Connected Vehicles: Filling in the Observation Gap for Data Assimilation

Sparse observation networks present a major limitation in obtaining the current state of the atmosphere for numerical weather prediction (NWP) applications. As computing capability increases, allowing operational forecasts to approach kilometer-scale horizontal grid spacing, finer resolution observations will be needed to accurately determine the initial atmospheric state. Connected vehicle technologies, which allow vehicles of all types to serve as weather-observing platforms, can be used to fill in these gaps in the surface weather observation network with thousands of additional data points available every few minutes across a statewide road network. The National Highway Traffic Safety Administration (NHTSA) is expected to issue a rule-making decision that could mandate that future vehicles transmit data relevant to road safety and meteorological applications. Whether through this rule-making or similar developments taking place via other means, there is a need for the meteorological community to study how best to prepare for and use that large volume of data to improve forecasts, both for surface transportation and other applications.

This study sought to assess the impact of assimilating vehicle-based observations into a high-resolution model. As connected vehicle technologies are not yet widely deployed, these observation datasets were simulated, and done so at varying spatial densities. Using Four-Dimensional Data Assimilation (FDDA), the datasets were assimilated into the Advanced Research Weather Research and Forecasting (WRF-ARW) model with 1.33-km horizontal grid spacing, and for select case studies with a range of weather conditions in Minnesota and Michigan. Model output from the vehicle data assimilation simulations and a baseline WRF simulation, without assimilating any vehicle data, were compared against observations. These comparisons were then used to assess the impact of variable-density vehicle observations on forecasts. Verification and evaluation was performed using the Model Evaluation Tools (MET) package.