Improved convective initiation forecasting in the Gulf of Mexico region through enhanced uses of NASA satellite assets and artificial intelligence

Short-term forecasts of convective initiation (CI) and early storm development are essential for providing decision makers adequate warning to mitigate aviation safety, lightning, flash flood, and other storm-related hazards. Such hazards have a particularly high impact in the Gulf of Mexico region, where convection is prevalent year-round, workers are exposed on ships and oil platforms and helicopter transportation is common. Operational weather prediction models are currently poor at pinpointing locations and timing of CI, while extrapolation techniques that work well for pre-existing storms do not apply to CI. The present study will fuse operational satellite and model data using artificial intelligence (AI) techniques to create improved CI forecasts over both land and water in the Gulf of Mexico. Our results will immediately be utilized by the Corpus Christi Weather Forecast Office (CCWFO), and this effort will serve as a feasibility study that, if successful, will lead to improvements in Convective Nowcast Oceanic (and Global Turbulence decision support systems, currently under development under NASA funding for use in the World Area Forecast System.

The proposed study will utilize Tropical Rainfall Measuring Mission and CloudSat to characterize storms to define “truth” for tuning and verification. Moderate Resolution Imaging Spectroradiometer land temperature gradients and Advanced Microwave Scanning Radiometer–Earth Observing System sea surface temperatures will be analyzed as potential predictor fields, as will Weather Research and Forecasting Rapid Refresh and/or Global Forecast System model data, both of which assimilate NASA satellite data products. AI techniques will be employed to identify new data for incorporation into the 0-1 hour SATellite Convection AnalySis and Tracking (SATCAST) CI nowcasting algorithm, so to optimize SATCAST and to create a probabilistic predictive model of early storm development. Synergy with NASA-funded research on CI, lightning initiation, verification, oceanic nowcasting, and global turbulence prediction will be exploited.

Presented will be (1) enhanced methods of using NASA and NOAA satellite data, along with non-satellite fields, to forecasts CI over oceanic and coastal regions, (2) enhance thunderstorm forecasts at the CCWFO produced by the Thunderstorm Artificial Neural Network, resulting in improved weather information for coastal residents and Gulf of Mexico transportation interests, and (3) improved understanding on how AI techniques can be applied to convective weather forecasting.