USE OF GROUND-BASED REMOTELY-SENSED DATA FOR SURFACE-ENERGY BALANCE EVALUATION OF A SEMIARID RANGELAND

An interdisciplinary field experiment was conducted to study the water and energy balance of a semiarid rangeland watershed in southeast Ari zona during the summer of 1990. Two subwatersheds, one grass dominated and the other shrub dominated, were selected for intensive study with ground-based remote sensing systems and hydrometeorological instrumen tation. Surface energy balance was evaluated at both sites using direc t and indirect measurements of the turbulent fluxes (eddy correlation, variance, and Bowen ratio methods) and using an aerodynamic approach based on remote measurements of surface reflectance and temperature an d conventional meteorological information. Estimates of net radiant fl ux density (R(n)), derived from measurements of air temperature, incom ing solar radiation, and surface temperature and radiance compared wel l with values measured using a net radiometer (mean absolute differenc e (MAD) congruent-to 50 W/m2 over a range from 115 to 670 W/m2). Soil heat flux density (G) was estimated using a relation between G/R(n) an d a spectral vegetation index computed from the red and near-infrared surface reflectance. These G estimates compared well with conventional measurements of G using buried soil heat flux plates (MAD congruent-t o 20 W/m2 over a range from - 13 to 213 W/m2). In order to account for the effects of sparse vegetation, semiempirical adjustments to the si ngle-layer bulk aerodynamic resistance approach were required for eval uation of sensible heat flux density (H). This yielded differences bet ween measurements and remote estimates of H of approximately 33 W/m2 o ver a range from 13 to 303 W/m2. The resulting estimates of latent hea t flux density, LE, were of the same magnitude and trend as measured v alues; however, a significant scatter was still observed: MAD congruen t-to 40 W/M2 over a range from 0 to 340 W/m2. Because LE was solved as a residual, there was a cumulative effect of errors associated with r emote estimates of R(n), G, and H.