AI supported forecasts of heat waves along the Pacific coast of the Americas
Abstract
Heat waves are extreme weather events that severely impact society due to increased demand for health services, diminished crop returns, and migration. Climate change is expected to significantly increase the frequency and intensity of heat waves in the followingdecades. Prediction algorithms and early warning systems are key technologies that enable emergency measures and public policies. The design of such computational techniques is challenging due to limited historical data and the complexity of climate and weather conditions that drive heat waves. Hence, specialized algorithms are necessary to predict future heat waves accurately. This project proposes developing an Artificial Intelligence (AI) supported platform for forecasting heat waves along the Pacific coast of the Americas.Designing efficient machine learning (ML) algorithms that accurately forecast heat waves requires fundamental research into mathematical models that detect the intricate climate patterns that cause theserare weather events. We will identify heat waves from temperature measurements at weather stations and characterize their extension and severity. The synoptic meteorological patterns accompanying them will be retrieved from global reanalysis models. Finding relevant predictors for extreme events from diverse data sources is a notoriously difficult task to perform. We propose a combination of automatically reducing the dimensionality to assess feature importances and manually labeling climate patterns such as the El Ni�o Southern Oscillation and atmospheric rivers that are knownto be influential. Different supervised ML algorithms will be designed, optimized, and extensively validated. We will primarily target mid- latitudes where the USA and Chile#s coastal climate and synoptic meteorology are remarkably similar, thus allowing us to train the algorithms on a more extensive database. We will explore transfer learning techniques to extend the AI platform#s capabilities to Central America and evaluate the importance of the Pacific climatology on heat-wave occurrence in this region.Applying the trained system to real-time weather information will provide an early-warning system for forecasting the probability of a heat wave in the coming days or weeks. In addition, we will apply our algorithms to future climate simulations to estimate the frequency of heatwave occurrence in the remaining decades of the 21st century.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- May 15, 2023
- Source ID
- N629092312033
Entities
People
- Elwin Van T Wout
Organizations
- Office of Naval Research
- Pontifical Catholic University of Chile
- United States Navy