The easiest way I found to produce True Color composites is through HDFLook software. I have compared the composites with the mosaics provided by LANCE. The quality is fairly equivalent (with some extra twick from ImageMagik), although strip lines can be seen on the image (those are much less apparent on LANCE composites, perhaps because of a smother interpolation algorithm than the bilinear used by HDFLook).
But my concern is actually with the projection. From the projections available on the software, I have only used the "Linear" projection, which is equirectangular projection with the Earth's modeled as a sphere of radius 6370997 m (a and b parameters of the PROJ4 string are equal in the metadata retrieved by gdalinfo). I have found some small but consistent displacement of these images from those produced by LANCE, and also with Natural Earth's coastlines shapefile.
I have not tested other projections because it would require me extra computing time to use gdalwarp for reprojection. I have to process a few thousand granules...
Does anyone have experience with the software and/or can suggest a solution?
I'm running it on Ubuntu 64 bits, 4Gb ram.
My best,
You can try to resample and reproject MODIS L1B or L2 data with MRT Swath, then compose RGB image.
https://lpdaac.usgs.gov/tools/modis_reprojection_tool_swath
Thank you Xianjun, I have tried this approach with the CREFL program to perform atmospheric correction, but I think that I didn't manage at the time to create 250m composites as two channels of MODIS are only available at 500m resolution. But it was a few months ago and I don't remember exactly why I didn't pursued this approach. Have you created 250m MODIS composites with MRT Swath?
I realised that my question was not very objective... could you recommend a way to check how good is a projection?
If you plan to compose true color image at 250m resolution, it is a little bit tricky. Please see the attached document about detailed steps. Please let me know if there are problems.
Yes, the Liam routines were the first approach I tried! But I couldn't run it. The virtual machine was not working and when I manage to test an IDL licence, I remember that there was one function in the routines that was not found... (perhaps an IDL version issue?). I tried the GDL language, which is a free IDL implementation but it hadn't at the time support for all needed functions. My next plan was to try the Fawlty Language (another free IDL implementation), but I found the HDFLook and have been using it since. Did the routines worked for you?
I am not familiar with IDL. I usually just compose true color image at 1km resolution with Matlab. MODIS data products just provides 1km geolocation data. For 250m and 500m data, need to calculate 250m/500m geolocation first swath by swath ( can see how to interpolate from the attachment of my previous post), then do map projection.
- Badges
- Science topic
- Similar topics
- Geoscience
- Remote Sensing