Cloud Detection and Cloud Top Height Assignment over the Poles Using MISR and MODIS Level 1 and 2 Data Products

Many preliminary case study investigations of the visible and near-infrared angular radiances recorded by the Multi-angle Imaging Spectro-Radiometer (MISR) indicate that these radiances have sufficient information both to separate many types of clouds from snow and ice surfaces and to assign heights to those clouds that are detected. These early studies are extended by analyzing a large set of colocated MISR and MODIS Level 1 and 2 data products over the Antarctic and Arctic with emphasis on understanding the additional information content that the MISR angular radiances adds to the MODerate Resolution Imaging Spectroradiometer (MODIS) nadir spectral radiances. The approach adopted in the current study consists of applying both supervised and unsupervised machine learning algorithms, as well as more traditional statistical analyses, to the MISR and MODIS Level 1 radiances for clear- and cloudy-sky image segmentation followed by comparison of the segemented images with the MISR and MODIS Level 2 standard cloud data products. Results from these comparisons are used to create a best-estimate clear- and cloudy- sky polar product using combined MISR and MODIS data. The clear-sky regions of the combined product are subsequently used to validate and improve MISR stereo cloud top height retrievals by improving geolocation of the nine MISR cameras. The MISR stereo-derived heights are finally compared with the MODIS CO2-slicing heights, demonstrating how the two techniques are complimentary and lead to improved estimates of cloud top heights over polar regions.