Augmenting the Tropical Cyclone Logistical Guidance for Genesis (TCLOGG) By Introducing a Most Likely Time of Genesis (MLTG)

This presentation details the implementation and evolution of the Tropical Cyclone (TC) Logistical Guidance for Genesis (TCLOGG) Most Likely Time of Genesis (MLTG) guidance. This new addition addresses a critical gap in TCLOGG, where the initial model determined genesis time (MDGT) exhibited biases concerning the timing of TC formation. Using a rolling five-year (2017-2021) developmental dataset, an ensemble-based machine learning regression model (Random Forest Regressor or RFR) is trained to produce MLTG forecasts for individual numerical weather prediction (NWP) models. To generate a consensus MLTG using all individual MLTG predictions, an additional ensemble-based machine learning regression model (XGBoost regressor) is employed as the statistical model. The implementation of the RFR and XGBoost-based MLTG guidance was completed for the 2023 season, and results from the 2023 and 2024 seasons are presented. These results showed the MLTG provides mixed improvements over the raw MDGT, with most improvements coming in the Joint Typhoon Warning Center’s area of responsibility.