Orthorectifying: Geoid or no?

Discussion created by Palebot on Jun 6, 2014
I have several tiles over a region. They are GeoEye-1 images, with red, green, blue, NIR, and pan. Many are from different dates, but when put together you get a complete coverage. The data came un-orthorectified. There is no real way I can find to create pansharpened composites and then orthorectify (you have to orthorectify each band individually before creating a composite and then pan-sharpening). I struggled to avoid that as long as I could, and I even tried to avoid orthorectifying by creating a large integrated mosaic and then draping it over a DEM to correct for topo errors. Of course, that did not work well nor does it really deal with what ortho-rectifying is about (given my limited knowledge). So to save time, I figured I didn't need to do each band and I could simply orthorectify the panchromatic images and use them only with analysis. That is fine. Worked. However, I was unclear on a couple things. For example, I decided to select the geoid option. I know what a geoid is but can't entirely get my head around how that is integrated into the orthorectification. The ESRI description in help doesn't, well, help. But for consistency, I selected it for each one. However, one thing I note is that now, unlike the un-orthorectified images, not all the images line up as they did before. What could be causing this? Is it a problem with the DEM? The only DEM I have is over a much larger area than each individual band coverage. Or is it a sensor issue due to the fact that many of the scenes were collected at different times between 2010 and 2013?