CIRRUS REMOVAL IN MULTISPECTRAL DATAWITHOUT 1.38μM SPECTRAL DATA
Keywords: Cirrus removal, haze removal, multispectral data, Landsat 8 OLI, Sentinel-2
Abstract. Cirrus is one of the most common artifacts in the remotely sensed optical data. Contrary to the low altitude (1-3 km) cloud the cirrus cloud (8-20 km) is semitransparent and the extinction (cirrus influence) of the upward reflected solar radiance can be compensated. The widely employed and almost ’de-facto’ method for cirrus compensation is based on the 1.38μm spectral channel measuring the upwelling radiance reflected by the cirrus cloud. The knowledge on the cirrus spatial distribution allows to estimate the per spectral channel cirrus attenuation and to compensate the spectral channels. A wide range of existing and expected sensors have no 1.38μm spectral channel. These sensors data can be corrected by the recently developed haze/cirrus removal method. The additive model of the estimated cirrus thickness map (CTM) is applicable for cirrus-conditioned extinction compensation. Numeric and statistic evaluation of the CTM-based cirrus removal on more than 80 Landsat-8 OLI and 30 Sentinel-2 scenes demonstrates a close agreement with the 1.38μm channel based cirrus removal.