Spatio-temporal surface mass redistribution in Amazon and Greenland recovered by satellite gravetry

Abstract

The Gravity Recovery and Climate Experiment (GRACE) satellite mission has provided monthly geopotential fields since its launch in March 2002. The monthly geopotential variations are closely related with Earths surface mass redistribution such as water transportations among oceans, atmosphere and land. Nominally, GRACE solutions exclude effects from tides, ocean dynamics and barometric pressure by incorporating geophysical models for them. However, those models are imperfect, and thus GRACE solutions include the residual gravity effects. Particularly, unmodeled gravity variations of sub-monthly or shorter time scale cause aliasing error, which produces peculiar longitudinal stripes. Therefore, difficulty in recovery of surface mass variations from GRACE lies in the error reduction with least signal loss, especially at sub-basin spatial scale. In this dissertation, Empirical Orthogonal Functions (EOF) method is examined to separate signals associated with surface mass variation from GRACE errors in Amazon Basin and Greenland Icesheet (GrIS). Because the two regions are different climatologically and geographically, spatio-temporal variability of signal and error in GRACE are distinct. Therefore, it is necessary to modify EOF method for different features of signal and error in the two regions. In this study, we develop novel methods based on rotated EOF and extended EOF for Amazon and GrIS, respectively. In Amazon Basin, the river discharge is estimated using water mass variation on the main stem. To achieve this, the terrestrial water storage (TWS) changes confined to the main stem have to be extracted not only from the GRACE error, but also from TWS variations adjacent to the main stem. In Greenland, detail month-to-month ice mass variations are estimated based on from separation between signal and aliasing error. The rotated EOF method over the Óbidos sub-basin successfully estimates river discharge in the basin. The resulting time series represents relative river discharge variations consistent to in-situ discharge estimate. However, the estimates are generally larger than in-situ data in high water seasons. This is likely due to detoured water in river pathway developed during flooding while GRACE observes integrated water mass variations in river channels. The rotated EOF method is extended to the entire Amazon Basin based on the results for Óbidos sub-basin. Therefore, the method provides total river discharge of Amazon basin, which has been unknown since in-situ observations are not available at the mouth of basin. The extended EOF provides higher temporal and spatial ice mass variations in GrIS, and this method is superior to the spatial filtering conventionally used in this region. In particular, results of extended EOF remarkably agree with surface mass balance (SMB) outputs of regional climate model during winter season when SMB is a major contributor to mass changes in GrIS. On the base of this consistency, GRACE observations and the regional climate model are used to retrieve ice discharge near the coastal region. The GRACE-based ice discharge remarkably agrees with results independently obtained from ice thickness and velocity survey in amplitude and spatial distribution.