Environmental factors affecting the accuracy of surface fluxes from a two-source model in Mediterranean drylands: Upscaling instantaneous to daytime estimates

The temperature-based two-source model (TSM) of Norman et al. (1995) has not been properly evaluated under the water stress conditions that are typical in natural Mediterranean drylands. In such areas, the asynchrony between precipitation and energy supply strongly reduces evapotranspiration, E (or latent heat flux, LE, if expressed in energy terms), making sensible heat flux (H) the dominant turbulent heat flux. In this study, we present a detailed analysis of the main environmental factors affecting the TSM effectiveness under such challenging conditions. The accuracy of the TSM, evaluated via errors in 15-min H estimates, was shown to have a diurnal variation. Accuracy was clearly reduced for solar elevation angles lower than 25° and during marginal hours of daytime, before 10am and after 3pm. The surface to air temperature difference (T_R-T_a) and the wind speed were the two environmental factors showing the strongest effect on the TSM accuracy. In contrast with results observed in other ecosystems, in this Mediterranean tussock grassland the TSM accuracy was not clearly reduced by cloudiness and it was improved under highly stressed vegetation conditions. The parallel resistances scheme of the TSM (TSM_P) showed overall lower errors and a lower tendency to underestimate at high H values, but the series resistances scheme of the TSM (TSM_S) increased the model accuracy under some specific circumstances such as low energy supply and atmospheric neutral conditions.Finally, two extrapolation methods to obtain daytime (Rn>55Wm^-2) turbulent fluxes from the 15-min estimates of TSM were compared: (i) assuming the self-preservation of the evaporative and the non-evaporative fraction (EF and NEF method) and (ii) averaging the total daytime instantaneous fluxes (Averaging method). Despite the assumption of daytime self-preservation of EF and NEF was showed consistent, this method retrieved less accurate daytime estimates of H, and E than the Averaging method as a result of inaccuracies affecting estimates of EF and NEF from the TSM at our site. Moreover, better daytime estimates of H and E were obtained when using instantaneous fluxes from the TSM_P than from the TSM_S. Thus, reliable daytime estimates of H were obtained from the TSM_P in a Mediterranean dryland, with mean errors of 20% and high correlations (R²=0.85). However, daytime E was strongly overestimated (125%) using the TSM by both methods, although a good correlation with eddy covariance measurements was found (R²=0.84).