An Investigation of the Failed Retrievals in the MODIS Cloud Product: Frequency and Potential Causes

In satellite-based passive remote sensing, the cloud optical thickness (τ) and droplet effective radius (re), which has a strong influence on cloud radiative effects, are derived simultaneously from reflectance measurements at two wavelengths, the so called bi-spectral method. Since biases in the τ and re of marine water clouds can lead errors in estimates of global radiative forcing, it is critical to identify the sources and magnitudes of errors in the retrieval. The look-up tables (LUTs) used in the bi-spectral method have upper and lower limits to τ and re. These limitations might be problematic for clouds having very large or very small radius droplets, and may be even worse for optically very thin clouds. Alternatively, retrievals near or outside the LUTs boundary might be indicative of retrieval artifacts. This study investigates the frequency of occurrence of the failed optical thickness and effective radius retrievals and their causes for Moderate Resolution Imaging Spectroradiometer (MODIS) onboard Aqua satellite together with Cloud Profiling Radar (CPR) observations on CloudSat satellite and Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) observations on CALIPSO satellite. First, a failed retrieval case study will be shown to identify the typical characteristics. We then investigate potential reasons for failed retrievals by demonstrating the dependency on the relevant parameters, such as horizontal homogeneity and reflectances. The difference between MODIS Collections 5.1 and 6 (newly released) will also be investigated. The CPR and CALIPSO lidar measurements collocated with MODIS can help to identify the cause of failed retrievals. Preliminary results show that a significant fraction of water clouds fails in the MODIS retrievals (about 23% for 2.1µm-derived retrievals and 14% for 3.7µm-derived retrievals in Collection 6).