Cirrus Cloud-Top Height Estimation Using Geostationary Satellite Split-Window Measurements Trained with CALIPSO and CloudSat Data

We developed a method for the estimation of cloud-top height using only split-window channels of infrared observations from geostationary satellites. Direct observations of almost the entire geometrical cloud profile and of cloud tops have been available from the Cloud Profiling Radar (CPR) onboard CloudSat and the Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) onboard CALIPSO. However, these observations are limited to a small footprint near the nadir of the satellites and they are temporally very sporadic. It is not possible to obtain information on any cloud systems beyond the nadir footprint or to trace specific cloud systems continuously using such direct observations alone. Observations performed by geostationary satellites have high temporal resolution and wide spatial coverage. Although cloud-top height is not observed directly by such platforms, it is estimated from infrared (IR) and/or visible radiances. The split-window method is based on the premise that difference of two brightness temperatures at different wavelengths in the atmospheric window range is large for semi-transparent clouds like cirriform clouds (Inoue, 1985, J. Geophys. Res.). The original version of the product (Hamada and Nishi 2010, J. Appl. Meteor. Climatol.) was made using the lookup tables (LUTs) constructed based on regression with direct observations of cloud top height from CPR. The dataset was effective for monitoring the activity of mesoscale convection and dense anvil clouds because the estimated cloud tops are close to the actual tops in the cases of precipitating clouds and/or the surrounding dense stratiform clouds. However, cirriform clouds generally have lower optical depths but may be geometrically thick; thus, the cloud-top heights observed by the CPR could be considerably different from the actual cloud-top heights, which can cause bias in the LUTs. In this study, LUTs were constructed based on regression with direct observations of cloud top height from CALIOP. By using CALIOP data, we succeeded in reducing the underestimation of the height of cirriform clouds. Although CALIOP can detect optically thin clouds around the tropopause, their top heights unfortunately cannot be estimated well using split-window observations. By defining the altitude at which the optical depth from the top had a specified value Τ_min(=0.2) as the cloud-top height, we could create a practical LUT. In the LUT, the underestimation of the heights of cirriform cloud was corrected substantially, while reducing the effect of the low sensitivity of split-window observations to thin tropopause cloud. The detail of the method is shown in Nishi et al. (2017, SOLA). Our products are accessible from http://database.rish.kyoto-u.ac.jp/arch/ctop/index_e.html.