Boundary Layer Evolution within a Canyonland Basin. Part II: Numerical Simulations of Nocturnal Flows and Heat Budgets

View More View Less
  • 1 Pacific Northwest National Laboratory, Richland Washington
© Get Permissions
Full access

Abstract

A mesoscale model is used to simulate the nocturnal evolution of the wind and temperature fields within a small, elliptical basin located in western Colorado that has a drainage area of about 84 km2. The numerical results are compared to observed profiles of wind and potential temperature. The thermal forcing of the basin wind system and the sources of air that support the local circulations are determined. Individual terms of the basin atmospheric heat budget are also calculated from the model results.

The model is able to reproduce key features of the observed potential temperature profiles over the basin floor and winds exiting the basin through the narrow canyon that drains the basin. Complex circulations are produced within the basin atmosphere as a result of the convergence of drainage flows from the basin sidewalls. The strength of the sidewall drainage flow varies around the basin and is a function of the source area above the basin, the local topography, and the ambient winds. Flows on the basin floor are affected primarly by the drainage winds from the northern part of the basin. The near-surface sidewall drainage flows converge within the southern portion of the basin, producing a counterclockwise eddy during most of the evening. Evaluation of the individual terms of the atmospheric heat budget show that the forcing due to advection and turbulent diffusion is significantly larger above the sidewalls than over the basin floor; therefore, measurements made over the basin floor would not be representative of the basin as a whole. The cooling in the center of the basin results from the local radiative flux divergence and the advection of cold air from the sidewalls, and the cooling above the basin sidewalls is due primarily to turbulent sensible heat flux divergence. A high rate of atmospheric cooling occurs within the basin throughout the evening, although the strongest cooling occurs in the early evening hours. Sensitivity tests show that the thermal structure, circulations, and rate of cooling can be significantly affected by ambient wind direction and, to a lesser extent, vegetation coverage.

Abstract

A mesoscale model is used to simulate the nocturnal evolution of the wind and temperature fields within a small, elliptical basin located in western Colorado that has a drainage area of about 84 km2. The numerical results are compared to observed profiles of wind and potential temperature. The thermal forcing of the basin wind system and the sources of air that support the local circulations are determined. Individual terms of the basin atmospheric heat budget are also calculated from the model results.

The model is able to reproduce key features of the observed potential temperature profiles over the basin floor and winds exiting the basin through the narrow canyon that drains the basin. Complex circulations are produced within the basin atmosphere as a result of the convergence of drainage flows from the basin sidewalls. The strength of the sidewall drainage flow varies around the basin and is a function of the source area above the basin, the local topography, and the ambient winds. Flows on the basin floor are affected primarly by the drainage winds from the northern part of the basin. The near-surface sidewall drainage flows converge within the southern portion of the basin, producing a counterclockwise eddy during most of the evening. Evaluation of the individual terms of the atmospheric heat budget show that the forcing due to advection and turbulent diffusion is significantly larger above the sidewalls than over the basin floor; therefore, measurements made over the basin floor would not be representative of the basin as a whole. The cooling in the center of the basin results from the local radiative flux divergence and the advection of cold air from the sidewalls, and the cooling above the basin sidewalls is due primarily to turbulent sensible heat flux divergence. A high rate of atmospheric cooling occurs within the basin throughout the evening, although the strongest cooling occurs in the early evening hours. Sensitivity tests show that the thermal structure, circulations, and rate of cooling can be significantly affected by ambient wind direction and, to a lesser extent, vegetation coverage.

2162 JOURNAL OF APPLIED METEOROLOGY VoLtns'm35Boundary Layer Evolution within a Canyonland Basin. Part II: Numerical Simulations of Nocturnal Flows and Heat Budgets$EROM~ D. FAST, SHIYUAN ZHONG, AND C. DAVID WHITEMANPacific Northwest National Laboratory, Richland, Washington(Manuscript received 11 December 1995, in final form 12 April 1996)ABSTRACT A mesoseale model is used to simulate the nocturnal evolution of the wind and temperature fields within asmall, elliptical basin located in western Colorado that has a drainage area of about 84 km~. The numericalresults are compared to observed profiles of wind and potential temperature. The thermal forcing of the basinwind system and the sources of air that support the local circulations are determined. Individual terms of thebasin atmospheric heat budget are also calculated from the model results. The model is able to reproduce key features of the observed potential temperature profiles over the basin floorand winds exiting the basin through the narrow canyon that drains the basin. Complex circulations are producedwithin the basin atmosphere as a result of the convergence of drainage flows from the basin sidewalls. Thestrength of the sidewall drainage flow varies around the basin and is a function of the source area above thebasin, the local topography, and the ambient winds. Flows on the basin floor are affected primarily by thedrainage winds from the northern part of the basin. The near-surface sidewall drainage flows converge withinthe southern portion of the basin, producing a counterclockwise eddy during most of the evening. Evaluation ofthe individual terms of the atmospheric heat budget show that the forcing due to advection and turbulent diffusionis significantly larger above the sidewalls than over the basin floor; therefore, measurements made over the basinfloor would not be representative of the basin as a whole. The cooling in the center of the basin results from thelocal radiative flux divergence and the advection of cold air from the sidewalls, and the cooling above the basinsidewalls is due primarily to turbulent sensible heat flux divergence. A high rate of atmospheric cooling occurswithin the basin throughout the evening, although the strongest cooling occum in the early evening hours.Sensitivity tests show that the thermal structure, circulations, and rate of cooling can be significantly affectedby ambient wind direction and, to a lesser extent, vegetation coverage.1. Introduction In a companion paper, Whiteman et al. ( 1996, hereafter referred to as W96) evaluated the mass, heat, andmoisture budgets for an atmospheric control volume inColorado's Sinbad Basin from tethered balloon observations and surface measurements during a 16.5-h period beginning at 1800 mountain standard time (MST)15 July 1988. A high rate of atmospheric cooling continued in this basin throughout the entire night that wasattributed primarily to turbulent sensible heat flux divergence and, to a lesser extent, radiative flux divergence. Several unexpected features appeared in their analysis. One unexpected feature was the large discrepancybetween the sensible heat flux divergence from the basin control volume, calculated as a residual of the atmospheric heat budget, and the sensible heat flux di Corresponding author address.' Dr. Jerome D. Fast, Battelle, Pacific Northwest National Laboratory, P.O. Box 999, M$IN K9-30,Richland, WA 99352.E-mail: jd_fast@pnl.govvergence estimated from surface energy budget measurements on the basin floor. The heat budget analysisindicated that a large sensible heat flux divergence mustoccur from the basin control volume during nighttime,but the observations showed that there was no appreciable downward turbulent sensible heat flux at the basin floor. Another unexpected feature was the observation of large mass fluxes through the narrow exit canyon. The Sinbad Basin was selected for the fieldexperiment partially because it was assumed that thebasin outflow would be weak; however, the mass fluxeswere comparable to fluxes observed in well-drainedvalleys. W96 assumed that the strong nocturnal coolingwithin the basin was caused by enhanced downwardturbulent sensible heat fluxes over the basin sidewallsassociated with downslope drainage flows and that theflow through the canyon was the result of a large horizontal pressure gradient that developed between thebasin and its surroundings due to stronger coolingwithin the basin. Unfortunately, no observations wereavailable on the basin sidewalls or at other locationsoutside of the basin to provide direct evidence to support these hypotheses. Understanding these features rec 1996 American Meteorological SocietyDECEMBER 1996 FAST ET AL. 2163A15.25 km ~I FIG. 1. Topography employed by the mesoscale model on (a) grid 1 and (b) grid 2 along with the locations oftethersonde and airsonde measurements taken at the basin and Salt Wash sites (filled circles) and tracer release locationsused by the dispersion model (open circles).quires measurements at many different sites, a difficultand expensive proposition for most basins or valleys.On the other hand, these features can easily be examined by dynamic models that can resolve the slopingsidewalls of the basin. Numerical experiments are capable of providing a new understanding of basin heatbudgets because datasets obtained from models represent dynamically balanced realizations of the flows. A high-resolution three-dimensional mesoscale numerical model is used in this study to simulate the evolution of the thermally driven flow system in the SinbadBasin for the case presented in W96. The predictedwind and temperature profiles are compared to the tethersonde observations. The wind and temperature fieldsat other locations are presented to illustrate the complexity of the flows, behavior of the sidewall drainageflows, and preferential flow patterns resulting from topographic forcing that may develop within the basin.The mechanisms responsible for the strong windswithin the narrow canyon will also be addressed. Asdescribed by W96, several assumptions are required toobtain the individual terms of the heat budget usingobservational data. However, assumptions regardingclosure of the heat budget are not necessary in this numerical modeling study. The terms of the heat budgetequation are calculated three-dimensionally during theevolution of the nocturnal boundary layer to provideinformation regarding the role of slope flows on thebasin heat budget. Some of the assumptions employedby W96 will be addressed by examining individualterms of the simulated heat budget. A Lagrangian particle dispersion model is also used in this study to depictthe circulations within the basin so that characteristicsof the drainage flows can be better understood. Moreover, this numerical modeling study will investigate theeffect of large-scale ambient winds and surface characteristics on the slope flows and the basin heat budget. The mesoscale model and the dispersion model aredescribed in section 2. Section 3 describes the numerical experiments, and section 4 compares some of themodel results with observations. Section 5 discusses thetemperature fields and the basin energy budget further,and section 6 presents the conclusions of this study.2. Model descriptiona. Atmospheric model The Regional Atmospheric Modeling System (RAMS,version 3a) described by Pielke et al. (1992) is usedto predict the small-scale flows within and around theSinbad Basin. In this study, the turbulence parameterization consists of a level 2.5 closure scheme witha prognostic turbulence kinetic energy equation asproposed by Mellor and Yamada (1982) and modified for the case of growing turbulence according toHelfand and Labraga (1988). Since cloud processesare not important for this case, the cumulus and microphysical parameterizations are not activated, sothat water vapor is treated as a passive scalar. TheMahrer and Pielke (1977) shortwave parameterization and the Chen and Cotton (1983) longwave parameterization are used to determine the heating orcooling due to radiative fluxes. A prognostic soilvegetation module calculates the diurnal variationsof temperature and moisture at the ground-atmospheric interface. Albedo, emissivity, leaf-area index, fractional coverage, roughness length, displace2164JOURNAL OF APPLIED METEOROLOGYTABLE 1. Summary of numerical experiments.VOLUME 35Name Initial wind profile Surface characteristicsPrimary experiments:Grand Junction sounding and basin site tethersonde at0000 UTC 15 July 1988Same as experiment 1 except wind speed set to 0.1 m s-~Sensitivity experiments: 3 Same as experiment 1 4 Same as experiment 1 5 Same as experiment 1 except northwest wind direction within 3.3 km of basin floor 6 Same as experiment I except southeast wind direction within 3.3 km of basin floor 7 Same as experiment 1 except northeast wind direction within 3.3 km of basin floorUniform vegetation type--desert shrub; uniform soil type-loamy sandSame as experiment 1Uniform vegetation type--desert; uniform soil type--loamysandMixed vegetation type--desert and desert shrub; mixed soiltype--loamy sand and claySame as experiment ISame as experiment ISame as experiment 1ment height, and root fraction all depend upon thetype of vegetation. The Biosphere-AtmosphereTransfer Scheme (BATS) vegetation classifications(Dickinson et al. 1986) are used. Turbulent sensibleheat, latent heat, and momentum fluxes in the surfacelayer are then based on similarity equations (Louis1979). Eleven soil levels are used down to a depthof 1 m below the surface with a variable grid spacing. The nested grid configuration employed to resolvethe small-scale flows within and around Sinbad Basinis shown in Fig. 1. A horizontal grid spacing of 750and 250 m is used for grids 1 and 2, respectively. Thetopography for both grids is derived from a 3" (about90 m) terrain dataset and is then smoothed with a silhouette-averaging scheme that preserves realistic topographic heights. Both grids employ a stretched, vertical, terrain-following coordinate with a grid spacingof 30 m near the surface that increases to 1000 m nearthe model top at an elevation of 12.4 km mean sea level(MSL). Due to the staggered coordinate system, thelowest grid point is located about 12.5 m above groundlevel (AGL). With this vertical resolution, 14 gridpoints are positioned within 1000 m of the ground toresolve the nocturnal slope flows. While the horizontalgrid spacing in this study is smaller than many mesoscale model applications, RAMS has been shown toreproduce many observed characteristics of flows incomplex terrain (Fast 1995; Poulos and Bossert 1995)using grid spacings slightly larger than most large-eddysimulation studies. As seen in Fig. 1, the Sinbad Basin is an ellipticalbasin that has only one exit, the deep and narrow SaltWash Canyon. The Salt Creek flows through the SaltWash Canyon and empties into the Dolores River thatruns from south to north through the eastern portion ofthe domain (Fig. la). Roc Creek flows down the narrow valley just south of the basin between SinbadRidge and Carpenter Ridge (Fig. lb). Since the terrainelevations are generally higher in the western portionof the domain, westerly nocturnal drainage winds maydevelop and flow into the lower terrain around the Dolores River when the synoptic forcing is weak. Thestrength and direction of these drainage winds are expected to be affected by the Sinbad Basin and the othersmall-scale terrain features in the area. It should benoted that the Salt Wash Canyon entrance in the model(Fig. lb) is wider and smoother than the actual entrance given on topographic maps (W96) because the3" terrain data does not resolve this feature well. Twoobservation sites (the basin and Salt Wash sites) arealso shown in Fig. lb.b. Dispersion model A Lagrangian particle dispersion model (LPDM) described by Fast (1995) is used to illustrate the behaviorof the nighttime drainage flows within the Sinbad Basin. Atmospheric dispersion is simulated by tracking alarge number of particles in which trajectories are basedon the mean velocity components produced by RAMSand subgrid-scale turbulent velocity components. Thesubgrid-scale turbulent velocities are computed bysolving the Langevin equation by a Markov chain formulation (Legg and Raupach 1982). The turbulent velocities are a function of the Lagrangian timescale, andthe turbulent velocity statistics are consistent with thesecond-order closure applied in RAMS. Particles areassumed to be nonbuoyant, and a perfect reflection ofparticles is a~sumed to occur at the ground. The mean velocity field employed by LPDM is basedon output produced by RAMS at 30-min intervals.LPDM linearly interpolates the mean velocity field intime during individual 30-min periods using a time stepof 5 s. Particles are released continuously between 1900DECEMBER 1996 FAST ET AL. 2165and 0500 MST from five locations along the basin sidewalls (Fig. lb) to illustrate the flow patterns, to trackair parcels originating from different slopes, and to determine the amount of mixing within the basin. Particleconcentration is then calculated by grid-cell averagingwithin grid 2 (Fig. lb) using a horizontal grid spacingof roughly 125 m.3. Experimental design The simulations performed to examine the nocturnalflows and the heat budget within the Sinbad Basin aresummarized in Table 1. The initial conditions for thecontrol simulation (experiment 1 ) shown in Fig. 2 areassumed to be horizontally homogeneous and are basedon a combination of the observations taken from anairsonde at the basin site and the rawinsonde soundingfrom Grand Junction, Colorado, (80 km to the northeast) at 0000 UTC (1700 MST) 15 July 1988. Windand temperature measurements from the airsonde areused up to approximately I km above the basin rim,and values from the Grand Junction sounding are usedabove that level. The vegetation type is specified asdesert shrub and the soil type is specified as sandy loamover the entire domain. The initial soil temperature profile for each column of soil grid points is set equal tothe measured profile at the basin floor at 1600 MST(W96) so that the temperature 1 m below the surfaceis 23-C. The initial vegetation temperature is set equalto the atmospheric temperature at the lowest model gridpoint. W96 found that latent heat fluxes were negligiblewithin the arid Sinbad Basin for this period; therefore,the initial soil moisture is set equal to a dry value ( 15%of saturation) throughout the domain. A 12-h forecastis then made using radiation lateral boundary conditions and a time step of 7.5 and 2.5 s on grids 1 and 2,respectively. Experiments 2-6 are similar to the control simulation, except in the treatment of the surface characteristics or the initial wind profile. An initial wind speedof 0.1 m s -~ is used throughout the domain in experiment 2 so that the effect of the ambient wind on thenocturnal drainage flows is removed. No vegetation(bare soil) is used throughout the domain in experiment3, and the distribution of vegetation in experiment 4(not shown) is similar to the actual vegetation coverage: bare soil in the center of the basin and along theupper slopes of the sidewalls. Experiments 3 and 4 areperformed because W96 speculated that vegetationmay affect the downward turbulent heat flux. Tethersonde and airsonde observations revealed that wind directions and speeds varied significantly with timeabove the basin rim; therefore, experiments 5, 6, and 7are performed with different initial wind directions upto 3.3 km above the basin floor to examine the effectsof ambient wind direction on the cooling within thebasin.~2~0 8 6 4 2 ~ specific humidity ( g kg-1 ) 0 2 4 6 8 .0 310 320 330 340 350 -- potential temperature (K) wind direction ( deg )90 180 270 $60 903 6 9 12 15 wind speed ( m s-1)FIG. 2. Initial vertical profile of potential temperature, specifichumidity, wind speed, and wind direction employed by experiment 1.4. Numerical results A comparison of the results from experi~nents 1 and2 with the tethersonde measurements is made first todemonstrate that the model reproduces key features ofthe observed temperature and wind profiles during theexperimental period. These key features include theamount of cooling in the center of the basin, light andvariable winds at the basin site, and strong winds at theSalt Wash site. Then the predicted evolution of theslope flows, thermal structure, and heat budget withinthe basin are examined to identify mechanisms that leadto the strong cooling within the basin.a. Comparison with observed profiles Figure 3 depicts the predicted potential temperatureprofiles at the basin site for experiments 1 and 2 alongwith the tethersonde measurements. The predicted temperature profile gradually cools with time throughoutthe depth of the basin, although the cooling rate slowsafter 0100 MST. The final temperature profile at 0500MST is nearly identical to the observed profile, indicating that the model correctly simulates the totalamount of cooling in the basin during the evening. Theobserved temperature profiles, however, have strongergradients within 100 m of the ground. This may becaused by insufficient vertical resolution or excessivevertical mixing near the surface. Experiment 1 also produces temperature profiles that are 2-5 K colder thanobserved above 200 m AGL between 2000 and 0100MST. Although this also appears to be due to verticalmixing, it is mostly the result of cold air advection fromthe basin sidewalls, as will be shown in the next section. Experiment 2 (without ambient winds) producestemperature profiles with stronger vertical gradientsnear the surface in the early evening hours ( 1900-21002166 JOURNAL OF APPLIED METEOROLOGY VOLt,ME35 (a) experiment 1 (ambient wind) (b)700 .... i .... , .... I .... ] . [' L 7006O0500,400.300.200.- 600'100. 0.... [ .... I .... I .... I .... 300 305 tethersonde observations.... I, ,,,I .... l, ,,,[~ ,,,IBasin site - 500 - 400 - 300 '- 200-' , ,"- 100 -' , ,, ,310 3i5 320 experiment 2 (no ambient wind) (C)- ;,z;;;;;"" .... i -+-19 i6oo~ ---o--- 20 j~ - ~. 21 /~l[[ --.-23 ':~oo ; . 2J ':,00] '- 200 ]0' .'..., .... ~ .... ~ .... ~... ' 0~2050295 300 305 310 315 320 295 300 305 310 315 320potential temperature ( K )FIG. 3. Vertical profiles of potential temperature at the basin site predicted by experiments 1 and 2 and observations from the tethersonde.MST) when compared to experiment 1. Nevertheless,the temperatures above 400 m AGL do not decreasesignificantly so that all of the cooling occurs in thelower part of the basin; the temperatures aloft are toowarm and the temperatures near the surface are too coolby 0500 MST. 318 ! , f - ~ , [ - ~ - I - I - ~ - [ - ~ , I , I ,tv 392: AGL~ 310~ ] a 12mAGL a '~. I o 15 m0 AGL 12mAGL--'~x,~.x [ 302 1 ' 392 m AGL '" ] model: ] experiment. 1_ (.,am 'eb .n~ t. wind!., experiment 2 (no arnbient wind) 298 17 18 19 20 21 22 23 00 01 02 03 04 05 time ( MST )FIG. 4. Potential temperature predicted by experiments 1 and 2 atthree selected heights along with observed values at lhe basin site. To examine the nocturnal cooling at the basin site inmore detail, the potential temperatures at three levelsare shown in Fig. 4. Experiments 1 and 2 producenearly the same temperature near the surface throughout the evening, indicating that the thermal forcingfrom the ground at the basin site is not affected by theambient winds. Both simulations produce temperaturesthat are usually within 1.5 K of the observations, butthe 392-m level temperatures from experiment 1 are 22.5 K too low at 2300 and 0100 MST, and the 150and 392-m level temperatures from experiment 2 areas much as 3.5 K too high at 0500 MST. Some horizontal variation in potential temperature is produced bythe model within 1 km of the basin site. The horizontaltemperature differences are as much as 2 K at 2300MST within a layer between 200 and 400 m AGL.Other potential temperature profiles within 1 km of thebasin site are sometimes in better agreement with thetethersonde values. The potential temperature profiles within the canyonat the Salt Wash site predicted by experiment 1 aresimilar to tethersonde observations (not shown); themodel also produces the transient neutral layers nearthe lower parts of the canyon described by W96. Theseneutral layers are due to strong winds and wind shearswithin the canyon predicted by the model. The temperature profiles from experiment 2 did not containneutral layers. Figure 5 depicts the predicted wind speed profile atthe Salt Wash site for experiments 1 and 2 along withthe tethersonde measurements. Experiment 1 producesDECEMBER 1996 FAST ET AL. 2167 (a) experiment 1 (ambient wind) (b) experiment 2 (no ambient wind (C) tethersonde observations - time ( MST ) ] Salt Wash site 500t :500: '. :500 & 0 ...~...~...f...~...,...,...' 0 ...f ............ ~...~... 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 wind speed ( m s-1 )FIG. 5. Vertical profiles of wind speed at the Salt Wash site predicted by experiments 1 and 2 and observations from the tethersonde.relatively strong winds exiting Sinbad Basin throughthe canyon with wind speeds of 4-5.5 m s -~ between2100 and 0100 MST. This agrees with the observedprofiles, although the observed maximum wind speedsare 0.5-1 m s -~ stronger. The weaker winds predictedby the model may be due to the differences between 5- o 4. 150 m AGL--! 12mAGL o J -o , 12 m AGL - --150 m AGL u) 2' tethersonde: ~ a 12 m AGL o 150 m AGL 1 model: experiment 1 (ambient wind) -- experiment 2 (no ambient wind) 17 18 19 20 21 22 23 O0 01 02 03 04 05 time ( MST ) FIG. 6. Wind speed predicted by experiments 1 and 2 at twoselected levels along with the observed values at the Salt Wash site.the modeled and observed temperature profiles withinthe basin as well as the representation of the canyontopography by the model (section 2a). The predictedwind speeds also vary within the canyon. For instance,the maximum wind speed 1 km further down the canyon and at 200 m AGL is 7.2 m s-~ at 2300 MST.Experiment 2 also produces strong winds within 200m of the ground after 2200 MST, but the winds aloftare significantly weaker than those in experiment 1 andthe observed profiles since all of the cooling within thebasin occurs near the surface (Fig. 3b) in this simulation. Although the observed wind speed profiles at the SaltWash site are quite complicated, the wind speedswithin 200 m of the ground tend to increase until 2300MST and decrease after this time. As shown in Fig. 6,this feature is produced only by experiment 1; experiment 2 produces wind speeds that increase with timeduring the whole evening. The peak wind speed in experiment 1 corresponds to the largest mass flux throughthe canyon and the largest pressure gradient computedbetween the Salt Wash site and the middle of the basinat a constant elevation of 2286 m MSL (not shown).Both the predicted pressure gradient and wind speeddecrease until 0100 MST and remain nearly constantfor the rest of the evening. At an elevation of 2286 mMSL, the potential temperature difference between thebasin and the canyon increases from 0 to 2.5 K between2000 and 2300 MST and then gradually decreases to1.5 K by 0500 MST. These temperature differencesproduce a thermally developed pressure gradient that2168 JOURNAL OF APPLIED METEOROLOGY VOLUME35drives the flow through the canyon. W96 found that thetemperature difference computed from the tethersondesat 2214 m MSL was about 1.2 K at midnight. It isunclear why the model produces lower wind speedswith a larger temperature difference, but the moclel canyon geometry' is probably one reason.To obtain a quantitative measure of the model per-Timeformance, the root-mean-square errors of potential tem- (MST)perature and wind speed, and the vector wind difference (VWD), are calculated at the tethersonde locations for experiments i and 2 and are given in Table 2. 1900 2O0OBoth speed and direction errors are taken into account 2tooby the VWD that is given by 2300 0100 VWD = [(u,~ - Uo)2 + (v,, - u0)2]-~5, (1) 0500 Meanwhere urn, Vm and uo, Vo are the model and observedhorizontal wind components. Even though the evolution of the stable boundary layer (Fig. 3) for experiments 1 and 2 is significantly different, the mean po- 2900 2000tential temperature errors during the evening are nearly 2100identical. The root-mean-square error for wind speed 2300and the VWD for these two experiments are about the moosame at the basin site, but the errors are smaller at the 0500 MeanSalt Wash site for experiment I when ambient windsare included by the model.As shown in W96, the winds in the middle of theSinbad Basin atmosphere were light and variable forthe evening of 15-16 July. This feature is reproducedwell by both experiments 1 and 2 near the surface. Thelight and variable winds observed by the tethersondeimply that eddies may form within the basin, as will beshown in the following section.b. Basin circulations The u component of the wind from experiment 1 onvertical cross section AA' (Fig. 1 ) is shown in Fig. 7for three times during tl~e evening. At 2100 MST,strong downslope flows up to 5.6 m s -~ are evident onthe western slope and weaker downslope flows up to-1.9 m s -~ occur on the eastern slope. The magnitudeof the downslope flow is also stronger on the westernslope of the basin in experiment 2, although the difference is not as great. For example, the maximum u component at 2300 MST is 3.3 and -2.4 m s -~ on the western and eastern slopes, respectively (not shown). Sincethere are no ambient winds in experiment 2, thestronger drainage winds over the western slopes areprobably due to the relatively large area surroundingPace Peak that is a source of cold air draining into thebasin. The eastern ridge of the basin (Sewemup Mesa)is relatively narrow so that drainage flows; developingon this slope are due almost entirely to local cooling.This suggests that the asymmetry of the u componentin experiment 1 occurs because the ambient winds enhance the drainage flows over the western slope, whileretarding their development on the eastern slope. Thiseffect is similar to the numerical results presented by TABLE 2. Summary of the root-mean-square errors for potentialtemperature (m~s 0) and wind speed (rms s) and the vector winddifference (VWD) of the model calculated at the tethersondeobservation locations.Experiment 1 Experiment 2rn~s 8 rms s VWD rms 8 rms s VWD(K) (m s-') (m s-~) (K) (m s-~) (ms-~) Basinsite0.8 0.8 2.8 0.6 2.4 3.20,9 1.9 3.1 1.0 1.8 2,61.1 1.5 2.6 1.1 1.2 2.31.6 1.3 1.5 0.8 0.8 1.11.8 1.1 1.2 1.0 1.0 1.l1.1 0.9 1.8 2.7 0.5 1.61.2 1.2 2.2 1.2 1.3 2.0 SaltWashsite1.2 1.9 3.5 0.6. 1.8 2.21.2 1.0 1.7 0.8 2.7 2.90.8 1,0 1.5 1.0 2.8 3.41.5 1.6 2.0 0.8 1.3 2.21.3 1.2 1.8 0.7 1.4 2.41.4 1.2 1,7 1.0 0.9 2.21.2 1.3 2.0 0.8- 1.8 2.5Doran (1991) in which the ambient wind crossed anidealized symmetrical valley. Cold air flows into the basin the entire evening,but a significant portion of the drainage flow becomeselevated by 2300 MST, as seen in Fig. 7b. A shallowlayer of drainage winds still follows the terrain to themiddle of the basin floor at this time. By 0100 MST,the intensity of the surface drainage winds decreasesand most of the air that enters the basin becomesdetached from the sidewalls. A significant portion ofthe drainage winds in experiment 2 also becomes elevated above the surface after 2300 MST, althoughthe magnitude of the wind speeds along cross sectionAA' is usually less than 2 m s-l. Also shown in Fig.7 is the vertical cross section of the particle plumereleased from the western sidewall site (Fig. lb).The particle plume initially remains near the surface,but it splits as the drainage flow becomes elevatedafter 2300 MST. While the maximum particle concentration is at the same level as the cross-basin maximum wind speed, a portion of the plume continuesto follow the terrain to the bottom of the basin duringthe whole simulation. The model predicts a complex wind field on the basinfloor by 2100 MST, as seen in Fig. 8. Drainage windsof 2-4 m s -~ follow the local slopes from Pace Peakdown into the western portion of the basin, and strongdownslope winds up to 6 m s-t occur on the west andsouthwest slopes of the basin at this time. The windsover the mesa top on the eastern side of the basin arealso southwesterly, but later in the evening these areasDECEMBER 1996 FAST ET AL. 2169(a) 3.5v 2.5g 2.0(b) 3.5o3 3.0 -> 2.0 -8 -6(C) 3.5 ......x~ 3.0-~ o0tZ._o~ Z0-m , .~ , .~ ,A west / east distance (km ) F~o. 7. Vertical cross section along AA' (Fig. Ib) of the predictedu component (contour interval 1 m s-~) and tracer concentrationgreater than 0 (shaded) at (a) 2100, (b) 2300, and (c) 0100 MST fromexperiment 1.winds from the northwestern sidewall flow south toform a counterclockwise eddy in the middle of thesouthern basin floor as shown in Fig. 8a. The eddyactually persists through much of the evening until0400 MST, but the center moves back and forthacross the southern portion of the basin. Other eddiesform near the surface on the basin floor and alongthe sidewalls during the evening, but they are smallerand more transient. Petkovsek (1978) employed ananalytical technique and observations in Slovenia todemonstrate that drainage flows within a closed basincan converge closer to the steepest sidewall. Hefound that downslope flows over a steep slope maynot penetrate far into the basin as a result of intenseadiabatic warming that decelerates the drainagewinds. In the Sinbad Basin, relatively steep slopesalso drain into the basin (not shown). The structure ofthe drainage winds along the northwestern sidewallsappears to be affected by the small-scale valleys andridges that run perpendicular to the basin rim, Maximum wind speeds of just over 4 m s -2 occur inside twoof these small-scale valleys, while weaker winds occuralong the ridges perpendicular to the basin rim. Thewind directions and areas of maximum wind speedsfrom experiment 2 at this time are very similar to thoseshown in Fig. 8, except that the wind speeds are weaker(the maximum wind speeds over the northwestern andsouthwestern sidewalls are 3.6 and 2.6 m s-l, respectively) and the drainage winds flow into the basin alongthe eastern sidewalls. This indicates that the northernslopes of the basin are largely unaffected by the ambient winds. The ambient winds increase the magnitudeof the downslope flows along the southwestern sidewalls and divert part of the drainage flow within thenarrow valley just south of Sinbad Ridge into the basin. Despite the strong downslope winds on the westand southwest slopes, the flows on the basin floorappear to be affected primarily by the drainage windsfrom the northern part of the basin. The drainage ~ 2100 m contour F~G. 8. Near-surface wind field at 12.5 m AGL predicted by experiment 1 at 2100 MST: (a) streamlines and isosurface of threedimensional tracer plumes projected onto the ground (shaded) and(b) wind speed (contours of 1 m s-~).2170 JOURNAL OF APPLIED METEOROLOGY VOL~IME35(a) 3.55-.~ 3.0E_,~~ 2.5e.o>II~(b) 3.5u~ 3.0:~E..~v~.o_>aeoY-8 -6 -4 -2 . 0 2 4 v~-8 -6 -4 -2 0 2 4 (C) 3.5 u3 3.0 :~ E ~ 2.5 c .o..~ _e 2.0 e -8 -6 -4 -2 0 2 4 A west / east distance ( km ) A' Fro. 9. Vertical cross section along AA' (Fig. lb) of the predicted potential temperature (contour interval l K) and TKE greater than 0.2 m2 s-2 (shaded) at (a) 2100; (b) 2300, and (c) 0100 MST from experiment 1.height of the eastern basin rim (not shown). Whileit is possible that the convergence of the downslopeflows in the center of the basin would result in positive vertical velocities that could transport the tracerabove the basin lid, this did not occur in this simulation. Instead the majority of the particles exit thebasin through the Salt Wash Canyon, while some ofthe particles at the top of the cloud are sheared offby the ambient winds and advected over the mesason either side of the canyon. The flows in Figs. 7 and 8 are coupled to the evolution of the temperature structure within the basinas shown in Fig. 9. Early in the evening, cold airgenerated over the slopes of Pace Peak is advectedby the drainage and ambient winds down the steepslopes of the basin, resulting in a hydraulic jump onthe western side of the basin by 2100 MST (Fig. 9a).Relatively high turbulence kinetic energy (TKE) values are produced near the surface and in the nearneutral areas. Hydraulic jumps that result from stratified flow over escarpments with slopes similar tothose within Sinbad Basin have also been observedand simulated by other investigators (Pitts and Lyons1989; Blockley and Lyons 1994). The hydraulicjump diminishes in magnitude by 2300 MST, butTKE generated along the sidewalls is now advectedinto the middle of the basin. While the whole basincools between 2100 and 2300 MST, the potentialtemperature gradient throughout the basin remainsnearly the same. The cooling aloft is due predominantly to the advection of cold air from the sidewalls(Fig. 7b), but there is sufficiently high TKE to alsosurround the relatively flat southern basin' floor,whereas the northern basin floor has gradual slopes(Fig. lb ); therefore, the mechanism proposed by Pet~kovsek (1978) may explain the near-surface convergence in the southern basin that persists for most ofthe evening. Most of the particle plume released from the western sidewall site advects directly through the basininto the canyon, but the shallow northerly flow alongthe basin floor also transports part of the plume intothe eddy at 2100 MST (Fig. 8a). Particles releasedfrom the southwestern sidewall site are advected intoand recirculated within the eddy before t'hey are advected out of the basin through the canyon 1 hourlater. The plumes from the southern and eastern sidewalls are largely-advected by the ambient winds atthis time, but they do drain into the basin by 2300MST. By 0200 MST, particles released from thenorthern sidewall are found over much of the basinfloor, but they are retarded by drainage flows comingdown the eastern sidewalls. By the end of the evening, particles fill up the basin volume below the(a) 3.5~ 2.5O - 2.0(b) 3.5~ 3.0~' 2.5'o >e 2.0'-8 -6 -4 -2 0 2 4 -8 -6 -4 -2 0 2 4A west / east distance (km ) A' FIG. 10. Position of 315-K isotherm along vertical cross sectionAA' (Fig. lb) as a function of time for (a) experiment 1 and (b)experiment 2.DECEMBER 1996 FAST ET AL. 2171(a) 16.08.0-8.0,~ -16.0--1~ 0.5E= 0.0O.-0.5-1.0 ~ ---o-- term B-8 -6 -4 -2 0 2 4(b) 4.02.0' 0.0,e~ -4.0 0.5 0.0-0.5 -1.0 -8 -6 -4 -2 0 2 4A west / east distance ( km ) A' A-8~ term----o--- term-6 -4 -2 0 2 4 west / east distance ( krn ) A'FIG. 11. Individual terms of the potential temperature tendency equation, Eq. (2) (averaged between 2000 and 0500 MST and normalized by h), along cross section AA' for (a) experiment 1 and (b) experiment 2.mix cooler temperatures upward. A strong temperature gradient cap develops just above the basin lid by2300 MST since the temperatures above the basin lidremain relatively constant. By 0100 MST, the vertical potential temperature gradient within the basinincreases and the amount of TKE advecting into thecenter of the basin decreases. The drainage flows become partially elevated after 2300 MST (Fig. 7b)because they are relatively warmer than the cold airadjacent to the basin floor. As the evening progresses, the drainage flows weaken and do not penetrate as deeply into the basin. The evolution of the thermal structure is given inFig. 10 by plotting the position of the 315-K isothermthrough the basin cross section as a function of timefor experiments 1 and 2. In experiment 2, the isotherm gradually rises with time until it reaches asteady-state height above the basin lid by 2300 MST.Since the height of the 315-K isotherm remainssteady while the basin continues to cool, the gradientincreases within the basin after 2300 MST. The potential temperature evolution is nearly uniform in experiment 2, although there is a tendency for slightlylower temperatures on the eastern side of the basin.In experiment 1, spatial inhomogeneity in the coolingis clearly evident. The hydraulic jump between 2000and 2300 MST produces downward vertical advection of warmer potential temperatures over the western side of the basin with upward vertical advectionof cooler air above the middle western slopes. By0000 MST the maximum height of the 315-K isotherm is reached, but it slowly sinks 100 m by themorning. Even though the cooling rate in the model at thebasin site differs from the tethersonde observationsin some respects, experiment 1 reproduces the temperature profile at 0500 MST quite well, suggestingthat the net cooling within the basin during the evening is reasonably simulated. Therefore, the modelresults can be analyzed further to provide information regarding the role of slope flows on the basinheat budget.c. Basin heat budget One of the advantages of applying a mesoscalemodel to simulate the flows in the Sinbad Basin is thatthe model produces a thermodynamically balanceddataset that can be used to examine the individual termsof the atmospheric heat budget (Zhong and Whiteman1995). The relative importance of each of these termswas determined for all the simulations in Table 1, although most of the discussion here will focus on experiments 1 and 2. The terms of the heat budget are2172 JOURNAL OF APPLIED METEOROLOGY VOLUME B5calculated by vertically integrating the thermodynamicequation to obtain--Ot dz = ~ -~7.(VO)dz+ - rp---~ (V-(R)&+ z~_X~-sdz (Kms., (2) b Advection and turbulent diffusion are also the dominant terms in Eq. (2) for experiment 2 (Fig. llb);however, their magnitude is smaller than in experiment1 because of the weaker slope flows. Since the drainageflows from experiment 2 are the strongest over thewestern basin slopes, these two terms are also the largest over the western slope. The storage term closelyfollows the underlying terrain as in experiment 1, butthe correlation coefficient of term A and the terrainelevation is now 0.99. The greatest heat loss in experiment 2 occurs in the center of the basin; therefore, themagnitude of term A in Fig. 1 lb is slightly larger thanin Fig. lla. Even though the average cooling depicted in Figs.lla and llb during the evening is rather simple, thewhere the local rate of change of potential temperaturein a vertical column (A) results from an imbalance between advection (B), radiative flux (C), and turbulentdiffusion (D). Equation (2) is integrated from the surface zx to a height h, approximately 800 m AGL. Thislevel was chosen because it is somewhat abow~ the basin lid where the contributions to Eq. (2) become small.The terms in Eq. (2) from experiments 1 and 2, averaged between 2000 and 0500 MST and normalized byh, are shown in Fig. 11 as a function of location alongbasin cross section AA '. A 9-h average is taken to depict the cooling mechanisms within the basin duringthe entire evening.In experiment 1, advection and turbulent diffusion {b)are the dominant terms in Eq. (2) along the sidewalls, ~with the largest values over the western slope (Fig. ~1 la). West of the basin, these two terms corttribute to m,...a net warming in the layer as warmer potential tem- ~ ~o~peratures are advected from the slopes surrounding .~Pace Peak. Cooling due to turbulent diffusion, is greater ~ ~than advective warming over the upper half of the west- ~o. ~ern sidewall and over most of the eastern sidewall during the evening. Along the lower western sidewallslopes and in the middle of the basin, turbulent diffusion becomes nearly zero so that the cooling is due to {e)local radiative cooling and the advection of cold airfrom the sidewalls. The measurements made by W96 .~_~. ~.0also showed virtually no turbulent sensible heat flux on ~ k: o.5the basin floor. Interestingly, the magnitude., of the av- ~ ~erage local change in potential temperature (term A) ~ o~ 0.0given by Eq. (2) nearly parallels the underlying terrain ~ 05 ~ ~ -0.selevation, with the strongest cooling in the center of the ~ o, O~basin. This happens because the lowest elevation in the o. -1.0center of the basin is where the coldest air accumulates,producing the greatest heat loss during the evening. -~.sSince the other terms in Eq. (2) are not uniform acrossthe basin, the largest value of term A, -l.17 K h-~,occurs at the base of the southern sidewall. The correlation coefficient of term A and the terrain elevationis 0.66, based on the values from all of the grid pointsbelow 2200 m MSL.(a) 1.5 ' I ' ~ ' ~ ' ~ ' ~ ' ~ ' ~ ' :~ ----,--- term A ~ term C t J ---=a---- term B+D ]:~ ~- 0.5 ....................................- o>, 0.0.~~'o -0.s ..................... -1,0 -1.5 . ~='='~'~=""~=-"- ~ , .~-8 -6 -4 -2 0 2 41.5. ~ I , I ~ I , I , I , I , I , - ----0--'-- term A ~ term C t_e_rm_ ~ ~~.o ............... =-+Y- ..........13.5 .....................................1.5'-8 -6 -4 -2 0 2 4 -8 -6 -4 -2 0 2 4A west / east distance ( km ) A' FIG. 12. Individual terms of the potential temperature tendencyequation, Eq. (2) (averaged over 30-min intervals and normalized byh), along cross section AA' for experiment I at (a) 2100, (b) 2300,and (c) 0100 MST.DECEMBER 1996 FAST ET AL. 2173rate of cooling within the basin varies significantly inspace and time. For example, the terms in Eq. (2) fromexperiment 1, averaged over 30-min intervals and normalized by h, are shown in Fig. 12 at three times duringthe evening. The radiation flux divergence does notvary significantly in space or time so that the complicated local rate of change of potential temperature atany one time is due to subtle differences between theadvection and turbulent diffusion terms. The complexcirculations within the basin (Fig. 8) no doubt contribute to the variation of the terms in Eq. (2) across thebasin floor. In general, relatively strong cooling ratesalong the sidewalls result from large turbulent diffusionvalues, while strong cooling rates in the middle of thebasin are usually due to advection of cold air from thesidewalls. Vertical advection is another mechanism thatcools the basin atmosphere at certain times during thesimulation period; low potential temperatures at thesurface are advected upward by both strong upwardvertical velocities within the hydraulic jump along thelower slopes of the western sidewall (Fig. 9a) and modest upward vertical velocities associated with the converging downslope flows early in the evening (Fig.7a). There are also mechanisms that occasionally resultin warming within the basin. For instance, horizontaladvection of a layer of higher potential temperatures bythe elevated drainage flow (Fig. 7b) warms the centerof the basin at 2300 MST (Fig. 9b). Radiative fluxdivergence and advection of cold air from the sidewallscontribute equally to cooling in the center of the basinafter 0100 MST (Fig. 12c). Integrating the thermodynamic equation over thevolume of the basin results in the atmospheric heatbudget equation given byO0 dV~(XT. R)dV fff+ pCp~'~I~l~dV (W), (3)which specifies that the change in heat storage (A)results from an imbalance between the convergenceof potential temperature flux by the mean wind (B),the convergence of radiative flux (C), and the convergence of the turbulent sensible heat flux (D) in theatmospheric volume. These terms are calculated fromthe model results using all the grid points below the2200-m contour (Fig. lb) and west of an arbitrary lineacross the canyon about 1 km west of the Salt Washsite. Before the values are integrated in Eq. (3), themodel results are interpolated to another vertical gridthat employs a 40-m grid spacing up to the basin lidheight that is defined here at 2200 m MSL (approximately the same height used by W96). The variation of the atmospheric heat budget termswith time, given by Eq. (3) and normalized by the basin area, for experiments 1 and 2 is shown in Fig. 13.Cooling is due to radiative and turbulent sensible heatflux during the evening, while advection warms the basin. Radiative flux divergence is nearly constant, at 35W m-2, but the magnitude increases slightly with timeduring the evening. Both experiments suggest that thestrongest cooling within the entire basin occurs in theearly evening between 1900 and 2300 MST. In experiment 1, advective warming increases slowly with time,while the turbulent sensible heat flux term remains relatively constant after 2300 MST so that the rate of heatstorage loss from the basin decreases later in the evening (Fig. 13a). The rate of the heat storage loss fromexperiment 2 also decreases after 2300 MST, but in thiscase it results from advective warming increasing fasterthan turbulent sensible heat flux cooling as the eveningprogresses (Fig. 13b). While the results from experiments 1 and 2 are similar, the ambient winds had asignificant impact on the heat budget within the basin.Table 3 gives the values of the terms in Eq. (3) averaged between 2000 and 0500 MST and indicate that(a) 200 -'''''' ~''''''''' ~ ' ~ ' ~ ' experiment 1 (ambient wind) ~ -200 -' ................................. -300' , .... . ,---~.,je.~,~, ,--~. ,,e~o 17 18 19 20 21 22 23 O0 01 02 03 04 05(b) 200 ........................i-100x -200' experiment 2 (no ambient wind) -300 -,. ~ . , - , - ,~. ~ - , - , - , - ~ - 17 18 19 20 21 22'23 O0 01 02 03 04 05 time ( MST ) FIG. 13. Individual terms of the heat budget equation, Eq. (3) (normalized by the basin area), as a function of time for (a) experiment1 and (b) experiment 2,21'74JOURNAL OF APPLIED METEOROLOGY VOLUME35TABLE 3. Summary of the heat budget terms (Wm-2) within the basin averaged between 2000 and 0500 MST. Storage Horizontal Vertical Advection Radiative flux Turbulent diffusionExperiment (term A) advection advection (term B) (term C) (term D)I -108.0 ~348.5 416.1 67.6 -32.4 -143.22 - 120.8 -93.4 158.1 64.7 -29.3 - 156.23 - 124.6 -263.8 336.4 72.6 -50.5 - 146.74 - 116.0 - 314~2 383.3 69.1 -43.7 - 141.45 -89.9 -292.6 376.4 83.8 -34.4 - 139.36 - 140.2 - 184.4 162.5 -21.9 -41.3 -77.07 - 149.2 - 183.1 141.1 -42.0 -36.6 -70.6the ambient winds reduced the overall cooling of thebasin when compared to experiment 2. To illustrate the possible effect of vegetation on thecooling rate within the basin, the variation of the normalized atmospheric heat budget terms with time fromexperiments 3 and 4 is shown in Fig. 14. In experiment3 (Fig. 14a), a uniform bare soil surface leads to aradiative flux divergence that is 1.5- 2 times larger thanthe radiative flux divergence from experiment 1 duringmost of the evening. When the model incorporates aheterogeneous vegetation distribution (Fig. 14b), theatmospheric heat budget terms are similar to those fromexperiment 1. The bare-soil areas on the basin floor inexperiment 4 increase the radiative flux divergencesomewhat and decrease the contribution from the tur(a)x(b)~DO200 i ' I , ! . ~ . i . i . I , I . I - I , I , I , -t-,00l 17 18 19 20 21 22 23 O0 O1 02 03 04 05200 ' ~ - ' - ~ -' ' ~ ' ~ - ~ - ~ ' ~ - ' - ~ - experiment 4 (ambient wind, variable vegotaion) 100 .........................-loo.-2oo. ~ term A -- a term C + term B ~ term D-~oo ......... 17 1~ 1~ 2b 2~ 2~ 2~ 0b 0~ '0~ 0~ '01 '05 time ( MST )FiG. 14, Same as Fig. 13 except for (a) experiment 3 and (b) experiment 4.bulent sensible heat flux, leading to cooling within thebasin that is slightly higher than for experiment 1 (Table 3). In contrast to experiments I and 3, the radiativeflux divergence from experiment 4 can vary by a factorof 2 across the basin (not shown). The direction of the ambient winds also modifies theheat storage within the basin, as seen in Table 3. Theambient northwest winds in experiment 5 produce lesshorizontal advection so that the basin becomes somewhat warmer than in experiment 1. This implies thatthe local topography will have an impact on the drainage flows that will ultimately affect the amount of cooling. When the ambient winds have an easterly component, as in experiments 6 and 7, very strong coolingoccurs within the basin. In these simulations, the surface temperatures at the basin site are similar to thosefrom experiment 1 (Fig. 4), but the potential temperature profiles are nearly neutral and the temperaturesaloft are much lower than those shown in Fig. 3a. Thelower potential temperatures aloft result from the relatively low magnitude of the vertical advection term.When the winds are from the west, the hydraulic jumpon the steep slopes of the basin (Fig. 9) brings higherpotential temperatures down from aloft resulting in anet warming of the basin atmosphere. This mechanismis not as effective when the winds are from the eastsince the eastern rim is lower in elevation and not assteep overall.5. Discussion Observations from the basin site tethersonde suggestthat pooling of cold air in the basin results in a strongstable layer near the surface early in the evening thatgrows in depth through the night. Above the stablelayer, a residual neutral layer exists that cools uniformly only by 1-2 K by 2300 MST. In experiment 1,the near-surface temperatures are well predicted, butthe boundary between the surface stable layer and theresidual layer within the basin is not as distinct as theobservations because the model cools the basin atmosphere in this volume too quickly. While the temperature profiles aloft remain nearly neutral, the whole profile decreases by as much as 5 K by 2300 MST, producing relatively weak vertical gradients near thesurface.DECEMBER 1996 FAST ET AL. 2175 As stated previously, the weaker stable layer near thesurface and the relatively strong rate of cooling in themiddle of the basin atmosphere may be due to a verticalgrid spacing that is not sufficient to resolve the shallowdrainage flows along the slopes of the basin or to vertical mixing that is too strong in the middle basin atmosphere. Additional test simulations were performedto examine these possibilities. Two two-dimensional simulations were performedusing topography along basin cross section AA' to evaluate the effect of vertical grid spacing. The same gridspacing as in experiment 1 (section 2a) was used forone simulation, and twice the vertical resolution and ahorizontal grid spacing of 150 m was used for the othersimulation. Decreasing the vertical grid spacing alsorequired that the horizontal grid spacing and the timestep be reduced. The depth and evolution of the drainage flows along the basin sidewalls from both the highand low-resolution simulations were qualitatively verysimilar. As in Fig. 7, a portion of the drainage flowbecame elevated in both simulations. Slightly strongervertical potential temperature gradients were producedin the middle of the basin by the high-resolution simulation; however, these gradients were still not nearlyas strong as those observed by the tethersonde. A threedimensional simulation with a higher resolution thanexperiment I is computationally expensive and thetwo-dimensional tests indicate that increasing the spatial resolution may not significantly improve the modelresults. Conversely, two-dimensional tests cannot fullydetermine whether three-dimensional features, such asthe canyon, need to be resolved with a smaller horizontal grid spacing. Two additional three-dimensional simulations wereperformed to examine the effect of turbulent diffusion,term D in Eq. (2), on the model results. The simulations were identical to experiment 1, except that theyemployed different length-scale formulations that reduced the TKE and vertical eddy exchange coefficientsin the middle basin atmosphere. The potential temperature profiles from these simulations were similar tothose in Fig. 3a, indicating that excessive vertical mixing may not be responsible for the weak surface temperature gradients or the high rate of cooling in themiddle basin atmosphere. Evaluation of individual terms of the heat budgetequation reveals that the simulated cooling aloft in themiddle of the basin is due primarily to advection ofcold air from the sidewalls rather than turbulent diffusion. The excessive cooling at these levels (Figs. 3aand 3c) thus suggests that the simulated elevated windsmay be too strong. For instance, the model producessouthwesterly wind speeds up to 4 m s-~ 400 m abovethe basin floor at 2300 MST (Fig. 7b) and the observedelevated maximum wind speed at the basin site is 2.8m s-~. The model wind speed decreases to 2.3 m s-~by 0100 MST, but the observed elevated wind speedmaximum is 1.5 m s -~ at that time. The observed windspeeds aloft at the basin site are more transient thanthose produced in experiment 1, and there are no otherobservations to confirm whether strong elevated windspeed maxima occur at other locations. Varying the ambient wind direction (experiments 5, 6, and 7) did affect the amount of cooling due to advection within thebasin, and it is likely that the amount of cooling is alsodependent upon the ambient wind speed. Higher ambient wind speeds would increase vertical motions overthe sidewalls and enhance the entrainment of air aloftinto the basin as described by Doran ( 1991 ). To furtherexamine the effect of ambient wind direction andspeed, an additional three-dimensional simulation wasperformed that employed four-dimensional data assimilation (Fast 1995) to incorporate the tethersondewinds above the basin rim elevation into experiment 1.As expected, the wind speed and direction errors at thebasin and Salt Wash sites were reduced somewhat inthis simulation. The amount of cooling due to advectionnear the basin rim elevation was smaller in the dataassimilation experiment than in experiment 1; nevertheless, the potential temperature profiles were nearlyidentical to those in Fig. 3a and the heat budget wassimilar to Fig. 13a. Despite the fact that the rate ofcooling in the model at the basin site differs from thetethersonde observations in some respects, the modelreproduces the temperature profile at 0500 MST quitewell, indicating that the net cooling within the basinduring the evening is reasonably simulated. One of the objectives of this study was to determinethe effect of the sidewall drainage flows on the basinheat budget by comparing the terms in Eq. (3) produced by the model with those computed from the observations alone (W96). The relative contributions ofthe individual terms in Eq. (3) for both the model andthe observational method are very similar when the results from experiment 1 (Fig. 13a) are compared withthe values given by Fig. 11 of W96; however, the magnitudes are significantly different. Both the model andthe observational method suggest that the magnitude ofthe heat storage change term is largest in the early evening around 2000 MST, but the model value is about2.5 times larger than the value obtained using the observational method. The difference in the heat storageterm gradually decreases to nearly zero by 0500 MST.At 0000 MST, the radiative flux divergence in themodel is about 1.5 times larger than for the model employed by W96, while advection in the model is about3.5 times larger than for the observational method.These differences may indeed suggest that it is important to account for the inhomogeneity in the flowswithin the basin (as shown in Figs. 7, 9, and 10) whencomputing the heat budget. Nevertheless, the flowfields along the sidewalls cannot be verified, so we canonly speculate that the model is producing a reasonableestimate of their contribution to the basin heat budget. The observational method in W96 assumes horizontal homogeneity of the meteorological fields based on2176 JOURNAL OF APPLIED METEOROLOGY VOLUME35measurements from the basin site, but these measurements are clearly not representative of the entire basin.However, the model results from experiment 1 suggestthat the assumptions employed by W96 for the verticaladvection term may be reasonable; the mean verticalvelocity on the 2200-m MSL surface is around 0.025m s-~ (nearly identical to the values used by W96),and the simulated potential temperatures are nearly horizontally homogeneous within the basin after midnight. The differences in the magnitude of the heat budgetterms between the model and the observational :methodcan also be attributed to model forecast errors since themodel does not produce a perfect forecast of the observed temperature profiles within the basin. The largest differences in the heat storage change term occurearly in the evening when the temperature errors at thebasin site are also the largest. The advantage of the model heat budget is that itincludes contributions from the complex circulationswithin the basin. While additional observations at different locations within the basin would increase theconfidence of the estimate of the observational heatbudget, other studies have shown that a very dense observational network would be required in complex terrain to adequately resolve the spatial and temporal meteorological fields. For instance, the heat budget equation computed by Bader and Horst (1990, 1992) didnot balance when observations from the relativelydense measurement network around the Brash CreekValley (Clements et al. 1989; Horst et al. 1987) wereused in their analysis. W96 developed a convenient way of comparing theheat budget of multiple valleys and basins by nondimensionalizing Eq. (3) to obtainA BC+D C+D--- 1. (4)Equation (4) provides a basic measure of the relativecontributions from heat storage or advection, and theresults from the experiments in this study are shown inFig. 15. Advection is larger than heat storage; for valueson the right side of the plot, corresponding to the heatbudgets of valleys. Heat storage is larger thm~ advectionfor values on the left side of the plot, corresponding tothe heat budgets of closed basins (W96). The resultsfrom experiments 1, 2, 3, and 4 fall on the left side ofthe plot, indicating large amounts of heat loss associated with a basin, but advection is large enough to suggest that the basin is not completely closed. These results seem more reasonable than the value computedby W96 because of the relatively large mass fluxthrough the canyon. The clustering of these points alsoindicates that vegetation did not significantly changethe relative contributions of the advection and heat storage terms within the basin. The results from experiment5, with northwest ambient winds, produce a larger advection term than the control simulation, suggesting(6,7) 1.00A / (C+D) trappers drainers .... ! .... ! .... !.,,, ,,. , I I [experiment [ ~~' , I= 0.75 .... --q o 2 [] 5 [ I 'Nx ' lO3 / li---%-,"-:"', , ,' 0.50 ...... ~ ..... computed ~ ~ ~ I I : 0.25 ~ ~ ~ I CSeo( % ', ', ~ N 0.00 .... I .... I .... I .... 0.0~ 0.25 0.50 0.75 1.00 -B / (C+D)FIG. 15. Categorization of experiments with respect tothe ratio of selected heat budget terns.coolerssteadystarersthat the basin behaves more like a valley when the ambient wind enhances the flow through the Salt WashCanyon. The two simulations that had easterly windsindicate that the basin is closed and traps the drainageflows. In fact, the northeast ambient winds in experiment 7 retarded the drainage winds within the SaltWash Canyon during most of the evening.6. Summary and conclusions A mesoscale model has been used to simulate theevolution of the boundary layer within Colorado's Sinbad Basin and the canyon that drains the basin duringthe evening of 15-16 July 1988. The predicted nocturnal potential temperature and wind profiles are compared to tethersonde measurements made at two different locations. Within the basin, the model reproduces the observed near-surface potential temperaturesand the profile at 0500 MST, but the model atmospherecools too quickly between 2000 and 0100 MST so thatthe near-surface potential temperature gradient is lessthan that observed. Both the observed and modeledwinds near the surface are light and variable. Withinthe canyon, the model reproduces the observed maximum wind speed around 2300 MST that correspondsto the largest temperature and pressure gradient be~tween the basin atmosphere and the canyon. The model predicts that the basin fills with cold airby 2200 MST because the outlet canyon is narrow anddoes not effectively drain the basin atmosphere. Astrong horizontal temperature gradient forms betweenthe basin and the Dolores River valley because the valley is much larger and does not cool as fast as the basin.This temperature gradient gradually diminishes by0500 MST as the Dolores Valley becomes filled withcold air, reducing the pressure gradient and the windDECEMBER 1996 FAST ET AL. 2177speeds within the canyon. Unlike larger basins or welldrained valleys, both the observations and the modelindicate that the atmosphere within the Sinbad Basincools throughout the entire evening. Drainage windsfrom the higher elevations around Pace Peak flow intothe basin during the entire evening as well, but the sidewall flows become weaker after 2300 MST as the stability near the basin floor increases. Then a portion ofthe drainage winds becomes elevated and flows overthe cold air pool and, subsequently, over the easternrim of the basin. The inflow of air into the basin wasnot anticipated by W96: they assumed that the sidewalldrainage flows resulted from local cooling alone andthat downward sinking motion compensated for themass flux through the canyon. Another unexpectedfinding was the convergence of the near-surface drainage flows in the southern portion of the basin throughout most of the evening. Even though the predicted temperature profiles in thecenter of the basin are not perfect, the model qualitatively captures the observed winds within the basin andthe canyon and reproduces the temperature profile atthe end of the period quite well. This suggests that thenet cooling within the basin during the evening is reasonably simulated so that the numerical results can beused to compute the basin heat budget. When the heatbudget is averaged between 2000 and 0500 MST, theheat loss within the basin can be attributed mostly toturbulent diffusion. Radiative flux divergence cools thebasin atmosphere to a lesser extent overall than turbulent diffusion, while advection produces warming. Turbulent diffusion and advection are both very large overthe basin sidewalls, but the cooling due to turbulentdiffusion is usually larger than advective warming inthose areas. Turbulent diffusion becomes nearly zeroin the middle of the basin and cooling in this regionresults from horizontal advection of cold air from thesidewalls and radiative cooling. Radiative flux divergence did not vary significantly across the basin unlessvariable surface characteristics were employed. Thecomplex flow field predicted by the model greatly complicates the cooling rate at any one location within thebasin; nevertheless, the average heat storage termacross the basin is almost directly proportional to theunderlying terrain elevation. Interestingly, the relative contributions of the termsin Eq. (3) are approximately the same for both themodel heat budget and the observational heat budgetdescribed by W96; however, the magnitudes of theterms are significantly different. One reason for the differences is that sidewall effects cannot be included inthe method used by W96 since there were no observations made along the basin sidewalls. On the otherhand, the drainage flows produced by the model alongthe sidewalls cannot be verified either. Forecast errorsat the basin site would also account for some of thedifferences in the methods. The results from this studyand W96 suggest that a basin heat budget based onsingle point measurements can contain relatively largeuncertainties in the magnitude of the individual heatbudget terms due to an inadequate representation ofadvection and the failure to resolve the slope flows. Wind direction had a profound effect on the nocturnal cooling. Since the higher ten'ain west of the basincan be a major source of cold air, the ambient winddirection affects the amount of cold air that eventuallydrains into the basin. The acceleration of the ambientwinds by the geometry of the underlying terrain willalso affect the horizontal and vertical temperature advection within the basin. The model indicates that thepresence of vegetation may be important in producingdetails in the wind and temperature fields at specificlocations, as well as affecting the overall basin heatbudget. A more complex treatment of vegetation, suchas a canopy parameterization, is probably necessary toadequately describe the generation of turbulence andthe cooling due to turbulent diffusion along the sidewalls. Still, the vegetation distributions employed inthis study had a relatively minor impact on the basinheat budget when compared to the ambient wind direction. Few three-dimensional dynamic model evaluations,such as the one in this study, have been made usingobservations from a small-scale basin. While the performance of the model is encouraging, the differencesbetween the model results and the observations demonstrate the difficulty in simulating the evolution of thenocturnal stable boundary layer within small-scale basins. Despite the preliminary results from the tests described in this study, improvements in the forecastsmay be achieved with higher spatial resolutions and/orimproved parameterizations of nocturnal turbulence incomplex terrain. An adequate representation of nocturnal turbulence in complex terrain is especially important since the turbulent diffusion is the dominant cooling mechanism in relatively small basins. A better understanding of basin meteorology is also important forair pollution applications since many urban areas arelocated within basins. Acknowledgments. The authors thank Drs. ChrisDoran and Rich Barchet for their comments on thismanuscript. This research was supported by the Environmental Sciences Division of the U.S. Department ofEnergy under Contract DE-AC06-76RLO 1830 at Pacific Northwest National Laboratory under the auspicesof the Atmospheric Studies in Complex Terrain program. Pacific Northwest National Laboratory is operated by Battelle Memorial Institute for the U.S. Department of Energy.REFERENCESBader, D. C., and T. W. Horst, 1990: Model evaluation of the as sumptions used in valley energy budgets. Preprints, Fifth Conf on Mountain Meteorology, Boulder, CO, Amer. Meteor. Soc., 277-280.2178 JOURNAL OF APPLIED METEOROLOGY VOLUME35 , and --, 1992: Scale analysis of the thermal energy budget in an idealized mountain valley. Preprints, Sixth Conf on Moun tain Meteorology, Portland, OR, Amer. Meteor. Soc., 243-247.Blockley, J. A., and T. J. Lyons, 1994: Airflow over a two-.dimen sional escarpment. III: Nonhydrostatic flow. Quart. J. Roy. Me teor. Soc., 120, 79-109.Chen, C., and W. R. Cotton, 1983: A one-dimensional simulation of the stratocumulus-capped mixed layer. Bound.-Layer Meteor., 25, 289-321.Clements, W. E., $. A. Archuleta, and P. H. Gudiksen, 1989: Exper imental design of the 1984 ASCOT field study. J. Appl. Meteor., 28, 405-413.Dickinson, R. E., A. Henderson-Sellers, P. J. Kennedy, and M. F. Wilson, 1986: Biosphere-Atmosphere Transfer Scheme (BATS) for the NCAR Community Climate Model. NCAR Tech. Note NCAR/TN-275+STR, National Center for Atmo spheric Research, Boulder, CO, 69 pp.Doran, J. C., 1991: The effects of ambient winds on valley drainage flows. Bound.-Layer Meteor., 55, 177-189.Fast, J. D., 1995: Mesoscale modeling in areas of highly complex terrain. J. Appl. Meteor., 34, 2762-2782.Helfand, H. M., and J. C. Labraga, 1988: Design of a nonsingularlevel 2.5 second-order closure model for the prediction of atmospheric turbulence. J. Atmos. Sci., 45, 113-132.Horst, T. W., K. J. Allwine, and C. D. Whiteman, 1987: A thermal energy budget for nocturnal drainage flow in a simple valley. Preprints, Fourth Conf on Mountain Meteorology, Seattle, WA, Amer. Meteor. Soc., 15-19.Legg, B. J., and M. Raupach, 1982: Markov chain simulation of particle dispersion in inhomogeneous flows: The mean drift ve locity induced by a gradient in Eulerian velocity variance. Bound.-Layer Meteor., 24, 3-13.Louis, J.-F., 1979: A parametric model of vertical eddy fluxes in the atmosphere. Bound.-Layer Meteor., 17, 187-202.Mahrer, Y., and R. A. Pielke, 1987: A numerical study of the airflow over irregular terrain. Contrib. Atmos. Phys., 50, 89 113.Mellor, G. L., and T. Yamada, 1982: Development of a turbulent closure model for geophysical fluid problems. Rev. Geophys. Space Phys., 20, 851-875.Petkovsek, Z., 1978: Zones of convergence in local air flow in valleys and basins. Preprints, 15 lnternationale Tagung Fuer Alpine Meteorologie, Grindelwald, Switzerland, Swiss Meteor. Inst., 92-96.Pielke, R. A, and Coauthors, 1992: A comprehensive meteorological modeling system--RAMS. Meteor. Atmos. Phys., 49, 69-91. Pitts, R. O., and T. J. Lyons, 1989: Airflow over a two-dimensional , escarpment. I: Observations. Quart. J. Roy. Meteor. Soc., 115, 965-981.Poulos, G. S., and J. E. Bossert, 1995: An observational and prog nostic numerical investigation of complex terrain dispersion. J. Appl. Meteor., 34, 650-669.Whiteman, C. D., T. B. McKee, and J. C. Doran, 1996: Boundary layer evolution within a canyonland basin. Part I: Mass, heat, and moisture budgets from observations. J. Appl. Meteor., 35, 2145-2161.Zhong; S., and C. D. Whiteman, 1995: A numerical study of atmo spheric heat budget terms over idealized valley topography. Pre prints, Seventh Conf on Mountain Meteorology, Breckenridge, CO, Amer. Meteor. Soc., 339-344.

Save