7B.5 Using Evolutionary Programming to Generate Improved Tropical Cyclone Intensity Forecasts

Tuesday, 17 April 2018: 2:30 PM
Masters ABCD (Sawgrass Marriott)
Jesse Schaffer, University of Wisconsin−Milwaukee, Milwaukee, WI; and P. J. Roebber and C. Evans
Manuscript (171.8 kB)

The difficulty of forecasting tropical cyclones has lead to the development of many different tropical cyclone forecasting models. These models can be dynamical, statistical, or a combination thereof. While these models have made notable advances in forecasting track, improvements in forecasting intensity have been comparatively lacking. A significant contribution to intensity errors is the difficulty of predicting times of rapid intensification (RI) and rapid weakening (RW). The goal of this research is to use evolutionary programming (EP) to construct improved forecasts for tropical cyclone intensity through a lead time of 120-h. Previously, EP-generated ensembles have been tested on 60-h minimum temperature forecasts as well as 500-hPa height forecasts and been shown to outperform traditional dynamical and multiple linear regression (MLR) ensembles. This improvement is a result of the flexible EP architecture that allows for adaptive forecasts. For this research, two similar but separate models will be developed, one for the north Atlantic basin and the other for the East/Central Pacific basin. These models will be developed using predictors sourced from the SHIPS predictor set, with data from 2000-2012 used for training and data from 2013-2016 reserved for testing. Due to the prior success of implementing EP it is believed that small but significant improvements to tropical cyclone intensity forecasts will be made. While there is also a hope that significant improvements will be made in forecasting RI and RW, it should be noted the added skill of EP relative to existing methods is highest amongst predictable variables and limited benefit is provided otherwise. Thus, predicting RI may still be difficult until a more complete dynamical understanding is reached. This presentation will summarize model development as well as model performance on data from 2013-2016.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner