Abstract
The impact of assimilating data from the 2011 Winter Storm Reconnaissance (WSR) program on numerical weather forecasts was assessed. Parallel sets of analyses and deterministic 120-h numerical forecasts were generated using the ECMWF four-dimensional variational data assimilation (4D-Var) and Integrated Forecast System. One set of analyses was generated with all of the normally assimilated data plus WSR targeted dropwindsonde data, the other with only the normally assimilated data. Forecasts were then generated from the two analyses. The comparison covered the period from 10 January to 28 March 2011, during which 98 flights and 776 total dropwindsondes were deployed from four different air bases in the Pacific basin and the United States. The dropwindsondes were deployed in situations where guidance indicated the potential for high-impact weather and/or the potential for large subsequent forecast errors. Downstream target verification regions where the high-impact weather was expected were identified for each case. Forecast errors around the target verification regions were evaluated using an approximation to the total-energy norm. Precipitation forecasts were also evaluated over the contiguous United States using the equitable threat score and bias.
Forecast impacts were generally neutral and thus smaller than reported in previous studies, most from over a decade ago, perhaps because of the improved forecast and assimilation system and the somewhat denser observation network. Target areas may also have been undersampled in this study. The neutral results from 2011 suggest that it may be more beneficial to explore other targeted observation concepts for the midlatitudes, such as assimilation of a denser set of cloud-drift winds and radiance data in dynamically sensitive regions.
