The "AI-enhanced" Advanced Dvorak Technique (AiDT) is a machine learning (ML) model that utilizes the current ADTV9.1 history file output parameters to adjust the ADTV9.1 intensity estimate. The model is a relatively simple regression-based loss model that has one hidden layer and one output layer. The single output layer provides the current intensity in terms of a continuous maximum sustain windspeed (MSW) estimate.
The model was developed in order to improve the ADT estimates while still basing the output estimate on the well-established and familiar ADT estimate of MSW. Significant improvement to the ADT MSW estimate errors has been achieved using this simple ML model, especially in situations where the ADT is less skilled, such as central dense overcast (CDO) scene type situations and during the dissipation stage of the tropical cyclone lifecycle.
The latest AiDT model performance and examples are provided in the presentation below (as presented at the 2022 AMS Tropical Cyclone Conference).
