Estimated evaporation of lakes by climate reanalysis data and artificial neural networks

Evaporation, together with precipitation, is the most important component of the hydrological cycle, and knowledge of the local values of lake evaporation has applications in reservoir design and management. The objective of this study was to estimate lake evaporation at locations without meteorological monitoring using ERA5 reanalysis data and artificial neural networks (ANNs). Data from 32 automatic stations in the state of Mato Grosso were used to estimate evaporation using the method of Penman (1948). The evaporation values were related to ERA5 data and radiation data at the top of the atmosphere using multilayer perceptron ANN models. The Mann-Kendall test was used for trend analysis in the estimated monthly evaporation series. From the analysis of the results, it is concluded that it is possible to quantify the spatial and temporal distribution of evaporation from lakes with data from ERA5 reanalysis and the use of ANNs. The historical evaporation series for the period 1980 to 2019 showed a positive trend in certain parts of the Brazilian Savanna and Amazon biomes. Isolated areas of the Pantanal biome also showed a positive trend for monthly evaporation. The proposed methodology allows for the precise and accurate estimation of evaporation from liquid surfaces at locations without meteorological monitoring.