Improving Atmospheric River Forecasts: the Role of Data Assimilation and New Observations

Atmospheric rivers (ARs) refer to elongated corridors of enhanced moisture transport, transferring water vapor from low latitudes to high latitudes. ARs are closely associated with weather and climate extremes in many populated regions globally. Landfalling ARs can produce up to 50% of the annual precipitation for the Western United States (U.S.) and are the cause of major flooding events. Accurate forecasts of landfalling AR events can enhance water management strategies and alleviate the risk of flooding. However, the forecasts of ARs, in particular, the landfall events, are challenging in numerical weather prediction models. Several constraints affect the accuracy of landfalling AR forecasts across the Western U.S.: 1) Sparse observations upstream over the Pacific Ocean; 2) Imperfect model physics and dynamics in representing water vapor budget and transport; 3) Absence of effective data assimilation (DA) techniques tailored to capture AR characteristics, including filamentary structures, sharp thermodynamical gradients, and their interactions with moisture sources from lower latitudes. The objectives of this study are twofold: firstly, to assess the impacts of DA methods on AR forecasts by employing different DA approaches in representative AR cases; secondly, to apply the optimized DA framework to gauge the influence of incorporating new observations on forecast accuracy.

Nine high-impact AR storms that AR Recon sampled during 2016, 2018, and 2019 were selected as representative AR cases for this study. For each case, data assimilation experiments were conducted using the Weather Research and Forecasting (WRF) model with the Gridpoint Statistical Interpolation (GSI) system. The advantages and disadvantages of each GSI DA method (i.e., 3DVar, the hybrid 3DEnVar and 4DEnVar) in assimilating the non-radiance observations have been investigated. Results show that DA methods have a significant impact on the AR analysis and subsequent forecasts. The 4DEnVar method typically changes the integrated vapor transport (IVT) amplitude by 10-15% compared to 3DEnVar at the analysis time. Positive impacts on analysis tend to be in the core and north side of the AR and in the sharp gradients (e.g., dry intrusion). They are closely associated with the time-evolving index of the background flow within the DA window, demonstrating that 4DEnVar is necessary for ARs that evolve rapidly.

The benefits of hybrid 4DEnVar method on the forecasts of IVT are maximized during the 6–48 h compared to 3DEnVar. The average skill of the 4DEnVar experiments is slightly higher during 48–108 h than that of 3DEnVar. However, 3DEnVar shows small benefits in 120–144 h. The influence on the precipitation is generally consistent with that on IVT. The impacts arising from DA methods are also sensitive to the shape, orientation and locations of ARs. An exceptional AR case that made landfall over Central California around the Valentine Day of 2019 shows the most robust skill improvement from 4DEnVar in all lead times. Positive impacts from 3DEnVar appear more in the two meridional AR cases that exhibited moisture advection directly from tropics. The results will also compare data impacts of new observations available for oceanic ARs, including the additional drifters and unique dropsondes taken during AR Reconnaissance (Recon) and GOES-West atmospheric wind vector (AMV). This study seeks to answer how different data assimilation approaches can impact AR initial conditions and forecasts, contributing to better understanding of the landfall AR physics and dynamics and improving its associated precipitation forecast in the WRF model.