Statistical downscaling for GFS precipitation forecast over Sahel region based on Meteosat Second Generation rainfall estimates
Francesca Guarnieri, National Research Council, Firenze, Italy; and M. Pasqui, L. Genesio, S. Melani, and P. Vignaroli
A reliable and detailed rainfall forecast system, during the monsoon season, over West Africa, plays an important role for managing agricultural systems. Many large scale forecasting systems have been developed and distributed such as Global Forecast System (GFS). As any global model it capability to represent correctly deep convection is very low. In order to improve the rainfall forecast skill produced GFS, we developed an operational statistical downscaling technique to better represent the spatial distribution. Such downscaling technique is based on rainfall satellite estimates coming from Meteosat Second Generation (MSG) and SSM/I through the Turk et al. algorithm. It merges the advantages of space resolution and refresh time of MSG data with rainfall estimation sensitivity of SSM/I MW observations. This procedure is based on the Turk's algorithm and it essentially starts building a kind of Look-Up Table, in order to match brightness temperature measured by MSG with corresponding (i.e. space-time co-located) rain rates retrieved by SSM/I MW observations. For downscaling the GFS rainfall (at 1 degree of spatial resolution), each day, we collect the precipitation forecasts for the following 72 hours and the satellite rainfall estimates for the previous 72 hours (upscaled to 0.1 degree of spatial resolution). In this way we integrate the forecast information, supplied by GFS, for the following three days, with the more detailed spatial information given by satellite rainfall estimates in the previous 3 days. In the hypothesis of a stationary rainfall regime, especially during July and August, such strategy could be adopted. On the basis of MSG rainfall estimates, a mask is built and the GFS forecasted rainfall is projected over that in order to introduce the small scale rainfall spatial variability. To build the mask, a weight has been assigned to each pixel of the rainfall estimates contained in a single GFS grid point. This weight is computed as the ratio between the rainfall given in that pixel and the cumulated rainfall occurred in all the pixels contained in that GFS grid point. The method consistency has been evaluated through classical statistical skill scores, with respect to the rainfall estimates obtained by daily global CMORPH provided by NOAA for a whole season with evidence of its reliability.
Poster Session 4, Operational Products
Wednesday, 1 February 2006, 2:30 PM-2:30 PM, Exhibit Hall A2
Previous paper Next paper
Browse or search entire meeting
AMS Home Page