The arid southwestern United States encompasses diverse ecosystems characterized by low precipitation, high temperatures, and unique desert ecosystems. The US-Mexico border region, inhabited by over 19 million people, including a large Latinx population, faces escalating water scarcity due to climate change and increasing multi-sectoral demands. Water allocation between the US and Mexico is primarily governed by the 1944 Water Treaty, which regulates the Rio Grande River and Colorado River basins. This study focuses on the critical water management of the Elephant Butte Reservoir in the Rio Grande basin. We have used remote sensing data and machine learning techniques to accurately estimate reservoir storage changes. Specifically, Random Forest (RF) and Long Short-Term Memory (LSTM) algorithms are utilized to identify important variables influencing reservoir storage and to capture temporal patterns and long-term trends in regional water availability. The analysis incorporates satellite-derived reservoir surface area data, with model performance evaluated using Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE) metrics. The developed generalized predictive models aim to enhance transboundary water resource management, potentially mitigating international water conflicts, supporting agricultural sustainability, and optimizing shared water resource allocation. Future research directions include exploring additional machine learning algorithms, integrating diverse remote sensing and climate data products, and expanding the approach to broader regional scales.

This presentation represents the research conducted by one of the I-GUIDE Summer School Teams (August 2024, Boulder, Colorado).