Uncertainty analysis for satellite derived sensible heat fluxes and scintillometer measurements over Savannah environment and comparison to mesoscale meteorological simulation results

Three methods for estimating instantaneous sensible heat flux (H) over Savannah environment in West Africa were compared: first, satellite derived estimations using the Surface Energy Balance Algorithm for Land (SEBAL) method [Bastiaanssen, W.G.M., Menenti, M., Feddes, R.A., Holtslag, A.A.M., 1998a. A remote sensing energy balance algorithm for land, SEBAL: 1. Formulation. J. Hydrol. 212¿213, 198¿212]; secondly, measurements at two test sites in Ghana with a large-aperture scintillometer (LAS); third, high resolution mesoscale meteorological simulations using the MM5 (5th-Generation Penn State/NCAR) mesoscale modelling system. Satellite-derived sensible heat flux was based on seven NOAA-16 AVHRR images that were processed for a 2-week period in December 2001 (dry season) and were compared to LAS-data and MM5 simulation results. A methodology based on Gaussian Error Propagation is presented to derive uncertainties in satellite derived sensible heat flux due to (a) input data, (b) coefficients to determine leaf area index (LAI) and (c) methodological differences in estimating surface temperature T0. Total computed relative uncertainty in H was 15% for the Tamale test site and 20% for the Ejura site. Uncertainties in instantaneous evapotranspiration ¿E, however, are much smaller than uncertainties of H. This results due to the same bias in H and Rn ¿ G. For LAS-data, an uncertainty analysis due to input data was performed which showed relative uncertainty of 8% for the Tamale site and 7% for Ejura. Satellite derived net radiation (Rn) was underestimated in comparison to ground measurements which finally caused an underestimation of H. Satellite estimates of H using spatially interpolated ground based measurements of net radiation showed good agreement to LAS data. MM5-computed latent heat flux showed very low values for the entire region. This caused a serious relative MM5-overestimation of sensible heat flux in comparison to LAS and satellite derived estimates. It could be shown that Gaussian Error Propagation can serve as an essential tool to asses the reliability of satellite derived sensible heat fluxes. The resulting uncertainties give information on sensitivities in estimating H and therefore provide a tool for validation purposes.