Modelagem da evapotranspiração por sensoriamento remoto: avaliação em diferentes biomas no mundo.

Abstract

Evapotranspiration (ET) plays a crucial role in the water cycle, influencing local climate and detrimental processes in the terrestrial environment, being essential for the water balance of ecosystems, in addition to representing a significant portion of the global energy balance. It is possible to estimate ET at a regional level using surface energy balance (SEB) models with Remote Sensing (RS) data. However, there are many uncertainties in the estimates of SEB models and several global products, which present poor performance when applied to non-agricultural environments. Modeling can generate diverse results, with significant variations according to the particularities of climate, vegetation and availability of water resources of each location. This study aims to evaluate the performance of SEB model parameters for estimating ET in different biomes. For this purpose, the performance of the STEEP, SEBAL and S-SEBI models were evaluated in different land uses and land cover in a sub-basin of the Upper Paraíba under varying humidity, with the inclusion of the MOD16 product, in 9 different vegetated biomes of the world under varying vegetation, humidity, soil and climate. The models were compared with each other and with local TE data collected, allowing the identification of the relevance of the complexity and interactions involved in each model. The SEB models were similar in locations of interesting ecosystems, evergreen and with water availability, such as pastures, the SEBAL and S-SEBI models resembled or surpassed the STEEP estimates, these models, due to their low algorithm complexity, carry less uncertainty. On the other hand, the application of SEB models in locations with heterogeneous vegetation, with phenological variation and water deficiencies, revealed that the STEEP model outperforms the other models, being able to represent the influence of seasonal dynamics and not overestimating ET in dry periods due to the incorporation of the seasonality of aerodynamic and surface variations in its environment. In addition, STEEP obtained good results in 6 of the 9 trained biomes due to the concentration of anchor pixels that proved to be efficient in several scenarios with an RMSE of 4.98 - 9.65 mm/8d. The application of STEEP was successful in presenting an advance in ET estimates in complex biomes that have a gap in accuracy for several global products. These results highlight the importance of considering the particularities of vegetation when modeling hydrological processes, especially in regions with contrasting climatic conditions. Understanding the differences between SEB models is crucial to improve ET estimates, providing more accurate information for the sustainable management of water resources in these regions.