Assessing a Multiphysics WRF Ensemble for Offshore Wind in the Northeastern United States Using LiDAR Data

Numerical weather prediction (NWP) systems remain the state of the art in middle to long-term forecasting for renewable energy resources including offshore wind. In recent years, constructing NWP ensembles has emerged as a leading strategy for characterizing the uncertainty associated with model boundary conditions and physical parameterizations within the model. However, NWP ensembles often exhibit both underdispersion and bias e.g., over-predicting wind speeds in the lower atmosphere where turbines extract power. This work leverages data collected by two LiDAR buoys launched off the coast of New York in 2019 by the New York State Energy Research and Development Authority (NYSERDA) to assess the performance of a multiphysics Weather Research and Forecasting (WRF) model ensemble over four one-week simulations with three nested domains. The largest has a 12 km spatial resolution covering the Northeastern United States, and the smallest has a 1.33 km resolution centered over New York City and the neighboring offshore region. For the first time, we carry out a spatiotemporal comparison of wind speeds produced by WRF with observations at multiple vertical levels, multiple times, and two geographic locations. We calculate and report common error metric at each of the two locations for each of five WRF physics setups appearing in recent offshore wind literature. We also report multivariate scoring metrics designed to capture the degree to which WRF accurately reproduces the spatiotemporal correlation pattern between the two LiDAR buoys. Finally, we discuss how this work supports New York States goal to develop 9000 MW of offshore wind by 2030.