Diagnosing water variations within the Amazon basin using satellite data

The components of the Amazon water budget and their spatiotemporal variability are diagnosed using monthly averaged remote sensing-based data products for the period September 2002-December 2006. The large Amazon basin is divided into 14 smaller watersheds, and for each of these sub-basins, fresh water discharge is estimated from the water balance equation using satellite data products. The purpose of this study is to learn how to apply satellite data with global coverage over the large tropical regions; therefore several combinations of remote sensing estimates including total water storage changes, precipitation and evapotranspiration. The results are compared to gauge-based measurements and the best spatiotemporal agreement between estimated and observed runoff is within 1 mm/d for the combination of precipitation from the GPCP and the Montana evapotranspiration product. Mean annual precipitation, evapotranspiration and runoff for the whole basin are estimated to be 6.3, 2.27 and 3.02 mm/d respectively but also show large spatial and temporal variations at sub-basin scale. Using the most consistent data combination, the seasonal dynamics of the water budget within the Amazon system are examined. Agreement between satellite based and in situ runoff is improved when lag-times between sub-basins are included (RMSE from 0.98 to 0.61 mm/d). We estimate these lag times based on satellite-inferred inundation extents. The results reveal not only variations of the basin forcing but also the complex response of the inter-connected sub-basin (SB) water budgets. Inter-annual and inter-SB variation of the water components are investigated and show large anomalies in northwestern and eastern downstream SBs; aggregate behavior of the whole Amazon is more complex than can be represented by a simple integral of the forcing over the whole river system.