5.6 EFFECTS OF SOIL MOISTURE HETEROGENEITY ON BOUNDARYLAYER FLOW WITH COUPLED GROUNDWATER, LAND-SURFACE,AND MESOSCALE ATMOSPHERIC MODELINGFotini Katopodes Chow1∗, Stefan J. Kollet2, Reed M. Maxwell2, and Qingyun Duan21Department of Civil and Environmental Engineering, University of California, Berkeley, California2Atmospheric, Earth, and Energy Department, Lawrence Livermore National Laboratory, Livermore, California1. Introduction and backgroundMesoscale atmospheric models currently rely on an in-tegrated land-surface model to provide fluxes of heat,momentum, and moisture from the land surface to theatmosphere. While improvements have been made bytuning land-surface models for a variety of test cases,current models are limited to vertical transport in a shal-low soil column. They are thus unable to capture lat-eral transport of soil moisture and limited in their abil-ity to provide spatial variability in predicted land sur-face fluxes. Current mesoscale atmospheric models aretherefore not provided with realistic fluxes at the surfacebecause land-use models cannot represent runoff andsubsurface lateral transport that is present when ter-rain or moisture gradients exist. This can lead to errorsin model predictions during periods when thermal forc-ing dominates the diurnal development of the boundarylayer.Starting with a so-called leaky-bucket parameteriza-tion (Manabe et al. 1965), climate models have steadilyevolved to use a more sophisticated lower boundarycondition into what is commonly known as the landsurface model (LSM). LSMs play an important role indetermining fluxes from the land surface to the atmo-spheric boundary layer (see e.g. the review by Bettset al. 1996). A large number of LSMs have beendeveloped, with differing parameterizations and lev-els of sophistication. This led to a number of inter-comparison studies, the Project for Intercomparison ofLand-Surface Parameterization Schemes (PILPS), fora range of climatic conditions (Henderson-Sellers andHenderson-Sellers 1995; Shao and Henderson-Sellers1996; Chen and Coauthors 1997; Qu and Coauthors1998; Lohmann and Coathors 1998; Pitman and Coau-thors 1999; Schlosser and Coathors 2000; Luo et al.2003). As LSMs ignore the deeper soil moisture pro-cesses and the saturated zone (i.e. groundwater), therehas been recent interest in incorporating a groundwatercomponent into LSMs (Liang et al. 2003; Maxwell andMiller 2005; Yeh and Eltahir 2005). While LSMs havegrown in sophistication, until this current study, lateralsubsurface and overland flow have not been explicitlyaccounted for.∗Corresponding author address: Department of Civil andEnvironmental Engineering, University of California, Berke-ley, MC 1710, Berkeley, CA 94720-1710, email: chow @ce.berkeley.eduThe focus of this work is to understand the influ-ence of soil moisture variability on atmospheric bound-ary layer forcing. To gain this understanding requiresthe development of a three-dimensional, fully-coupledgroundwater-atmospheric flow model, as described inthis work. That soil moisture and temperature vari-ability affects atmospheric boundary layer developmenthas been shown through previous studies. Chow et al.(2006) found that soil moisture initialization was a cru-cial factor in accurate simulations of thermally-forcedvalley wind systems in the Swiss Alps using the ARPSmesoscale model. An off-line hydrologic model wasused to provide improved spatial variability to the soilmoisture in the valley, significantly improving predic-tion of wind transitions in the valley. Desai et al.(2005) examined the effect of improved soil moisturedata and land-surface physics representations on sim-ulations of the atmospheric boundary layer during July1997 (SGP97), with mixed results in the extent of theinfluence on dry sunny days. York et al. (2002) investi-gated the effects of a single-column atmospheric modelconnected to a single layer ModFlow-based groundwa-ter model through a reservoir-type land surface scheme.They investigated a small watershed in Kansas andfound an effect of aquifer levels on evapotranspiration.Despite previous work in this area, determining theeffect of land-surface heterogeneity on the developmentof the atmospheric boundary layer remains unresolved.Is the effect of land-surface heterogeneity reflected inatmospheric heterogeneity? On what time scales is theeffect of soil moisture variations felt in the atmosphere?How do land-surface changes affect local precipitationevents? How can we best represent these processes fornumerical simulations of atmospheric flow and transportover a watershed, and eventually over a larger region?To address these questions and others, this paperdescribes the development of a 3D variably-saturatedgroundwater flow model with surface runoff capabil-ities dynamically coupled within a mesoscale atmo-spheric model to investigate the effects of soil mois-ture heterogeneity on boundary layer processes. Inparticular, we have coupled ParFlow, a 3D parallelunsaturated/saturated groundwater flow model (Ashbyand Falgout 1996; Jones and Woodward 2001), withthe Advanced Regional Prediction System (ARPS), amesoscale atmospheric model (Xue et al. 2000, 2001).ParFlow also includes a fully-coupled overland flow orrunoff component (Kollet and Maxwell 2006), and thusprovides ARPS with soil moisture information that in-17th Symposium on Boundary Layers and Turbulence, American Meteorological Society, 2006cludes the effects of ponding, runoff, and seepage. Inturn, ARPS, through its land-surface model, providesParFlow with precipitation and evapotranspiration in-puts. This leads to a fully coupled model which can rep-resent spatial variations in land-surface forcing drivenby 3D atmospheric and subsurface components.Our test case is the Little Washita watershed inOklahoma, which has been the subject of numerousstudies and provides a unique source of subsurface,surface, and atmospheric data for validation. Herewe report on the coupling procedure and preliminarysimulation results for the coupled groundwater/land-surface/atmospheric model.2. Little Washita watershed, SGP99The Little Washita watershed is located in central Ok-lahoma and has been the focus of several studies (e.g.Jackson et al. 1999; Vine et al. 2001; Guha et al. 2003),with the result that it is the source of an extensive ob-servational dataset. We focus on the data from theSouthern Great Plains (SGP) experiment of 1999, fromwhich there are special remote sensing observationsof soil