Advanced Journal of Environmental Science and Technology

ISSN 2756-3251

Advanced Journal of Environmental Science and Technology | Vol. 17 (11), pp. 61-75, November, 2026
DOI: 10.46882/2026.AJEST.171105

Research Article

Deep-Learning Downscaling of Grace Satellite Terrestrial Water Storage Anomalies Quantifies Groundwater Depletion in Transboundary Agricultural Aquifers

Arjun Deshpande¹*, Nadia El-Khoury², Liam MacKenzie³, Siti Aminah⁴, Diego Vargas⁵

¹Interdisciplinary Programme in Climate Studies, Indian Institute of Technology Bombay, Mumbai, India
²Department of Civil and Environmental Engineering, American University of Beirut, Beirut, Lebanon
³Global Institute for Water Security, University of Saskatchewan, Saskatoon, Canada
⁴Department of Geoinformation, Universiti Teknologi Malaysia, Johor Bahru, Malaysia
⁵Department of Water Resources, Pontifical Catholic University of Chile, Santiago, Chile

Corresponding Author: Arjun Deshpande — [email protected]

Abstract

Coarse spatial resolution (~300 km) limits the utility of Gravity Recovery and Climate Experiment (GRACE) satellite observations for local groundwater management in heavily irrigated basins. We trained a physics-informed convolutional long short-term memory (ConvLSTM) network integrating soil moisture, evapotranspiration, and land-use variables to downscale GRACE-FO terrestrial water storage anomalies to 5 km resolution from 2018 to 2026. Validation against 1,420 monitoring wells yielded a Nash-Sutcliffe efficiency of 0.87 and root-mean-square error of 1.9 cm equivalent water height. Results revealed an accelerating mean groundwater depletion rate of -3.4 ± 0.4 cm/yr across double-cropped alluvial plains.

Keywords: groundwater depletion; GRACE satellite downscaling; deep learning; transboundary aquifers; terrestrial water storage; hydrogeology

Manuscript Timeline: Received 06 October, 2026; Revised 08 November, 2026; Accepted 19 November, 2026; Published 20 November, 2026.

Citation: Deshpande, A., El-Khoury, N., MacKenzie, L., Aminah, S., and Vargas, D. (2026). Deep-Learning Downscaling of Grace Satellite Terrestrial Water Storage Anomalies Quantifies Groundwater Depletion in Transboundary Agricultural Aquifers. Advanced Journal of Environmental Science and Technology, 17(11), pp. 61-75.