WATERRESOURCES RESEARCH, VOL. 30,NO. 5, PAGES 1363-1373, MAY 1994 Use of remote sensing and referencesite measurements to estimate instantaneous surface energy balance components over a semiarid rangeland watershed K. S. Humesand W. P. Kustas Hydrology Laboratory, Agricultural Research Service, U.S.Department ofAgriculture, Beltsville, Maryland NI. S. Moran U.S. Water Conservation Laboratory, Agricultural Research Service, U.S. Department of Agriculture Phoenix, Arizona Abstract.A primary motivation for using remotely sensed datato estimate components of the surface energy balance is to quantify surface energy fluxes in a spatially distributed manner over various spatial scales. However, all models which utilize remotely sensed data to estimate surface fluxes alsorequire inputvariables and parameters which cannot be estimated on a spatially distributed basis withremotely sensed data. In this analysis, datafrom the Monsoon '90 experiment were usedto evaluate the limitations in spatially extending a relatively simple energy balance model with remotelysensed data over a semiarid rangeland watershed. Using one ground- based meteorological and flux stationas a reference site, aircragt-based remotely sensed data (surface temperatures and reflectances) were usedto compute energy balance components for seven other locations within the watershed. The results indicated that for clear sky conditions, all components of the surface energybalance could be estimated to within approximately the samelevel of uncertaintywith which the fluxes were measuredwith ground-based flux instrumentation. However, under •partly cloudy conditions the variability in incoming solar radiation across the watershed significantly degraded the estimation of distributed values of net radiation (Rnet). If ground-based estimates of incoming solarradiation are usedto calculate R ne t from remotely sensed data, then the spatial extent over which that measurement is valid limits the area over which accuratespatiallydistributed valuesof R net can be estimated. Additionally, the resultsof sensitivity analyses indicatethat the level of uncertainty to which the roughness lengthfor momentum, or Zorn, is typically known forspatially distributedvalues in an area of naturally variable vegetation can give rise to significant uncertaintiesin the estimationof sensible heat flux. For areas where the spatial variationin roughness parameters is of the order of several centimeters, the error associated with assuming constant values for the roughness length for momentum issimilar in size to the errors associated with temperature variationsof the order of several degrees. In order to utilize radiometric temperatures to reliably estimate spatially distributedvalues of sensible heat flux, techniques suchas those explored by Menenti and Ritchie (this issue) are needed to providespatially distributed information on surface roughness parameters. Introduction A primary motivation for using remotelysensed data to estimate components of the surface energy balance is to quantify surfaceenergy fluxes in a spatiallydistributed manner over various scales. Spatially distributed estimates ofsurface energy fluxes would be a useful tool for better understanding the interactions between the landsurface and lower atmosphere, as wellas thebehavior of surface energy fluxes atdifferent spatial scales. The estimation ofinstanta- neous and daily values of surfaceenergy balancecompo- nents has been demonstrated using local-scaleremotely This paper is not subject toU.S. copyright. Published in1994 by the American Geophysical Union. l•per number 93WR03082. senseddata and ground-basedmeteorological data. These techniqueshave been particularly successful for surfaces with nearly complete vegetation cover [Seguin and Itier, 1983; Reginato et al., 1985; Jackson, 1985; Huband and Monteith, 1986a, b; Jackson et al., 1987] and have more recently been investigatedover surfaces with incomplete vegetation cover [Kustas et al., 1989;Kustas and Daughtry, 1990; Moran et al., this issue]. However, not all of the surface and near-surface parame- ters and variablesrequired to compute surfaceenergy bal- ancecomponents canbe derivedfrom remotely sensed data. Most models which estimate energy fluxes using spatially distributed surface parameters derivedfrom remotely sensed data (e.g., surface temperature and albedo) also require as input parameters and variables which must either be known a priori or measured with groundinstruments (e.g., surface !363