A trispectral technique is developed for detecting tropical upper tropospheric opaque (tau > 6) and nonopaque (tau < 6) ice clouds over ocean based on the brightness temperature differences between the MODIS 8.5 and 11 micron bands and between the 11 and 12 micron bands together with the MODIS detected cloud thermodynamic phase. The brightness temperature differences provide robust information for classifying ice clouds, as illustrated by the observations made by a lidar, a radar, and the MODIS Airborne Simulator over tropical ice anvil systems during the Cirrus Regional Study of Tropical Anvils and Cirrus Layers‐Florida Area Cirrus Experiment. The trispectral technique for detecting tropical upper tropospheric opaque and nonopaque ice clouds is developed based on the analysis of 1 year of data, including MODIS infrared brightness temperatures at 8.5, 11, and 12 micron bands, MODIS‐derived ice cloud optical thicknesses, and cloud top heights from CALIPSO and CloudSat over a region (140°E–180°E, 0°N–20°N) in the Western Pacific Warm Pool. The accuracy of the present trispectral technique is above 80%. A 27 July 2007 MODIS granule over the chosen region is used to verify the trispectral technique. It is found that the classification from the trispectral technique is consistent with a classification based directly on the MODIS ice cloud optical thicknesses. The effects of the variations in the MODIS viewing zenith angle on the detection are found to be negligible. The CALIPSO and CloudSat observations used to develop the classification are more sensitive than MODIS to the height and presence of optically thin cirrus. These differences in cloud heights were found to have a negligible impact on the final detection results.