Abstract
In this paper, we have examined the possibility of minimizing the number of geostationary Very High Resolution Radiometer (VHRR) images required for the estimation of rainfall on large time and space scales using the Arkin's approach. In the selection of appropriate images we are guided by our knowledge of the pattern of diurnal variability of cloudiness/rainfall over the region of interest. For the present work, INSAT-VHRR thermal band images over the Indian region for the month of June 1986 are utilized. Monthly average brightness temperatures (Tb) over 2·5° by 2·5° regions, derived from afternoon (0900 UTC) and post-midnight (2100UTC) INSAT-VHRR thermal infrared band images, separately and in conjunction, have been compared with the Arkin et at. monthly average rainfall based on 3–hourly INSAT images, as well as with ground based measurements.. The analysis indicates that even one image taken at 0900 UTC daily is able to locate the regions of high convection almost as well as depicted by Arkin's analysis based on 3–hourly images. Inclusion of 2100 UTC images results in marginal improvement in the spatial distribution of rainfall. It is also observed that the present results based on 0900 and 2100 UTC VHRR data are somewhat better correlated with groundbased rainfall measurements than the results from Arkin et al.