An interpretation of large-scale GRACE observations of terrestrial water storage over the Amazon

The seasonal cycles of GRACE land water storage observations are analyzed with various land surface model (LSM) outputs including evapotranspiration (ET), runoff (R), and soil moisture (S). A simple yet effective runoff routing scheme based on a continuity equation is implemented in a spectral domain to obtain total land water storage from the LSM outputs. We present quite distinct spatial patterns in soil moisture and river storage distributions. The soil moisture storage tends to disperse with smaller magnitude while the river storage is more intense only along river and channel. Such characteristics are important to interpret large scale GRACE observations of total land water storage, particularly to assess the LSM runoff outputs. Various effective velocities, yet to be constant over the basin and time, for runoff routing are tested against the GRACE observations. Runoff in a LSM can be improved with more accurate modeling of ET since those two fluxes are substantially inter-related. The quality of ET from a LSM can be inferred indirectly by the quality of R that can be verified from GRACE observations after routing. We assess each of water cycle parameters from various LSMs including NOAH, CLM, VIC, Mosaic, and MATSIRO based on GRACE total storage observations and give possible physical interpretation. Particularly, the Amazon ET inferred from the combined analysis of water balance, LSM and routing simulations, and GRACE observations will be discussed.