Atmospheric River Analysis and Forecast System (AR-AFS): Impacts of Physics and Domain Configurations on Precipitation and AR Landfall Forecasts

Atmospheric rivers (ARs) are key features in the global water cycle and are tied to both flood hazards and beneficial water supply, particularly in the western North America. The rising frequency and impacts of ARs in recent years call for better Numerical Weather Prediction (NWP) models for forecasting AR landfalls and associated precipitation. This presentation describes the latest development of the Atmospheric River Analysis and Forecast System (AR-AFS), which is a stand-alone high-resolution regional model with a fine horizontal resolution of 3 km. AR-AFS is based on the FV3 dynamical core and uses initial and boundary conditions from the NCEP operational Global Forecast System (GFS). The current AR-AFS has shown a large dry bias in precipitation prediction. Given the precipitation forecast sensitivity to microphysics and PBL schemes, this study will further explore the impact of different physics configurations. Specifically, we will examine the use of Thompson microphysics and turbulence kinetic energy (TKE)-based moist eddy-diffusivity mass-flux (EDMF-TKE) PBL scheme for both convective and non-convective processes, and test various choices from the Common Community Physics Package (CCPP) on the prediction of AR landfalls and associated precipitation. Experiments with new domain configurations are carried out to cover the US East Coast and Guam. The impacts of various domain sizes (scaled by stretch factors, 1.0 and 0.8) and vertical resolutions (64, 81 and 127 levels) of the model are being examined. The AR-AFS forecasts will be evaluated against Stage IV observations and GDAS reanalysis data sets using standard metrics such as Mean Absolute Error (MAE), Equitable Threat Score (ETS), and the Method for Object-Based Diagnostic Evaluation (MODE) objects. This study highlights the role of physics parameterization and domain configuration in AR prediction and provides a valuable reference for the Research to Operations (R2O) process of NWP models in forecasting AR landfalls and associated precipitation.