MODIS observations for the evaluation of clouds in global models This document provides a short description of observational data sets derived from MODIS and intended for the evaluation of clouds produced by large-scale models.

These files contain a subset of the standard MODIS Level-3 Monthly data that has been reformatted as netCDF 3.4 files and combined across the Terra (morning) and Aqua (afternoon) platforms to produce a single file for each month in which both platforms are available. The underlying MODIS data are presently the so-called “Collection 5.1,” as the file names indicate. As with the ISCCP data, the boundaries between high, middle, and low clouds are 440 and 680 hPa respectively.

Unlike most other satellite observations of clouds, MODIS first identifies pixels which are likely to contain clouds, but produces retrievals of cloud optical depth and cloud particle size only for that subset of pixels which are likely to be entirely cloud filled. The first population is summarized in the “Mask” estimates of cloud fraction and cloud top pressure, while the second is summarized in the “Retrieval” estimates of cloud fraction and most other variables.

Many further details are available in the reference.


- Pincus, R., S. Platnick, S. A. Ackerman, R. S. Hemler, and Robert J. Patrick Hofmann, 2011: Reconciling simulated and observed views of clouds: MODIS, ISCCP, and the limits of instrument simulators. Submitted to Journal of Climate, May, 2011. Available from

Reading the files:

The files are CF-compliant netCDF files and can be read with many standard tools. The names of individual data sets correspond exactly to the names of the outputs from the MODIS simulator, although in some cases two estimates of the same quantity are provided (“mask” and “retrieval” estimates as described above).


Please cite the reference paper and doi:10.1109/TGRS.2002.808226 if you publish results using these files.


Ftp data access: MODIS


Please contact Dr. Robert Pincus if you have any question or comment about MODIS dataset.