Publications

Minnis, P; Hong, G; Sun-Mack, S; Smith, WL; Chen, Y; Miller, SD (2016). Estimating nocturnal opaque ice cloud optical depth from MODIS multispectral infrared radiances using a neural network method. JOURNAL OF GEOPHYSICAL RESEARCH-ATMOSPHERES, 121(9), 4907-4932.

Abstract
Retrieval of ice cloud properties using IR measurements has a distinct advantage over the visible and near-IR techniques by providing consistent monitoring regardless of solar illumination conditions. Historically, the IR bands at 3.7, 6.7, 11.0, and 12.0 mu m have been used to infer ice cloud parameters by various methods, but the reliable retrieval of ice cloud optical depth t is limited to nonopaque cirrus with tau < 8. The Ice Cloud Optical Depth from Infrared using a Neural network ( ICODIN) method is developed in this paper by training Moderate Resolution Imaging Spectroradiometer ( MODIS) radiances at 3.7, 6.7, 11.0, and 12.0 similar to m against CloudSat-estimated tau during the nighttime using 2 months of matched global data from 2007. An independent data set comprising observations from the same 2 months of 2008 was used to validate the ICODIN. One 4-channel and three 3-channel versions of the ICODIN were tested. The training and validation results show that IR channels can be used to estimate ice cloud tau up to 150 with correlations above 78% and 69% for all clouds and only opaque ice clouds, respectively. However, tau for the deepest clouds is still underestimated in many instances. The corresponding RMS differences relative to CloudSat are similar to 100 and similar to 72%. If the opaque clouds are properly identified with the IR methods, the RMS differences in the retrieved optical depths are similar to 62%. The 3.7 mu m channel appears to be most sensitive to optical depth changes but is constrained by poor precision at low temperatures. A method for estimating total optical depth is explored for estimation of cloudwater path in the future. Factors affecting the uncertainties and potential improvements are discussed. With improved techniques for discriminating between opaque and semitransparent ice clouds, the method can ultimately improve cloud property monitoring over the entire diurnal cycle.

DOI:
10.1002/2015JD024456

ISSN:
2169-897X