Impact of Assimilating AR Reconnaissance Dropsonde Observations on Landfalling Atmospheric River Forecasts

Atmospheric rivers (ARs) bring moisture across the northeastern Pacific Ocean and into the western United States, at times leading to extreme weather and water events. Improving our ability to predict them can be beneficial to the decision-making process related to the management of water supply and flood events. In order to improve the forecasts of ARs and their associated precipitations, a field campaign called AR Reconnaissance was initiated in Feb 2016. Dropsondes were deployed from the research aircraft to sample the AR core and other key features. The dropsondes collect high resolution profiles of temperature, pressure, humidity, and winds. This work focuses on the impact of assimilating the dropsonde on AR forecasts, by closely examining fifteen Intensive Observational Periods (IOPs) that took place during early 2016, 2018, and 2019.

Fifteen paired experiments are performed to evaluate the dropsonde impact by employing the Weather Research and Forecasting (WRF) model and the Community Gridpoint Statistical Interpolation (GSI) hybrid 4-dimensional (4DEnVar) data assimilation system. The “WithDROP” (“NoDROP”) runs assimilate all the conventional data (including the AMVs and GPS RO) with (without) AR Reconnaissance dropsonde data. The 6-day outputs for these two sets of experiments are compared to the ERA5 data. A case study of one IOP on February 3^rd, 2018 is analyzed in depth. It was found that the dropsonde data could reduce the forecast error for integrated water vapor (IVT) by about 25% in the first two days. There was an overall error reduction for IVT when dropsonde data were assimilated in 12 IOPs. Significant and consistent error reduction for IVT were found from day 1-3 and day 6.