Satellite-based detection of deep convective clouds: the sensitivity of infrared methods, and implications for cloud climatology

Abstract. Reliable deep convective cloud (DCC) climatology relies heavily on accurate detection. Infrared-based algorithms play a critical role, as they are the only ones that can be applied to the 6.7 μm water vapour (WV) absorption band, and the 11 μm infrared (IR) window band. For over 40 years, the latter have been the only daytime/nighttime channels used in satellite cloud imaging. The study presents the first global validation of three, commonly-used DCC detection methods, which use brightness temperature (T_b) in WV and IR bands. These methods are: the infrared window method (IRW; T_b11), the brightness temperature difference method (BTD; T_b6.7−T_b11), and the temperature difference method with the tropopause method (TROPO; T_b11−T_tropo). All methods were applied to one year (2007) of Moderate Resolution Imaging Spectroradiometer (MODIS) observations, and validated against collocated CloudSat-CALIPSO lidar-radar cloud classifications. Results indicate that even with optimal parameter configurations, DCC detection accuracy remains moderate, and below 75 % (Cohen’s κ < 0.4) for all methods. Global accuracy ranged from 56.6 % (for TROPO) to 72.8 % (for BTD) using an optimal threshold of −2 K. Regionally, the BTD method performs best, with accuracy of 72.9 % over Europe, and 67.9 % over Africa. Misclassifications are common with clouds such as Nimbostratus and Altostratus (single-layer cloud regimes) and Cirrus and Altostratus (multi- layer cloud regimes). Overall, the BTD method slightly outperforms the others, while TROPO is least effective. Our study highlighted the high sensitivity of these methods to threshold selection. Even a ±1 K change in the threshold resulted in a 10– 40 % variance in DCC frequency. The latter finding is of particular importance for the construction of homogenous DCC datasets, whether as global mosaics, or as time series spanning multiple generations of satellite instruments.