Evaporation from a large lowland reservoir-(dis)agreement between evaporation models from hourly to decadal timescales

Accurate monitoring and prediction of surface evaporation become more crucial for adequate water management in a changing climate. Given the distinct differences between characteristics of a land surface and a water body, evaporation from water bodies requires a different parameterization in hydrological models. Here we compare six commonly used evaporation methods that are sensitive to different drivers of evaporation, brought about by a different choice of parameterization. We characterize the (dis)agreement between the methods at various temporal scales ranging from hourly to 10-yearly periods, and we evaluate how this reflects in differences in simulated water losses through evaporation of Lake IJssel in the Netherlands. At smaller timescales the methods correlate less (r D 0:72) than at larger timescales (r D 0:97). The disagreement at the hourly timescale results in distinct diurnal cycles of simulated evaporation for each method. Although the methods agree more at larger timescales (i.e. yearly and 10-yearly), there are still large differences in the projected evaporation trends, showing a positive trend to a more (i.e. Penman, De Bruin-Keijman, Makkink, and Hargreaves) or lesser extent (i.e. Granger-Hedstrom and FLake). The resulting discrepancy between the methods in simulated water losses of the Lake IJssel region due to evaporation ranges from-4 mm (Granger-Hedstrom) to-94 mm (Penman) between the methods. This difference emphasizes the importance and consequence of the evaporation method selection for water managers in their decision making.