Unsupervised Detection of Groundwater Storage Anomalies in Ghana Using GRACE Satellite Data

2026-08-10

AI Summary

This study addresses the long-standing scarcity of in situ groundwater observations in Ghana, which has hindered effective monitoring of storage anomalies. By integrating GRACE satellite-derived terrestrial water storage estimates from 2004 to 2024, the work introduces an ensemble Isolation Forest—an unsupervised machine learning approach—into regional groundwater anomaly detection for the first time. The method is combined with Z-score normalization and spatiotemporal statistical analysis to identify subtle anomalies often missed by conventional threshold-based techniques. It successfully detects 12 anomalous months (5 deficit and 7 surplus events), revealing a prolonged groundwater deficit during 2004–2009 and an intensified surplus trend after 2018. A distinct spatial pattern emerges, with deficits predominating in the north and surpluses in the south, demonstrating the method’s effectiveness and innovation in data-scarce regions.

Abstract

Groundwater variability in Ghana remains poorly characterized due to limited long-term in-situ observations. This study investigates groundwater storage anomalies using GRACE-derived data from 2004-2024 combined with statistical analysis and unsupervised machine learning. Groundwater anomalies were standardized using Z-scores, while an ensemble-based Isolation Forest framework was applied for anomaly detection. The results revealed substantial temporal variability, with persistent groundwater deficits during 2004-2009 followed by increasing positive anomalies after 2018. A total of 12 anomalous months were identified, comprising 5 deficit and 7 surplus events, with the strongest anomalies associated with groundwater deficits. Spatial analysis showed more frequent deficit anomalies in northern Ghana and stronger surplus occurrence in southern regions. Comparison with statistical thresholds further indicated that the machine learning framework captured additional subtle deviations beyond conventional threshold-based methods. Overall, the integration of GRACE observations with unsupervised anomaly detection provides a practical framework for groundwater monitoring in data-scarce environments.

Problem

• groundwater storage anomalies
• GRACE satellite data
• unsupervised detection
• data-scarce environments
• groundwater variability

Innovation

• unsupervised machine learning
• Isolation Forest
• GRACE satellite data
• groundwater anomaly detection
• data-scarce environments