Do field observations agree with satellite-based evaporation products in the Miombo Woodland of Southern Africa?

Evaporation is a major constraining factor of water availability at the land surface which makes its assessment a highly significant prerequisite for application in hydrological, agricultural, climate studies and many other disciplines at various scales. However, its importance and calculation procedures have largely been crafted around and often limited to crop productivity. The overarching consequence of this is inaccurate estimates of evaporation for other land surfaces and particularly for forest systems. Due to limited field evaporation observations attention has been focused on the application of satellite-based products. However, in the case of Africa, and the Miombo ecosystem in particular, the number of flux towers is extremely limited (very few if any) which makes it extremely difficult to evaluate available satellite-based evaporation products. In this study we used the energy balance Bowen ratio approach to estimate field evaporation in a dense Miombo Woodland which we then used to evaluate four energy balance evaporation models. The models evaluated included the MOD16, SEBS, SSEBop and WaPOR. Furthermore, cluster analysis was used to assess the similarity of the models in simulating evaporation. The results show that at daily and dekadal scale the simulated evaporation by the four models significantly varied from field evaporation observations. However, less variations were observed at monthly scale.  Furthermore, all four models overestimated evaporation during the dry season (June-September) with RMSE ranges between 0.21 – 0.38 mm.day^-1 and 6.64 - 9.91 mm.month^-1. Based on the RMSE and biases the MOD16 (RMSE = 6.64 mm.month^-1; Bias = 2.04 mm.month^-1), SEBS (RMSE = 8.69 mm.month^-1; Bias = 5.72 mm.month^-1) and WaPOR (RMSE = 7.44 mm.month^-1; Bias = 6.67 mm.month^-1) ranked higher than the SSEBop (RMSE = 9.91 mm.month^-1; Bias = 9.84 mm.month^-1) in simulating evaporation in the Miombo Woodland. Three clusters were observed with the SEBS and WaPOR grouped together indicating their close similarity in simulating evaporation in the Miombo ecosystem while the MOD16 and SSEBop were each grouped separately. Results of this study could aid the interpretation of these evaporation models in Miombo Woodland covered basins such as the Zambezi River Basin in Southern Africa. This could help in monitoring basin water availability and ecosystem reactions and feedbacks to climate change and anthropogenic impacts.