Impact of Ensemble Design on Vapor Transport and Atmospheric Rivers

Abstract:

As much as 90% of extratropical water vapor transport occurs in localized narrow filaments of strong vertically integrated water vapor transport (IVT), referred to as atmospheric rivers (ARs, e.g., Zhu and Newell 1998). ARs are therefore of primary importance to the global hydrological cycle and of keen interest to forecasters and decision makers, as they are often associated with high-impact weather.

We examine the simulation and prediction of IVT and ARs in the Navy Earth System Prediction Capability (Navy ESPC, Barton et al. 2020), a global coupled forecast system planned for operational implementation in 2020. The Navy ESPC is composed of the Navy Global Environmental Model (NAVGEM) atmospheric model coupled to the Global Ocean Forecast System (GOFS), which consists of the Hybrid Coordinate Ocean Model (HYCOM) and the Los Alamos Community Ice Code (CICE). Here we consider coupled forecasts with an atmospheric model resolution of approximately 37 km and an ocean and sea ice model resolutions of 1/4^o (coarser than the 1/12^o operational implementation for computational reasons). We examine IVT biases in Navy ESPC 30-day 7-member ensemble forecasts run once per week during 2017 and compare two ensemble formations. The control ensemble (CTL) experiment does not include a method to account for model uncertainty. In the second ensemble experiment, the Analysis Correction-based Additive Inflation (ACAI, Crawford et al. 2020; Bowler et al. 2017) method is incorporated to correct for model bias and model uncertainty. ERA5 reanalyses are used as verification. As IVT has both a moisture and wind component, we examine both separately. We also compute IVT forecast biases using analyzed winds and forecast moisture in one experiment, and analyzed moisture and forecast winds in another experiment, to determine which component dominates the IVT error. We find that that ACAI is very effective at reducing biases over the CTL experiment, such that by forecast day 10, the global average absolute value of the bias is reduced by over 30% for moisture, over 10% for low level winds, and over 18% for IVT. These reductions are substantially higher in some regions, particularly in the tropics. ACAI also reduces ensemble mean RMSE of these fields, but the impact is smaller. Wind biases are the dominant component in the IVT biases, but this is location dependent. Wind biases are particularly prominent in regions where there is a positive IVT bias, while both wind and moisture biases contribute to the negative IVT bias that occurs over many ocean regions. Wind biases also dominate ensemble mean RMSE of IVT. We will also present results examining the impact of ACAI on AR forecast biases and probabilistic forecast skill.