Last post, I mentioned that we had acquired (or at least were in the process of acquiring!) quad-polarized images from the Phase Array L-Band Synthetic Aperture Radar (PALSAR) on Japan's Advanced Land Observing Satellite (ALOS). The images have been downloaded, and processing near completed! I've made multilooked and terrain corrected HH Intensity images and CPR images.
However, there is a bit of a problem.
If you look closely at some of the acquisitions, you can see that there is some kind of image artifact running East-West through some of them. This artifact does not exist in the raw level 1.1 images, and it only seems to appear after the terrain correction step.
I consulted Mike, an experienced post-doc in the Neish lab research group, about the artifacts and he suggested I start troubleshooting by looking at the DEM I'm using for terrain correction. This is the also same DEM that I've been using for the RADARSAT-2 images, and I've subsequently noticed the same error in these images as well. Despite the DEM being a mosiac, there does not appear to be any visible seam in the image which could have caused the trouble. Catherine thinks there is a misregistration issue, which could be manually resolved if other troubleshooting methods fail. We might also try terrain correction with a different DEM.
I'm hesitant to report specific CPR data analysis until after this is fixed, but we can make some preliminary observations from what we have.
Salty surfaces are seem to be a little bit rougher in PALSAR's L-band radar than in RADARSAT-2's C-band radar, with salt domes having average CPR values of 0.45 (average CPR of domes - taking the average of the mean CPR for measured salt domes) or 0.51 (average pixel value for all measured salt domes). In comparison, the RADARSAT-2 has average CPR values of ~0.40 (for both average of domes, and average per pixel). The difference between CPRs of 0.4-0.5 might not be too significant, but we can at least see that the diapirs are appearing rough over a range of scales.
The average CPR for remobilized salt deposits appears significantly rougher in L-band than C-band. This, however, appears to be attributed to two anomalous deposits. Of 14 measured deposits measured in L-band, two appear to have average CPR > 1, whereas the remaining deposits have a diverse range of averages from 0.14-0.56. In comparison, salt deposits in C-band show far more constrained average values between 0.23-0.38, with the mean of average deposits (0.28) being closer to mean pixel value (0.26). This might mean that a few of the deposit areas are not smooth sands as previously interpreted, but are actually filled with cobble-sized clasts that area being noticed at the longer L-band wavelength and not the shorter C-band. It would be interesting to field check these different deposits to validate this hypothesis.
I'm out of town next week (wooo! I'm visiting Québec City for the Winter Carnival!) but we will return to this problem once I'm back.