Evaluation of satellite‐based upper troposphere cloud top height retrievals in multilayer cloud conditions during TC4

Upper troposphere cloud top heights (CTHs), restricted to cloud top pressures (CTPs) < 500 hPa, inferred using four satellite retrieval methods applied to Twelfth Geostationary Operational Environmental Satellite (GOES‐12) data are evaluated using measurements during the July–August 2007 Tropical Composition, Cloud and Climate Coupling Experiment (TC4). The four methods are the single‐layer CO2‐absorption technique (SCO2AT), a modified CO2‐absorption technique (MCO2AT) developed for improving both single‐layered and multilayered cloud retrievals, a standard version of the Visible Infrared Solar‐infrared Split‐window Technique (old VISST), and a new version of VISST (new VISST) recently developed to improve cloud property retrievals. They are evaluated by comparing with ER‐2 aircraft‐based Cloud Physics Lidar (CPL) data taken during 9 days having extensive upper troposphere cirrus, anvil, and convective clouds. Compared to the 89% coverage by upper tropospheric clouds detected by the CPL, the SCO2AT, MCO2AT, old VISST, and new VISST retrieved CTPs < 500 hPa in 76, 76, 69, and 74% of the matched pixels, respectively. Most of the differences are due to subvisible and optically thin cirrus clouds occurring near the tropopause that were detected only by the CPL. The mean upper tropospheric CTHs for the 9 days are 14.2 (±2.1) km from the CPL and 10.7 (±2.1), 12.1 (±1.6), 9.7 (±2.9), and 11.4 (±2.8) km from the SCO2AT, MCO2AT, old VISST, and new VISST, respectively. Compared to the CPL, the MCO2AT CTHs had the smallest mean biases for semitransparent high clouds in both single‐layered and multilayered situations whereas the new VISST CTHs had the smallest mean biases when upper clouds were opaque and optically thick. The biases for all techniques increased with increasing numbers of cloud layers. The transparency of the upper layer clouds tends to increase with the numbers of cloud layers.