Estimativa de precipitação mediante satélite: um estudo de aplicação ao Estado do Ceará.

Abstract

The present study investigates behavior and limitations of the CST technique ("Convective-Stratiform Technique": ADLER and NEGRI, 1988) when adapted to daily precipitation totals. Only hourly images of the infrared channel of the METEOSAT-4 for the State of Ceara (seasons: March 1991 and January 1992); Daily totals of at least 150 posts belonging to the Rainforest Network managed by FUNCEME (Cearense Foundation of Meteorology and Water Resources) provide the "terrestrial truth". Regressions obtained based on associated cold nuclei to one pixel showed that the original CST parameters do not allow reliable estimates of daily totals. The correlation coefficients obtained were less than 0.30. Possible causes of this result are discussed. The method does not lead reliable estimates but allow for acceptable delineation precipitation areas in the region. The estimation method was reformulated, considering spatial and textural characteristics of the images, as well as targets (areas) instead of isolated pixels. Bivariate frequency histograms (in relation to temperature and its laplacian) show the existence of two classes of nuclei, associated in principle of light precipitation (PF), with temperatures T2 between 240K and 270K, and intense (PI), with T1 temperatures below 240K. To characterize PF and PI targets, variables associated with the mean temperature deviation from two T1 and T2 thresholds were defined. The bivariate linear adjustment of the average daily precipitation with the accumulated values ​​of these variables increased the correlation to 0.6. The results suggest that even better adjustments could be obtained with polynomial functions. However, correlation coefficients of the order of 0.56 are already obtained by simple linear regression by defining an associated variable T1 only, between 240-250K. It is recommended to carry out regional studies, complemented with images of the VIS and water vapor channels, aiming at characterize the evolution of local systems. In any case, it is advisable to have a network that provides total regional rainfall schedules.