Comparison of cloud-top height retrievals from ground-based 35 GHz MMCR and GMS-5 satellite observations at ARM TWP Manus site

Retrievals of cloud-top heights from the ARM 35 GHz Millimeter Wave Cloud Radar (MMCR) located on Manus Island are compared to those from the GMS-5 satellite as a means to evaluate the accuracy of both MMCR and GMS-5 retrievals, as well as to ascertain their limitations. Comparisons are carried out for retrievals of both single-layer and multilayer clouds as seen by radar, but only for satellite-detected clouds with 100% amount within a 0.3 x 0.3degrees domain centered at the ARM site of one cloud type (i.e., low, middle, or high). Mean differences, with 95% confidence limits, between radar- and satellite-retrieved cloud-top heights (i.e., radar-retrieved cloud-top heights-satellite-retrieved cloud-top heights) are 0.3 +/- 0.3 km for single-layer clouds and -0.7 +/- 0.3 km for multilayer clouds. The study reveals that for thick clouds (i.e., cloud base less than or equal to 1 km and cloud thickness greater than or equal to 10 km), which are representative of convective towers with no/light precipitation as well as thick anvil clouds, retrievals from the MMCR agree well with those from satellite with mean differences of 0.0 +/- 0.4 and -0.2 +/- 0.3 km for single-layer and multilayer clouds, respectively. For clouds of lesser thickness, mean cloud-top heights derived from satellite are lower than those derived from radar by as much as 2.0 km. It is also shown that for convective clouds with heavy precipitation, MMCR retrievals underestimate the cloud-top heights significantly. (C) 2004 Elsevier B.V. All rights reserved.