Abstract
There is a growing demand for regional-scale climate predictions and assessments. Quantifying the impacts of uncertainty in initial conditions and lateral boundary forcing data on regional model simulations can potentially add value to the usefulness of regional climate modeling. Results from a regional model depend on the realism of the driving data from either global model outputs or global analyses; therefore, any biases in the driving data will be carried through to the regional model. This study used four popular global analyses and achieved 16 driving datasets by using different interpolation procedures. The spread of the 16 datasets represents a possible range of driving data based on analyses to the regional model. This spread is smaller than typically associated with global climate model realizations of the Arctic climate. Three groups of 16 realizations were conducted using the fifth-generation Pennsylvania State University–National Center for Atmospheric Research (PSU–NCAR) Mesoscale Model (MM5) in an Arctic domain, varying both initial and lateral boundary conditions, varying lateral boundary forcing only, and varying initial conditions only. The response of monthly mean atmospheric states to the variations in initial and lateral driving data was investigated.
Uncertainty in the regional model is induced by the interaction between biases from different sources. Because of the nonlinearity of the problem, contributions from initial and lateral boundary conditions are not additive. For monthly mean atmospheric states, biases in lateral boundary conditions generally contribute more to the overall uncertainty than biases in the initial conditions. The impact of initial condition variations decreases with the simulation length while the impact of variations in lateral boundary forcing shows no clear trend. This suggests that the representativeness of the lateral boundary forcing plays a critical role in long-term regional climate modeling. The extent of impact of the driving data uncertainties on regional climate modeling is variable dependent. For some sensitive variables (e.g., precipitation, boundary layer height), even the interior of the model may be significantly affected.
Use of Sensor Imagery Data for Surface Boundary Conditions in Regional Climate Modeling