In semi-arid areas, where vegetation is sparse and clumped, models used to estimate evapotranspiration (λE) consider soil and plants as different sources of evaporation. When working at higher scales of heterogeneity, the modelling of surface fluxes introduces effective parameters that enclose the sub-grid heterogeneity. In this work we used both approaches to estimate the λE of a sparse-vegetation patch of Retama sphaerocarpa (L.) Boiss in a semi-arid area in southeast Spain. Firstly, we used a multi-layer model considering plant, soil under plant and bare soil, each with its own surface and aerodynamic resistances and available energy, interacting at a within canopy height. Secondly, we used a single-layer model that uses the effective surface and aerodynamic resistances of the patch, calculated by different aggregations of the soil and plant resistances considered in the multi-layer model. The estimates of λE were compared with measured values obtained by an Eddy covariance system. Results show that the use of effective surface resistances aggregated in parallel and effective aerodynamic resistances aggregated in series in a single-layer model produced similar estimates of λE as a multi-layer model. When compared to the measured values, the estimates of the single-layer model were even more accurate than the estimates of the multi-layer model. The results of this paper show that, in areas with low vegetation cover, a simple model, where patch-scale heterogeneity is solved by the use of effective aggregated resistances, is adequate to estimate the patch-scale evapotranspiration. © 2007 Elsevier B.V. All rights reserved.