报告题目：Harnessing AI Techniques for Enhanced Hydrometeorological Forecasting: Experiences and Case Studies

报 告 人：Gerald Corzo Perez, Associate Professor, IHE Delft Institute for Water Education

时    间：2024年5月28日（星期二）  14:30—15:30

地    点：地理资源所A901

报告摘要：

Over the past decade, the frequency and intensity of extreme weather events have raised significant concerns, primarily attributed to climate change and inadequate infrastructure preparedness. This has necessitated a global effort to enhance flood and drought preparedness at all levels, with forecasting emerging as a crucial element in this endeavor. The advent of AI and data-driven technologies has further expanded the possibilities for accurate and timely predictions.

This presentation explores the experiences garnered from various research projects focused on the analysis and prediction of floods and droughts, spanning from urban environments to expansive river basins. Key projects include:

·         Forecasting Sub-seasonal to Seasonal (S2S) Weather Patterns: Utilizing committee models to enhance ECMWF forecasts in Italy, the Netherlands, and Spain.

·         Flood Inundation Prediction in Drainage Systems: Employing multiple machine learning models combined with Monte Carlo simulations to generate rainfall scenarios, enabling the identification of flood timings and locations.

·         Flow Forecasting in the Amazon River Basins: Integrating river modeling, discharge data, land use changes, and water vapor transport analysis to improve predictions for Colombia.

·         Water Level Prediction in the Everglades, USA: Using Convolutional LSTM models (Deep Learning) for accurate forecasting.

·         Intelligent Machine Learning for Optimal Rule Generation: Developing models to create optimal water management rules in the Dominican Republic.

These case studies highlight the diverse applications and effectiveness of modern AI and machine learning techniques in mitigating the impacts of extreme hydrometeorological phenomena.

报告人简介：

Dr. Gerald A. Corzo is an Associate Professor at IHE Delft in the Netherlands, specializing in machine learning applications for water resource systems. His research focuses on analyzing and forecasting hydrological extremes, such as identifying patterns, tracking spatial events, and predicting flows. Notably, he developed a four-dimensional rainfall object representation to track storms and analyze catchment responses, which is now used in the Mekong River's forecasting system. He also pioneered drought analysis methods linking anomalies to critical crop yield changes.

In 2023, Dr. Corzo published the book "Hydroinformatics: Machine Learning Applications to Water Resources Problems." He has collaborated with the USGS on Deep Learning techniques for forecasting water levels in the Everglades, Florida, and researched atmospheric water transport in South America using machine learning.

Dr. Corzo holds a Ph.D. from the Technical University of Delft and has postdoctoral experience at Wageningen University. He has managed data-driven projects and high-performance computing services at UNESCO-IHE, developed data integration modules for Delft-FEWS, and conducted spatiotemporal drought analysis with global hydrological models. His international collaborations span Colombia, Mexico, China, and Norway. He has chaired geo-statistics sessions at the European Geoscience Union and led the LatinAqua network for water research in Latin America. Currently, he coordinates the Laboratory of Hydroinformatics at IHE Delft, supervising PhD and MSc students and overseeing the Hydroinformatics specialization.