Evaluation of AIRS Arctic Cloud Phase Classification against Combined CloudSat-CALIPSO Observations

The phase of Arctic clouds, or whether they are composed of liquid or ice or both, is a significant piece of information in our understanding of Arctic surface-atmosphere radiative coupling. Arctic cloud phase is complex, and the seasonal and spatial distributions of mixed phase clouds are still not well understood. To improve our knowledge of Arctic cloud phase, satellite measurements are needed. The CALIPSO lidar is considered the "gold standard" of cloud phase retrievals, but its observations are spatially limited along a narrow path. Passive remote sensors like AIRS have a much wider field of view (FOV) than CALIPSO. The current AIRS algorithm employs a series of multi-channel brightness temperature difference tests to determine cloud phase, which is particularly useful in the Arctic winter where sunlight is scarce. With 17 years of data collected so far, AIRS has potential to help improve our knowledge of Arctic cloud phase climatology. However, AIRS Arctic phase retrievals have never been validated using active sensors.

In this study, we evaluate AIRS cloud phase retrievals in the Arctic using combined CloudSat-CALIPSO (CCL) observations. Near-nadir AIRS FOVs and CCL FOVs within 60-83°N from 2007 to 2010 are collocated, where 15-19 CCL FOVs usually lie within a single AIRS FOV. The impacts of sub-AIRS FOV phase composition on AIRS phase classification are evaluated and so are the impacts of multiple cloud layers with different phases.

Preliminary results show that AIRS and CCL generally agree on the relative frequencies of occurrence of ice and liquid cloud scenes over Arctic sea ice and open ocean, respectively. AIRS scenes with unknown phase and CCL scenes identified as mixed phase clouds are prevalent. The frequency of occurrence of AIRS phase classification for a given set of 1-layer CCL FOVs with the same phase is not generally dependent on the surface type. AIRS classifies 75% of 1-layer CCL ice cloud scenes as ice clouds and the rest as unknown. For 1-layer CCL liquid cloud scenes, AIRS classifies 60% as liquid clouds and the rest as unknown. For all 1-layer CCL FOVs identified as mixed phase clouds, AIRS classifies 50% of them as liquid clouds, 40% as unknown, and 10% as ice clouds. The most common 2-layer CCL phase combinations are liquid above ice and mixed phase above ice, both of which AIRS tends to classify as unknown.