A Case Study of the Variational Assimilation of GPS Zenith Delay Observations into a Mesoscale Model

Abstract
Results from a case study of the four-dimensional variational assimilation of total zenith delay (TZD) observations from a dense global positioning system (GPS) network into the Pennsylvania State University–National Center for Atmospheric Research Fifth-Generation Mesoscale Model are reported. TZD is made up of the rescaled pressure and precipitable water at the site of the GPS receiver. Profiler-wind and radio acoustic sounding system (RASS) virtual temperature observations are also included in the assimilation experiments. Four experiments are performed. The study targets the 12-h period from 0000 to 1200 UTC 6 December 1997, characterized by the passage of a frontal system that produced intense rainfall over southern California. Forecasts prior to data assimilation underestimate the observed 6- and 12-h accumulated rainfall for most of the domain. The (sole) assimilation of TZD observations is found to have a small but beneficial impact on the short-range precipitation forecast. Measured against the control forecast, area-mean improvements of up to 33.15% and 25.08% are found in the 6- and 12-h accumulated rainfall in Los Angeles County. The inclusion of profiler-wind observations is found to have a significant impact on the model precipitation, with improvements in the 6- and 12-h accumulated precipitation as high as 88.26% and 32.53%, respectively. However, these increments are noticeably reduced when the TZD data are excluded from the assimilation experiments. Further improvements are achieved when the TZD and profiler-wind data are assimilated along with the RASS virtual temperature data. Increases of up to 93.21% and 50.58% are found in the 6- and 12-h accumulated precipitation, respectively. Because the virtual temperature also contains information on the three-dimensional moisture field, these findings point to the potential benefit that may result from the future assimilation of GPS slant-path delay data.