I try to study the vegetation change using NDVI from landsat 7 and landsat 8 wherein the maximum NDVI value for NDVI from landsat 7 is 0.8 whereas for the NDVI from Landsat 8 is only 0.6 but practically there is not change in vegetation health or cover. I am using the same month data and working directly on DN values only. how to deal with this inconsistency?
The differences are mainly due to spectral response differences in the near-infrared band of ETM+ and OLI sensors. Generally, you can intercalibrate both NDVIs using linear or qudratic equations. Several case studies can be found in Remote Sensing journal. Or a simple review can be found in our recent publication. Hope this may help you.
I suggest you to conduct atmospheric correction for both of your images and work with reflectance values
I have tried doing some atmospheric correction using the flaash module of envi as well as by the normal dark object subtraction technique also. But using this corrected data for ndvi doesnt generate the pixel values ranging from -1 to +1 and I just couldnt figure out whether i need to resample them to show the -1 to +1 scaling and in that case how to do it?
If you are using the 'raw' DN values from both Landsat 7 and 8 then this will certainly cause an inter-satellite comparison problem. Landsat 7 DN values are in an 8-bit range while Landsat 8 values have an 11-bit range so very different gain and offset values are used to convert the TOA radiance to DNs. Elif's suggestion to work with reflectance values could help solve the problem. This means converting to surface reflectance which includes an atmospheric correction.
You could also do a simple image-to-image relative calibration and match histograms of the two images (I would suggest that you make the Landsat 8 image the master and the Landsat 7 image the slave). Since these results would still be in 'raw' DN format the -1 to +1 range will probably not be valid. The only way to get the values within that range is to do the radiometric corrections and calibration that would be done by converting the DN values to surface reflectance values (again,as suggested by Elif).
Good Luck
Pat
Dear Vincent, I suggest that you perform dual-sensor NDVI comparison after radiometric correction (DN to top-of-atmospheric reflectance) or atmospheric correction (to surface reflectance), as suggested by Elif and Prof. Chavez. Recently, I have compared NDVI data from collocated ETM+ and ALI (one of the near-infrared bands resembles the OLI near-infrared band) and concluded that quadratic equation is better to correct for NDVI differences. Linear equation performs poorly for low NDVI values.
Dear Vincent,
After performing atmospheric corrections and sensors inter-calibrations, the difference in NDVI could also be due to phenological / seasonal differences.
- Badges
- Science topic
- Similar topics
- Geoscience
- Remote Sensing