Utilizing a retrospective cost adaptation (RCA) algorithm to achieve data-driven, adaptive, real-time (DART) precision meteorological forecasts

Highly-localized weather forecasts are an emerging need for the operational flight safety of uncrewed aerial vehicles (UAVs), particularly given the expected growth of UAV shipments and uses. Highly-localized forecasts will also serve to improve the tracking of toxic or contaminant clouds associated with accidents or incidents such as a train derailment or chemical plant fire. Therefore, with the ultimate goal of creating a data-driven, adaptive, real-time (DART) solution, the Weather Research and Forecasting (WRF) model has been augmented with a retrospective cost adaptation (RCA) algorithm as a potential means for improving precision meteorological (sub-1 km) forecasts. The RCA algorithm serves to adapt the simulated horizontal wind speed component values to observed measurements collected by appropriately instrumented UAVs at specified geographic locations. Horizontal wind speed data that was incidentally gathered from a UAV formation flight control experiment at the University of Kentucky flight field was employed to evaluate and validate the DART solution. The results indicate that when the RCA algorithm is employed with the WRF model, there is a 14% reduction in both the u and v wind speed component mean absolute errors compared to the baseline WRF model. The next step is to further validate the WRF model augmented with the RCA algorithm via flight test data collected specifically for this purpose in August 2023 at the University of Kentucky flight field. Once the promise of the DART approach is fully proven and operational, the ensuing forecasts will benefit the UAV community and contaminant incident first responders.