A network of meteorological stations was installed during the Monsoon '90 field campaign in the Walnut Gulch experimental watershed. The study area has a fairly complex surface. The vegetation cover is heterogeneous and sparse, and the terrain is mildly hilly, but dissected by ephemeral channels. Besides measurements of some of the standard weather data such as wind speed, air temperature, and solar radiation, these sites also contained instruments for estimating the local surface energy balance. The approach utilized measurements of net radiation (Rn), soil heat flux (G) and Monin-Obukhov similarity theory applied to first- and second-order turbulent statistics of wind speed and temperature for determining the sensible heat flux (H). The latent heat flux (LE) was solved as a residual in the surface energy balance equation, namely, LE=-(Rn+G+H). This procedure (VAR-RESID) for estimating the energy fluxes satisfied monetary constraints and the requirement for low maintenance and continued operation through the harsh environmental conditions experienced in semiarid regions. Comparison of energy fluxes using this approach with more traditional eddy correlation techniques showed differences were within 20% under unstable conditions. Similar variability in flux estimates over the study area was present in the eddy correlation data. Hence, estimates of H and LE using the VAR-RESID approach under unstable conditions were considered satisfactory. Also, with second-order statistics of vertical velocity collected at several sites, the local momentum roughness length was estimated. This is an important parameter used in modeling the turbulent transfer of momentum and sensible heat fluxes across the surface-atmosphere interface. ¿ American Geophysical Union 1994