Using initial and boundary condition perturbations in medium-range regional ensemble forecasting with two nested domains

Simulated evolution of climate and weather is sensitive to the specification of their initial state. Small errors in the initial state could lead the forecast into a different direction. It is essential to estimate the impact of the uncertainty in initial conditions on the forecast accuracy. For limited-area or regional forecasting, lateral boundary conditions also have considerable influence on the development of mesoscale or local-scale phenomena. Strong lateral boundary conditions derived from a larger scale environment could significantly alter or even remove local-scale components. This study investigates the impact of uncertainty in initial and lateral boundary conditions on medium-range regional forecasting using the Advanced Weather Research and Forecasting (WRF) model. The WRF model was configured with two nested domains: the parent domain has a 108 km horizontal resolution, and a nested domain with 36 km resolution covers the western U.S. The ensemble forecasting was conducted with 50 ensemble members using random perturbations in the initial conditions (ICs) and lateral boundary conditions (LBCs). A case period of 15 days in December 2008 is chosen, during which two intense frontal passages occurred in the western U.S. Results show that, applying only IC perturbations, the contribution from the IC perturbations to the ensemble spread decreases with time. Using both randomly perturbed LBCs and ICs from the coarser domain, the inner nested domain shows a wider ensemble spread. The resulting ensemble forecasting can be interpreted as a probabilistic prediction for wind energy, especially for wind gust and wind turbine operational cut-off. The analysis also includes an efficiency comparison of using coarser ensemble forecasting vs. a higher resolution single control run.