Hydroclimatological variables in the South America CORDEX domain for the Amazon River Basin

Abstract

Introduction: With 7.0 x 106 km2 the Amazon River basin represents the most important hydrological region in South America. The Amazon discharges approximately 200,000 m3 s-1 becoming the main source of fresh water from the continent to the Atlantic Ocean. Methodology and data: Our aim is to study the representation of precipitation and evapotranspiration of the Regional Climate Model RCA4 v3 driven by the nine Global Climate Models available for the South America domain in CORDEX. For that purpose, we selected as comparison period the verification period of the IPCC model verification, between 1986 and 2006 at a monthly time scale. The performance of the models is compared with two datasets, the one from GPCC for monthly precipitation and the one from GLEAM for monthly evapotranspiration. Results: Accounting for the long term means of the variables in the period of analysis, models underestimate precipitation by 19% and underestimate evapotranspiration by 4%. When analyzed by wet (January) and dry (July) seasons we find that for precipitation there is underestimation during the dry season that can reach up to 70% and underestimation in the order of 60% for the wet season consistently in model CSIRO. In the case of evapotranspiration, the higher underestimations are of approximately 30% in model CSIRO. However, as for the spatial distribution of the RMSE, the northeast of the Amazonpresents the highest errors for evapotranspiration and the north of the Amazon for precipitation. Finally, projections present increase in annual amplitude for both variables under the most severe climate change scenario. Conclusions: With the results so far we can conclude that CORDEX models can be used for assessment of hydroclimatological conditions of the basin taking into account that there is a general underestimation of both variables and that CSIRO is amongst the nine models the one with the lower performance in order to represent the analyzed variables.