracks T l and L7, respectively, on 4 August 1993. The locations of these tracks are depicted in Fig. 3.1b. (a) -10 km Fig. 3.2. Flight tracks L7 (a) and T l (b) on 4 August 1993. The gray scale is proportional to lidar backscatter with light colors indicating high backscatter intensities. The depression in the A L height is indicated by the dotted line in (a). The arrow in (b) indicates a region with relatively shallow aerosol layer heights, corresponding to the location where flight tracks L7 and T l intersect. 63 Flight track T l was heading west from 1351 to 1406 PST while L7 was heading north from 1445 to 1458 PST. The shading on the cross sections represents the intensity of backscattered radiation. Aerosol backscatter is proportional to aerosol concentration (see Appendix 9.1) so that dark shading represents clean air and light shading represents aerosol-laden air. The terrain surface is indicated in black. The aerosol layer (AL) height is clearly defined by the boundary between high and low backscatter intensities. A feature that can be seen in both Figs. 3.2a and b is the small and irregularly spaced peaks indicating individual thermal plumes penetrating the inversion which caps the C B L , a well-known feature of the C B L (e.g., Stull, 1988). The backscatter field during track L7 (Fig. 3.2a) shows that the A L is typically 500 to 600 m deep in the southern part of this region of the Lower Fraser Valley. Further to the north, the A L first increases slightly in height but then, rather suddenly, starts to decrease at about 5 km from the base of the Coast Mountains, eventually reaching a height of only 300 m at the mountain base. The dotted line in Fig. 3.2a indicates this depression in the A L height. The backscatter field during flight track T l (Fig. 3.2b) shows A L heights of 700 to 800 m on the eastern and western part of the flight track with smaller A L heights in between. The region where the A L is smaller (indicated by the arrow) corresponds with the location where flight tracks L7 and T l intersect, just south of the steep southeast-facing slope of the Coast Mountains. Tethered balloon profiles up to about 1 km agl were obtained at Harris Road, roughly 10 km south of the base of the Coast Mountains as depicted by the ' H ' in Fig. 3.1b. Additional sounding data are available from Langley in the center of the valley, roughly 20 km south of Harris Road ( 'L ' in Fig. 3.1b). Harris Road is located near sea level while Langley's elevation is roughly 80 m asl. To facilitate further analysis of the sounding data, they are interpolated to regular height intervals of 25 m (with respect to mean sea level). Shown in Fig. 3.3 are temperature soundings at Harris Road at 1452 PST and Langley at 1600 PST. A C B L height of 500 to 600 m can be detected, comparable with what is derived from the lidar backscatter field. This provides evidence that the A L height and C B L height are of comparable magnitude over the plains, as mentioned in 64 chapter 1. This equivalence of C B L and A L heights is further supported by the thermal plumes visible in the lidar backscatter field. In chapter 4, it wil l be shown that the equivalence of C B L and A L heights that occurs over flat terrain breaks down over mountain ranges. -Harris Road 1452 PST -Langley 1600 PST Potential Temperature (K) Fig. 3.3. Vertical profiles of potential temperature at Harris Road and Langley on 4 August 1993 at 1452 and 1600 PST. C B L height is indicated with a horizontal dashed line. The black circle near the surface is the surface temperature measured at Harris Road at 1500 PST. In the Fraser Valley case, C B L heights at Langley and Harris Road were of comparable magnitude and the depression in C B L height appears as a localized feature occurring over a horizontal scale of roughly 5 km just north of Harris Road but before the terrain elevation starts to increase (see horizontal dotted line in Figure lb). It was noted by Hayden et al. (1997), who did a thorough analysis of C B L heights in the Lower Fraser Valley during Pacific'93, that C B L heights at Harris Road were generally lower than those further to the south. This could indicate that the horizontal scale of the depression in C B L height varies and is not necessarily as confined to the mountain base as is seen in the example above. Although only one example of the C B L height depression near a mountain base is presented, several other lidar cross sections on other days during Pacific'93 show a similar feature (Hoff et al., 1997). 65 Other observational evidence Relatively low C B L heights near a mountain base compared to the nearby plains were also observed during the TRACT field study (Transport of Ai r Pollutants over Complex Terrain) carried out in the Rhine Valley, Germany, in 1992 (Fiedler, 1992). The slopes characterizing the transition from the Rhine valley (approximately 50 km wide here) to the mountainous Black Forest region are less steep than those characterizing the transition from the Lower Fraser Valley to the Coast Mountains. Another difference is the absence of a coastline in the Rhine valley. Figure 3.4 shows potential temperature profiles along a northwest-southeast aircraft track in the Rhine Valley on 22 September 1992. In the northwestern part of the region, CBLs extend up to 1000 m asl while further to the southeast near the mountain base, CBLs only extend up to roughly 500 m. Except for foggy conditions in the morning hours in the Rhine valley, convective weather conditions prevailed during the main part of this day and winds in the C B L were light (< 5 m s"1). Comparing the depressions in the C B L heights in Pacific'93 and TRACT, the horizontal scale of the phenomenon appears larger in the Rhine Valley (note the different horizontal scales in Fig. 3.2a and 3.4). In the following sections, the depression of the C B L heights in the Pacific'93 field study will be examined further. Fig. 3.4. Vertical profiles of potential temperature on a cross section of the Rhine Valley at 1300 CEST on 22 September 1992 during the T R A C T field study (adapted from De Wekker (1995)). 66 3.3. MODEL ING OF A 'CBL HEIGHT DEPRESSION' An idealized numerical study of the plain-to-basin flow by De Wekker et al. (1998) indicated that modeled C B L heights near a mountain base show a pronounced depression in the afternoon (see their Fig. 10). Cai et al. (2000) also remarked on the presence of anomalously low C B L heights near a mountain base in their realistic three-dimensional simulations of the meteorological conditions during Pacific'93. The question now is whether the depressed C B L height also occurs in a numerical model i f it is initialized with the atmospheric conditions encountered on 4 August 1993. An attempt is made to keep the approach as simple as possible. Two-dimensional simulations rather than three-dimensional simulations are performed so that there are no complicating effects as a consequence of valley flows or flows around mountain ridges, for example. From the three-dimensional realistic simulation in chapter 2, it became clear that C B L morphology in mountainous terrain can become very complex and that isolating the processes associated with this morphology may be difficult or, i f nonlinear interactions occur between these processes, impossible. Idealized two-dimensional simulations can help to reduce the complexity and facilitate the isolation of atmospheric processes producing a certain phenomenon. The idea is that i f simple two-dimensional simulations are able to simulate a depression in C B L heights, the model can be used further to investigate this phenomenon. 3.3.1. M O D E L SETUP The idealized two dimensional simulations are carried out with the Regional Atmospheric Modeling System (RAMS). Information about this mesoscale numerical model can be found in Appendix 1.1. The two-dimensional simulations are performed in a north-south cross section (resembling flight leg L7, see Fig. lb) with a horizontal grid spacing of 1 km. The vertical resolution used in the simulations ranges from 25 m near the surface to 100 m at 1000 m agl to 1000 m near the upper boundary. The model is initialized on 4 August 1993 at 1200 UTC (0400 PST). The initial temperature field is horizontally homogeneous and is based on a sounding at Langley (indicated by ' L ' in Fig. 3.1b) in the center of the Lower Fraser Valley on 4 August 1993, 67 at 0400 PST. Further initial conditions used in this study are calm winds (0.01 m s"1), invariant soil (sandy loam) and vegetation (short grass) properties, and a relative humidity of 20% that is constant with height. The volumetric soil moisture content is set to 0.3. An idealized triangular shaped mountain ridge surrounded by flat terrain is used for the simulations. The mountain height is 1 km and the simulations are carried out for slope angles varying between 3 and 12 degrees. To minimize numerical errors and to keep from changing the horizontal and vertical grid spacing, simulations with slopes steeper than 12 degrees are not carried out. See Appendix 1.3 for a detailed discussion of horizontal and vertical grid spacing issues. The lateral boundaries are set 50 km away from the mountain base. 3.3.2. C B L HEIGHTS F R O M A N U M E R I C A L M O D E L C B L heights are determined from simulations by analyzing vertical profiles of TKE. The C B L height is taken as the height where the T K E has dropped to a value of 0.03 mV2 following Cai and Steyn (1993). A comparison with other methods for the determination of the C B L height will be shown shortly. The C B L height as a function of the distance to the mountain base at 1200, 1500 and 1700 PST and as a function of slope steepness at 1600 PST is shown in Fig. 3.5a and 3.5b, respectively. Fig. 3.5a shows that the model produces CBLs over the plains far away from the mountain base that evolve gradually and eventually reach a height of about 600 m. This is of comparable magnitude to the C B L heights in the Lower Fraser Valley on 4 August (recall that the model was initialized with the 0400 PST temperature sounding at Langley on 4 August). Near the mountain base, C B L growth is significantly reduced. There, CBLs remain relatively low over the day and even decrease in height late in the afternoon. The horizontal extent of the depression increases as time progresses. As shown in Fig. 3.5b, the steep slope simulations produce a C B L that is about 200 m lower near the mountain base than the shallow slope simulations. Furthermore, the horizontal scale of the depression increases as slope steepness increases, although there still is a confined region near the mountain base where the depression is more pronounced on a small horizontal scale. 68 1000 800-600-< £ 400-CD I 200-0-( 1000-800-00 600-< 400-.55 OJ I 200-(b) Horizontal Distance (km) — • — 2.9degr. —o— 5.7 degr. —A— 11.3 degr. I l i m B W i i i i i i i i i i u m i i i I-^JJ i i I ^ A A A A A A A A A A A A A A A A A A A A A A A A A A / ^ ^ O O C C O O A / / 10 20 30 40 Horizontal Distance (km) 50 Fig. 3.5. Variation of C B L heights with horizontal distance from the mountain base as a function of time for a 11.5° slope (a) and as a function of slope steepness (b). C B L heights are determined by the T K E method. Figure 3.6 compares three different methods for determining C B L heights for 1600 PST. The .Kz'-mefhod is explained in Appendix 5. The parcel method determines the level of neutral buoyancy of an air parcel rising from the surface. No mechanical effects are included in the parcel method, which explains the lower C B L heights compared to those from the i?/-method. Both the C B L heights from the parcel and /?/-method are somewhat larger than C B L heights from the TKE method. A l l three methods show lower C B L heights near the mountain base, although the horizontal scale of the feature varies with the method. 69 X Horizontal Distance (km) Fig. 3.6. CBL heights on a cross valley section as determined by different methods using model output from 1700 PST. A simulation with a Lagrangian particle dispersion model (HYPACT, see Appendix 1.2) investigated whether the height of the particles is bounded by the C B L height and thus confined to relatively low heights near the mountain base. To this end particles were released continuously over the plains. As demonstrated in Fig. 3.7 for 1700 PST, particles indeed reach lower heights near the mountain base. 0 10 20 30 40 50 horizontal distance (km) Fig. 3.7. Particle distribution at 1500 PST. The CBL height determined by the TKE method from model output is indicated with the heavy black line. Particles were released continuously from the surface. 70 Comparing the simulations with the lidar observations, it can be concluded that the simple two-dimensional idealized simulations capture the observed phenomenon to some extent, even though the observed depression in C B L height is more pronounced than the modeled one. This is an encouraging result that stimulates further use of both observational data and the numerical model to investigate the phenomenon. 3.4. COMPARISON BETWEEN OBSERVATIONS AND NUMERICAL MODEL In the previous two sections, evidence was provided of the presence of depressed C B L heights near a mountain base from observations and numerical modeling. Additional analyses using Pacific'93 data will now be presented in an attempt to clarify the mechanisms that produce this depression and to determine whether the phenomenon is produced by similar mechanisms in the model and in the observations. Unfortunately, temperature soundings are not available at the location where the depression in C B L heights occurred (see Fig. 3.2a). Harris Road is located just south of the observed depression (see horizontal dotted line in Fig. 3.1b). However, boundary-layer processes occurring at Harris Road can be expected to be affected by the proximity of the mountains in a similar though less enhanced way than in the region directly adjacent to the mountain base. To isolate the mountain proximity effect, the following approach is taken. First, observed vertical profiles of potential temperature at Harris Road and Langley are used to calculate the vertical profiles of heating rate at both locations between approximately 1000 and 1300 PST and between 1300 and 1600 PST. Then the heating rate at Langley between 1000 and 1300 PST is subtracted from the heating rate at Harris Road between 1000 and 1300 PST. The same is done for the time period between 1300 and 1600 PST. The result of this analysis is seen in Fig. 3.8 for 1000-1300 (a, 'morning') and 1300-1600 (b, 'afternoon'). Clearly, the heating rate at Harris Road is larger than at Langley at virtually every height both in the morning and afternoon. The difference in heating rate is a minimum at about 300 m asl in the morning and 400 m asl in the afternoon (shown by the lower arrows). Above this minimum, the difference increases significantly and shows a maximum at 500 m asl in the morning and 600 m asl in the afternoon (shown by the upper arrows). Thus, the atmospheric heating rate near the 71 mountain base is not only larger inside the C B L but also above, when compared to the heating rate far away from the mountain base (recall that the maximum C B L height was 500-600 m asl at Harris Road on 4 August). (a) morning (b) afternoon 800 600 ^ < E 400 A X 200 -0.0002 0.0000 Ks' 1 0.0002 0.0004 -0.0002 0.0000 K s 1 0.0002 0.0004 Fig. 3.8. Differences in heating rates as a function of height between Harris Road and Langley, averaged between roughly 1000 and 1300 (a, 'morning'), and 1300 and 1600 PST (b, 'afternoon') for 4 August 1993. Arrows indicate the minimum and maximum differences in heating rates. Error bars indicate a measurement error of 1 K. In an attempt to understand this behavior, the model output is analyzed in a similar way. The approach is depicted in Fig. 3.9. The heating rates in a column of air near the mountain base, column B, and a column of air near the edge of the modeling domain, column A , are compared. Column averaged values are obtained by averaging over a horizontal distance of 5 km. Subsequently, the heating rates in column A are subtracted from the heating rates in column B for the same time periods as the observations, that is, between 1000 and 1300 PST and between 1300 and 1600 PST. A 50 km Fig. 3.9. Idealized topography used in the model simulations showing the columns A and B whose heat budget terms were compared. A slope steepness of 11.5° is used in the simulation. 72 The result of this analysis is presented in Fig. 3.10 as a function of height. It can be seen that there is a good qualitative agreement between the model and the observations. Noticeable in the vertical profile of total heating rate are the minimum and maximum (indicated by arrows) that were also seen in the observations. 0.0004 Fig. 3.10. Differences in heating rates as a function of height between the atmospheric columns B and A (depicted in Fig. 3.9), as averaged between 1000 and 1300 (a, 'morning'), and 1300 and 1600 PST (b, 'afternoon'). Arrows indicate the minimum and maximum differences in heating rates. Several points should be considered when comparing the results of the idealized numerical simulations with the observations. For example, the larger values of the heating rates in the C B L at Harris Road might be explained by differences in surface properties, which are not considered in the numerical simulations. A simple calculation indicates that the sensible heat flux would have to be more than ~50 Wm' larger at Harris Road continuously to account for the difference in heating rate in the C B L . Given the similar surface properties between Harris Road and Langley, it is unlikely that the difference in heating rate can be attributed to this effect. Three-dimensional channeling effects, valley flow effects, and sea breeze effects are also not accounted for in the idealized simulations and the simple topography used in the simulations is only a very rough approximation of the real topography. Despite these considerations the agreement is encouraging and implies that the idealized two-dimensional model simulations capture some important process affecting C B L growth near a mountain base. 73 Since the numerical model produces a thermodynamically balanced data set, it is possible to examine the relative contributions of the different terms in the temperature tendency equation. This equation can be written as: 60 80 60 80 8 „ 80 n — = -u v w— + — K — + R , dt dx 8y dz 8xt 8xt i " Ti TTi ~w ' v which states that the total heating rate (T) is due to horizontal temperature advection (II), vertical temperature advection (III), turbulent diffusion (IV), and radiation (V). The contribution of these processes to the total heating rate is shown for the afternoon case in Fig. 3.11. Total —V— Radiation —A— Turbulent Diffusion —o— Horizontal Advection —•— Vertical Advection Fig. 3.11. Observed differences in heating rates as a function of height between Harris Road and Langley, averaged between roughly 1300 and 1600 PST for 4 August 1993 (a). The corresponding modeled difference in heating rate as a function of height is shown by the bold line in (b). The contribution of the individual terms to the total heating rate is also shown. Arrows indicate the corresponding locations of minimum and maximum differences in heating rates. Error bars in (a) indicate a measurement error of 1 K. Near the mountain base, heating due to turbulent diffusion is larger while cooling is enhanced by horizontal temperature advection. This advective cooling is the result of a thermally-driven plain-to-mountain flow (De Wekker et al., 1998). Aloft, entrainment processes cause a net cooling by turbulent diffusion, while there is more warming due to 74 horizontal and vertical temperature advection near the mountain base. These two advection terms are both large where the maximum occurs at 500 m. Thus, the model results suggest that there is enhanced vertical and horizontal advection of warm air near the mountain base, increasing the heating rate above the CBL. Modeled wind fields imply that the thermally driven upslope flows establish this advection. As an example, the modeled wind field at 1500 PST is shown in Fig. 3.12. The upslope flows seem to affect the wind field in a large part of the modeling domain and are associated with significant sinking motions near the mountain base. 0 10 20 30 40 50 horizontal distance (km) Fig. 3.12. Cross section of potential temperature and wind field at 1500 PST. The C B L height determined by T K E method from model output is indicated with the heavy black line If the depression in C B L height near a mountain base is established by sinking motions associated with the upslope flows, then the phenomenon might only occur along mountain bases with strongly developed upslope flows. The mountain base where the phenomenon was observed on 4 August is indeed characterized by strong upslope flows, as indicated by the presence of vigorous mountain venting. This can be seen in Fig. 3.2 and is further demonstrated by Rucker et al. (1998) who describe mountain venting characteristics during Pacific'93. 75 It was discussed in the previous chapter that sinking motions associated with upslope flows have also been shown to play an important role in inversion breakup and the subsequent development of the C B L in narrow mountain valleys (Whiteman, 1990). Until now, the importance of these sinking motions in the vicinity of mountain bases in very wide valleys or mountain bases of single ridges has not been considered. The subsidence in the narrow valleys has been attributed to mass compensation for upslope flows that transport mass out of the valley. These compensatory flows certainly also happen in the model and are associated with the depressed C B L height. The reduced C B L height at the mountain base might be enhanced by subsidence generated there by horizontal wind divergence at the interface between upslope flows over the slopes and weak horizontal winds over the adjacent plain. Wind- and temperature data in a much higher spatial resolution than obtained in Pacific'93 is needed to investigate this. The lack of data also hinders the examination of other processes that may contribute to the spatial variability of the C B L height in regions of complex terrain. Examples include sea breeze and valley wind effects, and mechanical effects generated as air flows over and around mountain ranges. The same analysis as presented above was carried out for 1 August 1993 and is shown in Fig. 3.13. < 1000 800 600 X 400 200 1000 800 J 600 J 400 2004 -0.0002 0.0000 0.0002 0.0004 K s 1 - T o t a l - Radiat ion -Turbulent Diffusion - Horizontal Advect ion -Ver t i ca l Advect ion -0.0002 0.0000 0.0002 0.0004 K s 1 Fig. 3.13. As Fig. 3.11 but for 1 August 1993. 7 6 The agreement between the observations and numerical model results is not as good with regard to the heights where minimum and maximum differences in heating rates occur. The overall shape of the vertical profile, however, corresponds well. Interestingly, CBLs were deeper on this day and, correspondingly, the maximum heating rate difference was found at higher elevations than on 4 August. The enhanced heating below the mountain height that was present on 4 August cannot be seen clearly on 1 August when atmospheric stability was lower. The C B L height depression was not observed on this day, implying that ambient stability and the C B L height relative to the average mountain height play a role. 3.5. CONCLUSIONS In this chapter, C B L morphology near a mountain base was investigated using data from the Pacific'93 field study and a mesoscale numerical model. In particular, it was shown from observations and mesoscale numerical simulations that the proximity of a mountain base sometimes affects the C B L morphology in such a way that a C B L height depression is produced near a mountain base. Idealized simulations with a single mountain ridge suggest that the depression evolves in time and is most pronounced in the late afternoon. Also, the depression becomes larger as slope steepness increases. Observations and modeling results show that the atmosphere near the mountain base is heated more intensely, not only inside the C B L but also above. A n analysis of the different terms in the temperature tendency equation indicates that vertical and horizontal advection of warm air, associated with the thermally driven circulation along the mountain slope, play a role in the enhanced heating aloft. Factors that affect this heating include the ambient stability and the height of the C B L relative to mountain height. The enhanced heating aloft and modeled sinking motions near the mountain base are associated with the C B L height depression. These modeling results suggest that this phenomenon only occurs near mountain slopes with a well-developed slope wind system. 77 4. C B L M O R P H O L O G Y A N D A E R O S O L L A Y E R S T R U C T U R E O V E R A M O U N T A I N R A N G E In the previous chapters, C B L structure and morphology over two archetypal topographic units was investigated. It was shown, among other things, that C B L heights are rather uniform over the valley floor and that a pronounced depression in C B L height can sometimes be present near a mountain base. In this chapter, C B L morphology over a mountain range which encompasses a multitude of archetypal topographic units, will be investigated. 4.1. I N T R O D U C T I O N During the S T A A A R T E '97 field study, a downlooking lidar was carried aboard an aircraft that flew over the European Alps to investigate the spatial variability of C B L heights over mountainous terrain. Preliminary results from this field study were presented by Nyeki et al. (2000). As in previous studies, the A L height derived from the lidar data was assumed to be equal to the C B L height. Based on the lidar observations, it was concluded that C B L heights did not follow topography. Unfortunately, no observations of atmospheric temperature structure were available in the region of interest to confirm this conclusion. In this chapter, the observed A L behaviour during S T A A A R T E '97 is investigated in more detail with a three dimensional non-hydrostatic mesoscale numerical model and Lagrangian particle dispersion model. It is shown among other things that there is a considerable difference between A L height and C B L height for this case study. It is argued that processes causing this difference are common over mountainous terrain and that, therefore, the C B L height may not be the relevant height scale for air-pollution dispersion in mountainous terrain. From the results obtained with the mesoscale model and L P D M , a new interpretation of the lidar data is given, providing a more complete picture than can be determined from observations alone. 78 Before the simulations are presented and discussed, the investigation area and available lidar data are described, followed by an analysis of the spatial structure and temporal development of the A L height from lidar data. 4.2. D A T A Under the European Union's "Scientific Training and Access to Aircraft for Atmospheric Research Throughout Europe" (STAAARTE) project, an airborne field study was conducted in an area of approximately 0.5° latitude x 0.5° longitude around the Jungfraujoch high-alpine research station (JFJ, 46.55° N , 7.98° E; 3454 m asl) on 30 July 1997. The field study was carried out in support of an ongoing aerosol field campaign that was begun at the JFJ in July 1995 (Baltensperger et al., 1997). A nadir-pointing aerosol lidar (Nd-YAG laser, output wavelength X=532 nm) aboard the German Aerospace Research Establishment (DLR) Falcon 20 jet aircraft (Kiemle et al., 1995) obtained an aerosol structure data set with high temporal and spatial resolution below 5 km. The aircraft was flying at an altitude of 8 km asl while taking lidar measurements. The JFJ is located in the northern part of the Swiss Alps in the Bernese Oberland and is situated in a saddle between the Monch (4099 m) and Jungfrau (4158 m) peaks. South of the JFJ the Aletsch glacier dominates the topography. Topography around the JFJ is shown in Fig. 4.1. Lidar measurements were made with a morning and afternoon flight pattern over the JFJ. Flight patterns consisted of transects within a 0.5° latitude x 0.5° longitude domain oriented parallel and perpendicular to the regional mountain divide (NE-SW). A total of 35 flight legs were flown. Seventeen legs were flown in the morning between 0628 and 0925 UTC, and 18 legs were flown in the afternoon between 1247 and 1533 UTC, each leg lasting between two and eight minutes. The daylight period at the JFJ on 30 July 1997 was from 0407 UTC to 1902 UTC so that the first flights took place more than two hours after local sunrise. The horizontal and vertical resolution of the lidar measurements was approximately 100 m and 15 m, respectively. The NW-SE transect 79 makes a sharp topographical transition from the pre-alpine foothills in the Emmental region (up to -2000 m) through JFJ and the surrounding massif (3000-4000 m) and down over the Aletsch glacier towards the Rhone valley. To the north of JFJ, the topography drops very steeply into a valley, resulting in a height difference of nearly 3 km between JFJ and the outlet of the valley into the Swiss plateau over a distance of only 15 km. Lidar backscatter ratio cross sections of all 35 flight tracks are shown in Appendix 9.2. Several of these will be discussed in this chapter. Dashed lines in Fig. 4.1 depict the approximate location of these flight tracks. Surface and 500 mb weather maps for 30 July 1997 are shown in Figs. 4.2a and b, respectively. 7.6 7.8 8.0 8.2 8.4 Longitude (°E) Fig. 4.1. Topographic map of the experimental area. Contour lines are drawn every 1000 m, the darkest shade representing terrain over 3000 m asl. The Jungfraujoch station is depicted by the white asterisk. The approximate locations of the flight tracks perpendicular (07, 016, 024, 034) and parallel (P6, P23, P30) to the mountain divide are depicted by the dashed lines. The inner square represents the innermost grid in the mesoscale model run described in section 4.4. 80 Fig. 4.2. Surface (a) and 500 mb (b) weather maps from the Swiss Meteorological Institute for 1200 U T C 30 July 1997. (c) A V H R R satellite image at 1419 U T C (channel 2 image, 0.725-1.lOum). Satellite image courtesy of the University of Dundee (http://www.sat.dundee.ac.ukJ). 81 A high-pressure ridge from southern France to Scandinavia determined the weather situation in the investigation area. Synoptic winds were weak to moderate (5 to 10 m s"1 at 500 mb) and from northwesterly directions. The weather situation was characterized by cloudless skies over large parts of Europe, as can be seen in the satellite image for 1325 UTC on this day (Fig. 4.2c). Global radiation data from selected Swiss A N E T Z network stations in the investigation area are shown in Fig. 4.3. Even though the data show no clear evidence of clouds, it was reported by observers and is evident from photographs that convective clouds developed intermittently over the Alps in the afternoon. The clouds are apparent from the lidar data as will be shown later. Fig. 4.3. Global radiation at various surface meteorological stations in and around the investigation area. The stations are: Payerne (46.82°N, 6.95°E); Ulrichen (46.50°N, 8.32°E); Adelboden (46.50°N, 7.57°E); Grimsel-Hospiz (46.57°N, 8.33°E); and Jungfraujoch (46.55°N, 7.98°E). 82 4.3. OBSERVATIONS OF THE AEROSOL LAYER HEIGHT Figure 4.4 shows cross sections of lidar backscatter ratio (referred to as lidar cross section in the remainder of this chapter) along flight tracks P6, P23, and P30 parallel to and northwest of the mountain divide. Flight legs P6 and P23 are identical cross sections while flight leg P30 is located closer to the mountain divide (see Fig. 4.1). Figure 4.5 shows lidar cross sections along flight tracks 07, 016, 024, and 034 perpendicular to the mountain divide and almost identical to each other. Variations in the topographical cross sections are due to slight deviations in the flight path. The approximate locations of the tracks in Figs. 4.2 and 4.3 are depicted in Fig. 4.1, and some additional details are given in Table 4.1. Table 4.1. Details of the flight tracks shown in Fig. 4.1. The A L heights (which are leg-averaged values) will be discussed in section 4.5. Leg Orientation wrt Time (UTC) objective Subjective mountain divide A L height (m) A L height (m) P6 Parallel 0653 - 0656 2524 2608 07 Perpendicular 0702 - 0707 2449 2689 016 Perpendicular 0913 -0917 2701 2976 P23 Parallel 1308 - 1311 3483 3614 024 Perpendicular 1317-1322 3903 4001 P30 Parallel 1409- 1413 3987 4056 034 Perpendicular 1520- 1525 4077 4129 Shown in the lidar cross sections are the topography (in white) and the backscatter ratio as determined from the downlooking lidar. This backscatter ratio is defined as: 5=[5a+5m]/[5m] where 5a is the scattering due to aerosols and 5m the molecular scattering. 5 is affected by aerosol concentration, aerosol size, lidar wavelength X, and relative humidity. 5 is predominantly influenced by the accumulation mode aerosols having diameters between 0.1 and 1.0 pm (Schwiesow, 1984). On some occasions, 5 shows significant variability within the AL caused primarily by aerosol growth when relative humidity rises above 85% (Hanel, 1976) rather than to increased aerosol concentrations from natural or anthropogenic activity. Since relative aerosol concentrations inside the AL are not examined here, this does not affect the investigations 83 in this chapter. It is assumed that the transition between high and low backscatter ratios that is clearly visible in many lidar cross sections corresponds to the top of the A L . Latitude fN) 46.60 4865 0 10 20 horizontal distance (km) Latitude (°N) 46.55 46.80 46.65 10 20 30 horizontal distance (km) Latitude fN) 46.50 46.55 46.80 46.85 46.70 0 10 20 30 horizontal distance (km) Fig. 4.4. Cross sections of lidar backscatter ratio for flight track P6 (a), P23(b), and P30(c) at 0653, 1308, and 1409 UTC, respectively. The color scale is proportional to lidar backscatter ratio. 84 Latitude fN) 46.30 46.35 46.40 4645 46.50 4655 46.60 46.66 46.70 0 10 20 30 40 50 horizontal distance (km) Latitude fN) 46.35 46.40 46.45 46.50 46.55 46.60 46.65 0 10 20 30 40 50 horizontal distance (km) Latitude fN) 46.35 46.40 46.45 46.50 46.55 46.60 46.65 0 10 20 30 40 50 horizontal distance (km) Latitude fN) 46.35 46.40 46.45 46.50 46.55 46.60 46.65 0 10 20 30 40 horizontal distance (km) Fig. 4.5. Cross sections of lidar backscatter ratio for flight tracks 07 (a), 016 (b), 024 (c), and 034 (d) at 0702, 0913, 1317, and 1520 UTC, respectively. The color scale is proportional to lidar backscatter ratio. The Rhone Valley, Aletsch Glacier, and Jungfraujoch station (JFJ) are indicated in (a). As an example of the results from the semi-objective method of determining A L height, see the black dots in (d). 85 By 3 hours after sunrise (-0700 UTC), the A L reaches up to about 2.5 km asl northwest of the mountain divide (Fig. 4.4a and Fig 4.5a, north of 46.55°N). Southeast of the mountain divide, where the general terrain elevation is higher, the A L reaches a height of approximately 3000 m. In the morning hours, the JFJ, seen in Fig 4.5 as the high peak at about 46.55° N , is above the top of the A L . Aerosol measurements taken at JFJ, shown later in this chapter, confirm this (also see Nyeki et al., 2000). Between 0702 and 0913 UTC, the A L has grown only slightly. Most of the A L growth seems to occur between 0913 and 1317 UTC. The difference in A L height between 0913 and 1317 UTC is about 1 km on both sides of the mountain divide and the maximum A L height reaches slightly above 4000 m asl (Fig. 4.5c). The lower A L heights northwest of the mountain divide are still visible. The difference in A L height between the northwest and southeast sides of the mountain divide is approximately 500 m. Between 1317 and 1520 UTC, the A L growth is small (cf. Figs. 4.5c, d). The difference in A L heights between 1308 and 1409 UTC in Figs 4.4b and c, partly reflects a temporal evolution. Additionally, it is caused by the generally higher terrain closer to the mountain divide. This can be seen more clearly from the difference in A L heights between 46.55 and 46.65° N in Figs. 4.5c and d. The presence of clouds on some afternoon cross sections is indicated by the white vertical lines. Most lidar cross sections show the presence of these clouds on southfacing slopes or above mountain ridges. Some cloud tops correspond with the top of the A L while others stay well below it. It is noted that the top of the A L is relatively smooth and diffuse compared to published A L heights from lidar cross sections over flat terrain (e.g. Kiemle et a l , 1995). Also recall that the A L heights shown on the lidar cross section from Pacific'93 in chapter 2.2 showed an irregular behaviour which indicated the presence of vigorous thermal eddies stirring air within the C B L . This is not so obvious on the lidar cross sections from STAAARTE. On Fig. 4.5c, a region with relatively low backscatter ratio can be seen between 46.50 and 46.55° N . Nyeki et al. (2000) associated this with the intrusion of clean air by katabatic flows over the Aletsch glacier (see Fig. 4.5a). Some observational data confirmed this but also indicated that this may not occur often and is just a temporary feature. Another explanation for the clean air intrusion may be the presence of mountain wave activity on the downwind (southeastern) side of the mountain, initiating vertical 86 downward motion in the region and associated transport of clean air from the free troposphere. Because of low northwesterly wind speeds and small static stability (with correspondingly high Froude numbers), however, this may not play an important role in this case. Generally, it can be seen that the A L height does not follow the individual ridges and valleys on a scale of a few kilometers. On a scale of 40 to 50 km, however, the cross sections in Fig. 4.5 indicate that the A L top does follow the topography to some extent. The examples shown here are representative of the other lidar cross sections, all of which are shown in Appendix 9.2. For reasons that will become clear later in this chapter, A L heights from the cross sections were also determined using a semi-objective method illustrated in Fig. 4.6. 0.6 0.8 1.0 1.2 1.4 1.6 Backscatter ratio (-) 1.8 2.0 Fig. 4.6. Illustration of the determination of the A L height with the semi-objective method. See text for details. 87 The figure shows a vertical profile of backscatter ratio values at 46.654 °N in leg 21 (see appendix 9.2). It can be seen clearly that the values in the lower layer of the atmosphere are much higher than the values higher up. The A L height determination method consisted of defining regions in the lower and upper parts of the atmosphere where the backscatter ratios are fairly uniform. The lower part goes from height B to 100 m from the surface. The first 100 m above the surface was excluded since backscatter ratio values near the surface were often erroneous. The upper part goes from height A to the top of the profile at 5000 m. The parameters A and B are chosen subjectively for each leg, but the chosen values have little effect on the A L height. The average value of the backscatter ratios found in the lower and upper part were taken as the threshold value. The location where this value was first exceeded starting from the top of the profile was defined as the A L height. After the method was applied to the data, further analysis included visual scanning of the A L heights to detect any failures of the algorithm. These failures generally occurred at locations where convective clouds were present. Not all 35 legs were suitable for determining the A L height. Some flight legs showed erroneous data that were not useful (e.g. leg 1, Appendix 9.2) and there were some cases in which an A L height was not well-defined (e.g. leg 3, Appendix 9.2). A more sophisticated scheme by Steyn et al. (1999) was also applied. However, this method was not able to handle the many backscatter profiles that were ill-defined. A L heights were also diagnosed subjectively by drawing a line at a visually determined A L top with a graphical interface program that recorded the heights. Subjectively determined A L heights were somewhat higher than semi-objectively determined C B L heights. Differences between the two analyses were small, however, with a maximum difference of about 100 m. The resulting values from the semi-objective method are shown by the black dots in Fig. 4.5d. The leg-averaged values for the flight legs in Figs. 4.4 and 4.5 are shown in Table 4.1. Temporal evolution of the leg-averaged A L heights is depicted in Fig. 4.7 for 7 out of the 17 morning flight tracks and 12 out of the 18 afternoon flight tracks. The other flight tracks were not well-suited for determination of A L height, as explained before. It is possible to make a distinction between the A L heights in the cross section parallel to the mountain divide (open circles) and perpendicular to the mountain divide (closed circles). There is no significant difference in average terrain elevation between parallel 88 and perpendicular cross sections, which explains the small difference in A L height between these two cross sections. It is evident that A L heights during the morning flight are somewhere between 2 and 3 km, while in the afternoon, they were between 3 and somewhat over 4 km. The relatively low values early in the morning and early in the afternoon correspond to flights over the terrain northwest of the mountain divide where terrain height is relatively low. Unfortunately, no measurements were made between 0900 and 1300 UTC, the time during which most of the A L growth seems to have occurred. O parallel to mountain divide • perpendicular to mountain divide 8 1 cP E oi O '» • • -1 * O _Q_ l o 12 ' 14 16 ' 18 Time (UTC) Fig. 4.7. Flight-leg-averaged A L heights for legs parallel to the mountain divide (open circles) and perpendicular to the mountain divide (closed circles). A mesoscale model was used to investigate the processes underlying the observed A L heights. The next section provides a brief description of the setup of the mesoscale model and an evaluation of the model with observational data. 4.4. M O D E L SETUP AND EVALUAT ION 4.4.1. M O D E L SETUP The mesoscale numerical model used is the Regional Atmospheric Modeling System (RAMS). More information on R A M S can be found in Appendix 1.1. Land use types were determined from 30 arc-second USGS data, based on A V H R R satellite 89 Table 4.2. Characteristics of the four gids used in S T A A A R T E modeling case study. N X and N Y are the number of grid points in the west-east, and north-south direction, respectively. GRID N X N Y Grid spacing W-E Size N-S Size AT(s) (km) (km) (km) 1 50 50 27 1350 1350 60 2 50 50 9 450 450 30 3 62 62 3 186 186 15 4 41 59 1 41 59 7.5 images. The presence of glaciers is accounted for in the land surface scheme of R A M S . The observed snowline in late summer at about 2500 m asl corresponds well with the USGS data where it depicts the land use type glacier/ice. Below 2500 m, land use consists largely of urban, agricultural and forested areas. The land use types according to USGS are shown in Appendix 6. Sandy loam was used as the soil type over the entire domain and the volumetric soil moisture content was set to 0.28. The three-dimensional simulations use two-way interactive, nested grids. The model domain consists of four nested grids with horizontal grid spacings of 27, 9, 3, and 1 km, respectively. The four grids are depicted in Fig. 4.8 and further information on grid characteristics are given in Table 4.2. 0 4 8 12 16 Longitude CE) Fig. 4.8. Map of Europe with the four R A M S grids represented by the trapezoids. The innermost grid is the area within the inner square in Fig. 4.1 and is also shown in Fig. 4.11. 90 The outermost grid covers central Europe including the Alps while the innermost grid is shown in more detail in Fig. 4.1. A l l four grids have 53 vertical levels, with a grid spacing increasing from 50 m near the surface to 160 m at 2000 m agl to 1000 m near the model top at about 16 km. Due to a staggered coordinate system, the lowest grid point is located about 25 m above ground level. The height levels up to 10 km are indicated in Fig. 4.9. The choice of the horizontal grid spacing of 1 km was based upon consideration of the dominant horizontal scales of topography in the innermost grid. A wavelet analysis (see Appendix 1.3) indicated that these scales were on the order of 5 km or larger. Given that features of 2Ax to 4Ax are resolved in the model (Pielke, 2002), where Ax is the horizontal grid spacing, a Ax of 1 km is an appropriate choice. Terrain heights were obtained from the USGS "TOPO30" data set. The simulations cover 36 hours, from 1200 UTC 29 July to 0000 UTC 30 July 1997. The five outermost lateral boundary points in the largest domain were nudged toward NCEP objective analysis fields and radiosonde data to allow changes in large-scale conditions to influence the model simulations. 4.4.2. E V A L U A T I O N OF M O D E L OUTPUT WITH OBSERVATIONS Model simulations were evaluated with observed temperature, humidity, and wind data from aircraft and selected radiosonde stations around the investigation area. The Payerne, Switzerland (46.82°N, 6.95°E, 490 m) and Milan, Italy (45.43°N, 9.28°E, 107 m) radiosonde stations were chosen since they are located close to the S T A A A R T E investigation area. With the northwesterly synoptic winds, Payerne is in the upwind direction from the Alps and Milan is located downwind of the Alps. Unfortunately, no soundings were taken in the investigation area except for an aircraft descent/ascent over the JFJ at around 0900 and 1500 UTC. A comparison between observations and simulations at Payerne is shown in Fig. 4.9. Temperatures are underestimated by roughly 1 -2 K by the model but, in general, it can be seen that the vertical structure corresponds well. Vertical structure of relative humidity, wind speed, and wind direction are also captured well. Wind speeds in the lowest 4000 m are small (< 5 m s"1) and somewhat variable between southwest and northwest. Larger wind speeds are observed and modeled aloft with a rather constant west-northwest wind direction. 91 Fig. 4.9. Observed (squares) and modeled (solid lines) vertical profiles of potential temperature (a), relative humidity (b), wind speed (c), and wind direction (d) for Payerne at 1200 U T C 30 July 1997. The vertical grid spacing in the model is indicated in (a) with the small horizontal lines. A comparison between simulated and observed potential temperature soundings using the Milan 1200 UTC radiosonde ascent and the ascent/descent of the aircraft above the JFJ at 0900 and 1500 UTC is shown in Fig. 4.10. Taking into account the difference in the nature of the observational data and model outputs, the correspondence of the stability in the free atmosphere is good. Differences in the potential temperature in the 92 free atmosphere between Milan and Payerne are minor both in the model output as in the observations, implying that synoptic scale temperature advection across the Alps was insignificant. This indicates that this day was undisturbed and that the influence of synoptic scale heating or cooling processes on C B L development do not need to be considered. 300 320 340 ' 300 320 340 300 320 340 Potential Temperature (K) Potential Temperature (K) Potential Temperature (K) Fig. 4.10. Observed (squares) and modeled (solid lines) vertical profiles of potential temperature for Milan at 1200 U T C (a), above the Jungfraujoch at 0900 UTC (b) and above the Jungfraujoch at 1500 U T C (c) on 30 July 1997. The good correspondence between observations and simulations encourages a further investigation of C B L morphology with the mesoscale numerical model. 4.5. L P D M S I M U L A T I O N A N D C O M P A R I S O N W I T H O B S E R V A T I O N S A Lagrangian particle dispersion model (LPDM) is used here to gain a better understanding of the observed aerosol structure over mountainous terrain. The L P D M used is the HYbrid Particle And Concentration Transport (HYPACT) model (Walko et al., 2001). A description of H Y P A C T is given in Appendix 1.2. The main objective of the L P D M simulation was to examine i f the A L height observed from the lidar data can be simulated by releasing particles at the surface. For this purpose, a simulation was made using the previously validated model output fields. Particles were emitted two meters above the ground in valley regions northwest and 93 southeast of the mountain range, simulating the main source of aerosols in this area. The surface elevation is well below 2000 m in those regions. This means that particles that are found above this height are being carried there by C B L growth or other processes. The two rectangular source regions are depicted in Fig. 4.11. A l l sources emitted 2000 particles per hour. The release of the particles started around sunrise at 0400 UTC. Particles undergo the same turbulent motions as the surrounding fluid; gravitational forces on mean particle motions are neglected. 0 10 20 30 40 x (km) Fig. 4.11. Topographic map of the innermost R A M S grid. Contour lines are drawn every 1000 m, the darkest shade representing terrain over 3000 m asl. The solid rectangles indicate the release areas of the particles. The solid line indicates the location of the cross section in Fig. 4.12. For the meaning of the dashed lines and the dotted rectangle, see the text. Figure 4.12 shows the distribution of the particles at 0900, 1200, and 1500 UTC in a six km wide band centered on both sides of the cross section across the mountain divide. This band is depicted in Fig. 4.11 by the dashed lines. The width of the band was chosen so that enough particles could be captured while keeping the variability of the terrain to a minimum. The cross sections of the topography on both edges of the band are also shown in Fig. 4.12a. It is obvious that there is significant variability in the 94 topography from one side of the band to the other. The cross section corresponds approximately to the cross sections shown in Fig. 4.5. This allows direct comparison of the results of the L P D M simulations with observations of the A L height, determined with the semi-objective method, which are also shown in Fig. 4.12. co co E (a) 0900 UTC y S ~? » \ *. '». ' .-.>J. 'i :5|- ,'-»»"» -30 -20 -10 0 horizontal distance (km) 10 20 00 CC (b) 1200 UTC -20 -10 0 horizontal distance (km) 5 _ 4 co E 3 c) 1500 UTC -30 -20 -10 0 10 horizontal distance (km) 20 Fig. 4.12. Cross section of the particle distribution at 0900 U T C (a), 1200 U T C (b), and 1500 U T C (c) in the area between the dashed lines in Fig. 4.11. The dashed line, solid line, and light dotted line in (a) are topography cross sections on the western side, center, and eastern side of this area, respectively (locations of cross sections are shown in Fig. 4.11). The heavy dotted lines in (a), (b), and (c) represent the A L height at 0913, 1317, and 1520 UTC, respectively (from Fig. 4.5). 95 It can be seen that at 0900 UTC, i.e., five hours after sunrise, a significant number of particles has already reached up to heights of somewhat over 3000 m southeast of the mountain divide and about 500 m lower northwest of the mountain divide. By 1200 UTC, the particle filled layer southeast of the mountain divide reaches up to somewhat over 4000 m and the difference between the height of the particle layers northwest and southeast of the mountain divide has increased to about 1000 m. Between 1200 and 1500 UTC, the differences are minor. Sensitivity tests showed that the location of the top of the particle layer was insensitive to the location of the source regions. Also, a simulation in which particles were released at homogeneously distributed surface locations everywhere in the domain resulted in a similar behaviour. Comparing the height of the particle filled layer with the observed A L height, it can be seen that the agreement is very good. This is especially the case southeast of the mountain divide. Northwest of the mountain divide, the modeled top of the particle layer stays a few hundred meters below the observed A L height. Overall, this is an encouraging result which allows us to further utilize the results of the L P D M simulation to illustrate the horizontal transport of particles released from the sources northwest and southeast of the mountain divide. Figure 4.13 shows the distribution of the particles in the innermost R A M S grid at 1530 UTC. Shown in red are the particles released northwest of the mountain divide, in yellow the particles released southeast of the mountain divide. Based on the evolution of the flow field in the area (not shown) and the trajectories of the particles, a schematic pattern of the major surface flow pattern in the afternoon can be constructed. The pattern is depicted by the solid arrows in Fig. 4.13. It can be seen that the flow field is mainly dominated by channeling of upvalley flows. Particles cross the mountain divide via upslope and upvalley flows which blow through the passes on either side of the mountain divide. Superpositioned on the thermally-driven flows is a synoptic northwesterly flow so that there is a general tendency for the particles to move towards the southeast. Note that upslope flows occur on both sides of the mountain divide. Some of these originate in the upvalley flow and are dynamically forced up the slope. Others are purely thermally driven. Particles located at the top of the particle layer in Fig. 4.12 are carried there by these flows. Thus, the surface flow pattern establishes a mean upward motion in the valleys and slopes on both 96 sides of the mountain divide which carries particles aloft. It is noticeable that as early as 0900 UTC (i.e., five hours after the release of the particles), particles are present up to a height of three km, indicating the efficiency of this venting process in the morning hours. Considerable aerosol concentrations were also present in the observations at that height at that time (see also Fig. 4.5b). This means that the aerosols there are not necessarily the remnant of a residual layer as was suggested by Nyeki et al. (2000). 0 10 20 30 40 x (km) Fig. 4.13. Horizontal distribution of particles at 1530 UTC. Red dots denote particles released in the rectangle north of JFJ, yellow dots denote particles released in the rectangle south of JFJ. For the location of the release areas, see Fig. 4.11. For the meaning of dashed and solid arrows, see the text. 97 The diurnal course of aerosol concentration for the JFJ under various synoptic flow conditions was calculated by Lugauer (1998). For synoptic winds from the west and northwest and under anticyclonic weather conditions similar to the situation during this case study, the diurnal course of the average aerosol concentration is shown in Fig. 4.14. 60 50 4 "o "-c CO Q - 30-o CD .a E 20 104 •west • northwest °'\ -40 CO o I 'E o N E • -20 60 24 Time (UTC) Fig. 4.14. Diurnal course of aerosol surface area per unit volume S a (um 2 cm"3) at the Jungfraujoch for 500 mbar synoptic flows from the west (circles) and northwest (squares), averaged over the seven year period 1991 to 1997 (Lugauer, 1998). The diamonds depict the diurnal range of particle numbers arriving inside the rectangle shown in Fig. 4.11. It can be seen that the aerosol concentration starts to increase around 1000 UTC, reaches a peak in the late afternoon, and then gradually decreases. Recall from Fig. 4.5 that the A L height had not reached JFJ by 0900. The aerosol concentration measurements corroborate this observation. The line connected by diamonds shows the diurnal course of the number of particles arriving in the box (indicated in Figs 4.11 and 4.13) of around 20 km centered around JFJ. The size of the box was based on two considerations. First, a significant number of particles needed to be captured. Second, the particles that are counted need to be close enough to the JFJ ( a few kilometers at most) so that it is realistic that they could 98 be sampled by a sensor positioned there. The first consideration would make a big box desirable while the second consideration restricts the size of the box. It can be seen that the timing of the increase corresponds well with the observations but that the peak occurs somewhat earlier. Given the relatively small number of particles and the fact that local conditions at the JFJ measurement site might have a considerable impact on the measurement of aerosol concentrations (Lugauer, 1998), the agreement is surprisingly good. The dashed arrows in Fig. 4.13 indicate paths through which particles arrive in the box. Upvalley and usplope flows play an important role in the transport of aerosols to JFJ. This is confirmed by observations of aerosols at JFJ for which the origin can be tracked (Lugauer, 1998). The good correspondence between the L P D M results and aerosol observations is encouraging and provides confidence in the simulation of the C B L structure by the mesoscale model. The C B L structure will now be investigated further by examining C B L heights from model simulations. 4.6. CBL HEIGHTS AND AL HEIGHTS 4.6.1. C B L HEIGHTS F R O M M O D E L SIMULATIONS C B L heights in the discussion that follows were determined with the i?/-method following Vogelezang and Holtslag (1996), a description of which is given in Appendix 5. For the sake of comparison, the parcel and T K E methods for diagnosing C B L heights were also applied. Figure 4.15 shows a comparison between the different methods for all the grid points in the innermost R A M S domain for 1200 UTC. In the parcel method, C B L height is determined as the equilibrium level of a hypothetical rising parcel of air representing a thermal. For this method, lower heights are found than for the .Kz'-method (Fig. 4.15a), because shear generated turbulence is neglected. In the T K E method, a cut-off value is usually taken to determine the C B L height. For example, Cai and Steyn (1993) found a cut-off value of 0.03 m 2 s"2 to work well for a case study in coastal complex terrain in which they compared C B L heights from model output with observations. Applying this cut-off value to the model output results in C B L heights that 99 are on average larger than those determined from the i?i'-method (Fig. 4.15b). The T K E method overestimates C B L height somewhat and the scatter is rather large for this case. Inspection of vertical profiles of T K E showed elevated maxima in T K E that caused the overestimation. They are found to be due to the presence of secondary wind circulations and related wind shear, as also mentioned by Gopalakrishnan et al. (2000). This can mean that determination of C B L heights with this method can sometimes become rather ambiguous i f these layers with wind shear are located above the C B L . In the remainder of this chapter, only C B L heights determined from the i?/-method will be shown. The Ri-method has been applied thoroughly to model simulations and observations and has been shown to work well under various conditions (Vogelezang and Holtslag, 1996; Seibert et al., 1998; 2000). D-f* , 1 , 1 , 1 , h l U U U - f , 1 , 1 , 1 . h 1000 2000 3000 4000 5000 1000 2000 3000 4000 5000 CBL height Ri-method (m) CBL height Ri-method (m) Fig. 4.15. C B L height from the parcel method plotted against C B L height from the /{/-method for all grid points in the innermost R A M S domain at 1200 U T C (a). C B L height from the T K E method plotted against C B L height from the ^/-method for all grid points in the innermost R A M S domain at 1200 U T C (b). 4.6.2. C B L HEIGHT A N D TOPOGRAPHY As discussed in chapter 1, the extent to which the C B L height is affected by topography varies from case to case. In general, morning CBLs are observed to be more terrain following than afternoon CBLs. Fig. 4.16 shows C B L heights from model output for all points in the innermost R A M S grid for 0800 and 1300 UTC as a function of 100 elevation. It can be seen that at 0800 UTC (four hours after sunrise), the C B L height is almost proportional to the elevation while at 1300 UTC, the effect of the terrain on the C B L height has diminished somewhat and the relationship exhibits more scatter. For a certain terrain elevation, a grid point can be located in a valley, on a slope, or on a ridge, resulting in different behaviours of the C B L growth during the day and therefore different C B L heights in the afternoon. This partly explains the large scatter. Furthermore, since C B L determination is most sensitive to atmospheric stability and surface sensible heat flux (Driedonks, 1982a; Seibert et al., 1998), variability of these parameters with terrain elevation causes variability of the C B L height. Observations of the C B L height in the mountainous terrain of southwestern Germany (Kalthoff et al., 1998) show a similar behaviour as in Fig. 4.16 (compare with their figures 12 and 14). In their observations, however, the effect of elevation on C B L height becomes even smaller than in the present study in the late afternoon. Fig. 4.16. C B L heights from the ^'-method in the innermost R A M S grid plotted against elevation for 0800 (a) and 1300 U T C (b). The values of the parameters T (={<Ttapo —0~h^jl <7/opo) and r (correlation coefficient) are also indicated. To investigate the extent to which the C B L height follows the topography, the parameters T= (<Jtgpo -o~h)/<Jlopo and r were calculated from all the grid points in the innermost R A M S grid, where atopo and <jh are the standard deviations of the topography and C B L height, respectively, and r is the correlation coefficient. For completely terrain 101 following behaviour, T= 0 and r = 1; for completely non-terrain-following behaviour, T=l and r = 0. The result is shown in Fig. 4.17. The parameter T is found to increase during the day with a maximum of 0.35 at 1300 UTC and a decrease afterwards. The correlation coefficient decreases from 0.95 near sunrise to about 0.78 at 1300 UTC and stays rather constant afterwards. T 6 8 10 12 14 16 18 Time (UTC) Fig. 4.17. Diurnal course of the parameters T and r as a function of time of day for all grid points in the innermost R A M S domain. 14 16 4.6.3. COMPARISON OF C B L HEIGHTS A N D L P D M RESULTS Figure 4.18 shows the C B L heights determined from the ^/-method along with the particle field at 0900, 1200, and 1500 UTC (particle field identical to Fig. 4.12). Note that at 0900 UTC, C B L heights could not be determined at some locations. At these locations, surface sensible heat fluxes were negative. Comparing the C B L heights with the particle field, it can be seen that already at 0900 UTC, a significant number of particles are found as much as 1-2 km above the C B L in the valley regions. As the C B L grows, this difference becomes smaller. Especially between 1200 UTC and 1500 UTC, the C B L grows faster than the layer with particles. Between 1400 and 1500 UTC, the maximum C B L height is reached, which is about the same time that the particles reach their maximum height. After this time, particle and C B L heights gradually decrease as convective processes become weaker and subsidence starts to dominate. 102 (a) 0900 UTC . • V:' - . . / , T = 0.005 r =0.95 0 -20 -10 0 10 20 horizontal distance (km) (b) 1200 UTC r =0.89 -30 -20 -10 0 10 20 horizontal distance (km) -20 -10 0 10 20 horizontal distance (km) Fig. 4.18. Cross section of the particle distribution at 0900 (a), 1200 (b), and 1500 U T C (c) in the area between the dashed lines in Fig. 4.10. C B L height from the /^/-method is indicated by a dashed line. Values of the parameters T and r are also indicated. Fig. 4.19 shows the potential temperature field and the C B L height from the Ri-method. Also shown are the wind vectors in this cross section. The horizontal length of the wind vector is proportional to the wind speed along the cross section (aligned northwest-southeast) while the vertical length is proportional to the vertical wind speed. 103 It can be seen that near the location of the Aletsch glacier (at -5 km) there is region where upvalley/upslope flows converge with the northwesterly synoptic flows. The associated rising motions establish a C B L height that is relatively deep in that area. In general, the wind field below the C B L height possesses significant vertical motions which are the result of upvalley and upslope flows in the area. Comparing Figs. 4.19 and 4.18b, it becomes clear that particles can be found in a stable atmosphere. -30 -20 -10 0 10 20 horizontal distance (km) Fig. 4.19. Cross section of potential temperature and windfield at 12 U T C with superimposed C B L heights as determined using the Ri-method. The cross section is the same as in Fig. 4.18. This all implies that mixing and transport processes in the investigation area are enhanced compared to those over flat and horizontally homogeneous terrain where the height up to which the particles reach is limited only by the C B L top. This latter situation can be investigated by performing a simulation in which the topography is removed and particles are released at exactly the same location as for the simulation with topography. The result is illustrated for 1200 UTC in Figure 4.20. Variability in the C B L height in this case is only due to differences in land surface characteristics. It is clear that so few particles are seen above the C B L height that one could use the maximum height of the particles to determine the C B L height, as is frequently done over flat terrain. Thus, Figs. 4.12 and 4.18 suggest that for mountainous terrain, the similarity of A L and C B L heights breaks down. 104 5 ^ 4 co £ 3 CD -30 -10 0 10 horizontal distance (km) 20 Fig. 4.20. Cross section of the particle distribution at 1200 U T C at the same location as in Fig. 4.16 but for a simulation with no topography. C B L height as determined using the /?j-method is indicated by a dashed line. 4.6.4. COMPARISON OF M O D E L E D C B L HEIGHT A N D OBSERVED A L HEIGHT A comparison of A L heights with C B L heights during the entire day is presented in Fig. 4.21. A L heights were determined semi-objectively for 7 out of the 17 morning flight tracks and 12 out of the 18 afternoon flight tracks. The black circles in Fig. 4.21 represent averages of all the observed heights in one flight track. C B L heights were 10 12 Time (UTC) 18 Fig. 4.21. Observed A L heights (closed circles) and C B L heights from model output. Squares are C B L heights from model output averaged over a 15 km wide band perpendicular to the mountain divide. Plusses are C B L heights from model output averaged over a 15 km wide band parallel to the mountain divide. 105 determined from R A M S output with the i?z'-method for each hour from 0800 to 1700 UTC. The squares and plusses in Fig. 4.21 represent averages that are representative for a flight track perpendicular and parallel to the mountain divide, respectively, so that a direct comparison with observed A L heights can be made. To this end, C B L heights were averaged over all the grid points within a band of about 15 km wide centered around a line (representing the flight track) perpendicular and parallel to the mountain divide. C B L heights perpendicular to the mountain divide are up to about 200 m greater than C B L heights parallel to the mountain divide, particularly in the afternoon. This is caused by a greater average elevation of the topography perpendicular to the mountain divide. Generally, however, it can be said that the differences are small as was also concluded for the difference in the observed A L heights between flight tracks parallel and perpendicular to the mountain divide (see Fig. 4.7). More importantly, it can be seen that, overall, A L heights are larger than C B L heights. Differences can be very significant, from a few hundred meters up to one kilometer. C B L heights from R A M S and A L heights from lidar data were also compared with C B L heights from the E C M W F model, representative for an area of about 50 x 50 km around JFJ. Results can be found in Appendix 7. It should be mentioned that over flat terrain, Coulter (1979) sometimes found A L heights from lidar data that were larger than C B L heights from temperature profiles. Differences in that study over flat terrain are not as large as in the current study, though. The discrepancy between A L and C B L heights in the studies over flat terrain is explained by the fact that the most energetic thermals can penetrate through the stable layer and thus carry aerosols above the C B L height. Obviously, mountainous terrain exerts a profound influence on aerosol distribution in the atmosphere. Various mechanisms play a role in transporting aerosols from the C B L to the free atmosphere over mountainous terrain. One of the mechanisms that has been given considerable attention in the literature is aerosol transport by slope flows (Edinger et al., 1972; Fast and Zhong, 1998; Fiedler et a l , 2000). Fig. 3.2a of chapter 3 and Rucker et al. (1998) show evidence of the existence of this mechanism in the Lower Fraser Valley in Canada. This mechanism is enhanced when slope flows converge above mountain crests, a process that has been called the 'chimney effect' (Fosberg, 1967; Lu and Turco, 1994), or when there is an additional sea breeze flow in 106 coastal terrain. (Wakimoto and McElroy, 1986, McKendry et al., 1997). Another mechanism that can become important over mountainous terrain is advective venting which occurs i f a wind vector crosses an inhomogeneous C B L top (KoBmann et al., 1999). These venting mechanisms have been investigated mainly in relation to elevated pollutant layers (McKendry and Lundgren, 2000). It is interesting that the structure of the aerosol distribution in the observations and the particle distribution do not indicate the existence of elevated layers during the S T A A A R T E field study. A mechanism that is not necessarily related to topography but can also be important in transporting aerosol aloft occurs when thermals reach lifting condensation level. This mechanism is called cloud venting (Ching et al., 1988). There was evidence for the existence of clouds in many of the 35 lidar cross sections in the present study, suggesting that cloud venting may be one of the possible mechanisms that transport aerosols above the C B L height. Thermals are known to be more vigorous over mountainous terrain (e.g. WMO, 1993) so that the cloud venting mechanism is facilitated. Note that in the model simulations in this study, cloud venting is not simulated. This indicates that advective venting and transport of aerosols aloft by mountain induced updrafts ('mountain venting') can establish, by themselves, an A L layer height that exceeds the C B L height. The A L height on the northern side of the mountain divide is several hundred meters above the maximum height of the particles transported by the L P D M (Fig. 4.12). This may be explained by cloud venting or by horizontal advection of aerosols from elsewhere. The difference in A L and C B L heights will be further discussed in the next chapter. For a brief discussion on the effects of a glacier and clouds on the C B L height in the S T A A A R T E investigation area, see Appendix 8. 4.7. C O N C L U S I O N S Observations from downlooking lidar on one day in the northern Alps show that A L heights are rather uniform over the mountain range and increase in height from 2-3 km in the morning to over 4 km in the afternoon. A L top behavior can be modeled well with an L P D M simulation that used previously evaluated model output and in which 107 particles were released from valleys in the area. The L P D M simulation also performs well in its prediction of the diurnal aerosol concentration at JFJ. C B L heights determined from model output show a terrain following behaviour in contrast to the uniform A L height. C B L heights become less terrain following in the afternoon in agreement with previous observations over mountainous terrain. An important conclusion from the combined observational and numerical study is that C B L heights are much lower than A L heights for this case. C B L growth alone cannot explain the modeled particle distribution. Mechanisms were suggested to explain aerosol transport to regions above the C B L height in mountainous terrain. Mountain venting processes are thought to play a significant role in the discrepancy between A L and C B L heights. These processes are common over mountainous terrain and it is therefore suggested that the discrepancy occurs more generally. The L P D M simulation was also useful in visualizing the surface wind fields in the area and showed that upvalley and upslope flows dominated the wind field. It is the vertical motion associated with these thermally driven flows that induces the mountain venting processes and the subsequent transport of particles above the C B L height. 108 5. D I S C U S S I O N In the previous chapters, C B L morphology was investigated in individual topographic settings. Some key findings will be summarized now and the results integrated into a conceptual picture of the afternoon C B L over mountainous terrain. Some data from a very recent field study will be shown to put the findings in a more general context. Finally, the definition of a 'mixing height' for mountainous terrain will be revisited. 5.1. C B L M O R P H O L O G Y I N A V A L L E Y Figure 5.1 summarizes some key findings of chapter 2. In a deep, narrow valley, it was found that horizontal potential temperature structure across the valley is rather homogeneous. Uniform C B L heights can occur while the horizontal wind structure is very inhomogeneous, with stronger winds on the eastern side of the valley than on the western side. In the along-valley direction (not shown here), horizontal wind structure is disorganized. Variable along-valley temperature gradients suggest that the winds in the Riviera Valley are not entirely thermally driven (on the scale of the Riviera Valley) and that the relatively strong upvalley flows on the eastern side of the valley have a partly dynamical origin. The Riviera Valley does not open into a plain but into a bifurcation zone, contributing to the complicated flow pattern. 2 km Valley CBL 2 km Fig. 5.1. Sketch of the C B L morphology on the afternoon of 25 August 1999. Dotted ellipses are isotachs of the along-valley wind. The dashed lines denote C B L heights. The lower dashed line is the conventional C B L height and the upper dashed line is the 'valley C B L height'. Arrows denote upslope flows. 109 Particles released at the surface are advected along the valley and tend to advect up the western sidewall easier than up the eastern sidewall. Particles on the eastern sidewall are advected mostly in the along-valley direction rather than up the slope. Particle and aerosol concentrations therefore have a tendency to be larger on the western sidewall than on the eastern sidewall. Also, significant amounts of particles and aerosols are found above the top of the C B L as determined from model simulations and observational data. They are carried there by upward motions in the valley atmosphere caused by upslope flows and convergence zones within the upvalley flow. The height up to which particles are transported in the valley atmosphere can be called the 'valley CBL ' height. This height, which is indicated by the upper dashed line in Fig. 5.1, would correspond with the A L height as determined by a downlooking lidar while the lower dashed line in Fig. 5.1 is the 'conventional' C B L height as determined from temperature soundings. The valley C B L height also corresponds to the height to which upvalley flows extend, while synoptic scale winds above are from the opposite direction. Clearly, heating from the valley floor and sidewalls affects the entire valley C B L , not just the conventional C B L . An idealized evolution of the vertical potential temperature structure on 25 August 1999 in the Riviera Valley is sketched in Fig. 5.2b. Fig. 5.2. Idealized vertical profiles of potential temperature for an inversion that breaks up by subsidence (a), and a schematic of potential temperature structure evolution in the Riviera Valley on 25 August 1999 (b). The approximate height of the ridge is indicated with a grey rectangle. (a) (b) z 110 Multiple layers are identified in the valley atmosphere. Shown is the suggested distinction between the conventional C B L and valley C B L in the afternoon. Shown in Fig. 5.2a is the typical subsidence-dominated evolution of the potential temperature structure in deep valleys observed by Whiteman (1982) in many valleys in Colorado. Whiteman concluded that subsidence played an important role in the inversion breakup, and attributed the subsidence to compensatory sinking motions caused by the removal of mass from the valley by upslope flows over the valley sidewalls. It is clear that the typical breakup pattern observed in the Colorado valleys is not observed in the Riviera Valley on 25 August 1999. In fact, a breakup of the valley inversion does not occur at all in the Riviera Valley - a common feature of Alpine valleys. For example, Freytag (1987) and Furger et al. (2000) published data from Alpine valleys in which a deep well-mixed C B L extending beyond the valley depth is not observed, in contrast to many case studies in Colorado valleys (Whiteman, 1982). Various reasons can be given for the deviation of Whiteman's conceptual model in the Riviera Valley such as a difference in scale of the valleys studied (horizontal and vertical scales of Colorado valleys are generally smaller) and a difference in climate setting, as will be discussed later. Furthermore, the complex wind field in the along-valley direction and the associated vertical motion field, which is not accounted for in the conceptual model, may contribute to the different behaviour. Analysis of the modeled tendency terms suggest that heating of the atmosphere above the C B L in the Riviera Valley was not due to subsiding motions but primarily to horizontal advection associated with an interaction between heating over the slopes and upvalley flows. The heat along the slopes is transported horizontally by the mean flow in the valley to regions in the center of the valley in a similar way as surface-released passive tracers are transported there. The L P D M simulation in chapter 2 showed that particles above the conventional C B L height become detached from the slope and are transported horizontally towards more central regions of the valley. Observations and model results indicate that the amount of sensible heat available for heating the slope boundary layer generally increases with elevation up to the snow line (Mannstein, 1989; Brehm, 1986; De Wekker et al., 1998; Noppel, 1999). Although the diabatic heating increases with elevation, the volume of air that can be heated in a horizontal layer of the valley atmosphere also increases with elevation. This can explain a 111 decreased heating rate with elevation in the valley C B L . It is clear that many processes acting simultaneously and over a range of spatial and temporal scales make it difficult to find an explanation of all the observed and modeled features in the Riviera Valley and to isolate the effect of individual factors on C B L structure and morphology. 5.2. CBL MORPHOLOGY NEAR A MOUNTAIN BASE Figure 5.3 summarizes the key finding of chapter 3. The proximity of a mountain base sometimes affects the C B L morphology in such a way that a C B L height depression forms near a mountain base. Also, the heating of the atmosphere near the mountain base is enhanced relative to the atmosphere away from the mountain base. Vertical and horizontal advection of warm air associated with the thermally driven circulation along the mountain slope play a role in an enhanced heating of the atmosphere above the C B L . The enhanced heating and relatively strong sinking motions near the mountain base are related to the modeled C B L height depression. These modeling results suggest that the presence of a well-developed slope wind system is required to produce this phenomenon. 1 km < • 5-10 km Fig. 5.3. Sketch of the depressed C B L height near a mountain base. The arrows denote vertical winds. The dashed line denotes the C B L height. The observed C B L height was somewhat lower than in the model simulations but there was good agreement in the 10-km horizontal scale of the feature. Even though there was qualitative agreement between the features seen in the observations and in the simulations, other mechanisms which are not represented by the idealized two-112 dimensional simulations may have caused the observed depression. In this case study near a mountain base, C B L heights derived from vertical profiles of potential temperature correspond well with A L heights derived from lidar data. 5.3. C B L M O R P H O L O G Y O V E R A M O U N T A I N R A N G E The picture that emerges from comparing the observational and modeling results in chapter 4 along with some possible mechanisms to explain the observed aerosol structure, is summarized in Fig. 5.4. The conceptual picture summarizes atmospheric processes for the late afternoon. The wavy solid line depicts the A L height that was found from the lidar data. The dashed line depicts the C B L height that was found from model output with the .Kz'-method. 0 1 0 2 0 3 0 4 0 5 0 horizontal distance (km) Fig. 5.4. Conceptual picture of the situation on the afternoon of 30 July 1997. h is the C B L height and h a the A L height. The depicted mechanisms are (1) mountain venting, (2) cloud venting, (3) advective venting, and (4) advection of aerosols from airmasses elsewhere. h a is referred to as the 'mountain CBL ' . 113 A L heights are relatively uniform and reach a maximum altitude of about 4 km in the afternoon. A L heights are somewhat higher in the higher-elevation southern part of the mountain range implying that the A L height follows the large-scale topography to some extent. The distribution of particles in the afternoon from an L P D M simulation shows a similar behaviour (see chapter 4). C B L heights are considerably lower than A L heights and show more spatial variability. C B L heights tend to follow the terrain more than A L heights although the extent to which the C B L height follows the terrain decreases during the day. Similarly as for the A L heights, average C B L heights are lower in the lower-elevation northern part of the mountain range. The question arises how a passive tracer is transported into the region between the C B L height and the A L height. The mechanisms that are suggested are 1) mountain venting, 2) cloud venting, 3) advective venting, and 4) horizontal advection of aerosols from airmasses upwind of the area of interest. The mechanisms are indicated in Fig. 5.4 and were partly discussed in chapter 4. Mountain venting and cloud venting can occur simultaneously when an orographically induced thermal reaches lifting condensation level. A l l of these mechanisms have been observed to occur in a variety of field studies. For an overview, see KoBmann et al. (1999). The L P D M simulation presented in chapter 4 suggests that a combination of mountain and advective venting processes (and C B L growth, of course) can explain to a large extent the observed aerosol structure. Particles that were released around sunrise were found at maximum heights comparable to the A L height. Advection of aerosols from elsewhere and cloud venting, mechanisms that were not simulated, are not necessarily needed to explain general A L height behaviour. The L P D M simulation also suggests that the A L grows more rapidly and attains its maximum height earlier than does the CBL. The A L height is rather constant during the afternoon while the C B L height continues to increase somewhat, although at a smaller rate than in the morning. Therefore, the difference between A L height and C B L height becomes smaller during the afternoon. The difference in the late afternoon is smallest in the southern half of the investigation area but still amounts to a few hundred meters, as indicated in Fig. 5.4. 114 A schematic of typical potential temperature profiles encountered in the northern and southern part of the investigation area in the afternoon is depicted in Fig. 5.5. The profiles were obtained by examining potential temperature profiles from model output and looking for significant inflection points (a modeled cross section of potential temperature against which these profiles can be compared was shown in chapter 4). 2 0 K Potential Temperature Fig. 5.5. Schematic of vertical potential temperature profiles in the S T A A A R T E investigation area south of JFJ (solid line) and north of JFJ (dashed line). The numbers correspond with the mechanisms indicated in Fig. 5.4. The subscripts V and 'n ' refer to south and north of JFJ, respectively. The thin arrows denote turbulent eddies. The diagonal hatching denotes the valley surface, han And h a s are referred to as the 'mountain C B L ' (see text). Three layers can be identified in both potential temperature profiles: a well-mixed C B L near the surface, a layer between the C B L and A L heights, and a layer above the A L height which is identical to the free atmosphere (i.e., that part of the atmosphere that is relatively unaffected by the underlying topography and whose characteristics are determined primarily by synoptic-scale weather). The lower atmosphere is clearly warmer in the southern part than in the northern part of the investigation area. The three layers are characterised by different stabilities, with an increasing stability with height in this case. The tops of the first and second layers (starting from the 115 surface) where stability starts to increase slightly (not shown in the figure) are the C B L and A L heights, respectively. These heights are somewhat greater in the southern half than in the northern half of the investigation area, consistent with the conceptual picture in Fig. 5.4. The venting mechanisms that were indicated in Fig. 5.4 are also indicated in Figure 5.5. The small curved arrows in the C B L are shown to emphasize the turbulent character of the C B L in comparison with the non-turbulent character of the layer between the C B L and A L heights. In the model output, T K E is negligible in that layer. One might argue that the layer between the C B L and A L heights is a deep entrainment layer extending up to the A L height. The fact that turbulence is negligible in the layer argues against this interpretation. Thus it is concluded once again, that the C B L height is lower than the A L height. Even in some sensitivity tests where sensible heat input from the surface was increased by decreasing the soil moisture content in the entire domain, the C B L height did not reach the A L height. Similar to the 'valley C B L ' in a valley, the layer up to the A L height over this mountain range may be termed 'mountain C B L ' . It is this entire layer that is affected by heating from the valley surface, sidewalls, and mountain ridges on a timescale of one to several hours. The conventional temperature structure of the C B L constitutes a large part of the mountain C B L but there are important differences as discussed above. So far, mainly the C B L structure in individual topographic settings has been considered. Next, an attempt will be made to integrate and generalize the findings. 5.4. TOWARDS AN INTEGRATIVE APPROACH TO CBL MORPHOLOGY IN MOUNTAINOUS TERRAIN The initial plan for the MAP-Riviera field study was to investigate the A L structure over this area with a downlooking lidar as was done during STAAARTE'97 . This would enable the integration and further investigation of the concepts that were previously developed for C B L structure and morphology in the individual topographic settings. Unfortunately, this plan was not carried out and no lidar data were obtained for the Riviera Valley. However, another field study called CHAPOP (Characterization of High Alpine Pollution Plumes) was performed recently (August 2001) in a valley just north of the 116 Riviera Valley. In this field study, an aircraft with a downlooking lidar was operated simultaneously with airborne measurements in a deep, narrow valley (Leventina valley). The operation of the downlooking lidar was not confined to this single valley but covered an extensive mountain range in which the Riviera Valley is embedded as well. A further investigation of the concepts developed for the individual topographic settings seems feasible with this data set. Therefore, preliminary results of this field study are shown next (data were provided by Dr. S. Nyeki and S. Henne, PSI). Figure 5.6a shows a lidar backscatter (wavelength 532 nm) cross section over the CHAPOP investigation area in the southern part of Switzerland on 28 August 2001, just north (~10 km) of the Riviera Valley. Fair weather conditions prevailed during the field study. Measurements were taken around 1335 UTC. Darker shades of blue represent higher backscatter ratios. The backscatter ratio is highest below 1.3 km and then again a few hundred meters below 4 km. In between, backscatter ratios are relatively low, implying low aerosol concentrations. The A L height is horizontally homogeneous at a height of about 4 km. The white vertical lines whose upper ends correspond with the top of the A L indicate the presence of clouds. Clearly, the A L height does not follow the topography. Figure 5.6b is a close-up of Fig. 5.6a for the region between 8.80° and 9.07° E, corresponding to a distance of 20 km. The 1052 nm lidar wavelength provides better aerosol structure detail than the 532 nm lidar wavelength used in S T A A A R T E . Yellow and red colors represent higher backscatter ratios. A clear distinction can be made between the top of an A L in the valley (at about 1 km) and the top of another A L above the valley (at about 4 km). This is in contrast with the observed aerosol structure during S T A A A R T E where no separation of aerosol layers was visible. Note, however, that the increase in backscatter ratio above about 2.5 km, is probably due to aerosol swelling and not to higher aerosol concentrations. It could therefore well be that aerosol concentrations are relatively constant between the top of the A L in the valley and the top of the A L aloft. This should be kept in mind when interpreting the backscatter ratios. Figure 5.6c shows the vertical profiles of potential temperature and relative humidity at 1400 UTC in the Leventina Valley (the valley on the left hand side of Fig. 5.6b). A schematic of the vertical temperature structure similar to Fig. 5.5 is shown in Fig. 5.7. 117 -+ • 20 km Fig. 5 . 6 . (a) Cross section of lidar backscatter ratio around 1335 UTC 28 August 2001 during CHAPOP. Lidar wavelength is 532 nm and the covered region is from 8.5° to 9.7°E. (b) Same as (a) but for a lidar wavelength of 1064 nm and between 8.8° and 9.07°E. (c) Vertical profiles of potential temperature (solid line) and mixing ratio (dotted line) at 1400 UTC 28 August 2001, in the Leventina Valley (the valley on the left hand side of Fig. 5.6b). A strong inversion near the surface caps the CBL. It is questionable whether the strong superadiabatic gradient near the surface is realistic or whether this is a measurement error. In any case, it is assumed that the layer adjacent to the surface is statically unstable. By comparing Figs. 5.6b and 5.6c (or 5.7), it can be seen that the high backscatter values near the surface are related to the presence of this CBL. However, a closer look reveals that the top of the AL in the valley, indicated by the wavy line in Fig. 5.7, is located closer to the top than to the bottom of the inversion that caps the CBL. Even though it is somewhat difficult to identify the height of the CBL from this single 118 sounding, this may also indicate that significant aerosol concentrations are present above the CBL height as was found in MAP-Riviera. These pollutants may have been carried there by venting processes. On the other hand, the high backscatter ratio in the valley may also represent pollutants that are trapped in the stable core remnants of the nocturnal stable layer. Further analysis is required to investigate this. Very clear in Fig. 5.6c is a jump in potential temperature and water vapor mixing ratio at about 4 km, that corresponds to the height of the elevated AL (and to the height of the highest cloud tops). The strong inversion at that height obviously prevents aerosols from being transported further aloft. The inversion is much stronger than in STAAARTE and could explain the more distinct top of the AL in CHAPOP (cf. Figs. 4 . 4 and 4 . 5 ) . As was noted in chapter 4 , AL heights were occasionally found to be rather diffuse during STAAARTE and could not be determined by an objective method. co 3 E XL CD ^ ml r ® h2 hal h 20 K Potential Tempera tu re Fig. 5.7. Schematic of vertical potential temperature profile in the Leventina Valley at 1400 U T C during CHAPOP. The numbers correspond with the mechanisms indicated in Fig. 5.4. The thin arrows denote turbulent eddies, h is the C B L height, h a , is the A L height in the valley, and h a 2 is the A L height over the mountains. The diagonal hatching denotes the valley surface. h a i and h a 2 could also be referred to as a 'valley C B L ' height and a 'mountain C B L ' height, respectively (see text). 1 1 9 The relatively low backscatter ratios between the top of the A L in the valley and about 3 km asl as well as the relatively low backscatter ratios over the slopes above the A L height, suggest that mountain venting processes that were induced in the deep and narrow valleys do not transport a significant number of aerosols above the mountain ridges. This implies that the elevated A L was generated mainly by the advection of aerosols from elsewhere. However, the presence of clouds obscures a significant part of the lidar cross section and transport of aerosols aloft may have occurred also through cloud and mountain venting. In that case, the elevated A L may represent an accumulation of aerosols vented out of individual valleys. Clearly more research is needed to explain the observed aerosol structure during CHAPOP. It is evident that preliminary CHAPOP observations have already provided an excellent means to bring together and evaluate some concepts that were developed in the context of the S T A A A R T E and MAP-Riviera studies. It is interesting to note that the L P D M simulation for the Riviera Valley in chapter 2 showed that most of the surface-released particles stayed below ridge height within the 'valley C B L ' . These results were supported by aircraft observations of aerosol concentration. An L P D M simulation for the S T A A A R T E investigation area however demonstrated that surface-released particles were found in large numbers at heights well above the mountain ridges in that area within the 'mountain C B L ' , in agreement with the aerosol lidar data. This could imply that mountain venting processes are much more intense in the valleys in the S T A A A R T E investigation area than in the Riviera Valley and surroundings. Of course, one should be careful with general statements like this based on a few case studies. From previous observations however, it is known that vigorous transport and mixing takes place in the Rhone Valley (Neininger and Liechti, 1984) which is situated in the southern part of the S T A A A R T E investigation area. Horizontal scales of valleys such as the Rhone Valley are also much larger than the scale of the Riviera Valley, resulting in differences in characteristics of C B L structure between these areas. Longer valleys, for example, are known to have better developed valley wind systems than shorter valleys such as the Riviera Valley (Wagner, 1938) and therefore 120 may induce more vigorous updrafts. This implies that the spatial scale of valleys has an important impact on C B L structure and morphology over a mountain range. If the mountainous area consists of valleys that have spatial scales similar to the Riviera Valley, the 'valley CBL ' may not grow beyond the mountain ridges. If valleys have larger spatial scales such as the Rhone valley (valley width ~5-10 km, valley length ~ 40 km, depth comparable with Riviera Valley depth), the valley C B L exceeds the mountain ridges and is blended within a larger scale mountain C B L . Interestingly, the former case (applicable to MAP-Riviera and CHAPOP) corresponds with the notion of a separate valley and mountain atmosphere in Fig. 1.1 by Ekhart (1948), while the latter case (applicable to STAAARTE) corresponds with the situation as described in Fig. 1.2 (Fiedler et al., 1987; Whiteman, 2000). The horizontal and vertical scale of valleys in a particular mountain range is only one of the factors that affect C B L structure and morphology in this particular mountain range. Other factors include ambient stability, climate setting, surface characteristics, and synoptic conditions. With regard to the climate factor, the shallow CBLs in deep, narrow valleys in humid regions of the Alps (e.g. the Riviera Valley) are in contrast with the deep CBLs in the deep, narrow valleys in the dry Colorado Rockies. In the dry Colorado climate, higher Bowen ratios produce much deeper CBLs (Holzworth, 1964). The dependency of C B L structure and morphology on the factors mentioned above in general and the difference between the conventional C B L and the 'valley' and 'mountain' CBLs specifically, is difficult to investigate with observational data. Many measurements would have to be taken under many different situations in which each of these factors change. Idealized numerical investigations such as the ones by Gopalakrishnan et al. (2000) or the ones in chapter 3 of this dissertation seem a suitable tool for this purpose. 5.5. A REVISIT OF THE CBL HEIGHT DEFINITION FOR MOUNTAINOUS TERRAIN AND SOME FINAL REMARKS In summary, the data and modeling results in this dissertation indicate that distinctions can be made between a conventional C B L and mountain and valley CBLs. 121 These distinctions require that we revisit the definition of 'mixing height' in mountainous terrain. In chapter 1, the conventional mixing height was defined as: "the height of the layer adjacent to the ground over which pollutants or any constituents emitted within this layer or entrained into it become vertically dispersed by convection or mechanical turbulence within a time scale of about an hour''' (Seibert et al., 2000). The mixing height is synonymous with the C B L height or 'conventional' C B L height. The conventional C B L height can be determined from methods used in this dissertation such as the i?z-mefhod. If mountain-induced venting processes as illustrated in Fig. 5.4 play an important role in the transport of pollutants aloft, the actual C B L height may be underestimated. For mountainous terrain, therefore, the definition would be more useful i f it included the effects of mountain-induced venting processes, which have a timescale of about an hour (as implied from L P D M simulations). The current study also indicates that A L heights may be more appropriate for air-pollution considerations than C B L heights since the A L height is the height up to which pollutants can be transported. As noted before, it is suggested that the A L height in valleys be termed the 'valley C B L ' and the A L height over a mountain range the 'mountain C B L ' to indicate that this entire layer is affected by convectively generated processes near the surface. The valley and mountain CBLs are equivalent to the A L . Figures 5.2, 5.5 and 5.7 show that the top of the valley and mountain C B L can also be identified on vertical profiles of potential temperature. Aerosol observations or L P D M simulations, however, will give a clearer indication of the existence of such a height. In contrast to the situation in a valley or over a mountain range, C B L heights correspond well with A L heights near a mountain base. It is interesting to note that some effort has been made in the last decade or so to describe C B L heights over mountainous terrain. For air-pollution studies, however, the height of the A L may be of more interest. If no information about the aerosol structure is available from aerosol lidar or other measurements, an L P D M simulation is a useful way to investigate the behaviour of the A L height, as demonstrated in this dissertation. Although an attempt has been made to make some general statements about the phenomena seen in the case studies in this dissertation, it is well-known in the mountain meteorology community that there are difficulties in doing this. Even in a single valley, 122 C B L growth can change from day to day because of different stability or surface conditions, for example. In this context, Whiteman (1990) points out that the valley meteorology problem is a continuum problem. "The continuum in topographic complexity and scale, above-valley flows, climate, valley energy budgets, and even the scales of the local circulations, ensures that generalizations will be difficult" (Whiteman, 1990). This does not apply only for valleys but also for mountainous terrain in general. In this dissertation, it was possible, for example, to identify mechanisms that produce the observed aerosol structure, but it was not possible to determine how frequently the observed aerosol structure occurs. The same applies for the depressed C B L height near a mountain base and the wind- and temperature structure of the Riviera Valley. Of course, the fact that generalizations cannot easily be made does not mean that the results are worth less than when this would be possible or that it would be nonsense to take measurements in mountainous terrain. In this context, Ekhart (1948) makes an appropriate comment on his analysis of temperature data: "Indeed it would be hazardous to seek to make general conclusions about the mountain atmosphere temperature variation by basing them on a single series of observations made at a single station. Yet, it would be just as arbitrary to consider an interesting result as due to pure chance" (Ekhart, 1948, translated from French by Whiteman and Dreiseitl, 1984). The joint investigation of the C B L morphology over individual topographic features on one hand, and in a complete mountain range on the other, has facilitated the development of the conceptual pictures in this chapter and has given a more complete understanding of processes affecting C B L morphology. It is clear that much more research is needed before the general concepts of C B L structure and morphology as presented in this chapter can be rejected or accepted. 123 6. GENERAL CONCLUSIONS AND RECOMMENDATIONS FOR FUTURE RESEARCH In this chapter, major conclusions of the research are summarized. More specific conclusions will not be repeated here and can be found in the conclusion sections of chapters 2, 3, and 4 and also in chapter 5 where the results were integrated. This chapter ends with recommendations for future research. 6.1. GENERAL CONCLUSIONS The temperature structure in the Riviera Valley on 25 August 1999 was characterized by three layers. The lower layer is the 'conventional C B L ' with characteristics similar to those over flat terrain. The valley atmosphere up to the top of the second layer where a relatively strong inversion prevails, is heated during the day. Because observations and model results suggest that this layer is affected by surface processes on a small timescale, it is appropriate to call the layer from the surface to the top of the second heated layer the 'valley C B L ' . Upvalley flows prevail in the valley C B L but the horizontal wind structure is inhomogeneous, with larger wind speeds on the eastern than on the western side of the valley. Aerosols can be found up to the height of the valley C B L . The layer above the valley C B L represents a transition zone to the free atmosphere. - The MAP-Riviera field study provided an excellent data set to evaluate the performance of a mesoscale modeling system in a very steep valley. C B L structure was captured surprisingly well given that surface layer and turbulence parameterization schemes are strictly valid only for flat and homogenous terrain. While previous studies found subsidence to play a major role in the heating of valleys above the conventional C B L , the deep and narrow valley studied in this dissertation showed a more complex behaviour. There was no evidence of a homogeneous subsidence field in the valley, presumably because of the inhomogeneous and rather disorganized nature of the along-valley flow. Modeling results indicate that horizontal 124 advection constituted the major contribution to the heating aloft. It is suggested that the heating along the slopes is transported horizontally by the mean flow in the valley to regions in the center of the valley. Surface turbulent sensible heat fluxes generally increase with elevation in the Riviera Valley. This facilitates the heating of the valley atmosphere at heights above the CBL. The effect of this process alone on the C B L structure in the Riviera Valley could not be evaluated. Evaluation of sensible heat fluxes with a mesoscale numerical model over complex terrain is subject to difficulties related to the large variability in azimuth and slope angles on a small spatial scale and uncertainties of turbulence measurements in complex terrain. No systematic over- or underestimation was found but there was a large scatter due to this variability. C B L heights from the T K E and i?z-methods were compared repeatedly in this dissertation and they were generally in good agreement. For the MAP-Riviera case study, these C B L heights also corresponded well with C B L heights determined from aircraft data with the i?z'-method. This was the case over the valley floor where the C B L reached a maximum depth of about 1000 m which is well below the average ridge height of -2000 m. C B L heights over the slopes could not be determined from the aircraft data. Observations and simulations show that the atmosphere near the mountain base is heated more intensely than further away from it, not only inside the boundary layer but also above it. The enhanced heating aloft is related to a reduced boundary-layer growth near the mountain base, resulting in a depression of the C B L height, as documented for the Pacific'93 field study. An analysis of the different terms in the temperature tendency equation indicates that vertical and horizontal advection of warm air associated with the thermally driven circulation along the mountain slope play a role in the enhanced heating aloft. This is different from the heating above the 125 C B L in the Riviera Valley in which vertical advection was not found to play an important role. Independent of the causes, heating above the C B L played a role in reducing C B L growth in both the Pacific'93 and MAP-Riviera case studies. The examination of the depressed C B L height showed that combining observational data analysis with idealized numerical simulations is a promising technique for the investigation of the mountain boundary layer. Lidar data, numerical simulations, and L P D M output suggest that A L heights did not correspond to C B L heights in the S T A A A R T E field study. An L P D M simulation showed a significant number of particles to be present above the C B L height. These particles are carried to these heights by venting mechanisms associated with the presence of thermally driven mesoscale flows. A n L P D M simulation of the Riviera Valley shows a similar feature which is corroborated by aircraft observations of aerosol concentration. Recent unpublished data from a new field study (CHAPOP) indicates that the discrepancy between A L and C B L heights may occur more often over mountainous terrain. In contrast, C B L heights over the plains near a mountain base, as in Pacific'93, correspond well with A L heights. Downlooking lidar show that the A L height did not follow small-scale topography during S T A A A R T E . As a consequence of the difference between A L height and C B L height in mountainous terrain, no conclusive statements can be made concerning the behaviour of C B L heights in mountainous terrain from downlooking lidar data only. Information about the vertical temperature structure in the boundary layer is also required to interpret lidar data in mountainous terrain. A mesoscale numerical model combined with an L P D M can provide useful information i f such data are not available, as was demonstrated in this dissertation. The atmosphere above the C B L height over a mountain range, like the atmosphere in a valley, is affected considerably by surface processes and resulting mesoscale flows. It would be appropriate in this case to define a mountain C B L height (which is 126 equivalent to the A L height) designating the height up to which particles can be transported. The 'valley C B L ' and 'mountain C B L ' heights are more important parameters for air-pollution studies than the C B L height i f venting processes induced by mountainous terrain are present. The valley- and mountain C B L represent the real mixing height. Under certain circumstances such as those encountered during S T A A A R T E , the differences in height between the conventional and mountain CBLs can be substantial. Determination of the factors causing the height difference could be an important goal for future follow-up studies. 6.2. RECOMMENDATIONS FOR FUTURE RESEARCH Following is a selection of suggestions and recommendations for future research, based on findings from this dissertation. - Future investigations could examine the factors that produce differences in A L and C B L heights. Sensitivity tests with a mesoscale numerical model and an L P D M could be performed to investigate the effects of ambient stability, land surface characteristics, and horizontal and vertical topography scales. The factors affecting the various venting mechanisms in mountainous terrain could also be investigated. - The C B L growth rate in a valley is affected by the heating of the valley atmosphere above the C B L height. The research in this dissertation indicated that horizontal advection plays an important role in this heating aloft but the precise way in which this occurs is still unknown. The role of the slope flows in the Riviera Valley was also not clear. Future investigations should obtain additional information about the vertical structure of the slope boundary layer so that a better assessment of its role in the heating of the valley atmosphere can be made. Soil moisture has a strong effect on model results and realistic initializations are required for case studies. Sensitivity tests in this dissertation showed that soil temperature initialization also had significant effects on the modeling results. This is 127 not surprising since the soil temperature affects the ground heat flux and therefore the energy budget at the surface. In future studies, the effect of an inhomogeneous soil temperature offset could be investigated in more detail. The turbulence parameterization schemes currently used in mesoscale models were developed for horizontally homogeneous terrain. It is rather surprising that, despite this, the model seems to perform well in very complex terrain. Future investigations could take a closer look at differences between observed and simulated turbulence parameters and try to design better turbulence parameterization schemes for complex terrain. The use of large eddy simulations (LES) in complex terrain is also an interesting topic of future research. LES resolves convective motions explicitly and could provide insight into the effect of these small-scale motions on the mean structure of the C B L in mountainous terrain. It will be difficult, though, to evaluate these models, given, for example, the difficulties of measuring turbulent quantities and making representative measurements in complex terrain. Furthermore, the application of LES in realistic complex terrain settings is problematic due, for example, to the difficulty of describing turbulent inflows at the lateral boundary (Moeng, 1998). In future field studies, it would be very useful to probe the atmosphere simultaneously in a small scale valley and over a larger mountain range in which this valley is embedded. Knowledge of C B L and aerosol structure on a larger scale is sometimes required to gain understanding of C B L and aerosol structure on a smaller scale and vice versa. The recently performed CHAPOP field study, from which some preliminary data were shown in this dissertation, provides a good example of such a field study. Further work on this data set is planned. 128 7. B I B L I O G R A P H Y A M S , 2000: Glossary of meteorology. 2" edition. American Meteorological Society, Boston, Massachusetts, USA, 855 pp. Andretta, M . , A . Weiss, N . Kljun, and M.W. Rotach, 2001: Near-surface turbulent momentum flux in an Alpine valley: observational results. M A P Newsletter, 15, 122-125. Atkinson, B.W., 1981: Mesoscale Atmospheric Circulations (chapter 6). Academic Press, London. 279 pp. Ball, F.K. 1960: Control of inversion height by surface heating. Quart. J. Roy. Meteor. Soc., 86, 483-494. Baltensperger, U . , H.W. Gaggeler, D.T. Jost, M . Lugauer, M . Schwikowski, E. Weingartner, and P. Seibert, 1997: Aerosol climatology at the high-Alpine site Jungfraujoch, Switzerland. J. Geophys. Res., 102, 19,707-19,715. Banta, R .M. , 1982: An observational and numerical study of mountain boundary-layer flow. Atmospheric Science Paper No. 350. 203 pp. Banta, R .M. , 1984: Daytime boundary-layer evolution over mountainous terrain. Part I: Observations of the dry circulations. Mon. Wea. Rev., 112, 340-356. Banta, R .M. , and W.R. Cotton, 1981: An analysis of the structure of local wind systems in a broad mountain basin. J. Appl. Meteor., 20, 1255-1266. Barnes, S.L., 1973: Mesoscale objective map analysis using weighted time series observations. N O A A Tech. Memo. E R L NSSL-62, 60pp. Barry, R.G., 1992: Mountain Weather and Climate. 2 n d edition, Routledge, 402 pp. Baxter, R.A., 1991: Determination of mixing heights from data collected during the 1985 S C C C A M P field program. J. Appl. Meteor., 30, 598-606. Baidya Roy, S., and R. Avissar, 2000: Scales of response of the convective boundary layer to land-surface heterogeneity. Geophys. Res. Lett., 27, 533-536. Bernhofer, C , 1992: Estimating forest evaporation at a non-ideal site. Agric. For. Meteorol, 60, 17-32. Beven, K.J . , and M.J . Kirkby, 1979: A physically based, variable contributing area model of basin hydrology. Hydrol. Sci. Bull., 24, 43-69. BFS (Bundesamt fur Statistik), 1993: Die Bodennutzung der Schweiz. Arealstatistik 1979/85. Bern. Binder, H.-J . , 1997: Tageszeitliche und raumliche Entwicklung der konvektiven Grenzschicht iiber stark gegliedertem Gelande. Dissertation, Institut fur Meteorologie und Klimaforschung Karlsruhe, 310 pp. Binder, P. and C. Schar (Eds.), 1996: The Mesoscale Alpine Programme: Design Proposal', 77 pp. [Available from M A P Programme Office, c/o Swiss Meteorological Institute, C H - 8044 Zurich]. Blumen, W. (ed.) 1990: Atmospheric processes over complex terrain. American Meteorological Society, 323p. Bossert, J.E., 1997: An investigation of flow regimes affecting the Mexico City region. . J. Appl. Meteor., 36, 119-140. Bossert, J. E., and W. R. Cotton, 1994: Regional-scale flows in mountainous terrain. Part I: A numerical and observational comparison. Mon. Wea. Rev., 122, 1449-1471. 129 Bougeault, P., P. Binder, A . Buzzi, R. Dirks, R. Houze, J. Kuettner, R.B. Smith, R. Steinacker, and H. Volkert: 2001, The M A P Special Observing Period. Bull. Amer. Meteor. Soc, 82, 433-462. Braham, R.R. and M . Draginis, 1960: Roots of orographic cumuli. J. Meteor., 17, 214-224. Brehm, M . , 1986: Experimentelle und numerische Untersuchungen der Hangwindschicht und ihrer Rolle bei der Erwarmung von Talern. Wiss. Mitt. Nr. 54, dissertation, Universitat Miinchen. Businger, J.A., Wyngaard, J .C , Izumi, Y . and Bradley, E.F. 1971: Flux-profile relationships in the atmospheric surface layer. d. Atmos. Sci., 28, 181-188. Cai, X . - M . , and D.G. Steyn, 1993: Mesoscale meteorological modelling study of the Lower Fraser Valley, B.C., Canada from July 17 to 20, 1985. Department of Geography Occasional Paper No. 40. University of British Columbia, Vancouver, B.C.. Cai, X . - M . , R.A.S. Hourston, and D.G. Steyn, 2000: A numerical study of meteorological conditions during PACIFIC'93. Atmosphere-Ocean, 38, 457-479. Carson, D.J. and F.B. Smith, 1974: Thermodynamic model for the development of a convectively unstable boundary layer. In: Advances in geophysics, volume 18a, 111-124. Chen, C , and W.R. Cotton, 1983: A one-dimensional simulation of the stratocumulus-capped mixed layer. Bound.-Layer Meteor., 25, 289-321. Ching, J.K.S., S.T. Shipley, and E.V. Browell, 1988: Evidence for cloud venting of mixed layer ozone and aerosols. Atmos. Environ., 22, 225-242. Clark, T., L., and R.D. Farley, 1984: Severe downslope windstorm calculations in two and three spatial dimensions using analeastic interactive grid nesting: A possible mechanism for gustiness. d. Atmos. Sci., 41, 329-350. Colette, A .G . , and R.L. Street, 2002: Inversion-layer breakup in steep valleys and the effects of topographic shading. Proceedings 10 t h Conference on Mountain Meteorology, 17-21 June 2002, Park City, Utah, 101-104. Coulter, R.L., 1979: A comparison of three methods for measuring mixing-layer height. d. Appl. Meteor., 8, 1495-1499. Cramer, O.P. and Lynott, R.E. 1961: Cross-section analysis in the study of windflow over mountainous terrain. Bull. Amer. Meteor. Soc, 42, 693-702. Cramer, O.P. 1972: Potential temperature analysis for mountainous terrain. J. Appl. Meteor., 11, 44-50. Davies, H.C., 1976: A lateral boundary formulation for multi-level prediction models. Quart, d. R. Met. Soc, 102, 405-418. Dayan, U . , R. Shenhav, and M . Graber, 1988: The spatial and temporal behavior of the mixed layer in Israel, d. Appl. Meteor., 27, 1382-1394. Deardorff, J.W., 1972: Parameterization of the planetary boundary layer for use in general circulation models. Mon. Wea. Rev., 100, 93-106. Deardorff, J.W., G.E. Willis, and B.H. Stockton, 1980: Laboratory studies of the entrainment zone of a convectively mixed layer, d. Fluid Mech., 100, 41-64. Defant, F., 1949: Zur Theorie der Hangwinde, nebst Bemerkungen zur Theorie der Berg-und Talwinde. Arch. Meteor. Geophy. Bioklimatol, A l , 421-450. 130 Defant, F., 1951: Local winds. Compendium of Meteorology, T .M. Malone (ed.), Boston, American Meteorol. Soc, 655-672. De Wekker, S.F.J., 1995: The behaviour of the convective boundary layer height over orographically complex terrain. MS thesis University of Karlsruhe, Germany / Wageningen Agricultural University, the Netherlands. 74 pp. De Wekker, S.F.J., M . KoBmann, and F. Fiedler, 1997: Observations of daytime mixed layer heights over mountainous terrain during the TRACT field campaign. 12th A M S Symposium on boundary layers and turbulence, 28 July - 1 August 1997, Vancou-ver, British Columbia, Canada, 498-499. De Wekker, S.F.J., S. Zhong, J.D. Fast, and C D . Whiteman, 1998: A numerical study of the thermally driven plain-to-basin wind over idealized basin topographies. J. Appl. Meteor., 37, 606-622. De Wekker, S.F.J., M . Rucker, and D. G. Steyn, 1998: Depressed mixed layer depths near a mountain base. 8th Conference on Mountain Meteorology, August 3-7, 1998, Flagstaff, Arizona, 373-379. Dickinson, R.E., A . Henderson-Sellers, P.J. Kennedy, and M.F. Wilson, 1986: Biosphere-Atmosphere Transfer Scheme (BATS) for the N C A R Community Climate Model. N C A R Technical Note NCAR/TN-275+STR, National Center for Atmospheric Research, Boulder, Co, 69 pp. Doran, J . C , and S. Zhong, 1994: Regional drainage flows in the Pacific Northwest. Mon. Wea. Rev., 122, 1158-1167. Driedonks, A . G . M . , 1982a: Sensitivity analysis of the equations for a convective mixed layer. Bound.-Layer Meteor., 22, 475-480. Driedonks, A . G . M . , 1982b: Models and observations of the growth of the convective boundary layer. Bound.-Layer Meteor., 23, 283-306. Edinger, J.G., M . H . McCutchan, P.R. Miller, B.C. Ryan, M.J . Schroeder, and J.V. Behar, 1972: Penetration and duration of oxidant air pollution in the south coast air basin of California. Journal of the Air Pollution Control Association 22, 882-886. Ekhart, E., 1948: De la structure thermique de l'atmosphere dans la montagne. [On the thermal structure of the mountain atmosphere]. La Meteorologie, 4, 3-26. [English translation: Whiteman, C D . , and E. Dreiseitl: 1984. Alpine Meteorology: Translations of Classic Contributions by A . Wagner, E. Ekhart and F. Defant. PNL-5141 / ASCOT-84-3. Pacific Northwest Laboratory, Richland, Washington, 121 pp]. Fast, J. D., 1995: Mesoscale modeling and four-dimensional data assimilation in areas of highly complex terrain. J. Appl. Meteor., 34, 2762-2782. Fast, J. D., and S. Zhong, 1998: Meteorological factors associated with inhomogeneous ozone concentrations within the Mexico City basin. J. Geophys. Res., 103, 18927-18946. Fiedler, F., 1983: Einige Charakteristika der Stromung im Oberrheingraben. Wissenschaftliche Berichte des Meteorologischen Instituts der Universitat Karlsruhe, Nr. 4. 113-123. Fiedler, F., 1992: TRACT operational plan. Institut fur Meteorologie und Klimaforschung Karlsruhe, 66 pp. Fiedler, F., I. Bischoff-GauB, N . Kalthoff, and G. Adrian, 2000: Modeling of the transport of a tracer in the Freiburg-Schauinsland area. J. Geophys. Res., D 105: 1599-1610. 131 Fiedler F., G. Adrian, and C.-P. Hugelmann, 1987: Beobachtete Phanomene wahrend des Tulla-Experimentes, 3 r d PEF Statuskolloquium 10.-12.3.1987 Karlsruhe, KfK-PEF 12, Bd. 2, 347-365. Fosberg, M.A . , 1967: Numerical analysis of convective motions over a mountain ridge. J. Appl. Meteor., 6, 889-904. Fournet, J., 1840: Des brises de jour et de nuit autour des montagnes. Ann. Chim. Et Phys., 74, 337-401. [Ueber die Morgen- und Abendwinde in Gebirgen, German translation in: Annalen der Physik und Chemie (1842), Erganzungsband 1, 490-511, 594-631] Freytag, C., 1987: Results from the M E R K U R experiment: mass budget and vertical motions in a large valley during mountain and valley wind. Meteorol. Atmos. Phys. 37, 129-140. Frei, C., and C. Schar, 1998: A precipitation climatology of the Alps from high-resolution rain-gauge observations. Int. J. Climatol., 18, 873-900. Furger, M . , J. Dommen, W.K. Graber, L. Poggio, A . Prevot, S. Emeis, G. Grell, T. Trickl, B. Gomiscek, B. Neininger, and G. Wotawa, 2000: The V O T A L P Mesolcina Valley Campaign 1996 - concept, background, and some highlights. Atmos. Environ., 34, 1395-1412. Gal-Chen, T., and R.C.J. Sommerville, 1975: On the use of coordinate transformation for the solution of the Navier-Stokes equations. J. Comput. Phys., 17, 209-228. Geiger, R., 1965: The Climate Near the Ground. Cambridge, Harvard University Press, 611 pp. [Translated by Scripta Technica, Inc. from the 4th German edition, 1961]. Goodin, W.R., G.J. McRae, and J. Seinfeld, 1979: A comparison of interpolation methods for sparse data: application to wind and concentration fields J. Appl. Meteor.,18, 761-771. Gopalakrishnan, S.G., S. Baidya Roy, and R. Avissar, 2000: A n evaluation of the scale at which topographical features affect the convective boundary layer using large eddy simulations. d. Atmos. Sci., 57, 334-351. Grasso, L.D. , 2000: A numerical simulation of dryline sensitivity to soil moisture. Mon. Wea. Rev., 128, 2816-2834. Hageli, P., D.G. Steyn, and K . B . Strawbridge, K . B , 2000: Spatial and temporal variability of mixed-layer depth and entrainment zone thickness. Bound.-Layer Meteor., 97, 47-71. Hanel, G., 1976: The properties of atmospheric aerosol particles as functions of the relative humidity at thermodynamic equilibrium with the surrounding moist air. Adv. Geophys., 19,73-188. Hayden, K . L . , K . G . Anlauf, R . M . Hoff, W.J. Strapp, J.W. Bottenheim, H.A. Wiebe, F.A. Froude, J.B. Martin, D.G. Steyn, and LG. McKendry, 1997: The vertical chemical and meteorological structure of the boundary layer in the Lower Fraser Valley during Pacific'93. Atmos. Environ., 31, 2089-2106. Hawkes, H.B., 1947: Mountain and valley winds with special reference to the diurnal mountain winds of the Great Salt Lake region. Dissertation, Ohio State University, 312 pp. Helfand, H . M . , and J.C. Labraga, 1988: design of a nonsingular level 2.5. second-order closure model for the prediction of atmospheric turbulence, d. Atmos. Sci., 45, 113-132. 132 Hennemuth, B. , 1985: Temperature field and energy budget of a small alpine valley. Beitr. Phys. Atmosph. 58, 545-559. Hennemuth, B., 1986: Thermal asymmetry and cross-valley circulation in a small Alpine valley. Bound.-Layer Meteor., 36, 371-394. Hewson, E.W., and G.C. Gi l l , 1944: Meteorological Investigations in Columbia River Valley Near Trail, B.C. Bur. Mines Bull., U.S. Department of Interior, 453, 23-228. Hoff, R .M. , M . Harwood, A . Sheppard, F. Froude, and J.B. Martin, 1997: Use of airborne lidar to determine aerosol sources and movement in the Lower Fraser Valley (LFV), BC. Atmos. Environ., 31, 2123-2134. Holtslag, A . A . M . , E. van Meijgaard, and W.C. de Rooy, 1995: A comparison of boundary layer diffusion schemes in unstable conditions over land. Bound.-Layer Meteor., 76, 69-95. Holzworth, G . C , 1964: Estimates of mean maximum mixing depths in the contiguous U.S. Mon. Wea. Rev., 92, 235-242. Holzworth, G . C , 1972: Mixing depths, wind speeds, and potential for urban pollution throughout the contiguous United States. EPA, Office of Air Programs Publ. AP-101, 118pp. Jacobson, M.Z. , 1998: Effects of soil moisture on temperatures, winds, and pollutant concentrations in Los Angeles. J. Appl. Meteor., 38, 607-616. Jackson, P.L., and D.G. Steyn, 1994: Gap winds in a fjord. Part I: Observations and numerical simulation. Mon. Wea. Rev., 122, 2645-2665. Jasper, K. , 2002: Hydrological modelling of alpine river catchments using output variables from atmospheric models", Diss. ETH No. 14385, Zurich. Kalthoff, N . , H.-J . Binder, M . KoBmann, R. Vdgtlin, U . Corsmeier, F. Fiedler, and H. Schlager, 1998: Temporal evolution and spatial variation of the boundary layer over complex terrain. Atmos. Environ., 32, 1179-1194. Kiemle, C , M . Kastner, and G. Ehret, 1995: The convective boundary layer structure from lidar and radiosonde measurements during the EFEDA'91 campaign. J. Atmos. Ocean. Technol, 12, 771-782. Kikas, U . , A . Reinart, M . Vaht, and U . Veismann., 2001: A case study of the impact of boundary layer aerosol size distrbution on the surface U V irradiance. Atmos. Environ., 35, 5041-5051. Killinger, D. K. , and N . Menyuk, 1987: Laser remote sensing of the atmosphere. Science, 235, 37-45. Klemp, J.B., and R.B. Wilhelmson, 1978a: The simulation of three-dimensional convective storm dynamics. J. Atmos. Sci., 35, 1070-1096. Klemp, J.B., and R.B. Wilhelmson, 1978b: Simulations of right- and left-moving storms produced through storm splitting. J. Atmos. Sci., 35, 1097-1110. KoBmann, M . , R. Vdgtlin, U . Corsmeier, B. Vogel, F. Fiedler, H. -J. Binder, N . Kalthoff, and F. Beyrich, 1998: Aspects of the convective boundary layer structure over complex terrain. Atmos. Environ., 32, 1323-1348. KoBmann M . , U . Corsmeier, S.FJ. De Wekker, F. Fiedler, R. Vdgtlin, N . Kalthoff, H . Giisten, and B. Neininger, 1999: Observations of handover processes between the atmospheric boundary layer and the free troposphere over mountainous terrain. Contr. Atmos. Phys., 72, 329-350. Kondratyev, J., 1969: Radiation in the atmosphere. Academic Press, New York, 912 pp. 133 Lagouvardos, K. , V . Kotroni, and G. Kallos, 1996: Exploring the effects of different types of model initialisation: Simulation of a severe air-pollution episode in Athens, Greece. Meteor. Appl., 3, 147-155. Lau, K . - M . , and H. Weng, 1995: Climate signal detection using wavelet transform: how to make a time series sing. Bull. Amer. Soc., 76, 2391-2402. Lee, S-H., and F. Kimura, 2001: Comparative studies in the local circulations induced by land-use and by topography. Bound.-Layer Meteor., 101, 157-182. Lee, X . , and T.A. Black, 1993: Atmospheric turbulence within and above a douglas-fir stand, Part II: Eddy fluxes of sensible heat and water vapour. Bound.-Layer Meteor., 64, 369-389. Lenschow, D.H., B.B. Stankov, and L. Mahrt, 1979: The rapid morning boundary-layer transition. J. Atmos. Sci., 36, 2108-2124. Lilly, D.K., 1968: Models of cloud-topped mixed layers under a strong inversion. Quart. J. Roy. Meteor. Soc, 94, 292-309. Louis, H . , 1975: Neugefasstes Hohendiagramm der Erde. Bayer. Akad. Wiss. (Mafh.-Naturwiss. Klasse), 305-326. Louis, J.F., 1979: A parametric model of vertical eddy fluxes in the atmosphere. Bound.-Layer Meteor., 17, 187-202. Lu, R. and R.P. Turco, 1994: Air pollutant transport in a coastal environment. Part 1: Two dimensional simulations of sea-breeze and mountain effects. J. Atmos. Sci., 51, 2285-2308 Lugauer, M . , 1998: Vertical transport of atmospheric trace species in the Alps. PhD dissertation Bern. 91 pp. Lyons, W. A. , R. A. Pielke, W. R. Cotton, C. J. Tremback, R. L. Walko, M . Uliasz, and J. I. Ibarra, 1994: Recent applications of the R A M S meteorological and the H Y P A C T dispersion models. Proceedings of the 20th ITM of N A T O / C C M S on Air Pollution Modeling and Its Application, S. Gryning and M . Millan, Eds., Plenum, 19-26. Mahrer, Y . , and R. A . Pielke, 1977: A numerical study of the airflow over irregular terrain. Beit. Phys. Atmos., 50, 98-113. Mannstein, H. , 1989: Die radiometrisch bestimmte Oberflachentemperatur im gebirge und die Ermittlung des Stroms fuhlbarer Warme. Dissertation University of Munich, pp. 128. Marsik, F.J., K .W. Fischer, T.D. McDonald, and P J . Samson, 1995: Comparison of methods for estimating mixing height used during the 1992 Atlanta Field Intensive. J. Appl. Meteor, 34, 1802-1814. Matzinger, N . , M . Andretta, E. van Gorsel, R. Vogt, A . Ohmura, and M.W. Rotach, 2002: surface radiation budget in an alpine valley, accepted Quarterly J. Roy. Meteorol. Soc. Maughan, R.A., A . M . Spanton, M.L . , and Williams, 1982: An analysis of the frequency distribution of sodar derived mixing heights classified by atmospheric stability. Atmos. Environ., 16, 1209-1218. McKendry, I.G., D.G, Steyn, J. Lundgren, R . M . Hoff, W. Strapp, K . Anlauf, F. Froude, B.A. Martin, R . M . Banta, and L.D. Olivier, 1997: Elevated pollution layers and vertical downmixing over the Lower Fraser Valley, B.C. Atmos. Environ., 31, 2135-2146. 134 McKendry, I.G., and J. Lundgren, 2000: tropospheric layering of ozone in regions of urbanized complex and/or coastal terrain: a review. Progress in Physical Geography, 24, 329-354. Melfi, S.H., J.D. Spinhirne, S.-H. Chou, and S.P. Palm, 1985: Lidar observation of vertically organized convection in the planetary boundary layer over the ocean. J. Clint. Appl. Meteorol, 24, 806-821. Mellor, G.L., and Yamada, T. 1974: A hierarchy of turbulence closure models for planetary boundary layers. J. Atmos. Sci., 31, 1791-1806. Mellor, G.L., and T. Yamada, 1982: Development of a turbulent closure model for geophysical fluid problems. Rev. Geophys. Space Phys., 20, 851-875. Moeng, C.-H.,1998: Large eddy simulation of atmospheric boundary layers.in: Clear and cloudy boundary layers, A . A . M . Holtslag and P.G. Duynkerke, Eds., Royal Netherlands Academy of Arts and Sciences, 67-83. Moll , E., 1935: Aerologische Untersuchungen periodischer Gebirgswinde in V-formigen Alpentalern. Beitr. Physikfr. Atmos., 22, 111-191. Myrick, R.H., S.K. Sakiyama, P.R. Angle, A.S. Sandhu, 1994: Seasonal mixing heights and inversions at Edmonton, Alberta. Atmos. Environ., 28A, 723-729. Neininger J3., and O. Liechti, 1984: Local winds in the upper Rhone valley. Geojournal, 8, 265-270. Neininger, B., W. Fuchs, M . Baumle, A.Volz-Thomas, A.S .H. Prevot, and J. Dommen (2001): A small aircraft for more than just ozone: MetAir's 'Dimona' after ten years of evolving development. Proceedings of the 11th Symposium on Meteorological Observations and Instrumentation. Albuquerque, N M , USA, 14-19 January 2001, pp 123-128. Noppel, FL, 1999: Untersuchung des vertikalen Warmetransports durch die Hangwindzirkulation auf regionaler Skala. PhD dissertation University of Karlsruhe, 162 pp. Norton, C.L.,' and G.B. Hoidale, 1976: The diurnal variation of mixing height by season over White Sands Missile Range, New Mexico. Mon. Wea. Rev., 104, 1317-1320. Nyeki, S., M . Kalberer, I. Colbeck, S.F.J. De Wekker, M . Furger, H . W. Gaggeler, M . Kossmann, M . Lugauer, D. Steyn, E. Weingartner, M . Wirth, and U . Baltensperger, 2000: Convective Boundary Layer Evolution to 4 km asl over High-Alpine Terrain: Airborne Lidar observations in the Alps. Geophys. Res. Lett., 27, 689-692. Panin, G.N., G. Tetzlaff, and A. Raabe, 1998: Inhomogeneity of the land surface and problems in the parameterization of surface fluxes in natural conditions. Theor. Appl. Climatol, 60, 163-178. Pielke, R.A., 2002: Mesoscale meteorological modeling. 2nd Edition, Academic Press, San Diego, C A , 676 pp. Pielke, R.A., W.R. Cotton. R.L. Walko, C.J. Tremback. W.A. Lyons, L .D. Grasso, M.E . Nicholls, M.D. Moran, D.A. Wesley, T.J. Lee and J.H. Copeland, 1992: A comprehensive meteorological modeling system - R A M S . Meteor. Atmos. Phys., 49, 69-91. Plate, E.J., E.E. Fedorovich, D.X. Viegas, J.C. Viegas, and J.C. Wyngaard (eds.), 1998: Buoyant convection in geophysical flows. N A T O ASI series. Series C: Mathematical and physical sciences. Vol . 513. Kluwer academic publishers, Dordrecht, The Netherlands. 491 pp. 135 Pottier, J.L., S.C. Pryor, and R.M. Banta, 1997: Synoptic variability related to boundary layer and surface features observed during Pacific'93. Atmos. Environ., 31, 2163-2173. Poulos, G.S., 1996: The interaction of katabatic winds and mountain waves. Ph.D. dissertation, Colorado State University, 297 pp. [Available from Los Alamos National Laboratory, Publication LA-13224-T, Los Alamos, N M 87545] Reiter, R., H. Miiller, R. Sladkovic, and K. Munzert, 1983: Aerologische Untersuchungen der tagesperiodischen Gebirgswinde unter besonderer Beriicksichtigung des Windfeldes im Talquerschnitt. Meteor. Rundsch., 36, 225-242. Rotach, M.W., P. Calanca, R. Vogt, D.G. Steyn, G. Graziani, M . Andretta, A. Christen, S. Cieslik, R. Conolly, S.F.J, de Wekker, S. Galmarini, E. van Gorsel, J. Gurtz, E. Kadygrov, V. Kadygrov, E. Miller, B. Neininger, M . Rucker, H. Weber, A. Weiss, M . Zappa, 2002: The turbulence structure and exchange processes in an Alpine valley: The Riviera project. Submitted to Bull. Amer. Soc. Rucker, M. , S.F.J, de Wekker, and D.G. Steyn, 1998: Analysis of mountain venting characteristics from airborne lidar data. 8th AMS Conference on Mountain Meteorology, 3-7 August 1998, Flagstaff, Arizona, 404-409. Schwiesow, R.L., 1984: Lidar measurements of boundary-layer variables. Probing the Atmospheric Boundary Layer, D.H. Lenschow, Ed., Amer. Meteor. Soc, 139-162. Seibert P., F. Beyrich, S.E. Gryning, S. Joffre, A. Rasmussen, P. Tercier, 1998: Mixing height determination for dispersion modelling. In: European Commission, COST Action 710 - Final Report. Harmonisation of the pre-processing of meteorological data for atmospheric dispersion models. EUR 18195 EN. Part WG 2 (120 pp.) Seibert P., F. Beyrich, S.E. Gryning, S. Joffre, A. Rasmussen, and P. Tercier, 2000: Review and intercomparsion of operational methods for the determination of the mixing height. Atmos. Environ., 34, 1001-1027. Steyn, D.G., and K.W. Ayotte, 1985: Application of two-dimensional terrain height spectra to mesoscale modeling, d. Atmos. Sci., 42, 2884-2887. Steyn, D.G., J.W. Bottenheim, and R.B. Thomson, 1997: Overview of tropospheric ozone in the Lower Fraser Valley, and the Pacific'93 field study. Atmos. Environ., 31, 2025-2036. Steyn, D.G., M . Baldi, and R. Hoff, 1999: Detection of mixed layer depth and entrainment zone thickness from lidar backscatter profiles. d. Atmos. Ocean. TechnoL, 16, 953-959. Strobach, K., 1991: Unser Planet Erde - Ursprung und Dynamik. Gebr. Borntrager, Berlin. Stull, R.B., 1973: Inversion rise model based on penetrative convection, d. Atmos. Sci., 30, 1092-1099. Stull, R.B., 1988: An introduction to boundary layer meteorology. Kluwer Academic Publishers, Dordrecht, The Netherlands, pp. 666. Stull, R.B., 1998: Studies of mountain climates, and a theory for mixed-layer-top levelness over complex topography. In: Proceedings, symposium on advances in regional climatology, Karlsruhe, Germany, October 1998. Sullivan, P.P., C-H. Moeng, B. Stevens, D.H. Lenschow, and S.D. Mayor, 1998: Structure of the entrainment zone capping the convective atmospheric boundary layer, d. Atmos. Sci., 55, 3042-3064. 136 Tennekes, H. 1973: A model for the dynamics of the inversion above a convective boundary layer. J. Atmos. Sci., 30, 558-567. Tennekes, H. and A.P. van Ulden, 1974: Short-term forecasts of temperature and mixing height on sunny days. Symposium on atmospheric diffusion and air pollution. Santa Barbara, Calif, 9-13 September 1974. Thyer, N.H., 1966: A theoretical explanation of mountain and valley winds by a numerical method. Arch. Met. Geophys. Biokl. A, 15, 318-347. Torrence, C , and G.P. Compo, 1998: A practical guide to wavelet analysis. Bull. Amer. Soc, 79, 61-78. Tremback, C.J., and R. Kessler, 1985: A surface temperature and moisture parameterization for use in mesoscale numerical models. Preprints 7th AMS Conference on Numerical Weather Prediction, 17-20 June 1985, Montreal, Canada. Tripoli, G.J., 1992: An explicit 3-dimensional nonhydrostatic numerical-simulation of a tropical cyclone. Meteorol. Atmos. Phys. 49, 229-254. Tripoli, G.J., and W.R. Cotton, 1982: The Colorado state university three-dimensional cloud/mesoscale model. Part 1: general theoretical framework and sensitivity experiments./. Rech. Atmos., 16, 185-219. Troen, I., and L. Mahrt, 1986: A simple model of the atmospheric bondary layer: sensitivity to surface evaporation. Bound.-Layer Meteor., 37, 129-148. Twine, T.E., W.P. Kustas, J.M. Norman, D.R. Cook, P.R. Houser, T.P. Meyers, J.H. Prueger, P.J. Starks, M.L. Wesely, 2000: Correcting eddycovariance flux underestimates over a grassland. Agric. For. Meteorol., 103, 279-300. USGS, 2001: GTOPO30 Documentation. http://edcdaac.usgs.gov/gtopo30/README.html (accessed: May 2002). Van den Broeke, M.R., 1997: Structure and diurnal variation of the atmospheric boundary layer over a mid-latitude glacier in summer. Bound.-Layer Meteor., 83, 183-205. Van Pul, W.A.J., A.A.M. Holtslag, and D.P.J. Swart, 1994: A comparison of ABL-heights inferred routinely from lidar and radiosondes at noontime. Bound.-Layer Meteor, 68, 173-191. Vogelezang, D.H.P., and A.A.M. Holtslag, 1996: Evaluation and model impacts of alternative boundary-layer height formulations. Bound.-Layer Meteor., 81, 245-269. Wagner, A., 1938: Theorie und Beobachtung der periodischen Gebirgswinde. Gerlands Beitr. Geophys., 52, 408-449. Wakimoto R.M. and J.L. McElroy, 1986: Lidar observation of elevated pollution layers over Los Angeles. J. Clim. Appl. Met., 25, 1583-1599. Walko, R.L., W.R. Cotton, and R.A. Pielke, 1992: Large eddy simulation of the effects of hilly terrain on the convective boundary layer. Bound.-Layer Meteor., 58, 133-150. Walko, R.L., C.J. Tremback, R.A. Pielke, and W.R. Cotton, 1995: An interactive nesting algorithm for stretched grids and variable nesting ratios. J. Appl. Meteor., 34, 994-999. Walko, R.L., L.E. Band, J. Baron, T.G.F. Kittel, R. Lammers, T.J. Lee, D. Ojima, R.A. Pielke Sr., C. Taylor, C. Tague, C.J. Tremback, and P.L. Vidale, 2000: Coupled atmosphere-biophysics-hydrology models for environmental modeling, J. Appl. Meteor., 39, 931-944. 137 Walko, R.L., C J . Tremback, and M.J. Bell, 2001: H Y P ACT. Hybrid Particle and concentration transport model. User's guide, 35 pp. [Available from ASTER Division, Mission Reasearch Corporation, P.O. Box 466, Fort Collins, CO 80525-0466] Watson, D.F., 1992: Contouring: A guide to the analysis and display of spatial data, Pergamon Press, Volume 10 in the Series: Computer methods in the geosciences, 321 pp. Whiteman, CD.,1982: Breakup of temperature inversions in deep mountain valleys: Part 1. Observations. J. Appl. Meteor., 21, 270-289. Whiteman, C D . , 1990: Observations of thermally developed wind systems in mountainous terrain. Chapter 2 in Atmospheric Processes Over Complex Terrain, (W. Blumen, Ed.), Meteor. Monogr., 23 (no. 45), Amer. Meteor. Soc, Boston, Massachusetts, 5-42. Whiteman, C D . , 2000: Mountain Meteorology. Oxford University Press, 355 pp. Whiteman, C D . , and E. Dreiseitl, 1984: Alpine Meteorology: Translations of Classic Contributions by A . Wagner, E. Ekhart and F. Defant. PNL-5141 / ASCOT-84-3. Pacific Northwest Laboratory, Richland, Washington, 121 pp. WMO, 1993: Handbook of Meteorological Forecasting for Soaring Flights. W M O Techn. Note Nr. 495. 2nd edition. Young, G.S., and R.A. Pielke, 1983: Application of terrain height variance spectra to mesoscale modeling. J. Atmos. Sci., 40, 2555-2560. Zangl, G., 2002: A n improved method for computing horizontal diffusion in a sigma-coordinate model and its application to simulations over mountainous topography. Mon. Wea. Rev., 130, 1423-1432. Zappa, M , N . Matzinger, and J. Gurtz, 2000: Hydrological and Meteorological Measurements at Claro (CH)- Lago Maggiore Target Area in the MAP-SOP 1999 RJVIERA experiment including first evaluation, in: Hydrological aspects in the Mesoscale Alpine Programme-SOP experiment, edited by B. Bacchi and R. Ranzi, Technical Report of the Dept. of Civi l Engineering-Univ. of Brescia, 10(2). Zhong, S., and J.C. Doran, 1994: An evaluation of R A M S radiation schemes by field measurements. Preprints second R A M S user's workshop. Fort Collins, CO, 15-17 February 1994, 77-80. 138 APPENDICES A l . MESOSCALE MODELING The objectives of the numerical modeling part of this dissertation are to evaluate the performance of a mesoscale modeling system in very complex terrain, to use the model to aid in the interpretation of the observational data, and to investigate the mechanisms producing the observed phenomena. In addition, a Lagrangian particle dispersion model, which is driven by output of the mesoscale modeling system, is used to investigate the net effect of modeled C B L morphology on the dispersion of a passive tracer and to illustrate flow patterns. Idealized simulations are made in chapter 3. Realistic simulations are made in chapters 2 and 4. In this appendix, a brief overview of the mesoscale modeling system and Lagrangian particle dispersion model are given in sections A l . l and A1.2, respectively. In section A l . l , some information about the initial and boundary conditions used in the simulations in this study is given as well. Some issues related to choosing an appropriate horizontal and vertical grid spacing are discussed in section A1.3. Finally, in section A1.4, the application of a correction factor in the shortwave radiation parameterization scheme of the modeling system is discussed. A l . l . RAMS DESCRIPTION The Regional Atmospheric Modeling System (RAMS), as described by Pielke et al. (1992), is used in this dissertation. The numerical code was developed by scientists at Colorado State University for simulating and forecasting meteorological phenomena. It is very versatile and is well-suited to research on many different scales and is used in both idealized and realistic modes. This makes the model very suitable for the research in the present dissertation. The ability of R A M S to model complex terrain phenomena has been demonstrated by previous studies (e.g., Bossert and Cotton, 1994; Doran and Zhong, 1994; Jackson and Steyn, 1994; Fast, 1995; Bossert, 1997; Fast and Zhong, 1998; Poulos, 1996, among others). Details of the model can be found in Pielke et al. (1992) and in the 139 technical manual, which can be found on the following website: http://atmet.com/html/documentation.shtml. In this appendix, a brief overview of R A M S is given including the treatment of the initial and boundary conditions used in the simulations in this dissertation. Extensive use is made of the above mentioned technical manual. The version of R A M S used in this dissertation is version 4.3. R A M S consists of a set of nonhydrostatic, compressible dynamic equations, a thermodynamic equation and a set of cloud microphysical equations. It predicts the three velocity components, potential temperature, mixing ratio, and subgrid-scale T K E in a terrain-following or 'sigma-z' coordinate system (Gal-Chen and Sommerville, 1975; Mahrer and Pielke, 1977). The vertical coordinate in this system is defined as: Z* = H — V"-'.). where z* is the height of a particular grid point in the terrain following coordinate system; zs is terrain elevation at that grid point; z is the untransformed vertical coordinate; and H is the height of the model top at which the z* coordinate surface becomes horizontal (approximately 16 km in the simulations in this study). The terrain following coordinate system makes it possible to represent topographical features without the need for introducing many model layers and has been adopted in many numerical models simulating flow fields over complex terrain. For the time differencing scheme, a hybrid scheme is used which consists of forward time differencing for the thermodynamic variables and leapfrog differencing for the velocity components and pressure (Tripoli, 1992). A non-hydrostatic model permits fast moving sound waves which normally means that the time step must be small so that the Courant-Friedrichs-Levy -stability criterion is not violated. This means that waves or flow features must travel no further than one grid distance per time step, otherwise the model is numerically unstable. For computational efficiency, the equations are integrated using a "time-splitting" technique (Klemp and Wilhelmson, 1978a,b). This technique consists of splitting the equations into sound-wave and gravity wave components which are evaluated with a small time step, and the remaining (slow moving) components (such as horizontal advection and the Coriolis force) which are evaluated with a large time step. 140 T U T U T V V V •[• u T IJ T V V V T U T IJ T Fig. Al.l. Arakawa type C grid stagger used in R A M S . T represents the location of thermodynamic variables; U represents west-east velocity; V represents north-south velocity. W, the vertical velocity, is not shown, but is located half a grid distance above and below T. The equations are solved on an Arakawa-C staggered grid. Variables are staggered on a grid so that velocity components are defined at different locations than the rest of the variables. Fig. A l . l illustrates this. If T, representing the thermodynamic variables, is located at the grid intersections, then U (east-west velocity) is located half a grid spacing to the east and west of T. V (north-south velocity) is located half a grid spacing above and below T. The vertical velocity W is located half a grid point above and below T. R A M S allows the use of nested grids. The grids communicate with the coarser-resolution parent grid via two -way interaction following a grid nesting technique described by Clark and Farley (1984), and Walko et al. (1995). In short, at the boundaries of a fine grid, values are interpolated from the coarse grid in which it is nested, whereas in the grid interior, fine mesh values are averaged to replace the value given on the surrounding coarse mesh. Soil, vegetation, and surface layer parameterization R A M S uses the Land Ecosystem Atmosphere Feedback Model, version 2 (LEAF-2, see Walko et a l , 2000). LEAF-2 is the soil-vegetation-atmosphere-transfer module which represents the land-surface processes in R A M S . It includes the influence of soil, lakes and oceans, and snowcover on each other and on the atmosphere. It also includes 141 the effects of freezing and thawing soil. LEAF-2 includes prognostic equations for soil temperature and moisture for multiple layers, vegetation temperature, and surface water including dew and intercepted rainfall, snowcover mass, and thermal energy for multiple layers, and temperature and water vapor mixing ratio of canopy air. It simulates vertical transport of heat and moisture by conductive, gravitational, turbulent, and radiative exchanges between these components and with the atmosphere. The turbulent exchange for the soil and vegetation is found with the Louis (1979) scheme. The Louis scheme is based on Monin-Obukhov similarity theory and approximates the profile function of Businger et al. (1971) with non-iterative analytic expressions. The computed fluxes for the soil and vegetation must be averaged to provide the grid-area averaged flux. These fluxes serve as the lower boundary layer for the sub-grid diffusion scheme for the atmosphere. Heat and moisture flux between soil layers is parameterized in LEAF-2 based on a multilayer soil model described by Tremback and Kessler (1985). Horizontal water flow is also taken into account in LEAF-2 by incorporating a modified form of the hydrological model TOPMODEL (Beven and Kirkby, 1979). There are a total of 29 land use categories that can be chosen in R A M S : 18 classes that are adapted from BATS (Biosphere-Atmosphere Transfer Scheme, classes 0-17 in Table A l . l , Dickinson et al., 1986) and 13 more land use classes from L D A S (Land Data Assimilation System, see http://ldas.gsfc.nasa.gov/, classes 18-30 in Table A l . l ) . Each class has its own characteristic values of surface emissivity, leaf area index, fractional vegetation coverage, aerodynamic roughness length, aerodynamic displacement height, and root depth, see Table A l . l . 142 Table A l . l . LEAF-2 biophysical parameters by land use class number used in R A M S . The parameters include albedo (a), surface emissivity (e), leaf area index (LAI), fractional vegetation coverage (vfrac), aerodynamic roughness length (z0), aerodynamic displacement height (zd), and root depth (root). class Land use a s L A I vfrac z0(m) zd(m) root(m) 0 Ocean 0.14 0.99 0 0 0 0.1 0 1 Lakes, rivers, streams (inland water) 0.14 0.99 0 0 0 0.1 0 2 Ice cap/glacier 0.4 0.82 0 0 0.01 0.1 0 3 Evergreen needleleaf tree 0.1 0.97 6 0.8 1 15 1.5 4 Deciduous needleleaf tree 0.1 0.95 6 0.8 1 20 1.5 5 Deciduous broadleaf tree 0.2 0.95 6 0.8 0.8 15 2 6 Evergreen broadleaf tree 0.15 0.95 6 0.9 2 20 1.5 7 Short grass 0.26 0.96 2 0.8 0.02 0.2 1 8 Tall grass 0.16 0.96 6 0.8 0.1 1 1 9 Desert 0.3 0.86 0 0 0.05 0.1 1 10 Semi-desert 0.25 0.96 6 0.1 0.1 0.5 1 11 Tundra 0.2 0.95 6 0.6 0.04 0.1 1 12 Evergreen shrub 0.1 0.97 6 0.8 0.1 1 1 13 Deciduous shrub 0.2 0.97 6 0.8 0.1 1 1 14 Mixed woodland 0.15 0.96 6 0.8 0.8 20 15 Crop/mixed farming 0.2 0.95 6 0.85 0.06 0.7 1 16 Irrigated crop 0.18 0.95 6 0.8 0.06 0.7 1 17 Bog or marsh 0.12 0.98 6 0.8 0.03 1 1 18 Evergreen needleleaf forest 0.06 0.97 6 0.8 0.98 10.2 1 19 Evergreen broadleaf forest 0.08 0.95 6 0.9 2.21 20.7 1.2 20 Deciduous needleleaf forest 0.06 0.95 6 0.8 0.92 9.2 1 21 Deciduous broadleaf forest 0.09 0.95 6 0.8 0.91 7.2 1.2 22 Mixed cover 0.07 0.96 6 0.8 0.87 6.5 1.1 23 Woodland 0.08 0.96 5.7 0.8 0.83 7.4 1 24 Wooded grassland 0.18 0.96 5 0.8 0.51 3.6 1 25 Closed shrubland 0.1 0.97 5.1 0.63 0.14 1.4 0.7 26 Open shrubland 0.12 0.97 6 0.22 0.08 0.2 0.6 27 Grassland 0.11 0.96 2.6 0.73 0.04 0.2 0.7 28 Cropland 0.1 0.95 6 0.84 0.11 0.2 0.7 29 Bare ground 0.16 0.86 0.7 0.07 0.05 0.2 0.5 30 Urban and built up 0.15 0.9 4.8 0.74 0.8 1.1 0.8 143 Miscellaneous parameterizations Sub-grid scale turbulent diffusion is parameterized using a Mellor-Yamada level 2.5 turbulence closure scheme (Mellor and Yamada,1974, 1982) with modifications for a case of growing turbulence (Helfand and Labraga, 1988). In this scheme, the mixing coefficients are evaluated using the turbulent kinetic energy equation, which is solved prognostically in the model. With the grid spacing used in the present dissertation (0.33-1 km), it is assumed that convective motions are not resolved and need to be parameterized. Cumulus and microphysical parameterizations are not activated in the simulations in this dissertation, so water vapor is treated as a passive scalar. The case studies in this dissertation were primarily cloudless. Including the effect of clouds would increase computational time significantly. The Chen and Cotton (1983) shortwave and longwave parameterization schemes are used to determine the heating and cooling caused by radiative flux divergence. Shading effects due to terrain slope are also accounted for in the shortwave parameterization scheme following Kondratyev (1969). Initial data and boundary conditions Realistic simulations: The data used to initialize and provide lateral boundary conditions for the outermost R A M S grid were interpolated from E C M W F analysis data for the M A P -Riviera case study and from NCEP F N L analysis data for the S T A A A R T E case study. These data sets incorporate real data, including surface, rawinsonde, and satellite data. Besides horizontally inhomogeneous initial conditions, R A M S permits time-dependent lateral boundary conditions. This means that fields on the lateral boundary of grid 1 are nudged (following Davies, 1976) towards the time interpolated values obtained from E C M W F or NCEP output. The data used for this purpose were at 6 hour intervals. 144 The E C M W F data used for initialization of the M A P case study were obtained at 0.5 latitude by 0.5 longitude spacing at the following pressure levels: 1000, 925, 850, 700, 500, 400, 300, 250, 200, 150, 100, 70, 50, 30, 10 hPa (15 levels). The data were interpolated to a regular latitude/longitude grid with 0.5° resolution and the entire domain of the E C M W F data set is from -20°W to 30°E longitude and 20°N to 70°N latitude. For the S T A A A R T E case study, R A M S was initialized with data from the so called 'FNL archive' (available at http://www.arl.noaa.gov). This F N L archive is based on data from NCEP's GDAS (Global Data Assimilation System). The GDAS is the final (FNL) run in the series of NCEP operational model runs and includes late arriving conventional and satellite data. It is run 4 times a day, at 00, 06, 12, and 18 UTC. F N L data are output on the following mandatory pressure surfaces: 1000, 925, 850, 700, 500, 400, 300, 250, 200, 150, 100, 50, and 20 hPa (13 levels). The data were interpolated to a regular latitude/longitude grid with 1 ° resolution. The domain covers the entire northern hemisphere. Rawinsonde data were retrieved from N O A A and British Atmospheric Data Centre web sites. R A M S includes a mesoscale isentropic data analysis package that allows the user to use observational data sets for initial conditions, lateral boundary tendencies and four dimensional data assimilation. The meteorological variables that are required as input into the data analysis package are: temperature, u-and v wind component, relative humidity, and pressure. The package includes a "hybrid" vertical coordinate, a mixture of isentropic and terrain-following coordinates on which the objective analysis is performed. This reduces the problems associated with the coarse resolution of isentropic coordinates in the boundary layer. The gridded data from E C M W F and F N L archives are interpolated to both the isentropic vertical coordinate and the terrain following (az) coordinate. Available rawinsonde observations are interpolated vertically to the same isentropic levels or a z levels as was the large scale data. A separate objective analysis is then performed on the isentropic and G z data sets. The Barnes (1973) objective analysis scheme is applied to the wind, pressure, and relative humidity on the isentropes and the wind, temperature, and relative humidity on the rj z levels. User-specified parameters control the smoothing 145 characteristics of the objective analysis and the relative importance of the rawinsonde and large-scale data. Also, a 'blending' layer between 4000 and 6000 m agl is defined (heights can be controlled by the user). From the surface to the bottom of this layer, the analysis is defined from the a z data. From the top of this layer to the top of the model domain, the analysis is defined from the isentropic data. In the blending layer, there is a simple weighted average of the isentropic and a z data where the weighting function is a linear function of height. The lateral boundary nudging is an implementation of the Davies (1976) scheme where a number of grid points (five in the case studies in this dissertation) in a boundary region (of only the coarsest grid in a nested grid run) are nudged towards the data analysis. Top boundary nudging is patterned after the Rayleigh friction absorbing layer (e.g. Tripoli and Cotton (1982)), in which a layer below the model top damps vertically propagating gravity waves, thus reducing wave reflections from the model top. No interior nudging was applied in the simulations. Idealized simulations: The upper boundary consists of a rigid lid, and radiative lateral boundary conditions are employed following Klemp and Wilhelmson (1978a, b). A damping scheme is applied to the three vertical layers below the upper boundary in which the amplitudes of vertically propagating gravity waves are gradually suppressed to reduce wave reflection from the rigid lid (Rayleigh friction). The initial temperature, humidity and wind fields are horizontally homogeneous and are based on rawinsonde data from Langley airport (see chapter 3). Topography initialization The land use and topography for the M A P and S T A A A R T E case studies were derived from 30 arc second data, or about 1 km resolution from USGS (United States Geological Survey). The USGS land use data set is based on 1-km Advanced Very High Resolution Radiometer (AVHRR) data spanning April 1992 through March 1993. 146 For the high resolution M A P case study, the terrain height- and land use data for the innermost grid were obtained from 100 m resolution data. Since terrain height is used in defining the model vertical coordinate, compatibility of terrain heights between different model grids is required for proper grid nesting communication. To ensure compatibility between different grids, the 10 grid points closest to boundaries of grid 4 were taken from the initial grid 4 terrain height data set that R A M S produces from the 30 arcsecond data set. A1.2. H Y P A C T D E S C R I P T I O N The HYbrid Particle And Concentration Transport (HYPACT) model (Walko et al., 2001) is used in this dissertation to illustrate the behaviour of mesoscale flows and mixing processes over mountainous terrain. H Y P A C T simulates the motion of atmospheric tracers under the influence of winds and turbulence. Its Lagrangian component enables representation of sources of any size and the maintenance of concentrated, narrow plumes until atmospheric dispersion dictates that they should broaden. The Lagrangian particles are moved through space and time based on the resolvable wind components plus the turbulent wind components, i.e.,: X(t + At) = X(t) + (u + u')At Y(t + At) = Y(t) + (v + v')At Z(t + At) = Z(t) + (w + W + wp )At where u,v, and w are the resolvable scale wind components, which are obtained directly from R A M S , u', v', and w' are the turbulent wind components, and w p is the vertical velocity resulting from external forces. The resolvable wind components are obtained at a given time by linearly interpolating the mean wind fields obtained directly from R A M S output at 1-h intervals. The turbulent wind components of particles representing a passive tracer are derived from a first order Markov scheme. The variables required by the Markov scheme (variances of wind components and Lagrangian time scales) are calculated diagnostically from R A M S output (u, v, w, potential temperature and TKE) using a second order closure technique developed by Mellor and Yamada (1974, 1982), 147 i.e., a level 2.5 scheme modified for a case of growing turbulence (Helfand and Labraga 1988). This is consistent with the turbulence parameterization used in R A M S . Recent applications of the combined R A M S and H Y P A C T model can be found in Lyons et al. (1994) and Lagouvardos et al. (1996). A1.3. S C A L E S O F I N T E R E S T A N D G R I D S P A C I N G C O N S I D E R A T I O N For the realistic simulation of complex terrain flows in mountainous terrain, a variety of horizontal scales has to be considered. First, there is the horizontal scale related to synoptic features that are used as the boundary conditions. Under high pressure, fair weather situations, a horizontal extent of 500-1000 km is assumed to be large enough to contain the synoptic scale atmospheric forcing. Then there is the dominant horizontal scale of the mountain-valley configuration. This will change from region to region. For the regions in this dissertation, it was found by wavelet analysis (see later in this appendix) that this scale is at least 4 km. If the atmospheric processes in a single valley are the main topic of investigation as in the M A P case study, an even smaller scale on the order of the valley width needs to be considered. This scale is on the order of 1-2 km. To determine an appropriate minimum horizontal grid spacing, it needs to be realized that only atmospheric structures with a wavelength of 2Ax to 4Ax (Ax is the horizontal grid spacing) can be resolved by the numerical model (Young and Pielke, 1983; Pielke, 2002). Therefore, in situations where the focus of interest is a mountain range, the minimum required grid spacing is of the order 1-2 km. In the case where the focus of interest is the Riviera Valley, the minimum required grid spacing is 200-500 m to resolve the flow features in the Riviera Valley. To cover both the synoptic scale and the small scales mentioned above, four grids were used in a nested grid configuration in the case studies in this dissertation. The R A M S user manual recommends making the grid size ratios less than 3-4:1 between the neighbouring grids, otherwise problems can occur at the grid boundaries. The grid configuration in the M A P and S T A A A R T E case studies is specified in Tables 2.2 and 4.1, respectively. 148 In the remainder of this section, an appropriate horizontal grid spacing for the investigation of atmospheric processes over a mountain range (as in the S T A A A R T E case study) is determined using wavelet analysis. This is followed by a discussion of how to choose an appropriate vertical grid spacing. Horizontal grid spacing issues Thermally driven flows develop in mountainous terrain under fair weather in light wind conditions. These flows are expected to affect the boundary layer structure and it is important that they are resolved well in a numerical model. It is known that thermally as well as mechanically driven flows are forced by the underlying topography (e.g., Whiteman, 2000), for example the scale of slope flow circulation is determined by topographic scales such as the length and steepness of the slope. Furthermore, as demonstrated by Baidya-Roy and Avissar (2000) the response of the atmosphere to a forcing with a certain wavelength (i.e., topography or sensible heat flux) is found on a similar scale. That is, a peak in the variance of an atmospheric variable can be found at a wavelength corresponding to the wavelength of the forcing. This is due to the existence of organized flow structures that contain most of the energy at a scale similar to the forcing. In mountainous terrain, it is these flow structures (associated with e.g. slope and valley flows) that need to be resolved. Thus, since the scale of the topographic forcing and the scale of the atmospheric response are comparable, examining the dominant scales of the topography can give some guidance on the minimum grid spacing required to resolve the flows in mountainous terrain. In the research in this dissertation, the interest is in scales of boundary layer features from a few kilometers to a few tens of kilometers, thus scales that are at least as large as the scale of the topographic forcing. Boundary layer structures induced by very small-scale hollows on a scale of several hundred meters are not considered. And neither are individual thermals on the scale of a few hundred meters up to the C B L height, which can induce variability in the C B L height on that scale. The effect of these thermals is assumed to be captured by the subgrid turbulence parameterization in the numerical model. It is realized that these subgrid parameterization schemes were developed for 1 4 9 horizontally homogeneous terrain and that effects due to subgrid-scale terrain variations are not taken into account. One may also argue that the horizontal grid spacing used in this dissertation approaches that of a large eddy simulation model. The performance of subgrid turbulence schemes for these models is still uncertain, even in the case of flat homogenous terrain (Stull, 1998). Future investigations may indicate that these schemes are not applicable for mountainous terrain. As demonstrated by Young and Pielke (1983) and Steyn and Ayotte (1985), the distribution of terrain height variance with wavelength along a cross section of the Earth's surface can be used to determine the dominant scales of topography. A similar approach was taken in the current dissertation by applying wavelet analysis to the topography of the S T A A A R T E investigation area. Wavelet analysis has become a popular tool in the atmospheric sciences in the last decade, particularly in the analysis of climate and turbulence data. It is usually applied to time series, and allows one to determine both the dominant modes of variability and how those modes vary in time (Torrence and Compo, 1998; Lau and Weng, 1995). These characteristics can be derived from a wavelet power spectrum in which frequency is plotted against time (also called scalogram). An average over all the times is referred to as a global wavelet spectrum and is comparable to a Fourier power spectrum (Torrence and Compo, 1998). Lau and Weng (1995) explain that a wavelet transform is a generalized form of the Fourier transform. Following Torrence and Compo (1998), the Morlet wavelet (which is a cosine wave multiplied by a Gaussian shape) is used for wavelet analysis in this dissertation. Only global wavelet spectra will be presented. 150 To test the wavelet analysis tool, the global wavelet spectrum was calculated for a sine function (C*sin (27r/L)) with a horizontal scale L of 4 km superimposed on a sine with L equal to 20 km. C was set to 500 m in both cases and the horizontal grid spacing 333 m. The resulting topography is shown in Fig. A1.2a with the spectrum presented in Fig. A 1.2b. The spectrum clearly indicates that 4 km and 20 km are the dominant horizontal scales of the topography. For topography in the S T A A A R T E investigation area, 30 arcsecond data were available. This translates to a resolution of 598 m in the east-west direction and 927 m in the north-south direction at a latitude of 50 degrees. The data were interpolated on a regular grid of 333 m x 333 m. The vertical accuracy of this data set is high with an R M S E of 18 m (USGS, 2001). Four (arbitrary) cross sections in the north-south and east-west direction cross sections were selected in the area of interest. The locations of the 4000 (a) 3000 H E. D) X 1000H 0 0 20 40 60 Horizontal distance (km) 80 100 4,0 km 20.0 km, (b) 0.5 1.0 2.0 4.0 8.0 16.0 32.0 64.0 scale (km) Fig. A1.2. Idealized topography of two superimposed sine waves (a) and the corresponding wavelet spectrum (b). Also shown are the 90% significance line (dotted line) and the position of the maxima at 4 and 20 km (dashed line). 151 cross sections are shown in Fig. A 1.3 and the cross sections themselves are provided in Figs A1.4a and A1.4b. Figures A1.4c and A1.4d show the corresponding global wavelet spectra. For both north-south and east-west cross sections, most of the terrain features vary significantly (90% significance level) over scales larger than 3-4 km. The same analysis was performed for all other cross sections in the investigation area and it can be concluded that the most dominant horizontal scales are found in the range 4 km and larger. Given the fact that atmospheric structures with a wavelength of 2Ax to 4Ax (Ax is the grid spacing) can be resolved by the numerical model (Young and Pielke, 1983; Pielke, 2002), the minimum required grid spacing is of the order 1-2 km. Because of the resolution of the data set, statements about the significance of scales smaller than 1 km can not be made. As mentioned earlier, it is mainly the individual thermals that are important for the boundary layer structure on scales < 1 km and we assume that the effects of these thermals are well taken into account by the subgrid scale turbulence 7.80 7.90 8.00 8.10 8.20 Longitude (°E) Fig. A1.3. S T A A A R T E investigation area. Contour interval is 1000 m. The darkest color represents terrain over 3000 m. The locations of the cross-sections used in the wavelet analysis are indicated by black lines. 152 parameterization. A computationally expensive simulation with 333 m horizontal resolution was also carried out for the S T A A A R T E case study (the size of the innermost grid was 89x140 grid points which equals 29.67 km x 46.67 km). In this way, the 30 arcsecond topography resolution is better represented and the effects of large scale eddies (of the order 1.5 times the C B L depth) are taken into account more explicitly. It was found that the boundary layer structure and boundary layer heights in the 1 km and 0.333 simulation looked very similar. Based on the wavelet analysis given above, a resolution of 1 km was used in S T A A A R T E simulations. This lower resolution made it faster and easier to perform multiple simulations in which the sensitivity of certain parameters was examined. scale (km) scale (km) Fig. A1.4. Topography of four N-S cross sections (a) and four E-W cross sections (b) with the corresponding wavelet spectra (c and d, respectively). The 90% significance lines corresponding to each spectrum are indicated in c, and d. 153 Steep slopes and vertical grid spacing Resolving the effects of very steep slopes in numerical models imposes a very challenging problem. Adequate resolution of the mountain-valley structure in the M A P case study requires a minimum horizontal grid spacing of about 200-500 m (see above). Due to expected strong gradients in the vertical direction, especially in slope flows, there is also a great need for resolving vertical structure. However, the R A M S user guide indicates that completion of simulations containing steep terrain could be difficult which indeed seemed to be true. The rule-of-thumb that governs this constraint is that the terrain height difference between adjacent grid cells with grid spacing AX should not exceed about 3-5 times the vertical grid spacing AZ near the surface. This constraint can be expressed as: ,f 3 ~ 5 A Z ^ a < tan I AX where a is the maximum slope angle. Poulos (1996) states that truncation errors in the pressure gradient force calculation become too large and dominate the model solution i f this constraint is not met. Given AX = 333 m and a = 40°, which approximates the maximum slope angle in the Riviera Valley, the vertical grid spacing near the surface is limited to 55-90 m. hi this study, 70 m proved to be a feasible vertical grid spacing. Attempts to decrease the vertical grid spacing failed. Very steep slope angles require the vertical grid spacing to become of the same order of magnitude as the horizontal grid spacing and large-eddy simulations may seem more appropriate for these cases. However, a large-eddy simulation for the Riviera Valley would require immense computer power and other problems would be encountered related to large eddy simulations in complex terrain. Large-eddy simulations were therefore avoided in the present dissertation. For the S T A A A R T E case study, 50 m was found to be a feasible vertical grid spacing near the surface. 154 A1.4. S H O R T W A V E R A D I A T I O N C O R R E C T I O N I N R A M S R A M S allows the user to choose from two options for the shortwave radiation parameterization: the Mahrer and Pielke (1977) scheme and the Chen and Cotton (1983, C-C) scheme. For the simulations in this dissertation, the C-C scheme was used. For several surface stations in and around the M A P investigation area, it was found that the modeled incoming shortwave radiation significantly overestimates observed incoming shortwave radiation. Interestingly, Zhong and Doran (1994) also found differences between model and observed incoming radiation, but they found an unexplained underestimation by the C-C scheme. One possible explanation of the overestimation for the M A P case study is the attenuation of the incoming radiation by the presence of aerosols in the valley atmosphere that is not accounted for in the radiation parameterization. E.g., Kikas et al. (2001) estimate a 6% reduction in incoming shortwave radiation due to aerosols in the boundary layer from six case studies. To correct for this error, a factor was determined that reduced the solar constant to obtain close agreement between model and observations at solar noon, i.e., the time of maximum incoming shortwave radiation. Because model output and observations are available at discrete times, usually not corresponding exactly to solar noon, the values at solar noon were determined from a second order polynomial function through the three highest values. This was done for three sites, one inside the Riviera Valley and two that are located approximately 50 km to the south and north (Locarno and Comprovasco, respectively, also located in valley regions). The factors for the three sites were 0.91, 0.88, and 0.90, respectively. Based on this, a reduction factor of 0.9 was applied. To assure that the modeled radiation was not affected by slope angle or azimuth (observed radiation was measured relative to a horizontal plane), an additional simulation was carried out in which surface height was at the height of the valley floor (250m) everywhere. The same reduction factors were found for this simulation. Another run with the Mahrer and Pielke (1977) scheme resulted in even larger values of the incoming shortwave radiation at noon and reduction factors of about 0.84. Zhong and Doran (1994) also found significantly larger values of the incoming shortwave radiation when the Mahrer and Pielke (1977) scheme was used. 155 For the site 'Bosco di Sotto' (site A l ) in the Riviera Valley, the observed and modeled radiation (with and without reduction factor) is shown in Fig. A 1 . 5 . Notice that in the few hours after sunrise and before sunset, the modeled radiation is overestimated, independent of the inclusion of a reduction factor. This can be explained by the fact that the parameterization does not take shadowing by the sidewalls into account - a potentially important factor in steep valleys. I i L Fig. A1.5. Incoming shortwave radiation R s h o r t as a function of time for Bosco di Sotto (MAP case study). Open squares are observed values, closed squares are modeled values without a reduction factor, and plusses are modeled values with a reduction factor. 1 5 6 Figure A1.6 shows the modeled and observed incoming solar radiation for the Jungfraujoch station. The simulations for this case study were performed after the M A P case study. Good agreement is seen between the observations and the model i f the reduction factor is retained. This suggests that there may be a problem with the solar radiation scheme in R A M S . This issue was not investigated further in this dissertation. Time (UTC) Fig. A1.6. Incoming shortwave radiation Rsh0rt a s a function of time for the Jungfraujoch station ( S T A A A R T E case study). Open squares are observed values, plusses are modeled values with reduction factor. 157 A2. SURFACE BOUNDARY CONDITIONS FOR MAP-RIVIERA CASE STUDY In order to make the surface boundary conditions for the MAP-Riviera case study (chapter 2) as realistic as possible, high resolution topography, land use, and soil moisture data were used for the innermost R A M S grid. The horizontal resolution of the topography and land use data was 100 m, the resolution of the soil moisture data 500 m. These data were interpolated to the model resolution of 333 m. The resulting land use and soil moisture distribution in the innermost grid in the MAP-Riviera case study is shown in Fig. A2.1 and A2.3, respectively. The soil moisture data were obtained from a hydrological model (see chapter 2). The USGS land use data set that R A M S uses is based on 1-km Advanced Very High Resolution Radiometer (AVHRR) data spanning April 1992 through March 1993. The Swiss data set is based on actual observations of land use (BFS, 1993). It was found that the standard USGS data set did not give a realistic representation of the land use in the Riviera Valley. This is illustrated in Fig. A3-2 for the two land use types short grass and crop/mixed farming. The black filled polygons are the locations where the two data sets correspond. The light grey filled polygons are locations where the USGS data set had short grass and crop/mixed farming but the Swiss land use data set did not. The dark grey filled polygons are location where the USGS data set did not have short grass and crop/mixed farming but the Swiss land use data set did. The meaning of the polygon colours is summarized in Table A2.1 Most land use classes in the Swiss data set (Table A2.3) did not match those in LEAF-2 (Table A l . l ) . Also, the Swiss data set contains many more classes (69), than LEAF-2 (31). Therefore, land use classes from the Swiss data set had to be cross-referenced to land use classes of LEAF-2. Details of the cross-referencing are given in Table A2.2. The Swiss data set did not distinguish between needleleaf and broadleaf trees. By inspecting resources such as photographs, and satellite derived vegetation patterns in the Riviera Valley (Rotach, van Gorsel, personal communication), the boundary between broadleaf and needleleaf trees was put on 1000 m asl with only needleleaf above and 158 broadleaf below that boundary. The resulting land use in the Riviera Valley as it was used in the mesoscale simulations is shown in Fig. A2.1. Table A2.1. Meaning of the polygon colors in Fig. A2.2. Swiss data set USGS data set Colour of Short.grass/mixed Short, grass/mixed polygon farming? farming? Yes Yes Black Yes No Dark grey No No White No Yes Light grey Table A2.2. Details of the cross-referencing between land use classes from the Swiss data set and land use classes according to LEAF-2 . Land use class Land use class Swiss data set (Table A2.3) LEAF-2 (Table A l . l ) 10-15,17,18,19 3 71,72,75-78 15 73,81,82 27 1-9,16,83-89,95-97 7 69,91,92 1 99 29 20-68,70 30 159 0 5 10 15 20 x ( km) Fig. A2.1. Distribution of land use classes in the Riviera Valley based on a combination of the Swiss land use data set and LEAF-2 land use classes (according to Table A2.2). Height contours are drawn every 400 m. 160 x (km) x (km) Fig. A2.2. Differences between the USGS data set and the land use classes used in the R A M S simulations (see Fig. A2.1) for a) short grass and b) crop/mixed farming. The black filled polygons are the locations where the two data sets correspond. For meanings of the other shading types, see Table A2.1. Height contours are drawn every 400 m. 161 0 5 10 15 20 x (km) Fig. A2.3. Distribution of soil moisture in the Riviera Valley on 25 August 1999. Height contours drawn every 400 m. 162 Table A2.3. Land use classes according to BFS (1993). •D CJ M 2 « ; J :f S « S u : . i : S . | 2 l . x . » S £ S lJ i?S 3.1 » c £ £r X» «a ^ o < o < oo a: 5*. S o ~ 5" 5" s? (M tO iO 4D 04 r r r i n r S-S 6 O O <A o o 5 J a 'gt ' o 55" o t» > • ^ < < 65 a fiS " <5"^ r* 5* ST" J_J_1 51 ° p ... Ji S3 « S i l l . rg to o co u_ a s?2 o -3-•a s o «> > • • 0 .c S3 •a S I ? *\ |1 73 » 1 S O ^ l a M i * ii s Is -2 U "5 " X s s ; O p M S 5? O M « V •? >: « > V -s? s f § B " o o u . •8 i? S 'I: •f | i a o ?! STjtjT-'*T rC " L J J J ~* 5 c* h «* - 6 o ' » u . a-1 CN T 1 ss-.a-f * *w z 3 ri «5. 2 r) 9, «q *• o t w VI •g s i ?. c to 163 A3. SENSIBLE HEAT FLUXES IN THE RIVIERA V A L L E Y In this appendix, the observed and modeled turbulent sensible heat fluxes at 10 sites in the Riviera Valley (listed in Table A3.1) on 25 August 1999 are compared (Fig. A3.1). Observations are shown with the squares, modeled sensible heat are shown with the dashed line. The shaded area indicates the range of modeled sensible heat fluxes at 9 grid points surrounding the observation site. For a discussion, see chapter 2. Kinematic rather than dynamic sensible heat fluxes are compared. The dynamic turbulent sensible heat flux H can be written as: H = pcp w'T', where p is the air density, cp the specific heat at constant pressure and w'T' is the kinematic sensible heat flux. The air density is dependent on air pressure and temperature while the specific heat is slightly dependent on the moisture content of air. Thus, a comparison of dynamic heat fluxes would also include possible errors due to differences between observed and simulated values of other variables. Besides, sonic anemometers measure kinematic heat flux, not dynamic heat flux. For these reasons, kinematic sensible heat flux was used in the comparisons. Table A3 .1 . Characteristics of the ten surface stations measuring turbulence. Site Location Measurement Surface characteristics lat(°N),lon(°E),elev (m asl) height (m agl) A l 46.2572, 9.0131, 250 3.56 Valley floor, mixed agriculture A2 46.2500, 9.0153, 250 1.15 Valley floor, mixed agriculture B 46.2647, 9.0311, 760 23.78 Slope; forest (mean height of trees ~ 15m) C 46.2494, 9.0056, 340 6.33 Slope; vineyard D 46.2467, 9.0269, 256 2.62 Valley floor; mixed agriculture El 46.2667, 9.0372, 1060 12.70 Slope; meadow E2 46.2706, 9.0364, 1030 22.68 Slope; forest (mean height of trees ~ 13m) Fl 46.2700, 9.0553, 1750 6.30 Slope; sparse vegetation F2 46.2728, 9.0608, 2110 1.30 Slope; shrub (~75m below ridgeline) G 46.2742, 9.0317, 870 5.25 Slope; forest (bridge over small tributary valley) 164 0.30- site A1 0.30- site A2 0.20-lux (Kms 0.20-0.10 s (ble heal 1 0.10 0.00 c ^ 'Sri B = B = O ^ B " ^ 1 0 00 -0.10 -0.10 1 1 1 1 1 1 1 1 1 • 1 1 • 1 1 1 1 1 1 1 • ' ' 10 12 14 time (UTC) 6 8 10 12 14 time (UTC) 16 18 20 | 0.10 £ a I o.oo -0.10 E — 0.20 0.10 0.00 -0.10 site B ---J / \ • / / K \ in a -g H g-g n u D^]— 4 6 8 10 12 14 time (UTC) 16 18 20 site D w s ^ E 0.20 X CD J= 0.10 a; JD c 3 0.00 site C 10 12 14 time (UTC) 16 18 20 10 12 14 time (UTC) 10 12 14 time (UTC) | 0 .10 S •35 I 0 0 0 -0 .10 - site E2 -- V / / _\ I 1 < 1 \ \ > ra 1 PI > ^ & t ^ t - B u u 10 12 14 time (UTC) 10 12 14 time (UTC) Fig. A3 .1 . Comparison between observed and modeled sensible heat fluxes at 10 different sites in the Riviera Valley (listed in Table A3.1) on 25 August 1999. Squares are observations, dashed lines are the model output at the grid point closest to the observation. The shaded area represents the range of sensible heat fluxes modeled at the nine model grid points surrounding the observation location. 165 A4. AIRCRAFT MEASUREMENTS DURING MAP-RIVIERA In this Appendix, cross sections of potential temperature, specific humidity, cross-valley wind component, and along-valley wind component wil l be presented from all aircraft data on 25 August 1999. Before the data could be visualized, the data had to be gridded. The gridding method that was applied will be discussed first. A4.1. GRIDDING METHOD There are a number of methods that can be used for gridding, that is, estimating the value at the nodes of a regular grid given nearby sample values on an irregular grid. Some of the more common methods are nearest neighbour interpolation, weighted interpolation, polynomial interpolation, kriging, and Delauney triangulation with linear interpolation. For a general overview of interpolation methods, Watson (1992) can be consulted. For an overview of some methods used in meteorological applications, see Goodin etal. (1979). The Delauney triangulation method with linear interpolation has been used to visualize the aircraft data in this dissertation. In brief, the method works by creating triangles (called Delaunay triangles) by drawing lines between data points. The original data points are connected in such a way that no triangle edges are intersected by other triangles. The result is a patchwork of triangular faces over the extent of the grid. Since Delaunay triangulations have the property that the circumcircle of any triangle in the triangulation contains no other sample point in its interior, interpolated values at the regular grid nodes (which have to be defined) are only computed from nearby points. One advantage of Delauney triangulation is that parameters that are sometimes difficult to determine such as a 'radius of influence' in weighted interpolation and 'range', 'nugget', and 'sil' in kriging are not required. One particular situation in which many other techniques perform poorly is when there is a mixture of regions of high and low density sampling. Triangulation based methods honour this situation by giving a large number of triangles and hence more detail to the highly sampled regions and large triangles, hence 166 less detail, to the regions with a few samples. More detailed information can be found in Watson (1992). Figures A4.1 and A4.2 demonstrate that this method performs well for the interpolation of data along a flight track. Fig A4.1a and A4.2a show interpolated cross sections of potential temperature and wind speed, respectively, from high resolution model output. The model output was sampled along a flight track and the Delauney triangulation method was applied to the resulting data set. Values were linearly interpolated to a regular grid, defined with 150 m horizontal grid spacing and 30 m vertical grid spacing. These values gave best results and were used in all the contour plots in this dissertation. The result is shown in Fig. A4.1b and A4.2b where the flight track is shown by the dotted line. From comparison of A4.1a and b and A4.2a and b, it can be seen that similar main features (such as a wind maximum on the eastern valley side) appear in the contour plots. This indicates that the Delauney triangulation method with linear interpolation does not introduce artificial results that may be misinterpreted. 0 i _ , — , — , — , — , — , — . — , — , — . — . — , — . — , — _ i 0 i — • — • — • — • — • — • — • — • — • — • — • — • — • — • — • — 1 -4000 -2000 0 2000 4000 -4000 -2000 0 2000 4000 horizontal distance (m) horizontal distance (m) Fig A4.1. Interpolated cross section of potential temperature from (a) model output and (b) model output sampled at locations along a flight track (shown by the dotted line) and interpolated according to the Delauney triangulatian method. Contour lines are in K and drawn every IK. -4000 -2000 0 2000 4000 -4000 -2000 0 2000 4000 horizontal distance (m) horizontal distance (m) Fig. A4.2. As Fig. A4.1 but for wind speed. Contour lines are in m s"1 and drawn every 1 m s"1. 167 A4.2. A IRCRAFT CROSS SECTIONS In this appendix, interpolated cross sections are shown for potential temperature, specific humidity, cross-valley wind component and along-valley wind component, based on aircraft data. Contour lines are drawn for intervals of 1 K (for temperature), 1 g kg"1 (for specific humidity), and 1 m s"1 (for wind speed). The 0 m s"1 isoline in the cross sections of the wind component is drawn with a thick line, dashed lines indicate contour lines with negative values. The abscissa indicates the horizontal distance from the point of reference at Claro (see Fig. 2.1). In the figure captions of the interpolated cross sections, only the starting and ending times of the aircraft traverses are indicated. Figures A4.4, A4.7, A 4 . l l , and A4.16 show the location of the data points used in the construction of the interpolated cross sections. The starting and ending times of the aircraft traverses, the orientations of the aircraft traverses (across/along-valley), and the figure numbers where each cross section can be found are shown in Table A4.1. For the cross-valley cross sections, data points up to 2 km north and south of the cross section were taken. For the along-valley cross sections, a distinction was made between measurements taken at the western and at the eastern side of the valley. For details, see chapter 2. Table A4 .1 . Specifics of the time and the orientation of the interpolated cross sections, and the figure number where each cross section can be found. Morning/Afternoon Orientation Time (decimal) Time (HHMM) Figure morning across 0710-0770 0706-0742 A4.1 across 0885-0900 0851-0900 A4.2 across 0914-0931 0908-0919 A4.3 along 0770-0880 0742-0848 A4.5+A4.6 afternoon across 1130-1230 1118-1218 A4.8 across 1350-1370 1330-1342 A4.9 across 1510-1580 1506-1548 A4.10 along 1237-1350 1222-1330 A4.12+A4.13 along 1373-1517 1344-1510 A4.14+A4.15 168 -4000 -2000 0 2000 4000 horizontal distance (m) -4000 (d) -2000 0 2000 horizontal distance (m) -2000 0 2000 horizontal distance (m) 4000 Fig. A4.3. Interpolated cross sections of potential temperature (K), specific humidity (g kg"1), cross-valley wind component and along-valley wind component ( m s"1) for 0706-0742 UTC. 1 6 9 -4000 -2000 0 2000 4000 -4000 -2000 0 2000 horizontal distance (m) horizontal distance (m) Fig. A4 .5. Same as Fig. A4.3 but for 0908-0919 UTC. o | o l -4000 -2000 0 2000 4000 -4000 -2000 0 2000 4000 horizontal distance (m) horizontal distance (m) (C) -4000 -2000 0 2000 4000 horizontal distance (m) Fig. A4.6. Location of the data points of the cross sections in Fig. A4.3 (a), Fig. A4.4 (b), and Fig. A4.5 (c). 170 -5000 0 5000 10000 -5000 0 5000 10000 horizontal distance (m) horizontal distance (m) -5000 0 5000 10000 -5000 0 5000 10000 horizontal distance (m) horizontal distance (m) Fig. A4.7. Same as Fig. A4.3 but for 0742-0848 UTC. Only data points on the western side of the valley are considered. horizontal distance (m) horizontal distance (m) Fig. A4.8. Same as Fig. A4.3 but for 0742-0848 UTC. Only data points on the eastern side of the valley are considered. 171 (a) WEST (b), EAST 0 5000 horizontal distance (m) 0 5000 horizontal distance (m) Fig. A4.9. Location of the data points of the cross sections in Fig. A4.7 (a), and Fig. A4.8 (b). 172 horizontal distance (m) horizontal distance (m) Fig. A4.10. Same as Fig. A4.3 but for 1118 -1218 UTC. Fig. A 4 . l l . Same as Fig. A4.3 but for 1330-1342 UTC. 173 Fig. A4.12. Same as Fig. A4.3 but for 1506 -1548 UTC. -4000 -2000 0 2000 4000 -4000 -2000 0 2000 4000 horizontal distance (m) horizontal distance (m) o I -4000 -2000 0 2000 4000 horizontal distance (m) Fig. A4.13. Location of the data points of the cross sections in Fig. A4.10 (a), Fig. A4.11 (b), and Fig. A4.12 (c). 174 Fig. A4.14. Same as Fig. A4.3 but for 1222-1330 UTC. Only data points on the western side of the valley are considered. 175 -5000 0 5000 10000 -5000 0 5000 10000 horizontal distance (m) horizontal distance (m) Fig. A4.17. Same as Fig. A4.3 but for 1344-1510 UTC. Only data points on the eastern side of the valley are considered. 176 (a) WEST (b) EAST 0 5000 10000 horizontal distance (m) 0 5000 10000 horizontal distance (m) (C), WEST (d), EAST -5000 0 5000 10000 horizontal distance (m) 0 5000 10000 horizontal distance (m) Fig. A4.18. Location of the data points of the cross sections in Fig. A4.14 (a), Fig. A4.15 (b), Fig. A4.16 (c), and A4.17 (d). 177 A5. CBL HEIGHT DETERMINATION WITH THE /{/-METHOD C B L heights were determined in the M A P and S T A A A R T E case studies from R A M S output using the Richardson number (Ri-) method described by Vogelezang and Holtslag (1996). Ri is calculated as: R l = (g/e ,)(P.-0(A-P ( A 5 i l ) (uh-us)2+(vh-vs)2+\00u2 where g is the acceleration due to gravity, 0 v / l , uh and vh are the virtual potential temperature and the wind speed components at the boundary layer height h (asl). 0 r a , us, and vs are the virtual potential temperature and the wind speed components at the first model level (~25 m). zs equals the height of the first model level asl. u, is the friction velocity at the surface and Q r a =QVS + A 9 where A 0 is a temperature excess following Troen and Mahrt (1986). A 0 is calculated as: A 0 = £ L ^ e A (A5 .2 ) W L where w'Q 'vs is the surface virtual sensible heat flux and wm is a turbulent velocity scale: wn = (ul + c2 wl ) (A5 .3 ) and where u* is the friction velocity and w„ is the convective scaling velocity defined by: w,=[(g/QJ(h-zs)w'Q ' v s ] 1 / 3 (A5 .4 ) For c i and c 2 in Eqs. A 5 . 2 and A 5 . 3 , values of 8.5 and 0.6 were used, respectively (Holtslag et a l , 1995). The temperature excess can be regarded as a measure of the strength of convective thermals (Vogelezang and Holtslag, 1996). Typical excess temperatures from the model output were between 0.1 and 0.7 K . Ri is calculated at each model level starting from the surface, and the C B L height is derived by linear interpolation between the level where Ri becomes larger than 0.25, and the level below. 178 A6. LAND USE IN THE STAAARTE'97 INVESTIGATION AREA Shown in Fig. A6.1 is the land use in the innermost domain in the STAAARTE'97 modeling case study of chapter 4. The data is based on 1-km USGS land use data. Fig. A 6 . 1 . Distribution of land use classes in the S T A A A R T E ' 9 7 investigation area 179 A7. COMPARISON OF ECMWF CBL HEIGHTS DURING STAAARTE WITH OBSERVED A L HEIGHTS AND CBL HEIGHTS FROM RAMS E C M W F has recently started to provide boundary layer heights (stable, neutral, and convective) determined using Vogelezang and Holtslag's (1996) i?z'-method. (Beljaars, personal communication). To investigate whether the E C M W F C B L heights give reasonable values, we compare these with the C B L heights determined from R A M S output (with the same i?/-method). The result can be seen in Fig. A7.1. 10 12 14 Time (UTC) Fig. A 7 . 1 . Development of the C B L in a 0.5° x 0.5° area around the JFJ from R A M S (squares) and E C M W F (diamonds). The stars depict the C B L height determined subjectively from E C M W F potential temperature profiles for 1200 and 1600 U T C (Fig. A7.2). The values for E C M W F are for a grid point located at 46.5°N, 8°E (ECMWF model resolution is 0.5°x 0.5°). The values from R A M S are averaged over the innermost R A M S grid (0.5°x 0.5°). E C M W F C B L heights are clearly larger than R A M S C B L heights with differences up to 1000 m in the mid-afternoon. To find the reason for this overestimation, a closer look was taken at the E C M W F potential temperature profiles valid for the grid point nearest the JFJ (8° E, 46.5° N). Examples of these profiles are shown in Fig. A7.2 for 1200 and 1600 UTC. The C B L heights that E C M W F provides are also indicated by the horizontal lines. It can be seen that these heights are much higher than one would get by determining the C B L heights subjectively (e.g. by locating the base of the inversion or via the parcel method). While the E C M W F C B L heights are 4000 180 m and 3740m at 12 and 16 -UTC, respectively, visually determined C B L heights are around 3300 and 3100, respectively. These heights are indicated in Figure A7.1 by the stars. These heights are in good agreement with R A M S C B L heights. Thus, the overestimation is probably due to an error in the E C M W F boundary layer height routine as confirmed by Beljaars (personal communication). 310 315 320 Potential Temperature (K) Fig A 7 . 2 . Vertical profiles of potential temperature from E C M W F at 8E, 46.5N at 1200 U T C (filled squares) and 1600 U T C (open squares) The C B L height from E C M W F is indicated by a dashed line for 1200 U T C and by a solid line at 1600 UTC. 181 A8. GLACIER AND CLOUD EFFECTS ON CBL HEIGHT DURING STAAARTE Sensitivity simulations were performed to investigate how the C B L heights in the S T A A A R T E investigation area (see chapter 4) are affected by the presence of glaciers and clouds. The presence of glaciers is taken into account as one of the land use types in the surface parameterization scheme. The model shows the existence of upslope flows and the development of a C B L in the area in which the glacier is embedded and it is somewhat questionable whether the effects of the glacier are correctly parameterized. To our knowledge, the performance R A M S for ice covered valley floors under daytime fair weather conditions has not been "evaluated. Also in the current study, no observations are available over the glacier for comparison with model output. Observations over glaciers are mainly concerned with katabatic flows over the area (e.g. Van den Broeke, 1997). These occur in a shallow layer close to the ice surface. On the other hand, glaciers are embedded within valleys in which thermally driven upvalley flows may prevail (Geiger, 1965), especially i f the sidewalls are snow free and i f there is a change of the topographic amplification factor (Whiteman, 1990) along the valley. Thus upvalley flows may occur above a shallow katabatic flow over glaciated surfaces. Furthermore, the protruding ice free ridges and sidewalls may trigger vigorous convection (WMO, 1993). The net results of snow- and ice covered surfaces combined with ice free sidewalls and ridges may be an overall positive sensible heat transfer. This would result in the development of a C B L over the area as predicted by the model. Some of the parameters that characterize the glacier surface (see Table A l . l ) , such as albedo, were changed (from 0.4 to 0.9) but these were found to have only a minor impact on the development of the C B L and the mesoscale circulations in the area. As noticed before, cumulus clouds developed during the experimental day, as indicated in some of the lidar images by white vertical lines (e.g., Fig 4.4b). To investigate the effects of these clouds, an extra simulation was made in which cumulus clouds were allowed to develop in the model. That is, water vapour was not treated as a passive scalar anymore and condensation was allowed. Nonetheless, relative humidities in the simulations at high elevation stayed well below 100%, condensation did not develop, and C B L heights were similar to those in the reference run. 182 A9. LIDAR MEASUREMENTS DURING STAAARTE In this appendix, some background information is given on lidar measurements and lidar data from all flight legs during S T A A A R T E are presented. A9.1. LIDAR BASICS Lidar is an acronym for light detecting and ranging. A lidar system transmits a short pulse of light into the atmosphere. Part of the light pulse is received by the lidar system after it is reflected from a distant target or from atmospheric constituents such as molecules, aerosols, clouds, or dust. The laser radiation interacts with these constituents, causing alterations in its intensity and wavelength according to the strength of the optical interaction and the concentration of the interacting species in the atmosphere. Consequently, information on the composition and physical state of the atmosphere can be deduced from the lidar data. Because the light pulse travels at the speed of light, it is possible to convert time to range and consider the lidar signal to be backscatter intensity as a function of range. (Schwiesow, 1984; Killinger and Menyuk, 1987). The aerosol backscatter signal ratio in the S T A A A R T E figures is defined as: B=[Ba+Bm]/[Bm] where Ba is the scattering due to aerosols and Bm the molecular scattering. B is affected by the aerosol concentration, aerosol size, lidar wavelength X, and relative humidity. In general, the variation of molecular backscatter with wavelength goes as: Bm cc A,"4 , while Ba decreases less rapidly with increasing wavelength. This means that the aerosol to molecular backscatter coefficient ratio increases with increasing wavelength. Therefore, aerosol lidars tend to operate better at longer wavelengths by avoiding the masking effects of molecular scatter. The lidar uses X =532 nm in S T A A A R T E and X -1064 nm in PACIFIC'93. The aerosol backscatter is dominated by aerosols with a diameter d = X/TI (Schwiesow, 1984), i.e., about 180 nm for X =532 nm. The aerosol sensitivity decreases rapidly with decreasing diameter (<x d"6) so that aerosols with diameter less than 100 nm 183 are hardly seen. Furthermore, because aerosols grow in size at relative humidities above about 85%, B is strongly affected by relative humidity. For more background information on lidars, see Schwiesow (1984) and Killinger and Menyuk (1987). A9.2. LIDAR CROSS SECTIONS FROM STAAARTE In this section, an overview of all available lidar cross sections during STAAARTE'97 is presented. In total, 35 flight legs were flown, 17 in the morning and 18 in the afternoon. The orientation of the flight path and the starting and ending times of each are shown in Table A9.1 for morning flights and A9.2 for afternoon flights. The table also shows the A L heights determined semi-objectively with the method presented and subjectively by tracing visually the boundary between high and low backscatter ratio (see chapter 4). The lidar cross sections from the morning and afternoon flights are shown in Figs. A9.1 and A9.3, respectively. The location of each individual flight leg during the morning and afternoon flights is shown in Figs. A9.2 and A9.4, respectively. 184 Table A9.1. Details of flight legs 1-17 (morning flights). Orientation is either parallel (P) or perpendicular (O) to the mountain divide. The time indicates the approximate starting and ending times of each flight leg. A L heights are leg-averaged values. These could not be determined for some flight legs. leg Orientation Time (UTC) Start - end objective A L height (m) subjective A L height (m) 1 P 0628 - 0631 - -2 P 0634 - 0636 2054 2077 3 P 0639-0641 - -4 P 0643 - 0646 2273 2193 5 P 0648 - 0652 - -6 P 0653 - 0656 2524 2608 7 0 0702 - 0707 2449 2689 8 0 0711 -0716 2765 2615 9 0 0718 - 0724 2436 2687 10 0 0726 - 0733 2571 2576 11 0 0735 -0741 2439 2797 12 0 0743 - 0749 2738 2709 13 P 0803 - 0806 - -14 P 0811 -0815 2884 2878 15 0 0822 - 0826 2661 2890 16 0 0913 -0917 2701 2976 17 0 0919 -0925 2749 2979 Table A9.2. As Table A9.1 but for flight legs 18-35 (afternoon flights). leg Orientation Time (UTC) objective subjective Start - end A L height (m) A L height (m) 18 P 1247- 1248 - -19 P 1250 - 1253 2994 3114 20 P 1256 - 1257 - -21 P 1259- 1302 3437 3487 22 P 1304- 1306 - -23 P 1308 - 1311 3483 3614 24 O 1317-1322 3903 4001 25 O 1324- 1330 3764 3839 26 o 1333 - 1339 3852 4005 27 o 1340- 1347 3879 4003 28 o 1350 - 1355 3979 4037 29 o 1358 - 1404 3937 4114 30 p 1409- 1413 3987 4056 31 p 1416-1420 4031 4106 32 p 1425 - 1428 4129 4161 33 0 1434 - 1438 3958 3981 34 0 1520 - 1525 4077 4129 35 0 1527 - 1533 4123 4212 185 horizontal distance (km) horizontal distance (km) Latitude fN) 46.30 46.35 46.40 46.45 46.50 46.55 46.60 46.65 46.70 20 30 40 horizontal distance (km) 20 30 40 horizontal distance (km) 20 2.5 Backscatter ratio {-) Fig. A9.1. Cross sections of lidar backscatter ratio for flight legs 1 to 17. These legs were flown in the morning of 30 July 1997. The times of each leg can be found in Table A9.1. The exact location of each leg is depicted in Fig. A9.2. 186 Latitude ("N) 48.30 46.35 48.40 46.45 48.50 48.55 48.60 46.85 46.70 Latitude fN) 46.40 46.45 46.50 46.55 46.60 46.65 46.70 48.75 46.80 i. ^^x-^-^^..^-^-^ 5 n leg 10 I less i ' ° - F ? 20 30 40 50 60 horizontal distance (km) 10 20 30 40 50 60 horizontal distance (km) 20 30 40 horizontal distance (km) 20 30 40 horizontal distance (km) 2.0 Backscatter ratio (-) Fig. A9.1. ctd. 187 20 30 40 horizontal distance (km) 1.0 15 2.0 2.5 3 0 Backscatter ratio (-) Fig. A9.1. ctd. 47.0 46.8 Height (m asl) • 0-1000 j 1=1 1000-2000 <j n 2000-3000 • >3000 8.0 8.2 Longitude (°E) 0 8.2 Longitude (°E) Fig. A9.2. Topography of the S T A A A R T E investigation area along with the location of flight legs 1-10 (a) and flight legs 11-17(b). 188 Latitude f N) 10 horizontal distance (km) 10 20 horizontal distance (km) Latitude f N) horizontal distance (km) horizontal distance (km) Latitude fN) 46.60 horizontal distance (km) 10 20 horizontal distance (km) Latitude f N) Latitude f N) 46.35 46.40 46.45 46.50 46.55 46.60 46.65 46.35 46.40 46.45 46.50 46.55 46.60 46.65 46.70 46.75 0 10 20 30 40 50 0 10 2 0 3 0 4 0 5 0 6 0 horizontal distance (km) horizontal distance (km) ^0 2.5 ratio (-) Fig. A9.3. Cross sections of lidar backscatter ratio for flight legs 18 to 34. These legs were flown in the afternoon of 30 July 1997. The times of each leg can be found in Table A9.2. The exact location of each leg is depicted in Fig. A9.4. 189 20 30 horizontal distance (km) Latitude f N) 46.35 46.40 46.45 46.50 46.55 46.60 46.65 46.70 20 30 horizontal distance (km) Latitude (°N) 46.30 46.35 46.40 46.45 46.50 46.55 46.60 46.65 46.70 n 'agZJ 20 30 40 horizontal distance (km) 20 30 40 horizontal distance (km) horizontal distance (km) 20 30 horizontal distance (km) Latitude CN) Latitude (*N) 46.50 46.55 46.60 46.65 46.40 46.45 46.50 46.55 46.60 46.65 46.70 horizontal distance (km) horizontal distance (km) Fig. A9.3. ctd. 190 20 30 horizontal distance (km) Fig. A9.3. ctd. 1.0 1.5 2.0 2.5 3.0 Backscatter ratio (-) 46.8 8.0 8.2 Longitude (°E) T 7.8 8.0 8.2 Longitude (°E) 84 Fig. A9.4. Topography of the S T A A A R T E investigation area along with the location of flight legs 18-27 (a) and flight legs 28-35 (b). 191 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

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:
https://iiif.library.ubc.ca/presentation/dsp.831.1-0052567/manifest

Comment

Related Items