Elsevier

Journal of Hydrology X

Volume 13, 1 December 2021, 100108
Journal of Hydrology X

Comparison of decadal water storage trends from common GRACE releases (RL05, RL06) using spatial diagnostics and a modified triple collocation approach

https://doi.org/10.1016/j.hydroa.2021.100108Get rights and content
Under a Creative Commons license
open access

Highlights:

  • 11 gridded GRACE products from RL05 and RL06 were analyzed in space domain.

  • A GWR scheme is proposed to modify the classic TCA uncertainty estimation.

  • Spatial autocorrelation and clustering declined significantly within 10-degree spatial radius.

  • The trend magnitude and the uncertainties follow the same clusters within a basin.

  • The GWR scheme underscored the “space” domain to analyze different GRACE products.

Abstract

GRACE (Gravity Recovery and Climate Experiment) and GRACE-FO (Follow-On) satellites have provided unique insights into the evolution of Terrestrial Water Storage (TWS) in space and time. Despite such advancements, various GRACE solutions produced by different data centers display uneven spatial attributes with varying associated uncertainties. Via spatial diagnostics tools and a modified triple collocation (MTC) approach, this research evaluates the TWS (terrestrial water storage) trend estimations “on the grid-scale” from 11 gridded GRACE products of RL05 and RL06 releases between 2002 and 2017. Distinct from classic TCA (triple collocation analysis), the MTC employs a GWR (geographically weighted regression) scaling scheme with distinctive spatial coefficients. The spatial diagnostics analyses identified different autocorrelation patterns, clustering tendencies of hot (positive) and cold (negative) spots agglomeration at varying spatial width, and unique frequency distributions. The results indicated that within a 10-degree spatial radius the SHs (Spherical Harmonics) of RL05 and RL06 are highly autocorrelated compared to the mascons (mass concentration blocks) solutions. The spatial clustering results revealed that many solutions agreed on the overall directions and distribution of the hot and cold spots. The clustering among mascon products, however, reflected more localized mass anomalies. At the scale of drainage basins, the trend magnitude, as well as their associated uncertainties appeared to be driven by the occurrence of spatial clusters within the basin area. The MTC results showed that the uncertainty patterns follow the same spatial extent within each cluster. The MTC analysis underscored the added benefits of cluster analysis and the GWR scaling over the classic OLS approach.

Keywords

GRACE
GRACE-FO
Triple collocation
GWR
OLS
Climate
Spatial statistics
Autocorrelation
Cluster analysis

Cited by (0)