Identification of smoke and polluted clouds based on polarized satellite images

Abstract

Smoke aerosol plays an important role in climate change and the Earth's environment. Using multispectral, multiangle, polarized data for detecting small particles has its unique advantages. In this study, two algorithms were developed to detect smoke aerosols and smoke-polluted clouds (SPC). Multidirectional polarized reflectance (Rp) of different aerosol types was simulated by the UNL-VRTM (Unified Linearized Vector Radiative Transfer Model). Compared to dust, urban, rural and oceanic aerosols, the Rp of smoke at 0.490 µm, 0.670 µm and 0.865 µm increased with the aerosol optical depth (AOD), while other types of aerosols showed the opposite trend. Based on the polarized characteristics of the different aerosol types at three single channels, we found that ratios between polarized reflectances Rp of smoke were more distinct from Rp ratios of other aerosol types for any aerosol optical depth. The polarized reflectance ratios at wavelengths of 0.490 µm, 0.670 µm and 0.865 µm were chosen to detect smoke. We analyzed the radiative characteristics of SPC and smoke-free clouds. In contrast to smoke-free clouds, the simulated total reflectance of SPC increases with wavelength in the spectral range from the visible (VIS) to the near infrared (NIR) by the UNL-VRTM. Additionally, Rp of SPC is enlarged compared to smoke-free clouds in the scattering angle range of 80°-120° at the wavelength 0.865 µm. Therefore, two slopes, the slope of the reflectance in the VIS to NIR spectrum and the slope of the Rp at the scattering angle range from 80° to 120°, were designed to identify polluted clouds.Finally, the results of the two proposed algorithms were verified in different biomass burning areas by the products of Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) and Moderate Resolution Imaging Spectroradiometer (MODIS), demonstrating that smoke and SPC can be well detected. In addition, the proposed algorithm can be applied to a polarized image with channels at 0.490 µm, 0.670 µm and 0.865 µm such as the Chinese Gaofen-5 (or GF-5) satellite.