Optimizing Dropwindsonde Levels for Data Assimilation

Dropwindsonde observations have traditionally been assimilated into numerical models from a subset of the data known as the Tempdrop message. These data include standard meteorological level data and the local extrema between those levels in a compact text format which can be easily transmitted directly from the reconnaissance aircraft. Since 2018 a high resolution binary form of the data, BUFR, is also available for transmission. The BUFR file contains the entire sounding parsed at 1 second time intervals. Unlike the Tempdrop messages, BUFR files provide time and location information for each observation, an important feature for tropical cyclone soundings. However it may not be feasible or even optimal to assimilate the full resolution data since the ~1hPa vertical resolution of the BUFR file is far greater than even advanced numerical models such as the Hurricane Weather and Forecast model (HWRF), which runs with a maximum of 75 vertical levels. In order to provide guidance for use of the BUFR files, we conduct a set of observing system experiments (OSE’s) using the Hurricane Ensemble Data Assimilation System (HEDAS) within HWRF, testing various means of sub-setting the data. The dropwindsonde data is assimilated either as Tempdrop, with time and location information, at full resolution, or as “super-obs” centered at height, pressure or model equivalent levels. The resulting analyses and forecasts are then compared for the cases of Hurricanes Michael (2018), Maria (2017) and Tropical Storm Karl (2016).