The objective of this study is to demonstrate that by use of simple physical techniques it is possible to obtain useful skill in retrieving sensible heat fluxes by the radiometric method but only on a site-specific basis. Sensible heat fluxes measured by Bowen ratio systems at two sites during the First International Satellite Land Surface Climatology Project (ISLSCP) Field Experiment (FIFE) 1987 and FIFE 1989 are compared to sensible heat fluxes estimated from radiometric measurements of skin temperature taken by pyrgeometers mounted 1.75 m above the surface. An experimental version of the biosphere-atmosphere transfer scheme (Ex-BATS), capable of reproducing statistically reliable estimates of the measured heat fluxes, is used to show that good agreement can be obtained between the model-diagnosed surface temperatures and the skin temperatures obtained from the pyrgeometers. However, because of biases in the radiometer values which are small in an absolute sense but large in terms of the differential temperatures between surface and atmosphere, use of such temperatures in modified bulk formula expressions gives rise to significant biases in the sensible heat fluxes. By reconciling radiometric estimates of the sensible heat fluxes and the fluxes measured by the Bowen ratio systems, a mean effective emissivity considered in the form of a site-specific calibration constant can be calculated for each of the two sites. Using the new emissivities, corrected radiometric surface temperatures are obtained that remove the biases at each site, demonstrating that the radiometric method is feasible but only on a site-specific basis.

The corrected skin temperatures are then compared to the advanced very high resolution radiometer (AVHRR)-derived split window estimates of skin temperature obtained on 30 clear-sky days during the 1987 experimental period. The two data sets are found to be well-correlated but also with an underlying bias. Differences between the satellite estimated temperatures and the corrected skin temperatures are attributed to imperfections in the coefficients used in the split window equation and residual cloud contamination, as well as considerable differences in the scenes viewed by the pyrgeometers and satellite. These differences lead to significant errors in retrieving heat fluxes from the raw satellite skin temperature estimates. The estimates can be improved after applying a regression between the corrected radiometric skin temperatures obtained for the radiometers and the satellite-derived temperatures, although because of the residual cloud effects and scene mismatch, the final RMS error is of the order of 65 W m-2. An independent application of the regression fit between the radiometer temperatures and the AVHRR-derived temperatures derived from the 1987 data to sensible heat flux calculations for four clear-sky days in 1989 indicates poor skill, reinforcing the notion that for satellite applications the radiometric approach is also only feasible on a site-specific basis. ¿ American Geophysical Union 1995