The importance of soil moisture inputs and improved model physics in the prediction of the daytime boundary-layer structure during the Southern Great Plains Hydrology Experiment 1997 (SGP97) is investigated using the non-hydrostatic fifth-generation Pennsylvania State University/National Center for Atmospheric Research (PSU/NCAR) Mesoscale Model MM5. This is Part II of a two-part study examining the relationship of surface heterogeneity to observed boundary-layer structure. Part I focuses on observations and utilizes a simple model while Part II uses observations and MM5 modelling. Soil moisture inputs tested include a lookup table based on soil type and season, output from an offline land-surface model (LSM) forced by atmospheric observations, and high-resolution (»≈ 800 m) airborne microwave remotely sensed data. Model physics improvements are investigated by comparing an LSM directly coupled with the MM5 to a simpler force-restore method at the surface. The scale of land surface heterogeneities is compared to the scale of their effects on boundary-layer structure.The use of more detailed soil moisture fields allowed the MM5 to better represent the large-scale (hundreds of km) and small-scale (tens of km) horizontal gradients in surface-layer weather and, to a lesser degree, the atmospheric boundary-layer (ABL) height, which was evaluated against observations measured by differential absorption lidar (DIAL). The benefits of coupling an LSM to the MM5 were not readily evident in this summertime case, with the model having particular difficulty simulating the timing of maximum surface fluxes while underestimating the depth of the mixed layer.