10th Conference on Satellite Meteorology and Oceanography

P2.12

Verification of the Operational GOES Infrared Rainfall Estimation Technique over the Upper Midwest

Dan A. Baumgardt, NOAA/NWS, La Crosse, WI; and G. R. Lussky and A. M. Elfessi

During 1998, the National Weather Service (NWS) in La Crosse, Wisconsin participated in a collaborative project with the National Environmental Satellite Data and Information Service (NESDIS), the Northcentral River Forecast Center (NCRFC), and the University of Wisconsin-La Crosse (UW-L) to assess the accuracy of the GOES infrared rainfall estimation algorithm over the Upper Midwest. The GOES infrared rainfall estimation algorithm is also known as the auto-estimator and used exclusively by NESDIS to support NWS hydrologic warning operations (Vicente et. al. 1998). The auto-estimator output was analyzed over a domain approximately 300 km square centered near La Crosse, Wisconsin. Specifically, 6-hour and 24-hour auto-estimator products generated by NESDIS for any rainfall between March and October in 1998 were compared directly with corresponding gage data collected by the NCRFC for the valid time frame. All gage and GOES auto-estimator data points which both reported values of zero were removed to not bias the algorithm towards a perfect forecast. Thus, this research is for "rain days" only. Using latitude and longitude coordinates, the 24-hour rainfall totals were compared for the gage and auto-estimator points, numbering over 4400. These comparisons were then classified by observed total precipitation amount (<0.25", 0.25<1.00", > 1.00", and total) and statistically analyzed. Further classification and comparison of the gage versus auto-estimator performance was done using 0-6 km ambient wind shear and warm versus cold cloud-top environments using observed rawinsonde data.

A statistical analysis of the 24-hour gage data versus the auto-estimator output for the entire data set, for all precipitation classes, yielded a favorable mean error of only 0.19". The GOES auto-estimator was overestimating by only 0.19" overall. However, the standard deviation was quite large (0.80"). When the entire data set was broken into precipitation classes, the mean error (standard deviation) was an underestimate of 0.28 (0.49"), an underestimate of 0.12 (0.87"), and an overestimate of 0.002 (1.39") for the less than 0.25", 0.25<1.00", and greater than 1.00" categories, respectively. Warm cloud-top environments yielded a mean underestimation of rainfall (0.02") however over 50% of the cases were in the lowest precipitation class. The highest precipitation class (>1.00") yielded a mean underestimation of 0.37" with a large standard deviation of 1.04". 75% of the data points in the greater than 1.00" class were underestimates with a mean underestimate error of 0.87". This confirms previous findings of Vicente et. al. (1998) where warm cloud-top precipitation was under forecast by the GOES auto-estimator. 0-6 km wind shear indicated high (low) shear environments caused the GOES algorithm to overestimate (underestimate) rainfall totals. This error increased with increasing observed precipitation amount.

Poster Session 2, OPERATIONAL APPLICATIONS OF SATELLITE OBSERVATIONS: Part IV
Monday, 10 January 2000, 3:00 PM-5:00 PM

Previous paper  Next paper

Browse or search entire meeting

AMS Home Page