The work to be presented describes the results of an experiment utilizing the EDAS to study the 00-hr sensitivity and 24-hr forecast impact of five remotely sensed and five in-situ data types during 11-day periods for three seasons (winter, spring and summer). Data availability precluded an autumnal case. The five satellite data types are SSM/I vertically integrated precipitable water, GOES three layer clear air precipitable water, TOVS temperature profiles down to cloud top, GOES infrared cloud drift winds, and GOES water vapor cloud top winds. The five in-situ data types are rawinsonde temperatures and moistures, rawinsonde winds, aircraft temperatures, aircraft winds and land surface observations. The three 11-day periods are 13-23 December 1998, 10-20 April 1999, and 13-23 July 1999. Sensitivity and forecast impact statistics are diagnosed for standard meteorological fields on levels reaching from near the earth's surface to jet stream level.
When considering the entire horizontal extent of the EDAS domain, the cumulative three season results demonstrate that a modest positive forecast impact is achieved from all ten data types. Rawinsonde temperature and moisture observations and GOES infrared cloud drift wind observations have the largest positive impact season to season, while SSM/I and GOES PW observations provide the largest positive contribution in July.
Diagnostics within a domain covering only the continental United States indicate that both rawinsonde data types are largest season to season. GOES infrared cloud drift wind observations have the next largest season to season contribution. However, as with the entire domain statistics, GOES three layer PW data provides the largest impact during July.
In conclusion, this study provides one of the first detailed assessments on the impact of remotely sensed versus in-situ data in an operational NWP model for an extended length of time. It stands to reason that improvements in forecast accuracy related to the available data are only going to be made if a more complete understanding of how to utilize existing and future data types is developed.