Identifying precipitating clouds in Greece using multispectral infrared Meteosat Second Generation satellite data

Abstract

The present study aimed to investigate the potential of possible rain area delineation schemes based on the enhanced infrared spectral resolution of the Meteosat Second Generation–Spinning Enhanced Visible and Infrared Imager. The proposed schemes use the brightness temperature (BT) Τ10.8 along with the brightness temperature differences (BTDs) Τ10.8 − Τ12.1, Τ8.7 − Τ10.8, and Τ6.2 − Τ10.8 as spectral cloud parameters. Two different methods were used to develop the rain area delineation models. The first is a common threshold technique in the multispectral space of the spectral cloud parameters, and the second is an algorithm based on the probability of rain (PoR) for each pixel of the satellite data. Both schemes were trained using as rain information gauge data from 41 stations in Greece for 107 rainy days, covering a period of 1 year. As a result, one single-infrared model (TB10), three two-dimensional (BTD10–12, BTD8–10, and BTD6–10), and two multidimensional models (BTDall and PoR) were constructed and verified against an independent sample of rain gauge data for four daily precipitation events. It was found that the introduction of BTDs as additional information to a model works in improving the discrimination of rain from no-rain events compared with the single-infrared model BT10. During the training phase, BTDall exhibited the best performance among the threshold techniques, while the PoR model outperformed all the threshold techniques, producing scores slightly better than those of BTDall model. When verifying against the independent dataset, all models exhibited the same performance with that of the dependent dataset according to the ETS score but less skill according to the HK score. The proposed techniques, however, still perform better than the single-infrared technique but with different ranking; BTall performs best followed by PoR and BTD10–12. Finally, two case studies are presented to gain a visual impression of the performance of the new developed rain area delineation schemes, showing the effectiveness of BTDall in delineating the thicker part of cirrus clouds and of PoR in detecting rain from the low-level thick clouds.