GRAiCE: Reconstructing terrestrial water storage anomalies with recurrent neural networks
Four global monthly terrestrial water storage anomaly reconstructions (1984-2021) at 0.5° using LSTM neural networks to extend the GRACE/GRACE-FO satellite record.
doi.org
At a glance
- Region
- Global
- Coverage
- Global
- Access
- Free
- Category
- Research
- Last updated
- 2025
What's inside
- Water Storage
- Machine Learning
What you can do with it
Typical work that data of this type supports.
- 01
Storage tracking
Follow storage and releases through the season.
Because it has: Water Storage
- 02
Change over time
See how lakes and wetlands have grown or shrunk.
- 03
No-gauge areas
Monitor places with no ground stations.
Where AI adds value
Ideas for machine learning and AI agents built on this kind of data.
- 01
Storage from space
Estimate volume from satellite water area and altimetry.
- 02
Release decisions
Optimise releases against inflow forecasts.
- 03
Segmentation
Map water, crops and land use pixel by pixel with deep learning.
Machine access and documentation
Data paper
GRAiCE: Reconstructing terrestrial water storage anomalies with recurrent neural networks
Scientific Data 2025
Read the paperRelated water data
- GTWS-MLrec: Global terrestrial water storage reconstruction by machine learning from 1940 to presentGlobal | Free
- ML-based reconstruction of long-term global terrestrial water storage anomaliesGlobal | Free
- A global multi-catchment synthesis for water fluxes and storage changesGlobal | Free
- C3S Satellite Lake Water LevelGlobal | Registration Required
- Dataset of stable isotopes of precipitation in the Eurasian continentGlobal | Free
- DeepExtremeCubes: Earth system data for assessing compound heatwave and drought impactsGlobal | Free
Find more water data
Search and compare 1,900+ water data sources by region, type and access.