Key points are not available for this paper at this time.
This paper describes a new method of estimating both tropical convective precipitation and stratiform precipitation (produced under the anvils of mature and decaying convective systems) from satellite infrared data. The method, denoted CST (Convective-stratiform Technique) locates, in an array of infrared data, all local minima in the brightness temperature field (Tmin. After an empirical screening to eliminate cirrus, these points are assumed to be convective centers. Rainrate and rain area are assigned to each minimum point as a function of its Tmin, based on one-dimensional cloud model results. A stratiform rain algorithm, using a brightness temperature threshold based on the mode temperature of thunderstorm anvils, completes the convective/stratiform rain estimation. Individual CST rain fields wore spatially most similar to the radar for young, isolated storms, and most dissimilar in capturing linear features such as squall lines. Some convective features were missed, while others (generally cirrus debris) were sometimes misrepresented as active convection. Stratiform estimates generally corresponded to the radar-derived 1 mm h−1 contour. The technique was tested for four south Florida cases during the second Florida Area Cumulus Experiment (FACE). Half-hourly estimates made in the FACE target area are verified against raingages and both unadjusted and gage-adjusted radar. When compared to three other infrared techniques applied to the same dataset, the CST had the lowest bias (−0.02 mm), lowest mean absolute difference (0.28 mm), lowest root mean square difference (0.39 mm), and lowest percent difference (41.2%) of any tested satellite technique. The evolution of the precipitation averaged over the FACE target (104 km2), was well represented by the CST, particularly in capturing peak rainfall and the transition early and overestimate late, when compared to the gage-adjusted radar. Area-averaged estimates were in agreement with radar-based analyses, and comprised 10%30% of the total rainfall.
Adler et al. (Fri,) studied this question.