Evaluation of the VIIRS Cloud Base Height Intermediate Product Retrieval Algorithm for Single-layer Water Clouds

Cloud base height (CBH) is an important factor for both aviation and climate research. Low-level water cloud (i.e. stratus) bases, in particular, can be detrimental to military and humanitarian operations, alike; especially in remote, data-sparse locations. Likewise, CBH plays a significant role in downward longwave emission in determining the surface radiation budget for climate research. One study found that a 100-millibar (mb) uncertainty in CBH at the 650-mb level leads to surface errors of approximately 5 W/m2. The biggest limitation for surface-based CBH retrieval using lidar is the lack of spatial coverage across the globe, therefore space-based retrieval by polar-orbiting satellites becomes essential, and both radar/lidar and visible/infrared methods have been employed. Radar and lidar have been combined to retrieve CBH using the synergistic observations of CloudSat's Cloud Profiling Radar (CPR) and the Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) instrument on-board the Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO). In 2009, evaluation of the CPR/CALIOP CBH retrieval algorithm demonstrated a clear bias, misclassifying low clouds as ground clutter in the lowest 500 meters (m) where CBH retrieval is most critical for aviation. In 2014, the Visible Infrared Imaging Radiometer Suite (VIIRS) CBH Environmental Data Record (EDR) was shown to perform best for single-layer water clouds, with a correlation of 0.811 when VIIRS cloud top height (CTH) was within specification. This previous validation of the CBH EDR has focused on a statistical relationship with the CPR/CALIOP product, with the lowest 1 kilometer (km) being ignored due to the ground clutter limitation. For this research, the VIIRS CBH Intermediate Product (IP) retrieval algorithm is being evaluated for single-layer water clouds at five Department of Energy Atmospheric Radiation Measurement (ARM) sites. The CBH IP provides pixel-level, horizontal resolution of approximately 750 m at nadir and 1.6 km at edge of scan (EOS), while the EDR is a lower-resolution, averaged product at a resolution of 6 km at both nadir and EOS. Ceilometer data from these ARM sites are used as CBH ground truth by matching Suomi National Polar-orbiting Partnership (NPP) overpasses with ceilometer measurement times from 1 June 2013 to 31 Oct 2015. Key components of the algorithm, including CTH and cloud optical thickness (COT), are also evaluated for accuracy. Two value-added products (VAPs), developed for deriving important cloud properties from ARM site measurements, are used as ground truth for these components: an algorithm that combines polarization lidar and millimeter-wave radar to determine CTH; and an algorithm that uses multi-filter rotating shadowband radiometer (MFRSR) irradiance values with Langley regression top-of-the-atmosphere (TOA) irradiance values to infer COT for optical depth values greater than approximately seven. Initial results for 20 daytime cases at the Lamont ARM site demonstrated a CBH correlation of 0.56, an average error of 4178 m, and a standard deviation (1-σ) of 6021 m. CTH had a correlation of 0.62, an average error of 5262 m, and a standard deviation of 6098 m. COT correlation was 0.91, average error was -15.44 (negative meaning that the truth value was larger than the VIIRS-calculated COT, on average), and standard deviation was 25.8. When VIIRS-derived CTH was replaced with ground truth CTH from the ARM VAP, CBH correlation improved to 0.94, average accuracy improved by 75% to 1059 m, and the variability decreased by 76% to 1438 m. This initial finding indicates that the largest source of error may be external to the VIIRS CBH algorithm, which uses parameterization of COT and effective particle size to estimate the cloud thickness. This trend is expected to continue as the dataset is expanded, and a robust error budget will be constructed from the results using sensitivity studies. These results will help to identify the largest sources of error in the algorithm, and will help focus future efforts on evaluation of the VIIRS CBH retrieval algorithm.