Abstract Submission: Extreme climate events have threatened human health, economic growth, and the longevity of natural and built environments around the globe. Over the past decades, many studies, highlighting the projected changes to weather and climate extremes, have been conducted using physical and statistical models which have drawbacks such as accuracy, weakness in uncertainty analysis, and high computation cost. Machine learning and deep learning methods many of which were developed to work with “big data” have been able to tackle these shortcomings very well through their efficient computation and intelligence. This project aims to experimentally assess the performance of 5 machine learning (ML) models: Decision Tree, Support Vector Machine, Random Forest, K-Nearest Neighbors and Artificial Neural Network on extreme precipitation datasets for classification and regression tasks. Observed precipitation dataset for the years 1948–today across the Eastern United States will be used to train and test the ML models. Well known and accepted evaluation metrices including accuracy, F-Score, R-Squared (R2), root mean square error (RMSE), and AUC will be used to examine the performance of the selected machine learning models. Amazon’s AWS cloud service will be used for data consolidation, data storage, machine learning modeling, data analytics, and visualization. Comparison of different machine and deep learning models from this project will elucidate the efficiency and the potential use of AI/ML approaches for the forecasting and classification of the climate extreme events.
