The NCEP data is a 4km composite of the 11 micron channels of the 5 geostationary weather satellites. I combined the data from the 1430 and 1500 UTC mosaics for this image. There are a few characteristics of the NCEP data that make it not ideal to use for visualizations:
the mosaic only extends from 60 north latitude to 60 south latitude
high altitude clouds are moved to their correct geometric position leaving blank 'cloud shadow' pixels where they have been moved from
the calibrated brightness temperatures for each satellite's 11 micron channel don't always match so there are sometimes visible seams between the data from different satellites.
I filled in most of the cloud shadow pixels and got rid of one seam, but there are still some cloud shadows left.
I ordered all the MODIS Calibrated Radiances L1B 1km data near both poles for June 21, 2000 from the Goddard DAAC and reprojected band 31 data into the same projection as the NCEP data. I applied a latitude dependent stretch to the reprojected data and used a lookup table (previously calculated using a 2-d hist between representative NCEP and MODIS data) to convert the MODIS data to the same approximate gray-scale range as the NCEP data. Probably in the future it would be better to simply calculate the brightness temperature from the radiances. I needed to combine about 8 polar over-passes to get a full 360 degree coverage at 60 degrees latitude. Some clouds had moved a bit too much in that time to make a linear average between over-passes look nice so I selectively masked the over-passes in Photoshop to get a prettier result.
Finally, I blended the MODIS and NCEP data and cleaned up a few more problems in Photoshop. Here are some remaining problems:
Ice-covered Greenland and Antarctica are white since they are as cold as the clouds. Since this is MODIS data at the high latitudes, I could possibly fix this problem using a MODIS cloud product.
Northern latitude waters are cold enough and the lands are warm enough that there are distinct land-water boundaries and the water is too white. Once again, this problem could be fixed using a MODIS cloud product.
The marine stratus west of the Americas is too dark.
The texture is in a equidistant cylindrical projection centered on 180 degrees longitude and 360/9896 degrees per pixel scaling.
If you fix any of these problems or make any significant improvements to this texture, I'd like a copy back.
Also, I'm not sure if MODIS/Terra and NCEP need credits if a work using this texture is published.
Here are a few images of the global texture mapped onto a sphere and rendered at 4096x4096, I only show the North and South poles, since that is where the most mapping and temperature artifacts are.
And a few zooms on the pole, at this point it is entirely MODIS data. The artifacts are mostly from the texture mapping and rendering in MAYA; they could be minimized by increasing tesselation on the sphere, but if you really need to get this close to the pole it would make more sense to swap in a polar centered texture. Maybe I'll add that as a future refinement:
Make matching North Pole and Soth Pole centered textures.
