The Impacts of Different Satellite Data on Rain Estimation Schemes

Abstract

Rain estimates for the Great Plains States were made for a one-month period, August 1979, using different combinations of satellite and other data. The data tested were as follows: 1) two satellite images per day without any other data, 2) four satellite images per day, 3) 24 images per day, 4) 24 images per day with hourly surface observations and two per day radiosonde soundings (excluding the 6 h raingage reports), 5) two images per day with the Service A 6 h raingage reports, 6) 24 images per day with the Service A raingage reports, and 7) an automatic rain estimate made from infrared temperatures without human intervention. Each method was applied to the same geographic area by the same meteorologists. Estimates produced from the seven data combinations were compared to a withheld data set of 538 hourly recording raingages. The rain estimates from all methods tested were very similar in their ability to locate rainfall and estimate the monthly patterns. The first two methods tested, using only satellite imagery at low-frequency sampling rates, gave slightly poorer skill scores than the more data-rich methods. Best scores were found for methods using the Service A raingage reports (Methods 5 and 6). The frequency of satellite imagery did not change the quality ofthe estimates when the raingages were included. The rain estimates made without the judgment of a meteorologist (Method 7) scored surprisingly close to the other methods tested. The additional effort of a meteorologist improved the rain estimates in all cases, but the level of improvement was small beyond that produced by a simple automated scheme.