Using Mobile Vehicle Observations to Improve Numerical Weather Prediction Forecasts

As Numerical Weather Prediction (NWP) models move toward kilometer-scale grid spacing in an effort to improve forecast skill, the current state of the atmosphere must also be sampled at higher resolution to properly resolve features at the same scale as the model. However, the lack of a dense surface observation network, particularly in rural areas, presents a major limitation in obtaining the necessary information for data assimilation. Connected vehicle technologies, where public, private, and commercial vehicles serve as weather-observing platforms, can be used to fill in these gaps in the surface observation network. To that end, the Federal Highway Administration developed the “Utilization of Vehicle Probe Data to Support Road Weather Hazard Diagnosis & Prediction” project in which a simulated vehicle observation data set was used to test the impacts of assimilating vehicle data on the performance of NWP model forecasts.

A simulated vehicle observation data set was created based on nearby surface stations and the Real-Time Mesoscale Analysis (RTMA) product. These data were assimilated into the Weather Research and Forecasting (WRF) model using its Four-Dimensional Data Assimilation (FDDA) system, and the resulting output of both higher and lower densities of vehicle observations were compared to the same WRF model output without vehicle data assimilation. Results showed improved performance with vehicle data assimilation in 2-m air temperature, 2-m dew point temperature, mean sea level pressure, and 10-m wind speed, and some instances of improved performance with vehicle data assimilation for precipitation forecasts. Using vehicle wiper status (as a proxy for precipitation) to forward error correct model output from the Road Weather Forecast System (RWFS) resulted in substantially improved QPF compared with the RWFS output without wiper status forward error correction.