1. IntroductionThere has been increasing realization recently thatrealistic simulation of surface fluxes is extremely im-portant for 3D mesoscale modeling. Particularly, theevolution and maximum depth of the planetaryboundary layer (PBL) is highly dependent on the par-titioning of the mailable net radiation into sensibleand latent heat fluxes. Clearly, this partitioning alsohas profound importance to low-level cloud formation(Wetzel and Argentini 1990). In addition, mesoscalefluxes of heat due to secondary circulations caused byspatially varying land use have been shown to be ofthe same order of magnitude as turbulent fluxes (Pielkeet al. 199 1 ). Therefore, adequate treatment of surface* On assignment to the Atmospheric Research and Exposure As-sessment Laboratory, US. Environmental Protection Agency, Re-search Triangle Park, North Carolina.Corresponding author addrexs: Dr. Jonathan E. Pleim, U.S. EPA,MD-80, Research Triangle Park, NC 277 I I.properties such as soil moisture and vegetation char-acteristics are essential for accurate simulations of PBLdevelopment and cloud coverage, both of which arekey factors influencing atmospheric chemistry and airquality.Several models have been developed in recent yearsfor the simulation of surface-vegetation-atmosphereexchange in meteorological models (Sellers et al. 1986;Dickinson et al. 1986; Wetzel and Chang 1987; Noilhanand Planton 1989). Most of these models evolved fromthe methods presented by Deardorf( 1978), who de-scribed a force-restore algorithm for both surface tem-perature and near-surface soil moisture. The key pro-cesses that need to be represented in these models in-clude short- and longwave radiation, turbulent surfacefluxes of heat and moisture, evapotranspiration, andheat and moisture fluxes within the soil. The sensitivityof these models to various vegetative and soil charac-teristics have been investigated by Wilson et al. ( 1987 )and Mihailovic et al. ( 1992). A key concern related tothe use of these models, which are essentially 1D andvery local in scale, in mesoscale or larger-scale gridmodels is the effects of subgrid heterogeneity of landJANUARY 1995 PLEIM AND XIU 17surface characteristics (primarily vegetation and soiltype). One aspect of this issue, which was studied byWetzel and Chang (1988), is the nonlinearity of hor-izontal averaging over varying surface conditionswithin each model grid cell. Another aspect ofthis issue,which was described by Avissar and Pielke ( 1989), isthe occurrence of subgrid circulations resulting fromcontrasting surface conditions.For many years the Pennsylvania State University-National Center for Atmospheric Research MesoscaleModel Version 4 (MM4) ( Anthes et al. 1987 ) has beenused as the meteorological driver for the Regional AcidDeposition Model (RADM) (Chang et al. 1987).However, the capability of this model for accuratecharacterization of the PBL, particularly the daytimePBL height, has been found to be somewhat lacking.For example, a model evaluation study of the MM4-RADM system conducted during the summer of 1988,which included aircraft measurements over a large areaof the eastern United States, showed consistently un-derpredicted PBL heights (Pleim and Ching 1993).Since the summer of 1988 was extraordinarily dry inthe eastern United States, the underprediction of PBLheights reflects the model's insensitivity to surfacemoisture conditions.The current MM4 system uses a force-restoreground temperature algorithm in the manner of Dear-dorff ( 1978). This scheme, however, does not includethe prognostic simulation of soil moisture but insteaduses a moisture availability factor to specify the effectivesoil moisture available for evaporation. A drawback ofthis approach is that these moisture availability factorsare functions of land use category only and thereforehave no ability to respond to changes in soil moistureconditions. Therefore, MM4 simulations of PBLheight, surface temperature, PBL temperature and hu-midity, and low-level cloud development may be sig-nificantly unrealistic whenever soil moisture conditionsdiffer from the assumed moisture availability. SinceRADM simulations of chemical concentrations dependenormously on PBL height, we decided to embark onan effort to upgrade this part of the MM4 system.In this paper, we describe the development of a sim-ple surface-PBL model for eventual inclusion into theMM4. The model consists of a land surface model forthe prognostic simulation of soil moisture and soiltemperature, a PBL model for the simulation of verticalturbulent transport of heat, moisture, and momentum,a flux-profile algorithm that couples the surface withthe atmosphere through the surface fluxes, and a simplesurface radiation model. The land surface portion ofthe model is based on a model developed by Noilhanand Planton ( 1989). The PBL module and flux-profilerelationships are the same as currently used in the latestversion of RADM. The PBL model is a hybrid of non-local closure, for convective conditions, and eddy dif-fusion. A preliminary version of the model, initially inID form, is applied to two field studies for comparison,where one case was in a dry sparsely vegetated desert(Wangara) and the other in a rather moist longgrass prairie [First ISLSCP (International SatelliteLand Surface Climatology Project ) Field Experiment(FIFE)]. Note that these case studies represent the ini-tial testing of a model that is still under development(e.g., the effects of condensation and clouds are yet tobe included). Last, a test is made of the sensitivity ofPBL height and surface temperature to soil moistureinitialization in order to demonstrate the range of errorslikely to result from the neglect of soil moisture vari-ation. The next stage of this effort will be to incorporatethe surface-PBL model described here into the MM4-MM5 system. This 3D model will then be extensivelytested for a variety of conditions and seasons to testthe model's ability to realistically respond to hetero-geneous vegetation and soil conditions as well as time-varying soil hydrology. These are all capabilities absentfrom the current version of MM4 / 5.2. Model descriptiona. Soil-vegetation modelThe soil-vegetation model used in this study wasdeveloped specifically for use in mesoscale meteorologymodels. It has been extensively documented and testedfirst in a 1D prototype (Noilhan and Planton 1989;Jacquemin and Noilhan 1990) and more recently in acomprehensive 3D meso-0-scale meteorology modelas described in a series of three papers (Bougeaultet al. 1991a; Bougeault et al. 1991b; Noilhan et al.199 1 ). In both cases, model results were compared toextensive field measurement data from the HAPEX-MOBILHY experiment (Andrk et al. 1986), which in-cluded both cropland and forest. The model is designedto simulate the essential processes involved in surface-atmosphere interactions with the fewest parameters andcomplexities. For example, one temperature is usedfor both vegetation and the soil surface unlike somecanopy resolving models (Sellers et al. 1986; Dickinsonet al. 1986), which simulate ground and canopy tem-peratures separately. Furthermore, compared to someof the more complex models, the additional data re-quirements are minimal. Specifically, incorporation ofthis model into the MM4 would require the additionof only soil texture classification, leaf area index (LAI),minimal stomatal resistance, and fractional vegetativecover to the model input database. Note that roughnesslength is also needed by this scheme but MM4 alreadyuses this parameter. Therefore, the combination ofsimplicity, minimal data requirements, and the model'sproven performance over a variety of land use condi-tions made it a logical choice for use in the MM4system.The soil-vegetation model includes prognosticequations for both soil and vegetation temperature andsoil moisture using a two-layer force-restore algorithm.The driving force for the prognostic simulation of sur-18 JOURNAL OF APPLlface temperature is the surface energy budget that in-cludes net radiation, sensible heat flux, latent heat flux,and heat flux into the ground. Similarly, the surfacesoil moisture is driven by the surface moisture budget,which includes precipitation and dew formation, directevaporation from the ground and canopy, evapotrans-piration, and moisture flux within the soil. Transpira-tion through vegetation directly from the lower soillayer (root zone) is modeled via a stomatal resistanceanalog algorithm. The details of the land-surface modelcan be found in both Noilhan and Planton ( 1989) andJacquemin and Noilhan ( 1990). Our version of themodel was coded directly from these papers. The modelis based on the following set of five partial differentialequations:= vegP - E,,aw,at(4)where Ts is the soil surface temperature (nominally 1cm) and T2 is the average temperature of the lowerlayer (usually 1 m), which acts as a slowly varying heatreservoir. Here w, is the volumetric soil moisture inthe top 1 cm (d, ) and w2 is the average volumetric soilmoisture down to about I m (d2). Here W, is theamount of water in the canopy, which is limited toW,,, = h vegLAI where h = 0.2 mm, and veg is thefractional coverage of vegetation; P is precipitation rateand P, is the precipitation rate reaching the ground;and Eg and E,, are the evaporation rates from theground and by transpiration, respectively. Note thatvolumetric soil moisture ( wg and w2) is limited to thesoil saturation point. If the net change in soil moisture(precipitation - evaporation) would result in soilmoisture greater than saturation, the excess is assumedto runoff.Equation ( 1 ) shows that local changes in soil surfacetemperature result from the residual of the surface en-ergy balance among net radiation R,, surface heat fluxH, latent heat flux LE, and soil heat flux, which isparameterized as a restoring force on the soil temper-ature back toward the diurnal average ( T is the timeconstant- 1 day). The coefficient CTin Eq. ( 1 ), whichrepresents the inverse of the bulk heat capacity of thesurface and vegetation, is a function of the deeper layersoil moisture w2 as shown in Noilhan and Planton( 1989). Expressions for the soil moisture coefficientsED METEOROLOGY VOLUME 34C, , C2, and for wgeq, which is essentially w2 modifiedto account for gravity, are given in the appendix ofJacquemin and Noilhan ( 1990). These are expressedin terms of soil parameters, such as field capacity, wilt-ing point, saturation, and various thermal and hy-draulic properties of the soil, which are specified ac-cording to the 1 I soil types of the U.S. Department ofAgriculture (USDA) textural classification (Clapp andHornberger 1978). Therefore, the only soil data re-quired for this model is the soil texture type. In ourversion of this model, Eqs. ( 1 ) - ( 5 ) are integrated usinga semi-implicit Crank-Nicolson technique.A notable variation of our model from the work ofNoilhan and Planton (1989) is in the coefficient CT(inverse of the bulk heat capacity), which they defineas the harmonic average of the value for soil and thevalue for vegetation weighted by the vegetative frac-tional coverage. Their rationale is that for areas com-pletely shielded by vegetation the relevant surface tem-perature for computing surface fluxes is that of the veg-etation and not the ground. The problem with thisapproach is that the value they use for the heat capacityof vegetation is practically negligible leading to virtuallyno heat storage at the surface. This goes against someobservational evidence that even in highly vegetatedareas, such as grasslands, there is a significant soil heatflux and lag to morning surface heating. For example,the daytime (8 h centered on local noon) soil flux av-eraged over all sites and measurement days during the1987 FIFE, where the ground was essentially com-pletely covered by tall grass, was about 50 W m-2,which is 13% of the averaged net radiation (Smithet al. 1992b). Therefore, for these case studies, ratherthan using the weighted average of soil and vegetationheat capacities, we simply used the soil heat capacityregardless of vegetative cover.Other investigators (Argentini et al. 1992; Wetzeland Chang 1988) have parameterized the heat capacityof vegetation according to the amount of liquid watercontained within the plants as well as any water in thecanopy from rain or dew. Thus, the heat capacity ofbiomass in their models is simply (b, + W,)C,, whereb, is the water equivalent biomass, W,. is dew and rain-water in the canopy, and C, is the heat capacity ofwater. In areas of dense vegetation, such as forests oragricultural crops the biomass heat capacity can sig-nificantly delay surface heating in the morning. How-ever, for the cases studies presented here the heat ca-pacity of the soil was probably much greater than thebiomass heat capacity.b. Planetary boundary layer modelThe PBL model is a hybrid of a simple nonlocalclosure scheme used during conditions of free convec-tion and an eddy diffusion scheme for all other con-ditions. The criterion currently used to define free con-vective conditions is h/L -3, where h is the heightJANUARY 1995 PLEIM AND XIU 19of the PBL and L is the Monin-Obukhov length. Thenonlocal closure model, referred to as the "asymmet-rical convective model" ( ACM ) , is designed to simulaterapid upward transport from the surface layer to alllevels within the convective boundary layer (CBL) byrapidly rising buoyant plumes and more gradualdownward transport by broad slow compensatory sub-sidence (Fig. 1 ). The ACM was developed from theconvective model of Blackadar ( 1978), which is cur-rently used in MM4. The ACM retains the direct non-local upward flux of the Blackadar model but replacesthe downward return flux with a local layer by layerapproach, thereby combining the rapid upward con-vection of the Blackadar model with diffusion-likesubsidence. The ACM has been shown, through com-parison to large-eddy simulations, to provide a morerealistic simulation of vertical fluxes in the CBL thaneither the Blackadar model or local eddy diffusionschemes without a significant increase in computa-tional expense (Pleim and Chang 1992 ) .For all conditions other than free convective a simpleeddy diffusion model is used. The eddy diffusivities arebased on surface-layer and boundary layer length andvelocity scaling in the manner of Holtslag and Nieuw-stadt (1986), Troen and Mahrt (1986), and others.The eddy diffusion model is described in the appendixof Pleim and Chang ( 1992). An important differencebetween this scheme and the model currently used inthe MM4 (Blackadar 1976) is that the eddy diffusivitiesK, within the PBL depend on the estimated height ofthe PBL. PBL height is estimated according to a bulkRichardson number method as suggested by Holtslaget al. ( 1990). In this scheme, the PBL bulk Richardsonnumber is computed,gzmRi -/I - e,[u(z)2 + v(z)2] 'where A& = &(z) - Bus and = 0.5 [O,(z) + O,,], atsuccessive heights z above the ground until Rib 2 Ri,(Ri, = 0.25). The top of the PBL is defined as theheight at which Rib first equals the critical Richardsonnumber.The governing equations for the PBL in 1D form(horizontal advection neglected) in a coordinates [ a- (P - Ptqp)(Psurf - PJ' and plop = 10 kPa1 wherecondensation is not considered are-IFIG. 1. Schematic representation of mixing in a 1 D column of airas simulated by the asymmetrical convective model (ACM). Linethicknesses are proportional to mixing rates.where 0" is virtual potential temperature, q is mixingratio, U and V are the horizontal wind components,and w is the vertical velocity in sigma coordinates (w= da/dt). Primed variables represent turbulent fluc-tuations from the mean. Here U, and V, are the geo-strophic wind components, which for the case studiesdescribed below were derived from observations andsupplied as inputs to the model. For the Wangara caseit was possible to resolve the geostrophic winds in timeand vertically, thereby incorporating some baroclinicforcing in the simulation. However, this model canperform well only under conditions where horizontaladvection and condensation are unimportant. There-fore, the case studies were selected for their lack ofclouds and relatively weak horizontal and vertical ad-vection.The surface model and the PBL model are linkedthrough their respective boundary conditions, namely,the surface fluxes of heat, moisture, and momentum.Heat and momentum fluxes are determined from windspeed and temperature differences between the surfaceand the lowest atmospheric level using surface layerand the PBL flux-profile relations developed by Byun( 1990, 199 1 ) based on Rossby number similarity andprofile matching. Humidity fluxes are represented by20 JOURNAL OF APPLIED METEOROLOGY VOLUME 34the following three parallel pathways: transpirationR, = Riw( 1 - a)~,, - c,aT; + c,aT;f,where 7sw is the shortwave transmissivity, cg is theemissivity of the surface, and tu is the emissivity of theair. Net radiation flux is computed similarly to the(15)(Etr),4sat( Ts) - 4a . 'r,, + r,( la)Etr = P, veg( 1 - u)-IMM4 (Anthes et al. 1987) but with some minor im-provements. Rather than using a constant surface al-evaporation from wet parts of the canopy (E,),4sat(T,) - 4,. r,bedo, as in the MM4, the surface albedo is defined aswork of Idso et al. ( 1975), such that the total albedo(a) isE, = pa Vega , ( b, a function of the solar zenith angle as suggested by theand direct evaporation form the ground ( Eg),whereand Hu is constrained to 1 if wg is greater than fieldcapacity ( wfc). Veg is the fractional vegetation coverage,and a is the wet fraction of the canopy. The surfaceresistance for evapotranspiration is computed as(13)rsmnr, =LA1 FI F2 F3 F4 'where leaf area index (LAI) and minimum stomatalresistance r,,,, are specified according to land use clas-sification. The fractional conductances Fare functionsof solar radiation, root-level soil moisture w2, air hu-midity deficit, and ambient air temperature. See Noil-han and Planton ( 1989) and Jacquemin and Noilhan( 1990) for the functional forms of these conductances.The aerodynamic resistance r, is computed from thedifference in virtual potential temperature between theair and the ground and the surface heat flux,a = a, + a,,(16)where a, is a solar zenith angle Z adjustment,az = 0.01 [e~p(O.O03286Z'.~) - 13, (17)and a, is the minimum albedo with a solar zenith angleof zero, which is specified according to surface type.The surface albedo a, is sometimes specified as a func-tion of soil moisture ( McCumber and Pielke 198 1 ),however, this is appropriate only for bare soil.In MM4 the direct downward solar irradiancereaching a horizontal surface of unit area at the top ofthe atmosphere ( R!w) is a function of the solar zenithangle only. This quantity, however, also depends onthe distance between the earth and the sun such thata more complete formulation iswhere a is the average distance from the earth to thesun and r is the distance to the sun as a function oftime of year and So is the solar constant. The ratio a2/r2 at any specific day of the year can be calculated(from Paltridge and Platt 1976) asa21 = 1.0001 IO + 0.034221 codo + 0.001280 sindor+ 0.000719 cos2do + 0.000077 sin2do, (19)where do = 2~m/365 and m is the day number startingwith 0 on 1 January and ending 364 on December 3 1.Note that the effect of the variation of the earth's dis-tance from the sun results in a maximum of 3.3% vari-When heat flux is near zero (H/paCp K m s-'),during the morning and afternoon transitions, r, is es-timated from neutral surface-layer similarity theory,r, = - in( :),( 14b)0.74u*kation in Rtw.where u, is the surface friction velocity, k is the vonShortwave transmissivity for multiple reflection isKgrmgn constant (Oe4 )> 'I is the center height Of thecomputed as in the MM4 ( Anthes et al. 1987). Clear-first model level, and zO is the roughness length- Theair transmissivities for direct and diffuse radiation dueto absorption and scattering, as well as backscatteringcoefficients, are determined as functions of precipitablewater and pathlength from a look-up table created frommodel. The expressions for transmissivitv and the look-surface and PBL models are integrated using operatorsplitting with a short time step to minimize oscillationsof the temperature and humidity in the lowest air layersurface fluxes.and surface soil layer that can propagate through thethe Carlson and Boland ( 1978) radiative transferup table are given in the MM4 model description(Anthes et al. 1987). Note that the effects of cloudsare not included in this study but will be in the nextphase of this development when the surface-PBLe. Radiation modelgiven byNet radiation flux [ R, in Eq. ( 1 )] at the surface isJANUARY 1995 PLElM AND XIU 21model described here is incorporated into the full3D MM4.The net longwave radiation at the surface is com-puted, as shown in Eq. ( 15), simply as the sum of theupward and downward components. In both cases theblackbody relationship of Stefan-Boltzman is usedwith the emissivity of the ground tg set to one and theemissivity of air e, computed as a function of precip-itable water (Monteith 196 1 ),where w,, is the precipitable water computed as the ver-tical integral of water vapor concentration. Further-more, the atmospheric temperature T, used in Eq. ( 15)is the ambient temperature at the level of the verticalcentriod of precipitable water.3. Modeling resultsThe initial testing of the combined surface-PBLmodel is the application of a 1 D prototype to the fieldstudies of Wangara (Clarke et al. 1971) and FIFE(Sellers et al. 1992a). The purpose of this effort is todetermine the model's ability to simulate real worldsurface fluxes and PBL characteristics with the eventualgoal of improving those aspects of the MM4 for use inair quality modeling. Therefore, the evaluation em-phasizes parameters that are most relevant to air qualitysuch as PBL height and surface temperature. However,since these parameters are interdependent with manyothers, the evaluation of sensible and latent heat fluxesas well as soil moisture are also quite relevant. In ad-dition, moisture fluxes are very important for themoisture loading of the PBL and therefore for clouddevelopment. Clouds have the potential to feedback tosurface fluxes of heat and moisture through the ob-struction of solar radiation and the moistening of thesurface by precipitation. Cloud effects will be studiedin the next phase of this work in the context of the3D MM4.a. WangaraThe first case study involved a 36-h simulation ofdays 33 and 34 of the Wangara Boundary Layer Ex-periment ( 16 and 17 August 1967) starting at 0900local time. These particular days have been used quiteoften as a test case for PBL models (Deardorff 1974;Wyngaard and Cot6 1974; Yamada and Mellor 1975;Binkowski 1983) because of clear skies all day and veryweak horizontal advection. The site was an arid regionof Australia with very sparse vegetation. Therefore, thiscase does not provide a good test of the soil moistureand evapotranspiration components of the model butis a good test of the PBL and heat flux algorithms.Table 1 lists the surface parameters used in this sim-ulation as well as for the FIFE simulations, which aredescribed in the next section.TABLE 1. Surface parameters for case studies.Parameter Wangara FIFE FIFEDate 16, 17 Aug 1967 11 July 1987 6 June 1987Soil type Loam Silty clay loam Silty clay loam20 (cm) 0.24 6.5 4.5LA1 0. I 2.8 1.9Veg ("/.) 5 99 99The model was run with 25 vertical levels up to about2 km using a u,, coordinate system. The lowest levelwas Au = 0.0025 or about 20 m thick. The secondlayer was Au = 0.0075 and each layer above was Au= 0.0 1 or about 80 m thick. Temperature and humidityprofiles were initialized by interpolation of the 0900radiosonde measurements and the winds were inter-polated from pilot balloon observations at the sametime. The soil surface temperature T, was initializedto the 0900 surface temperature measurement (279K). Since there was no direct measurement of the deepsoil temperature, the initial value had to be inferredfrom measurements of ground heat flux and surfacetemperature. At 0800 the ground heat flux was aboutzero (Clarke et al. 197 1 ) indicating that the deep-layertemperature T2 was approximately the same as thesurface temperature. Therefore T2 was initialized tothe 0800 surface temperature (273 K). Both shallowand deep soil moisture were set to the wilting point assuggested by Clarke et al. ( 197 1 ) since it had not rainedfor many days.Surface geostrophic winds were derived every 3 hfrom five closely spaced field measurement stations and14 Bureau of Meteorology stations. In addition, ther-mal winds in two layers from 0 to 1 km and 1 to 2 kmwere estimated twice daily (0900 and 2 100 LT) fromthe Bureau of Meteorology synoptic radiosonde net-work. Hourly estimates of thermal winds were linearlyinterpolated from the twice daily measurements. Ver-tical profiles of geostrophic winds were derived from aparabolic fit to the thermal wind data as suggested byYamada and Mellor ( 1975 ). The resulting geostrophicwind estimates as functions of both height and timewere supplied to the model as input.Figure 2 shows both measured and modeled surfacefluxes of net radiation, sensible heat, latent heat, andground heat starting at 0900 LT on day 33 ( 16 August)and ending at midnight on day 34. Measured net ra-diation and ground fluxes were reported by Clarkeet al. ( 197 1 ). Sensible heat fluxes were derived ac-cording to surface-layer similarity theory from 1- to4-m temperature and wind speed differences by Hicks( 198 1 ) , hereafter referred to as the Hicks method.These calculations were made from 3-h running av-erages of the difference measurements in order tosmooth out some of the noise. An alternative to theHicks calculations of sensible heat flux is the differencebetween the net radiation and the ground heat flux22 JOURNAL OF APPLIED METEOROLOGY VOLUME 34(R, - G) . Under steady-state conditions, this quantityshould represent the sum of the sensible and latentheat fluxes if all other sources of heat, such as thermaladvection and condensation, are negligible. If latentheat flux is also negligible, which was likely during thesevery dry days, R, - G can be used as an approximationof sensible heat flux. The similarity between the Hicksmethod for calculating sensible heat flux and R, - Gfor the first day supports the validity of both ap-proaches.In general, the modeled fluxes compare very well tothe observations on both days. The magnitudes of thedaily peaks and their timing are well simulated suchthat the peak of the sensible heat flux lags the peak inthe net radiation by 1 h. Also, in both the model andthe observations the sensible heat flux curves cross thenet radiation curves in the afternoon at hours 7 and3 1 ( 1600 LT). The net radiation is almost perfectlysimulated on both days, which merely shows that sim-ple radiation calculations are sufficient under such dry,clean conditions when aerosol scattering is negligible.Note that we specified the minimum surface albedo a,for the Wangara experiment, which was 0.15 accordingto Edson ( 1980), and then computed the total surfacealbedo according to zenith angle as in Eq. ( 16).The peak values of the model simulated sensible andground heat fluxes are about the same on the two days,whereas the observed values differed. The measuredground heat flux was slightly higher than the modeledflux on the first day, whereas the modeled and mea-sured values on the second day were about the same.Relative to the values computed by the Hicks method,the modeled sensible heat flux was a little low on thefirst day and high on the second day. However, on thesecond day R, - G was considerably higher than theHicks method heat flux, which casts some doubt onthe accuracy of the Hicks method estimates for thisday. Note that the modeled peak value is between theHicks method and R, - G.Figure 3 shows modeled and observed PBL heightsfor both days of Wangara. Observations were derivedfrom 3-h radiosonde measurements as reported by Ya-mada and Mellor ( 1975). Only daytime PBL heightswere used for this comparison so nighttime values inFig. 3 are not meaningful for either the model or ob-servations. The observations at hours 9 and 33 of thesimulation ( 1800 LT) actually represent the top of theresidual layer rather than active PBLs since ground-based inversions had developed by these times. We in-cluded these points, even though the actual PBL heightsat these times were much lower, so that the peak ofthe PBL development, which probably occurred some-where between the 1500 and the 1800 LT radiosondes,is better represented. Clearly, the model and observa-tions compare very well in terms of the peak PBL heightas well as the timing of the rise. Even the increasedPBL height on the second day was accurately simulated.NE5YxY-hEBh\vXObserved Fluxes - Wangara days 33 and 340 6 12 18 24 30 36Time (hrs)-200 '''''~"~''1'''"'''~"''"''''''''0 6 12 18 24 30 36Time (hrs)FIG. 2. Observed (top) and modeled (bottom) surface fluxes ofnet radiation (I?"), sensible heat (H), latent heat (LH), and groundheat (G flux) for Wangara days 33 and 34. Hour 0 is 0900 LT.More details of the model simulations can be seenin the soundings of temperature and winds. Figure 4shows modeled and measured vertical profiles of virtualpotential temperature at four times during day 33. The0900 sounding is the initialization so the slight differ-ences between the two plots is just due to the inter-polation of the sounding measurements to the model'svertical grid. By 1200 the surface inversion dissipatedand a convective layer formed in both the modeledand measured profiles. The model, however, showsJANUARY 1995 PLEIM AND XIU 23Wangara PBL Height1500--0 9 18 27 36 Time (hrs)FIG. 3. Observed and modeled PBL heights for Wangara days 33 and 34.lower temperature near the ground due to the limitedvertical resolution of the model (layer 1 is about 20 mthick) compared to the measured profile, which startswith the temperature at screen height. Also, the pre-dicted top of the convective layer is about 200 m toolow. By 1500 these differences have lessened such thatthe height of the convective layer is about 1200 m inboth the modeled and measured profiles but the lowestlayer modeled temperature is still underestimated byabout 1C. The 1800 LT profiles both show a well-Model Thetav Profiles - Day 3320001600-5 1200W4vLL ulI'5 8004000272 276 280 284 2aa 292Thetav (K)mixed residual layer still up to about 1200 m with adeveloping surface inversion. The surface temperatureis now overestimated, again due to limited model res-olution.Figures 5 and 6 show modeled and measured profilesof U (eastward) and V (northward) wind componentsat four times during day 33 and the early morning ofday 34. Again, the 0900 LT profiles are the initializa-tion. At 1800, measurements show that the winds havebecome quite uniform in the PBL at moderate speeds(about 6 m s-') from the east southeast. The modelsimulates the U component quite well but overesti-mates the V component with a gradient that decreaseswith height. The 0300 and 0600 soundings show a well-developed nocturnal jet peaking from the east northeastat 13-14 m s-' in a layer from about 200 to 400 mAGL. The modeled profiles show a nocturnal jet ofsimilar magnitude (- 12 m s-') but directed a littlemore from the north and more spread out vertically.Note that winds are controlled by a combination ofgeostrophic forcing (pressure gradient force and Co-riolis force) and vertical turbulent momentum fluxes.Since the thermal winds used to define the vertical pro-files of the geostrophic winds were resolved only in two1 -km-thick layers, the modeled profiles tend to haverather uniform gradients in each of the two layers.In summary, the model performed extremely wellin simulating this case from the Wangara experiment.The surface fluxes as well as the PBL height and verticalprofiles of winds and temperatures compared very wellwith the observations. Success in this case is only abeginning, however, since the arid, clean clear sky,sparsely vegetated conditions are the simplest to model.20001600hE 1200.iW4vLcI.g aoo40002Observed Thetav Profiles - Day 33! 276 280 284 2Thetav (K)1 3 292FIG. 4. Modeled (left) and measured (right) virtual potential temperature profiles at 0900, 1200, 1500, and 1800 LT on day 33.24 JOURNAL OF APPLIED METEOROLOGY VOLUME 34Model Eastward Wind Component-14 -12 -10 -8 -6 -4 -2 0 2U (m/s)20001600h5 1200c7vYrm8004000Observed Eastward Wind ComponentIIIIII I-14 -12 -10 -8 -6 -4 -2 0 2U (m/s)FIG. 5. Modeled (left) and measured (right) eastward wind component (U) profiles at 0900 and 1800 on day 33 and 0300 and 0600 on day 34.Clearly, this case does not test the soil moisture, mois-ture flux, and vegetation-related aspects of the model.Therefore, the testing continues with two case studiesfrom the FIFE.b. FIFEThe second stage of model testing involves simula-tion of two case studies from the FIFE. The field studyarea was a 15 km X 15 km area of predominantly tallModel Northward Wind Component20001600hEff4j 1200Yc-c mI'5 8004000-a -6 -4 -2 o 2 4V (m/s)grass prairie near Manhattan, Kansas. An extensivemonitoring program of satellite, meteorological, bio-physical, and hydrological measurements was madeduring the growing seasons of 1987 and 1989. Therewere four 12-20-day intensive field campaigns ( IFCs)during 1987 and one in 1989. During the IFCs, surface-based measurements were coordinated with airborneand satellite measurements (see Sellers et al. 1988,1992a for an overview of FIFE). For this study, surface-based flux and radiosonde profile measurements wereObserved Northward Wind Component20001600hF 12001ff4../Y.c.$ 800I4000-8 -6 -4 -2 0 2 4V (m/s)FIG. 6. Modeled (left) and measured (right) northward wind component (V) profiles at 0900 and 1800 on day 33 and 0300 and 0600 on day 34.JANUARY 1995 PLEIM AND XIU 25used for comparison to the model simulations. Theexperiment included measurements of sensible and la-tent heat fluxes made by both eddy correlation andBowen ratio energy balance methods and ground heatflux by heat flux plates at 20 sites. In addition, therewere 12 automatic meteorological stations ( AMs) thatmeasured such parameters as winds, temperature, hu-midity, pressure, precipitation, soil moisture, soil tem-perature, surface temperature, and the components ofthe radiation budget. Rather than averaging over allthe sites in the study area we chose to model one cen-trally located site that included both types of flux mea-surements and AMS measurements. The specific siteused in this study (collocated AMS site 1 1, eddy fluxsite 16, and Bowen ratio site 18) is the same one usedby Kim and Verma ( 1990) in their study of surfaceenergy balance and Verma et al. ( 1992) in their studyof momentum, water vapor, and carbon dioxide ex-change.1) 11 JULY 1987The first FIFE case study was 11 July 1987, the lastday of the second IFC, which was identified as a "goldenday" meaning that it was the best day of data collectionfor that IFC. The day was mostly clear and very windyuntil about 1700 CDT when significant cloud coverdeveloped. Therefore, the model comparison breaksdown at that point since the model does not yet includeparameterizations for the effects of clouds. This casewill be revisited when the new surface-PBL model istested in the full 3D MM4, which includes cloud coverparameterizations. The soil moisture conditions werebeginning to dry out after a very wet late June andearly July. The soil at the site is predominantly Dwightsilty clay loam (Kim and Verma 1990), which wasmodeled as silty clay loam in the USDA classificationsystem. According to Stewart and Gay ( 1989) zo was6.5 cm, and according to Kim and Verma ( 1990) andVerma et al. ( 1992) LA1 was 2.8 at this site. The min-imum surface albedo [at Z = 0, a, in Eq. ( 16)] wasset to 0.2, which is a value generally used for grassland.Surface parameters are summarized in Table 1.The model used the same vertical grid as used forthe Wangara simulation except that it was extended to35 levels, up to almost 4 km above the ground. Theprofiles of wind, temperature, and humidity were ini-tialized according to the 1200 UTC radiosonde sound-ing (0700 CDT). The surface soil temperature was ini-tialized to the 1200 UTC measurement of skin tem-perature (22C). The deep-layer soil temperature TZ,which represents the average temperature in the top 1m of soil, was estimated to be 24C from the 10-cmmeasurement of 24.5"C and the 50-cm temperature of23C. Soil moisture in the upper layer was estimatedto be 27% based on the average of gravimetric mea-surements made at 10 sampling locations around thesite at a depth of 25 mm. The deep soil moisture (1-m average) was more difficult to determine since thedaily gravemetric measurements went down only to75 mm and the neutron probe measurements, whichwere made at many depths down to 2 m, were infre-quent such that the nearest measurements were from8 July. The data on that day showed no significantgradient between 1 m and 20 cm with soil moisturemeasurements mostly from 26% to 28%. Therefore,the lower-layer soil moisture was estimated from thegravemetric measurements at 75 mm averaged overthe 10 locations, which was 25.5%.The geostrophic winds were estimated from sevenradiosonde profiles during the day and were interpo-lated for intermediate hours. A very simple schemewas used in which only two values were input to themodel. The observed winds near the top of the PBLwere used to represent the geostrophic winds through-out the PBL where they are assumed to be constantwith height. Likewise, the observed winds at about 2500m were used to represent the geostrophic winds at andabove this level. Between the top of the PBL and 2500m, the geostrophic winds were linearly interpolated.Clearly, this scheme results in a far more crude esti-mation of the geostrophic wind profile than was usedin the Wangara simulations, where a high-density spa-tial network of radiosondes was available.Figure 7 shows surface fluxes measured at site 16(eddy correlation), site 18 (Bowen ratio techniques),and model-simulated fluxes. The measurements fromsites 16 and 18, which are collocated, differ only intheir method of deriving the latent and sensible heatfluxes. The Bowen ratio energy balance method (BR)relies on measurements of net radiation flux and soilheat flux to define the available energy (R, - G), whichis then partitioned into sensible and latent heat fluxesaccording to the Bowen ratio. The Bowen ratio wascalculated from temperature and specific humiditymeasurements made at two heights within the surfacelayer. Eddy correlation measurements (EC) were madeusing fast response instruments (sonic anemometers,fine wire thermocouples, and fast response optical hy-grometers) to derive eddy flux covariances of heat andmoisture. Each method has its pros and cons such asthe assumptions of flux-profile proportionality (first-order closure of the equations of motion) and similarity(ICEI = ICE) in the BR method. On the other hand, BRuses relatively inexpensive and simple instrumentation.Whereas the EC method is a more direct measurementof turbulent fluxes, it is more susceptible to problemsin its fast response instrumentation and interferencefrom the tower and associated structures. However, therelative agreement between the two methods supportstheir accuracy. Ground fluxes were derived from theaverage of two heat flux plates buried at a depth of 5cm and a calculation of the soil heat storage based onthe measured soil temperature for the top 5-cm soillayer. See Kanemasu et al. ( 1992) and Smith et al.26 JOURNAL OF APPLIED METEOROLOGY VOLUME 34Observed Fluxes - July 11, 1987hNE5Y-1007 9 11 13 15 17 19Time (LT)Model Fluxes - July 11, 19877OOL. ,. I, I I I I I 1 I ., . . ,,E\zXY--100 t ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ,-l7 9 11 13 15 17 19Time (LT)FIG. 7. Observed and modeled surface fluxes of net radiation (Rn),sensible heat (H), latent heat (LH), and ground heat (C) for 1 1 July1987 of FIFE. H-18 and LH-18 were measured by Bowen ratiomethod, whereas H- 16 and LH-16 were measured by eddy correlation.( 1992a) for detailed descriptions of the flux measure-ments.The most obvious difference between the model andthe measurements is that the model computed highernet radiation flux. In the late afternoon substantialcloud cover was observed, according to the FIFE cloudcamera data, which obstructed much of the incomingsolar radiation. However, before about 1600 LT thesky was essentially clear. Therefore, the overestimationof the peak net radiation, by about 8.5%, is due toeither the underestimation of surface albedo or insuf-ficient accounting for aerosol effects. Measurements attwo sites in the FIFE study area (unfortunately not thesite we modeled) showed minimum albedos of 0.15and 0.18, which are less than the model minimum al-bedo of 0.20. Therefore, the overestimation of net ra-diation by the model may be due to underestimationof the effects of aerosols in the model, particularly atthe high relative humidities observed in the PBL (87%at 1400). While the accuracy of net radiation mea-surements is generally estimated to be about 5%, in-tercomparisons reported by Smith et al. ( 1992a) showthat midday differences between instruments were lessthan 10 W mP2.The peak sensible and latent heat fluxes are alsooverestimated compared to the observed fluxes mea-sured by both methods. The modeled Bowen ratio,however, which is the ratio of sensible heat flux to latentheat flux, is quite similar between the model and themeasurements. For example, at the peak of the netradiation curve, the modeled Bowen ratio was 0.34while the measured Bowen ratio was 0.33 by the Bowenratio method and 0.30 by the eddy correlation method.The peak modeled soil flux of 56 W m-2 comparedvery well to an observed peak of 59 W m-2, althoughthe model was higher than the measurements duringthe most of the morning. This overestimation of soilheat flux in the morning reflects a disparity in the initialturbulent fluxes, particularly sensible heat flux, betweenthe model and the observations. This results from thecombination of measurement sources used to initializethe model at 1200 UTC. Specifically, surface temper-ature was measured at site 18 while air temperatureswere measured by a radiosonde that was released at1143 UTC near the northern edge of the FIFE studyarea. Therefore, the initial air temperature in the lowestmodel level is not entirely consistent, either temporallyor spatially, with the surface temperature. Since sensibleheat flux in the model is derived from the temperaturedifference between the surface and the lowest modellevel air temperature, it is not surprising that the initialmodel computed sensible heat flux differs from thesensible heat flux measured at the site.Figure 8 shows observed and modeled surface andsoil temperatures. The most relevant comparison isbetween the observed and modeled surface tempera-tures T,. The model results compare very well to themeasurements as they both increase nearly linearlyuntil about 1 100 when the observed temperature startsto increase a little faster while the modeled temperaturestarts to curve downward. As a result, the modeledpeak is about 1 "C cooler than the measured peak (30Cvs 3 1 "C) . The rapid decrease in the measured surfacetemperature at about 1630 reflects the increasingcloudiness. The T2 curve is the modeled deeper soiltemperature that can be considered as an average ofthe top 1 m of soil. Measured ground temperatures at10 cm ( TgI) and 50 cm ( Tg2) are shown for compar-JANUARY 1995 PLEIM AND XIU~27Ground TemperaturesTs obs32 ,',,I "1 , , , , , , ,\ _. .--- -- -.22""""'"''""""''7 9 11 13 15 17 19Time (LT)FIG. 8. Measured ground surface temperature (T, obs), soil tem-perature at 10 cm ( Tg1 obs), and 50 cm ( TC2 obs), and modeledground surface temperature ( T, model ) and average soil temperaturein top I m (Tz model) for 11 July 1987 of FIFE.ison. The T2 curve fits nicely between the two observedcurves.One of the main reasons for developing this modelis to improve the PBL height simulations in MM4,which are so crucial to air quality modeling. Figure 9shows the modeled PBL heights compared to estimatesof the PBL height derived from six radiosondes. Boththe modeled and observed PBL heights were estimatedby the bulk Richardson number method [Eq. (6)].The model was initialized with the 1200 UTC sound-ing. The model compares quite well at the PBL max-imum in the afternoon. However, during the time ofmost rapid PBL development, the model lags the ob-servations by almost 2 h. A possible reason for thisdiscrepancy is the model's inability to sufficiently ac-count for both shear induced turbulence and buoyancyinduced turbulence at the same time. As describedabove, the PBL model is a hybrid of a PBL scalingeddy diffusivity model and a nonlocal closure-freeconvective model ( ACM) . Usually during clear sum-mer days, buoyancy dominates PBL turbulence andthe ACM performs quite well. However, this case wasvery windy (about 15-20 m s-' in the PBL), such thatshear turbulence was also very important. In fact, h/L was about -2 for much of the day, which indicatesthat buoyancy and shear were equally important at z= h/2. Therefore, neither the ACM, which includesno shear effects, nor the eddy diffusion model, whichcannot adequately simulate convective transport, aresufficient for this case. This suggests the need for moregeneralized transilient models of the type developedby Stull and Driedonks (1987). The challenge is todevelop a model that is sufficiently general and simpleenough to run quickly in 3D Eulerian grid models.2) 6 JUNE 1987The second case study from the FIFE was 6 June1987, which was the last day of IFC 1 and also a "goldenday." The sky was completely clear of clouds for thewhole day and the winds were more moderate thanthe previous case at about 10-12 m sC1 from the southin the PBL. Since it was earlier in the growing season,the LA1 was 1.9 (Verma et al. 1992) and zo was about4.5 cm (Sellers et al. 1992b). Soil moisture conditionswere again rather moderate after three days of clearweather during which time the soil was drying fromnear-saturated conditions the previous week.The model profiles of winds, temperature, and hu-midity were initialized to the 1200 UTC (0700 CDT)sounding. The surface and lower-layer soil tempera-tures were initialized to the 1200 UTC measurements,which were 16.0"C and 20.2"C, respectively. The initialsoil moisture for the top 1 cm was set to 23%, whichis just slightly less than the 10 sample average of grav-emetric measurements at a depth of 25 mm, which was24.3%. Since this day was in the midst of a drying trend,the shallower I-cm model layer should be a little dryerthan the deeper measurements. The initial moisture inthe 1-m model soil layer was again more difficult toestimate. The average of the gravemetric measurementsat 75 mm was 26% and the neutron probe data from3 June show a slight decreasing gradient with depth inthe top 1-m layer. Therefore, 25% was used as the initialsoil moisture for the deeper ( 1 m) model layer.Figure 10 shows surface fluxes measured at site 18(BR) and computed by the model. Eddy correlationPBL Height - July 11, 19871200 -t4002000--L"l"'l'" """'1"'7 9 11 13 15 17 19 Time (LT)FIG. 9. Observed and modeled PBL heights for 1 I July 1987 of FIFE.28 JOURNAL OF APPLIED METEOROLOGY VOLUME 34Observed Fluxes - June 6, 1987, Site 18700,. I. I, .. ,. . , ,. I I. I, I7 9 11 13 15 17 19 Time (LT)Model Fluxes - June 6, 1987700, I I I I I I I I ,. . I I.. I I I I I-1 007 9 11 13 15 17 19Time (LT)FIG. 10. Observed and modeled surface fluxes of net radiation(&), sensible heat (H), latefit heat (LH), and ground heat (C) for6 June 1987 of FIFE.measurements were not available for this day. Themodel again overestimates the maximum net radiationat the surface compared to the measurements but bya lesser degree than for the 11 July case. The reasonfor the better radiation simulation (about 5% differenceat the peak) for this case may be the clearer skies andlower humidity in the PBL. The peak values of thelatent heat and sensible heat fluxes are also overesti-mated by the model by a somewhat greater degree thanthe net radiation. The greater amount of energy avail-able for the turbulent surface fluxes results from acombination of the overprediction of net radiation andunderprediction of ground heat flux. The peak modeledground heat flux was 82 W m-', whereas the peakmeasured ground heat flux was 1 10 W m-'. In theearly afternoon ( 1300-1400) the underprediction waseven greater since the modeled ground heat flux peakedearlier in the day (around 1100) than the measuredground heat flux. The Bowen ratio of the model resultsat peak net radiation is slightly higher (0.37 ) than theobservations (0.32). The model, however, is very sen-sitive to the initial specification of the deep-layer soilmoisture, which is not well known. For example,changing the initialization from 25% to 26% changesthe Bowen ratio from 0.37 to 0.30.The kinks in the modeled sensible heat flux andground heat flux curves (Fig. 10 ) at 1000 and again atabout 1700 are caused by the transition from the eddydiffusion model to the convective model ( ACM) in themorning and then back to the eddy diffusion model inthe afternoon. These transitions cause slight oscillationsin the surface temperatures and heat fluxes that veryquickly damp out.Time series of modeled and measured soil temper-atures (shown in Fig. 1 1 ) indicate that the model sig-nificantly underestimates surface temperaturethroughout the middle of the day. For the first twohours of the simulation (0700-0900), the model andthe measurements compare quite well and again forthe last two hours ( 1700- 1900). However, in betweenthe measurements peak almost 4.5 "C higher than themodel. Underprediction of surface temperature couldresult from too large surface heat capacity in the model.Recall that our modification of the Noilhan and Plan-ton ( 1989) model effectively increased the bulk surfaceheat capacity to be more like the value for soil. There--uvzenc5kGround Temperatures35 k15~"""'"'''"''""'~7 9 11 13 15 17 19Time (LT)FIG. 1 1. Measured ground surface temperature (T, obs), soil tem-perature at IO cm ( Tgl obs), and 50 cm (T,, obs), and modeledground surface temperature ( T, model) and average soil temperaturein top 1 m (T, model) for 6 June 1987 of FIFE.JANUARY 1995 PLEIM AND XIU 29400035003000E 2500$ 2000v-1WCIcalcn._15001000500018 GMTI I - I,r,,VI- - -Thetav-Model300 305 310 31 5 320Virtual Potential Temperature (K)FIG. 12. Modeled and measured virtual potential temperature profiles at 1800 UTC (1300 LT) 6 June 1987.fore, we tested the effect of this modification by rerun-ning the model with a much smaller heat capacity (byabout two orders of magnitude). The result was anincrease in peak surface temperature of less than 1 "Cand a decrease of the soil heat flux to a peak of about1 W m-2. Clearly, the discrepancy in the surface tem-perature was not due to heat capacity.The underprediction of soil flux is consistent withthe underprediction of surface temperature because ofthe smaller temperature gradient between the surfaceand deeper soil. Also, the underprediction of soil fluxresults in the underprediction of surface temperaturesince less energy in available for heating the soil. Theunderprediction of surface temperature, however, isnot consistent with the overprediction of sensible heatflux. In the absence of unaccounted for heat sourcesin the PBL the model's lower surface temperatureshould lead to a lesser temperature gradient betweenthe surface and the air and hence smaller sensible heatflux. Therefore, the overprediction of sensible heat fluxsuggests that the air temperature was also underpre-dicted. Comparison of virtual potential temperatureprofiles at 1300 (Fig. 12) show that the modeled airtemperature in the PBL is indeed less than observed.The only conclusion that fits these facts is that the airin the PBL was heated by some process missing fromthe model formulation, most probably warm advectionon the predominantly southerly winds. Warm advec-tion in the PBL is not simulated by the 1D model,which therefore underestimates the temperature andheight of the PBL and therefore also the surface tem-perature and soil flux.The comparison of observed and modeled PBLheights (Fig. 13) also indicates the possibility of warmadvection since the model underestimates the observedPBL heights at all times after the initialization. Thevirtual potential temperature profiles at 1800 UTC( 1300 LT), shown in Fig. 12, demonstrate that theshallower mixed layer in the model is associated withcooler PBL temperatures (by about 1 "C). The simi-larity between the observed and predicted profilesabove the PBL indicates that significant thermal ad-vection was not occurring at these levels.c. Sensitivity to soil moistureThe sensitivity of PBL height and surface tempera-ture, which are critically important parameters for airquality simulation, to variations in soil moisture istested. The 6 June 1987 FIFE case is used to illustratehow the simulation would change when only the initialsoil moisture is varied from wilting point to field ca-pacity. Figures 14 and 15 show the PBL heights andsurface temperatures resulting from three simulations:1 ) the FIFE simulation as described above with initialwg = 23% and w2 = 25%; 2) the initial soil moistureset to field capacity ( wg = w2 = 32.2%); and 3) theinitial soil moisture set to wilting point (w, = w2= 2 1.8%). Note that the entire range from wilting pointto field capacity is only about 10% for silty clay loam.Furthermore, the difference between the FIFE initial-ization and the wilting point is only about 3%, whichis less than the typical standard deviation of soil mois-ture (us) for field sizes on the order of 0.1 km2 ( Wetzeland Chang 1988). This small change in soil moisture,PBL Height - June 6, 19871600 L~ " 1 '' ' 1 " ' 1 " ' 1 " 1 " '4140012001000aoo600400t*01,, 0 , , , , , , , , , ,I7 9 11 13 15 17 19 Time (LT)FIG. 13. Observed and modeled PBL heights for 6 June 1987 of FIFE.JOURNAL OF APPLIED METEOROLOGYVOLUME 34PBL Heightt i/-cWilting PointField Capt'I 1*f/-/1500. tI1000 1teI f I///--06 8 10 12 14 16 18 20Time (LT)FIG. 14. Sensitivity of modeled PBL heights to initial soil moisture for 6 June 1987 of FIFE.however, results in a huge change in PBL height, froma maximum for the FIFE initialization of 1270 to 2 100m for the wilting point run (Fig. 14). Similarly, Fig.15 shows that the maximum surface temperature forthe wilting point simulation is 8.5"C higher than forthe FIFE initialization. The differences between theFIFE initialization and the field capacity run are muchsmaller even though the change in initial soil moisturewas greater. This exercise demonstrates that a modelthat is insensitive to soil moisture conditions is proneto significant errors in PBL and surface temperaturesimulations.The dramatic effect of changing the initial soil mois-ture to the wilting point is due to the complete cessationof evapotranspiration. The model parameterizes sto-matal conductance as a linear function of root zonesoil moisture w2 between wilting point and field ca-pacity. This effect is so extreme because of the almostcomplete coverage of the ground by vegetation in theFIFE study area. Note that the model drastically sim-plifies the situation, particularly by neglecting the highdegree of spatial variation inherent in soil moisture. Inreality, there is always a distribution of soil moistureeven over very small spatial scales. For example, Wetzeland Chang ( 1988) estimate fl8 - 7% for grid sizes onthe order of 100 km2. Therefore, when the average soilmoisture is near wilting point some parts of the distri-bution dip below wilting point, whereas other partsremain above wilting point enabling some evapotrans-piration to continue. Consequently, spatial averagingof soil moisture, as in Eulerian grid models, can leadto significant errors, particularly near wilting point orfield capacity. Wetzel and Chang ( 1988) have devel-oped a scheme that uses subgrid distributions of soilmoisture thus allowing evapotranspiration to occur atdifferent degrees of water stress in the same grid cell.We plan to include a similar scheme for subgrid het-erogeneity as part of our MM4 modifications.4. Summary and conclusionsInitial tests of the model in 1D form are comparedto field studies from Wangara in Australia (dry, sparsevegetation) and FIFE in northeast Kansas (moist tallgrass prairie). Preliminary results from the simulationof days 33 and 34 of the Wangara experiment showvery good agreement with observations of surfacefluxes, PBL height, and profiles of temperature andwinds. Whereas the Wangara simulation provided agood test case for the PBL model and the flux profilealgorithms, it did little to test the soil moisture modeland the partitioning of sensible and latent heat fluxessince the site was so add. Therefore, two cases fromthe FIFE are also studied so that the vegetation andsoil moisture components of the model could be tested.The FIFE simulations show that the model tends tooverestimate net radiation when humidity is high, sug-gesting that aerosol scattering may be underpredicted.The model also overestimated sensible and latent heatfluxes by a small amount in both cases. The groundflux predictions were quite close to the measurementsin the 11 July case but underpredicted in the 6 Junecase. Surface temperature was slightly underpredictedin the 11 July case and greatly underpredicted in the6 June case, which was probably due to warm advec-tion.The most positive result of the FIFE simulationswas the good agreement between the modeled and ob-Surface Temperature40Wilting Point35el 30al t/t/f15- " 'I " " ' " '7 9 11 13 15 17 19 Time (LT)FIG. 15. Sensitivity of modeled surface temperature to initial soil moisture for 6 June 1987 of FIFE.JANUARY 1995 PLElM AND XIU 31served Bowen ratios. This shows that the model is ca-pable of realistic simulation of moisture fluxes pri-marily by transpiration through vegetation. Also, themaximum PBL height simulation for 1 1 July was quiteclose to the observation even though the rise was tooslow. The slow rise of the PBL could be an indicationthat the PBL model cannot adequately simulate PBLmixing and growth on very windy but clear days whenboth buoyancy and shear-induced turbulence are im-portant. In addition, part of the difficulty in trying tosimulate surface fluxes and PBL processes in a coupledfashion may be that the surface fluxes are respondingto turbulence induced by vegetative roughness, whilethe PBL turbulence may be more associated with ter-rain.In general, the simulation of 1 1 July was better thanthe 6 June case. The main reason for this is probablythat horizontal warm advection was occurring in thePBL during the middle of the day on 6 June. We areplanning to revisit both of these cases with the 3D MM4modified to include this new surface-PBL model. Withthe 3D model we should be able to see if warm advec-tion was indeed an important factor on 6 June.The sensitivity tests of PBL height and surface tem-perature to soil moisture demonstrates the importanceof the realistic simulation and initialization of soilmoisture in meteorological models. These tests alsosuggest the need for representation of soil moisturesubgrid heterogeneity in Eulerian grid models. Usinggrid cell averages of soil moisture tends to exaggeratethe effects of extreme conditions, near wilting point orfield capacity. Therefore, some scheme for the repre-sentation of subgrid heterogeneity will be applied tothe MM4 modifications.The FIFE database contains an enormous wealth ofinformation that will be very valuable for continuingthe development of surface-PBL modeling techniquesfor mesoscale modeling. For example, the extensiveradiation measurements will help construct and im-prove radiation algorithms for both clear and cloudyskies. Also, the variety of flux measurements from thesurface sites and aircraft may help resolve some of theissues related to the scale of roughness appropriate forgrid cell average calculations as well as the influenceof subgrid heterogeneity.The next phase of this work, which has already be-gun, is to incorporate the surface-PBL model discussedhere into the MM4-MM5 system. The modified MM4results will be extensively compared to the currentmodel values and also to field study data. The key fea-tures of this model, which represent a significant ad-vance over the current model in the MM4-MM5 sys-tem, are the prognostic simulation of soil moisture andthe parameterization of vegetative evapotranspiration.This allows the model to be more responsive to chang-ing moisture conditions due to precipitation and evap-oration, thereby enabling a better simulation of thepartitioning of surface energy into latent and sensibleheat flux components. As a result, the simulation ofthe evolution of PBL heights, which are extremely im-portant to air pollutant concentration simulations, andsurface temperatures, which are key factors determin-ing biogenic emissions, should be significantly im-proved.The incorporation of this model into the MM4-MM5 system requires additional data inputs as well asmodifications to the initialization and 4D data assim-ilation techniques. The additional data includes soiltexture type and some additional land use parameterssuch as leaf area index and fractional vegetation cov-erage. High resolution ( - 1 km) datasets for both soiland land use parameters are currently under devel-opment and should be available for this work.Disclaimer. The information in this document hasbeen funded wholly or in part by the United StatesEnvironmental Protection Agency. It has been sub-jected to agency review and approved for publica-tion. Mention of trade names or commercial productsdoes not constitute endorsement or recommendationfor use.Acknowledgements. The authors appreciate the workaccomplished by the entire FIFE team and particularlythe FIFE Information System (FIS) staff, headed byDon Strebel, for making this remarkable dataset soeasily accessible.REFERENCESAndrt, J. C., J. P. Goutorbe, and A. Perrier, 1986: HAPEX-MOB- ILHY: A hydrologic atmospheric experiment for the study of water budget and evaporation flux at the climatic scale. Bull. Amer. Meteor. Soc., 67, 138-144.Anthes, R. A,, E.-Y. Hsie, and Y.-H. Kuo, 1987: Description of the Penn State/NCAR Mesoscale Model Version 4 (MM4). NCAR Tech. Note, NCAR/TN-282+STR, 66 pp.Argentini, S., P. J. Wetzel, and V. M. Karyampudi, 1992: Testing a detailed biophysical parameterization for land-air exchange in a high-resolution boundary-layer model. J. Appl. Meteor., 31,Avissar, R., and R. Pielke, 1989: A parameterization of heterogeneous land surfaces for atmospheric numerical models and its impact on regional meteorology. Mon. Wea. Rev., 117, 2 1 13-2 136.Binkowski, F. S., 1983: A simple model for the diurnal variation of the mixing depth and transport flow. Bound.-Layer Meteor., 27,Blackadar, A. K., 1976: Modeling the nocturnal boundary layer. Third Symp. on Atmospheric Turbulence, Diffusion and Air Quality, Raleigh, NC, Amer. Meteor. SOC., 46-49.- , 1978: Modeling pollutant transfer during daytime convection, Preprints, Fourth Symp. on Atmospheric Turbulence, Diffusion, and Air Quality, Reno, NV, Amer. Meteor. SOC., 443-447.Bougeault, P., J. Noilhan, P. Lacarkre, and P. Mascart, 1991a: An experiment with an advanced surface parameterization in a me- sobeta-scale model. Part I: Implementation. Mon. Wea. Rev.,-, B. Bret, P. Lacarrere, and J. Noilhan, 1991b: An experiment with an advanced surface parameterization in a mesobeta-scale model. Part 11: The 16 June 1986 simulation. Mon. Wea. Rev.,142-156.2 17-236.119,2358-2373.119,2374-2392.32 JOURNAL OF APPLIED METEOROLOGY VOLUME 34Byun, D. W., 1990: On the analytical solutions of flux-profile rela- tionships for the atmospheric surface layer. J. Appl. Meteor.,- , 1991: Determination of similarity functions of the resistance laws for the planetary boundary layer using surface-layer simi- larity functions. Bound.-Layer Meteor., 57, 17-48.Carlson, T. N., and F. E. Boland, 1978: Analysis of urban-rural canopy using surface heat flux/temperature model. J. Appl. Meteor., 17, 998-1013.Chang, J. S., R. A. Brost, 1. S. A. Isaksen, S. Madronich, P. Middleton, W. R. Stockwell, and C. J. Walcek, 1987: A three-dimensional Eulerian acid deposition model: Physical concepts and formu- lation. J. Geophys. Res., 92, 14 681-14 700.Clapp, R. B., and G. M. Hornberger, 1978: Empirical equations for some soil hydraulic properties. Water Resour. Res., 14, 601- 604.Clarke, R. H., A. J. Dyer, R. R. Brook, D. G. Reid, and A. J. Troup, 197 1 : The Wangara experiment: Boundary layer data. Tech. Paper No. 19, CSIRO, Division of Meteorological Physics, As- pendale, Australia, 362 pp.Deardorff, J., 1974: Three-dimensional numerical study ofthe height and mean structure of a heated planetary boundary layer. Bound.-Layer Meteor., 7, 81-106.- , 1978: Efficient prediction of ground surface temperature and moisture, with inclusion of a layer of vegetation. J. Geophys. Res., 83, 1889-1903.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,Edson, R. T., 1980: Parameterization of net radiation at the surface using data from the Wangara experiment. Environmental Re- search Papers, Colorado State University, Fort Collins, CO, 26 PP.Hicks, B. B., 198 1: An analysis of Wangara micrometeorology: Surface stress, sensible heat, evaporation, and dewfall. NOAA Tech. Memo. ERL-104, 36 pp.Holtslag, A. A. A,, and F. T. M. Nieuwstadt, 1986: Scaling the at- mospheric boundary layer. Bound.-Layer Meteorol., 36, 201- 209.- , E. 1. F. de Bruijn, and H.-L. Pan, 1990: A high resolution air mass transformation model for short-range weather forecasting. Mon. Wea. Rev.. 118, 1561-1575.Idso, S., R. Jackson, B. IOmball, and F. Nakayama, 1975: The de- pendence of bare soil albedo on soil water content. J. Appl. Meteor., 14, 109-1 13.Jacquemin, B., and J. Noilhan, 1990: Sensitivity study and validation of a land surface parameterization using the HAPEX-MOB- ILHY data set. Bound.-Layer Meteor., 52, 93-134.Kanemasu, E. T., S. B. Verma, E. A. Smith, L. J. Fritschen, M. Wesely, R. T. Field, W. P. Kustas, H. Weaver, J. B. Stewart, R. Gurney, G. Panin, and J. B. Moncrieff, 1992: Surface flux mea- surements in FIFE: An overview. J. Geophys. Res., 97, 18 547- 18 555.hm, J., and S. B. Verma, 1990: Components of surface energy bal- ance in a temperate grassland ecosystem. Bound.-Layer Meteor., 51,401-417.McCumber, M. C., and R. A. Pielke, 1981: Simulation ofthe effects of surface fluxes of heat and moisture in a mesoscale numerical model. Part I: Soil layer. J. Geophys. Res., 86, 9929-9938.Mihailovic, D. T., H. A. R. de Bruin, M. Jeftic, and A. van Dijken, 1992: A study of the sensitivity of land surface parameterizations to the inclusion of different fractional covers and soil textures. J. Appl. Meteor., 31, 1477-1487.Monteith, J. L., 196 1: An empirical method for estimating long wave radiation exchanges in the British Isles. Quarl. J. Roy. Meteor. Soc., 87, 171-179.Noilhan, J., and S. Planton, 1989: A simple parameterization of land surface processes for meteorological models. Mon. Wea. Rev.,29,652-657.NCAR/TN-275+STR, 69 pp.117,536-549.-, P. Lacarrere, and P. Bougeault, I99 1: An experiment with an advanced surface parameterization in a mesobeta-scale model. Part 111: Comparison with the HAPEX-MOBILHY dataset. Mon. Wea. Rev., 119, 2393-2413.Paltridge, G. W., and C. M. R. Platt, 1976: Radiative Processes in Meteorology and Climatology. Elsevier, 57-63.Pielke, R. A,, G. A. Dah, J. S. Snook, T. J. Lee, and T. G. F. Kittel, 199 1 : Nonlinear influence of mesoscale land use on weather and climate. J. Climate, 4, 1053-1069.Pleim, J. E., and J. S. Chang, 1992: A non-local closure model for vertical mixing in the convective boundary layer. Atmos. En- viron., 26A, 965-98 1.-, and J. K. S. Ching, 1993: Interpretive analysis of observed and modeled mesoscale ozone photochemistry in areas with nu- merous point sources. Atmos. Environ., 27A, 999-101 7.Sellers, P. J., Y. Mintz, Y. C. Sud, and A. Dalcher, 1986: A simple biosphere model (SiB) for use within general circulation models. J. Atmos. Sci., 43, 505-53 1.-, F. G. Hall, G. Asrar, D. E. Strebel, and R. E. Murphy, 1988: The First ISLSCP Field Experiment (FIFE). Bull. Amer. Meteor. Soc., 69, 22-27. , , , , and -, 1992a: An overview of the First International Satellite Land Surface Climatology Project (ISLSCP) Field Experiment (FIFE). J. Geophys. Res., 97,-, M. D. Heiser, and F. G. Hall, 1992b: Relations between surface conductance and spectral vegetation indices at intermediate ( 100 m2 to 15 km2) length scales. J. Geophys. Res., 97, 19 033-19 059.Smith, E. A,, W. L. Crosson, and B. D. Tanner, 1992a: Estimation of surface heat and moisture fluxes over a prairie grassland. 1: In situ energy budget measurements incorporating a cooled mirror dew point hygrometer. J. Geophjx Res., 97, 18 557- 18 582.-, A. Y. Hsu, W. L. Crosson, R. T. Field, L. J. Fritschen, R. J. Gurney, E. T. Kanemasu, W. P. Kustas, D. Nie, W. J. Shuttle- worth, J. B. Stewart, S. B. Verma, H. L. Weaver, and M. L. Wesely, 1992b: Area-averaged surface fluxes and their time- space variability over the FIFE experimental domain. J. Geo- phys. Res., 97, 18 599-18 622.Stewart, J. B., and L. W. Gay, 1989: Preliminary modeling of tran- spiration from the FIFE site in Kansas. Agric. For. Meteor., 48,Stull, R. B., and A. G. M. Driedonks, 1987: Applications of the tran- silient turbulence parameterization to atmospheric boundary- layer simulations. Bound.-Layer Meteor., 40, 209-239.Troen, I., and L. Mahrt, 1986: A simple model of the atmospheric boundary layer; sensitivity to surface evaporation. Bound.-Layer Meteor., 37, 129-148.Verma, S. B., J. Kim, and R. J. Clement, 1992: Momentum, water vapor, and carbon dioxide exchange at a centrally located prairie site during FIFE. J. Geophys. Res., 97, 18 629-18 639.Wetzel, P. J., and J.-T. Chang, 1987: Concerning the relationship between evapotranspiration and soil moisture. J. Climate Appl. Meteor., 26, 18-27.-, and -, 1988: Evapotranspiration from nonuniform surfaces: A first approach for short-term numerical weather prediction. Mon. Wea. Rev., 116,600-621.-, and S. Argentini, 1990: The sensitivity of daytime low cloud amount to vegetation cover, soil moisture and predawn sound- ing-Two case studies. Preprints, Eighth Conf: on Hydrome- teorology, Kananaskis Park, Canada, Amer. Meteor. SOC., 12- 17.Wilson, M. F., A. Henderson-Sellers, R. E. Dickinson, and P. J. Ken- nedy, 1987: Sensitivity of the Biosphere-Atmospheric Transfer Scheme (BATS) to the inclusion of variable soil characteristics. J. Climate Appl. Meteor., 26, 341-362.Wyngaard, J. C., and 0. R. CotC, 1974: The evolution ofa convective boundary layer-A higher-order-closure model study. Bound.- Layer Meteor., 7, 289-308.Yamada, T., and G. Mellor, 1975: A simulation of the Wangara atmospheric boundary layer data. J. Atmos. Sci., 32,2309-2329.18 345-18 371.305-3 15.