Research Highlights

Harnessing Machine Learning to Decode Rainfall in the Desert: UAEU Leads Breakthrough Study on Predicting Precipitation in Arid Climates

Amid rising climate unpredictability and the increasing frequency of extreme weather events, researchers at the United Arab Emirates University (UAEU) are turning to artificial intelligence to tackle one of the region’s most pressing challenges: forecasting rainfall in a hyper-arid environment.

The UAE, known for its dry landscapes and minimal annual precipitation, faces significant hurdles in managing its water resources. However, the recent surge in abnormal rainfall patterns, including intense storms and flash floods, has amplified the urgency for innovative forecasting tools that can aid decision-makers in water conservation, agriculture, and infrastructure planning.

In response, a groundbreaking study led by Professor Mohsen Sherif, Provost at UAEU explored how Machine Learning (ML) can revolutionize rainfall prediction in the region.

“Accurate rainfall prediction is no longer a luxury in arid regions - it is a necessity. Through the integration of machine learning and environmental data, we’re opening new frontiers in water sustainability and climate resilience for the UAE and beyond.”   Prof. Mohsen Sherif, Project Lead

Conducted by Doctoral Scholar Faisal Baig, in collaboration with UAEU’s College of Information Technology and Colorado State University, the study analyzed over 30 years of climate data using advanced ML models including XGBoost, LSTM, Random Forest, and Support Vector Machines.

Early experiments using rainfall as the sole predictor yielded limited results. However, predictive accuracy significantly improved when the models were expanded to include key meteorological parameters such as temperature, humidity, wind speed, and evapotranspiration. Notably, XGBoost’s correlation coefficient rose from 0.45 to 0.76, and LSTM models improved from 0.21 to 0.71 during testing.

A sensitivity analysis using LASSO regression identified wind speed and minimum temperature as critical variables in forecasting monthly rainfall, providing valuable insights for resource planners and policymakers.

Building on these promising results, the team has launched a follow-up project focused on the bias correction of satellite-based rainfall estimates. By incorporating topographic and atmospheric variables, this next phase aims to enhance the accuracy of satellite data, further supporting proactive water management strategies across the UAE. Further, the team is also working on utilizing ML to enhance the prediction accuracy of the Global Climate Models (GCMs) over the Arabian Peninsula as a core theme of the post-doctoral research project awarded under the Strategic Research Grant of UAE University to the PI, Professor Dr. Mohsen Sherif.

This cutting-edge research not only reinforces UAEU’s leadership in environmental innovation but also contributes meaningfully to the UAE’s broader goals for climate resilience and sustainable development.

---

To read more about the research: https://www.sciencedirect.com/science/article/abs/pii/S0022169424004359