Abstract:The Amazon area is the largest water shed on Earth. Thus, it is of great importance to observe the water levels regularly. The satellite mission Gravity Recovery and Climate Experiment (GRACE) allows, since its launch in 2002, a monthly global overview of the water distribution on Earth, in particular floods and droughts. In recent years, the Amazon area has experienced a number of extreme weather situations in late summer (July through October), explicitly a drought in 2005 and one in 2010. Furthermore, one can identify the remains of a flood in spring 2009 in the summer season of 2009 as well, where the names of the seasons refer here to the northern hemisphere, though some events were also located on the southern hemisphere. Here we present corresponding results with respect to a recently introduced localized method called the RFMP (Regularized Functional Matching Pursuit) that can be applied to ill-posed inverse problems. In comparison to the usual processing of GRACE data as well as other data types (i.e., the volumetric soil moisture content given by the NOAA-CIRES Twentieth Century Global Reanalysis Version II and the average layer 1 soil moisture given by the GLDAS Noah Land Surface Model L4), we gain an improved spatial resolution with the novel method. We also observe that it is very difficult to validate inverted GRACE data with hydrological models due to, e.g., discrepancies among these models.