Improving winter storm forecasts with Observing System Simulation Experiments (OSSEs): Part 2, Evaluating a satellite gap with idealized and targeted dropsondes
Numerous satellites utilized in numerical weather prediction are operating beyond their nominal lifetime and their replacements are not yet operational. We investigate the impacts of a loss of US‐based microwave and infrared satellite data and the addition of dropsonde data on forecast skill by conducting Observing System Simulation Experiments (OSSEs) with the ECMWF T511 Nature Run and the NCEP GFS Model. Removing all US‐based microwave and infrared satellite data increases GFS analysis error, global forecast error, and forecast error during the first 36h of three winter storms that impact the USA. Data from Suomi‐NPP contributes roughly one‐third of the total satellite impacts. Assimilating “idealized” dropsondes (sampling over a large region of the Pacific/Arctic Oceans) significantly improves global forecasts and forecasts for all three storms. Assimilating targeted dropsonde flight paths using the Ensemble Transform Sensitivity method for 15 verification dates/locations for the three storms improves roughly 80% of forecasts relative to the control and 50% of forecasts relative to their corresponding experiments without dropsondes. However, removing satellite data degrades only 30% of targeted domain forecasts relative to the control. These results suggest that targeted dropsondes cannot compensate for a gap in satellite data regarding global average forecasts, but may be able to compensate for specific targeted storms. However, as with any study of specific weather events, results are variable and more cases are needed to conclude whether targeted observations – as well as satellite data – can be expected to improve forecasts of specific weather events.