An efficient and accurate python tool for uncertainty quantification of GRACE based TWS

Background

GRACE-derived TWS (terrestrial water storage) uncertainty provides crucial stochastic information for geophysical studies that assimilate GRACE data. The recent advances of assimilation studies demand to quantifying GRACE-derived TWS at a finer spatial resolution. However, traditional methods like the rigorous error propagation and the empirical modelling have their own limitations at either efficiency or flexibility towards the increasing demand of relevant assimilation studies. The Monte Carlo method, which has never been applied for GRACE uncertainty quantification before, is surprisingly found to achieve high flexibility, satisfied accuracy and unprecedented efficiency. The simplicity and generality of this method also allow one to easily customize it to accommodate user-specific GRACE study without the need to be professional to GRACE data processing. This method expects to bridge the gap and advance the data assimilation with GRACE in hydrological models.

Overview

1. software-MC-GRACE: this includes all the python codes and auxiliary data for running MC estimation of the TWS uncertainty. Two demo are provided for simple use, one of which indicates how to de-code sinex-data of ITSG-2018 normal equations and extract the variance-covariance of gravity Stokes coefficients, and another of which indicates how to run MC process for output the variance-covariance of the global map. Notice: to run the model one needs to download ITSG-2018 sinex data in advance.
2. PCA analysis: The first EOF/PC and the second EOF/PC are provided (from 2002-04 to 2023-08), so that one can reconstruct the uncertainty field easily. This data set is given in terms of 'HDF5', of which the key words includes: 'Mean', 'EOF-1', 'PC-1', 'EOF-2', 'PC-2' , 'Mask', 'latlon_info' and 'time_epoch'. The 'mask' indicates the 1 degree land mask (1-land, 0-ocean/polars) that we use to generate PCA result.
3. Time series of basin-averaged TWS: The time period covers from 2002-04 to 2023-08 for all basins. The data is given in terms of 'HDF5', of which the key words include 'Amazon' (the name of the basin) etc., and 'time_epoch' (year-month).
4. Mask file of all 36 basins. The spherical harmonic expansion of the mask is given as well.