Global Terrestrial Water Storage Time Series Reconstructed from 1979 to 2020

A 42-year global time series (1979-2020) of terrestrial water storage changes, reconstructed by Siyou Xu to address gaps in GRACE/GRACE-FO satellite data. The dataset was created using a method combining principal component analysis, seasonal-trend decomposition, and a genetic algorithm-optimized neural network with spatial mode correction. It is validated against observed satellite data and independent hydrological models.

Use Cases

• Analyzing long-term trends in global freshwater availability based on the reconstructed 1979-2020 time series.
• Validating and calibrating hydrological models using the basin-scale terrestrial water storage data.
• Studying the impact of climate variability on water resources during the pre-GRACE observation period.
• Bridging the data gap between GRACE and GRACE-FO missions for continuous water storage monitoring.

Strengths

• Covers a 42-year period from 1979 to 2020, providing a long-term perspective.
• Validation shows high consistency with GRACE/GRACE-FO data over 25 major river basins.
• Reported average correlation coefficient exceeding 0.80 with independent models during the GRACE/GRACE-FO gap.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data is stored in MAT format, which may require specific software (e.g., MATLAB) for access.

Provenance

Source

figshare, author Siyou Xu.

Collection Method

Reconstructed using a novel method integrating PCA, STL decomposition, and a Genetic algorithm–optimized ANN with spatial mode correction.

Time Range

1979 to 2020

Freshness

Last updated 2026-06-02 12:44:46; freshness should be verified.

Geography

Global, with validation over 25 major river basins.