Operational Assimilation of GOES Imager Water Vapor Channel at MSC Nicolas Wagneur and Louis Garand Meteorological Service of Canada, Dorval, Québec, Canada Abstract The assimilation of GOES imager radiance data at 6.7 micron (channel 3) was made operational in the Canadian Meteorological Center (CMC) global 3Dvar analysis system June 19th 2003. GOES imager channel 3 is sensible to moisture in the atmosphere at 500 to 250 mb. The MSCFAST physical radiative transfer model is used as forward operator (Garand et al. 1999). This implementation of GOES water vapor radiances was made in the same time as ATOVS AMSU B radiance data were added in the assimilation system. Impact on water vapor fields of these two data sets assimilated separately is quite similar for overlapping areas in the 500 to 250 mb layer. The improvement of adding independently GOES imager channel 3 on moisture analysis and forecasts is shown. Positive impact on the quality of forecasted moisture fields is present up to 48 hours. Quantitative precipitation forecasts improvement over North America is also noted. The monitoring suite of data treatment shows stable innovations statistics. Furthermore geographical maps of monthly averages of innovations show no viewing angle bias problems yet significant local monthly mean corrections. These data are to be implemented in the regional system as the new main frame computer becomes operational. GOES radiance data are readily available for the CMC’s regional system despite a short cutoff time. This is not necessarily the case for AMSU B data where large data voids are noted. Similarity of impacts of GOES imager water vapor and AMSU B radiances In order to evaluate the effect of a new data type on the quality of the assimilation system an experimental cycle is launched with the addition of this data. Here two experiments were run with each the addition of GOES channel or ATOVS AMSU B data only. Maps of 6 weeks averaged differences to the control of dew point depression (T-Td) analysis are shown for both experiments (see Fig. 1). Theses averaged “corrections” to the control are remarkably similar. This is a strong validation of both data sets as they show almost the same behavior when tested independently. Impact on analysis and forecast quality An objective evaluation of both cycles was done. The improvements noted were mainly on moisture fields, but mass fields were also somewhat better. Figure 2 shows objective scores of the forecast issued from the cycle with added GOES data using radiosonde measurements from South America as reference observations. Even if the analysis does not agree as well with radiosondes, the 6 hour forecasted moisture is greatly improved over the control run and is still somewhat present up to 48 hours. The scores are more neutral over North America. However a significant gain over that area is noted in precipitation scores when goes radiances are added (see Fig. 3).