The importance of accurate cloud parameterization is well established. With decades of atmospheric measurements from space, there is no shortage of data. However, differences in instrument domain definitions (e.g. equatorial crossing time, swath width, footprint size, repeat cycle, etc.) introduce complexities in multi-instrument comparisons of retrieval results or algorithms. So much so that efforts in long-term trend analysis are often limited to retrievals from single instruments or groups of instruments with similar domain definitions.

We introduce a statistical (or gridding) algorithm with which to project data from their unique instrument domain to a uniform space-time domain. Once data are in the same domain, inter-comparisons are simplified. The algorithm has two phases. First, the snap-to-grid routine is employed to project cloud retrievals to an equal-angle latitude-longitude grid (the size of which can vary depending on the application). Snap-to-grid is a simple method for indexing geographic data into nearest neighbor clusters without any distance calculation. Second, a weighted monthly mean is calculated as the average of daily averages for each grid cell to neutralize differences in temporal sampling.

The sensitivity of the gridding algorithm is demonstrated for a month (1-31 August 2009) of Level 2 Terra/MODIS (Moderate Resolution Imaging Spectroradiometer) cloud top pressure (CTP) retrievals (MOD06, collection 5). Analysis is limited to high cloud retrievals (CTP < 440 hPa) from near nadir measurements (instrument viewing angle less than 32º). Algorithm sensitivity to grid size (0.5º versus 1.0º), diurnal definition (variations in sun zenith angle thresholds), and daily mean definition (variations in minimum number of observations per grid cell per day) is tested.

Lastly, we demonstrate the usefulness of the gridding algorithm by projecting a month (1-31 August 2009) of high CTP near-nadir retrievals from two polar-orbiting imagers, MODIS and AVHRR (Advanced Very High Resolution Radiometer), and two polar-orbiting sounders, AIRS (Atmospheric Infrared Sounder) and IASI (Infrared Atmospheric Sounding Interferometer), to a uniform 1.0º global grid. With the data projected onto a single grid, differences in CTP retrieval algorithms are highlighted. These are discussed in some detail. We conclude that the gridding algorithm greatly facilitates the inter-comparison of CTP retrieval products and algorithms. It is sensitive to the definition of space (grid size) but robust to the definition of time (diurnal and daily mean thresholds). Its simplicity lends it transparency in understanding and implementation thereby making it useful for both research and operational use.