Mosaic versus dual source approaches for modelling the surface energy balance of a semi-arid land

Two-layer parameterisation of the surface energy budget proves to be realistic for sparse but homogeneously distributed vegetation. For semi-arid land surfaces however, sparse vegetation is usually interspersed by large patches of unshaded bare soil which may interact directly with the atmosphere with little interference with the vegetation. Therefore such surfaces might not be realistically represented by a two-layer parameterisation. The objective of this study is to investigate the issue of representing water and energy transfer processes in arid and semi-arid regions. Two different surface schemes, namely the classic two layer (one-compartment) approach and a two adjacent compartment ('mosaic') approach are used. The performance of both schemes is documented using data sets collected over two sparsely vegetated surfaces in the San Pedro river basin: homogeneously distributed grassland and heterogeneously distributed shrubs. In the latter case the mosaic scheme seems to be more realistic given the quality of the temperature estimates. But no clear statement can be made on the efficiency of both schemes for the total fluxes. Over each site, we investigate the possibility of artificially modifying some of the surface parameters in order to get the surface fluxes simulated by the one-compartment scheme to reproduce the two-compartment ones. The 'cost' associated with this process in terms of surface temperature estimates is eventually discussed.