Combining Optical and Microwave Remote Sensing for Mapping Energy Fluxes in a Semiarid Watershed

A dual-source model treating the energy balance of the than measured. An empirical model was developed to reduce this bias, but it is not known how generally applicasoil/substrate and vegetation that was developed to use radiometric surface temperature observations is revised to ble it will be. Model sensitivity to typical uncertainties in remotely sensed leaf area index (LAI) and near-surface use remotely sensed near-surface moisture from a passive microwave sensor for estimating the soil surface energy (0–5 cm) water content, W, was quantified. The variation in flux predictions caused by errors in prescribing leaf balance. With remotely sensed images of near-surface soil moisture, land cover classification, and leaf area index, the area index and W was less than 30%. More tests with this model over different landscapes are necessary to model is applied over a semiarid area in the Walnut Gulch Watershed in southern Arizona. The spatial and temporal evaluate its potential for predicting regional fluxes. In particular, microwave and radiometric surface temperavariation of the Bowen ratio (i.e., the ratio of the turbulent fluxes, sensible, and latent heat) “maps” generated by the ture observations are needed under drought conditions for evaluating if the model formulation of vegetation model were similar to the changes in near-surface moisture fields caused by recent precipitation events in the study transpiration can properly adjust to this extreme and very important environmental condition. Published by area. The estimated fluxes at the time of the microwave observations (i.e., “instantaneous” estimates) and those Elsevier Science Inc., 1998 simulated over the daytime period are compared with the ground observations within the watershed. Differences between predicted and observed “instantaneous” fluxes INTRODUCTION were usually comparable to the measurement uncertaintA dual-source model treating the energy balance of the ies, namely, 5% for net radiation and 20–30% for soil, soil/substrate and vegetation that was developed to use sensible and latent heat fluxes, except when there was radiometric surface temperature observations at zenith large temporal and spatial variations in solar radiation view angle u, TR(u), (Norman et al., 1995) has been sucacross the study area. However, by running the model cessfully applied on a pixel by pixel basis using TR(u), over the daytime period, this variability in solar radiavegetation index, and land-use maps in order to compute tion proved to have a minor effect on computed daytime spatially distributed surface energy fluxes over a semiarid totals. In fact, differences with observed heat fluxes were basin (Kustas and Humes, 1996). This approach has adsignificantly less (i.e., around 15%) than when comparing vantages over earlier attempts at mapping surface energy “instantaneous” values. Model predictions of the total soil fluxes with remote sensing that treat the surface as a sinheat flux over the daytime period were generally higher gle source and/or do not account for variation in surface resistances due to variation in vegetation cover and surface roughness. Furthermore, spatially distributed flux * USDA-ARS Hydrology Laboratory, Beltsville, Maryland maps obtained by this approach will be useful in inter† University of Maryland, Department of Geography, College Park preting aggregated flux estimates using bulk atmospheric Address correspondence to W. P. Kustas, USDA-ARS Hydrology boundary layer approaches over heterogeneous surfaces Lab., Bldg. 007, BARC-WEST, Beltsville, MD 20705. E-mail: bkustas (e.g., Brutsaert and Sugita, 1992; Hipps et al., 1994; [email protected] Received 18 August 1997; revised 1 December 1997. tas et al., 1995).