An experiment in the use of geosynchronous satellite data for cloud detection and summer convective rainfall estimation.
Seymour, M.L., 1998.
M.S. Thesis, Louisiana State University, p. 60.
This thesis presents first steps in developing a rainfall retrieval algorithm that combines the high temporal and spatial resolution of the GOES-8 geostationary satellite and available rain guage measurements. Using a discriminant analysis technique, pixels of the training sample were classified as either "cloud" or "other." The results were then applied to validation data sets. Analyses of data from 17 June 1997 demonstrated clear separation of the data into two classes. Reflectance values peaked at 36% and 84%, which are consistent with the expected values for land and cloud peaks respectively. A distinct peak in temperature at 210 K is attributed to cloud top measurements. Data in the validation set that were categorized as cloud were then used to calibrate and test the rainfall estimation technique. No rain was likely for cloud tops warmer than 240.6 K. Using stepwise regression, a two variable model was devloped. To estimate precipitation in millimeters, the normalized visible and infrared data were entered into the equation: PRECIP = 6.87141173 + 0.00753048 * NVIS - 0.02929830 * IR. From discriminant analysis, equations were derived to distinguish raining and non-raining clouds. The bivariate frequency distribution from the discriminant analysis failed to show clear distinction between thw two classes. Most of the raining clouds occurred at temperatures between 200 K and 230 K, with a substantial peak at 210 K. In general, raining clouds occurred most often at reflectance between 25% and 98% and temperatures between 200 K and 220 K. Only 19% of the total data set was misclassified as raining, while 77% of the total set was successfully classified as either raining or non-raining. The determination of raining cloud area and estimation of precipitation rates are particularly important parameters as early warning mechanisms for approaching areas of potentially severe weather. Thus, more research is justified.