Estimation of Amazon River discharge based on EOF analysis of GRACEgravity data

Abstract

River discharge is a critical component for understanding hydrological processes and sustainable management of water resources. The importance of discharge observation has increased due to its potential extreme variation resulting from the projected climate change and stronger variability of precipitation and temperature in some large basins. However, inherent difficulties in ground-based observations and decreasing number of gauge stations hinder accurate measurement of global river discharge and its spatio-temporal variations. Various remote sensing methods have been examined as alternatives, however, they require ground measurements to convert their proxy measurements into the actual river discharge. In this study, we estimate the discharge at the O？bidos station and the mouth of the Amazon basin using the water storage variations derived from GRACE gravity data without relying on any auxiliary ground observations. We extract the water mass signal along the main stem of the river by applying the Empirical Orthogonal Function (EOF) for water storage variations over the basin. The relative water storage variations along the main stem derived from the EOF decomposition are highly correlated with in-situ discharge at the O？bidos. However, in high water season, the GRACE-based discharge is estimated larger than the in-situ observations, and the difference is particularly significant during the 2009 extreme flood season. We argue that the in-situ river discharge in 2009 was underestimated due to the missed water volume for the flow detouring around the O？bidos gauge station during the high-flow event. Net river discharge of the Amazon Basin to Atlantic Ocean is also estimated, and its annual discharge is about 23% larger than that of the O？bidos. In particular, 2009 river discharge to Atlantic Oceans is estimated as 1050 Gton.