Subseasonal Cyclogenesis Number Predictions in the East Pacific Using Random Forests
Abstract
Trustworthy subseasonal forecasts of cyclone activity are crucial for providing vital lead time to prepare for catastrophic storms. Previous research by Slade and Maloney (2013) used logistic regression to build an intraseasonal prediction model of Atlantic and East Pacific cyclogenesis with a seven-week forecast lead time. We expand upon this work in our present study, in which we use nine more years of hurricane data, and a machine learning technique known as a Random Forest. A Random Forest consists of an ensemble of Decision Trees, which can be used for numerous classes of prediction problems. We develop and employ a Random Forest algorithm to predict the number of weekly hurricanes in the East Pacific two to three weeks in advance. To make these predictions, we use information about the state of the El Niño Southern Oscillation and Madden-Julian Oscillation, as well as the historical hurricane climatology as predictors. This combination of predictors allows the Random Forest to learn different pathways to tropical cyclogenesis. A comparison of performance to previous models is presented.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2022
- Bibcode:
- 2022AGUFM.A22F1721C