Comparison of hybrid 4DEnVar and 3DEnVar data assimilation methods for atmospheric river forecasts in the WRF model

Landfalling Atmospheric Rivers (ARs) can bring large environmental and societal impacts. They can produce between 30-50% of the annual precipitation for the Western United States (U.S.) and are cause of the major flooding events. Accurate forecasts of a landfalling AR can improve water management decisions and reduce the flooding risks. However, the forecasts of ARs, in particular, the landfall events, are challenging in numerical models. For example, the landfall position error is on the order of +/- 400 km at 3 days lead time in NCEP GFS model. There are three limiting factors for the landfall AR forecast over the Western U.S.: 1) the relatively sparse observations upstream over the Pacific Ocean; 2) the imperfect model physics and dynamics in resolving water vapor budget and vapor transport; 3) a lack of effective data assimilation methods tuned for representing the characteristics of an AR such as its filament features and the connection with low-latitude moisture source. Of them, the optimized data assimilation strategy is of paramount importance because it can take the best advantage of the available upstream observations and compensate the model errors by correcting the model towards observations. This study compares two data assimilation approaches, the hybrid 3D-EnVar and 4D-EnVar, for the forecast of two different AR cases in the WRF model. The advantages and disadvantages of each method in assimilating the major data types over the ocean, e.g., the atmospheric wind vector (AMV) and the unique dropsonde observations taken during the 2016 AR Reconnaissance (Recon), will be investigated.

Two AR storms from the 2016 AR Recon are chosen as cases studies. Four experiments are conducted using the WRF model with the Gridpoint Statistical Interpolation (GSI) system, including 1) the ALL_hyb3d run assimilating the Recon dropsonde data and other conventional data using hybrid 3DEnVAR, 2) the ALL_hyb4d assimilating the same data in the All_hyb3d run but using hybrid 4DEnVAR, 3) the NoDROP_hyb3d run assimilating the same data as in All_hyb3d but removing Recon dropsonde data, and 4) the NoDROP_hyb3d same as NoDROP_hyb3d but using hybrid 4DEnVAR. Comparisons in the four runs show that the initial analysis in All_hyb4d has the least errors in integrated vapor transport (IVT) among the four runs when compared with the high-resolution ECMWF operational analysis. Hybrid 4DEnVAR can better assimilate the moisture field and the meridional wind field that the hybrid 3DEnVAR does in both cases. In case 2, the hybrid 4DEnVAR also improves the representation of temperature and zonal wind fields in initial conditions. The differences in case 1 and case 2 are mainly due to the different flow patterns in two cases represented by the persistence index. Case 1 has a well-organized AR with higher persistence index while a developing upstream AR and a decaying AR coexist in case 2 with lower persistence index. The forecast validation shows that the All_hyb4d and All_hyb3d runs reduce the IVT errors more during 6h-48h forecast for both cases than the other two runs, suggesting the runs with dropsonde data can significantly reduce moisture transport errors in an AR. In case 1, All_hyb4d run shows the least IVT error during 84h-120h among the four runs, indicating the additional advantage of using hybrid 4DEnVAR approach. In case 2, the errors of all four experiments are comparable beyond 48h. The precipitation forecast skill for case 1 show that the All-hyd4d run has the least bias and RMSE during 24-48h forecast among the four runs. The differences in the four runs for case 2 are smaller, with the All_hyb3d and All_hyb4d has slightly better forecast skill during 42-54h. Note that the landfall AR in case 2 originated from the tropical moisture plume, which was not sampled in AR Recon. The poor predictability beyond day 2 could be partly due to the lack of observations for the tropical moisture plume. This work seeks to answer the question of how different data assimilation approaches can impact AR initial conditions and forecasts, contributing to better understand the landfall AR physics and dynamics and improve its associated precipitation forecast in the WRF model.