The hydrological cycle over the Amazon: how well is it assimilated in ERA-40?

Abstract General Circulation Models (GCMs) reanalyses have been widely used as the best representation of reality derived from a model. Reanalyses are used to initialize regional models or describe climatological features of a certain region, being especially useful in areas where the observational network is sparse or absent, like the Amazon River basin. In this study the hydrological cycle variables in ERA40 for the Amazon River basin are evaluated in both temporal and spatial scales. An analysis of the hydrological cycle in annual time scales revealed that ERA40 basin averaged precipitation agrees well with observations, whereas runoff is underestimated by approximately 23% possibly due to an overestimation in interception loss. ERA40 rain rate is lower than that estimated by TRMM-GPI, indicating that the reanalysis is, most likely, overestimating the amount of water intercepted by the forest canopy. Despite the fact precipitation averaged over the entire Amazon basin agrees well with observations, its spatial distribution does not suggesting that basin averaged variables hide regional inaccuracies in ERA40. A similar feature is seen in the reanalysis spatial distribution of runoff, which more closely resembles precipitation instead of UNH-GRDC runoff composite fields. The seasonal cycle of ERA40 runoff is similar to precipitation in the model, whereas the observed discharge shows a cycle that lags precipitation by two months. The delay in the observed discharge is likely due to the time rainfall takes to actually reach the stream, a feature simulated by river routing in models but absent in ERA40. Using two other models, IBIS, that simulates runoff, and HYDRA, that routes IBIS runoff, it could be verified that the best representation of observation is obtained when local runoff is routed, resulting in an annual cycle very similar to observations for the Amazon. Our results suggest that basin wide estimates may not be the most adequate approach to evaluate the hydrological cycle over large river basins like the Amazon.