Publications related to the GRACE Missions (no abstracts)

• Sorted by Date • Sorted by Last Name of First Author •

Improving the Spatial Resolution of GRACE–Derived GFZ G3P Groundwater Storage Anomaly Product Through Unsupervised Deep Learning Downscaling

Farías, Celina Anael and Goracci, Gilberto, 2026. Improving the Spatial Resolution of GRACE–Derived GFZ G3P Groundwater Storage Anomaly Product Through Unsupervised Deep Learning Downscaling. Remote Sensing, 18(18):3119, doi:10.3390/rs18183119.

Downloads

• from the NASA Astrophysics Data System  • by the DOI System  •

@ARTICLE{2026RemS...18.3119F,
       author = {{Far{\'\i}as}, Celina Anael and {Goracci}, Gilberto},
        title = "{Improving the Spatial Resolution of GRACE-Derived GFZ G3P Groundwater Storage Anomaly Product Through Unsupervised Deep Learning Downscaling}",
      journal = {Remote Sensing},
     keywords = {deep learning, GRACE, downscaling, groundwater, G3P, U-Net},
         year = 2026,
        month = sep,
       volume = {18},
       number = {18},
          eid = {3119},
        pages = {3119},
     abstract = "{What are the main findings? The unsupervised U-Net-based convolutional
        neural network framework tested in this study successfully
        generated high-resolution (0.1{\textdegree}) groundwater storage
        anomaly (GWSA) maps for the test area. Validation against IGRAC
        groundwater monitoring wells showed that the downscaled product
        improves the accuracy of the original GFZ G3P GWSA product at
        the basin level, while maintaining it at the localized well
        level. The unsupervised U-Net-based convolutional neural network
        framework tested in this study successfully generated high-
        resolution (0.1{\textdegree}) groundwater storage anomaly (GWSA)
        maps for the test area. Validation against IGRAC groundwater
        monitoring wells showed that the downscaled product improves the
        accuracy of the original GFZ G3P GWSA product at the basin
        level, while maintaining it at the localized well level. What
        are the implications of the main findings? The study
        demonstrates the potential and limitations of unsupervised deep
        learning for the downscaling of GRACE-derived products, without
        requiring high-resolution target data or in situ observations
        during model training. Built exclusively on globally available
        open datasets, the framework is potentially transferable to
        data-scarce regions worldwide. The study demonstrates the
        potential and limitations of unsupervised deep learning for the
        downscaling of GRACE-derived products, without requiring high-
        resolution target data or in situ observations during model
        training. Built exclusively on globally available open datasets,
        the framework is potentially transferable to data-scarce regions
        worldwide. While groundwater is the second largest freshwater
        reservoir on Earth, it remains difficult to monitor, relying
        mainly on sparse and often unavailable well measurements. The
        GRACE/GRACE--FO missions enabled the estimation of regional
        terrestrial water storage changes, although their coarse spatial
        resolution remains a major limitation. In this study, an
        unsupervised deep learning approach was exploited to downscale
        groundwater storage anomalies (GWSA) from the GRACE-derived GFZ
        G3P product from 0.5{\textdegree} to 0.1{\textdegree} spatial
        resolution, covering from 2003 to 2023. Thirteen ERA5 climatic
        variables and a digital elevation model were used as predictors.
        The framework was tested over France and validated in the Paris
        Basin using piezometric level information from the IGRAC GGMN
        open database. Results show good agreement between the
        downscaled and the original G3P product, with average temporal
        and spatial correlations of r = 0.99. When compared with in situ
        measurements, the downscaled product outperforms the original
        G3P data at the basin level (r = 0.67 vs. r = 0.62), while
        maintaining the accuracy at the local well scale (r = 0.39 vs. r
        = 0.38, respectively). The results highlight the potential of
        machine learning for GRACE super--resolution, while the main
        limitations are discussed.}",
          doi = {10.3390/rs18183119},
       adsurl = {https://ui.adsabs.harvard.edu/abs/2026RemS...18.3119F},
      adsnote = {Provided by the SAO/NASA Astrophysics Data System}
}

Generated by bib2html_grace.pl (written by Patrick Riley modified for this page by Volker Klemann) on Thu Oct 08, 2026 16:44:00

---

Thu Oct 08, F. Flechtner