Hi there,
I would like to make a survey in northern parts of Scandinavia of vegetation biomass evolution according to Landsat NDVI (Landsat 5 and Landsat 8).
Since the spectral and radiometric resolution are not the same between sensors (TM and TIRS), what are the GIS/remote sensing software tools I could use to preprocess/normalize/making comparable 3 images taken in july 1990, july 2006 and july 2017 (+/- same tile)?
Thanks,
You have first to resample your three scene un a same pixel resolultion (30m is the most common in Landsat).
Then turn the 3 scenes from DN values (defaut) to reflectance value
ENVI>Basic tools>Calibrations utilities>Landsat Calibration
One "NIR" and "R" Bands are in reflectance, the NDVI most be the most comparable possible.
Kind regards
Romain Courault, since you are comparing multi-date (1990, 2006, 2017) satellite images, you would have to anyway perform atmospheric correction of Landsat 8 OLI and Landsat 5 TM satellite images. To make your task easier, USGS provides ready-to-use atmospherically corrected - Level 2 Surface Reflectance products of OLI and TM that can be ordered and downloaded for free. In addition, they also provide vegetation indices such as NDVI that is derived from Level 2 SR data. So if you wish, you can use USGS provided NDVI or choose to atmospherically correct Level 1 OLI and TM data yourself using ENVI or some other open source Remote Sensing/GIS Software/Tool.
I think even before pre-processing OLI and TM data, you should also consider some other factors, such as the processing level and cloud cover of Landsat data 'suitable for pixel-level time series analysis'. For example, L1T processed Level 1 Landsat data are Radiometrically calibrated and orthorectified using ground control points and DEM data to correct for relief displacement. Also check the Collection Category (Tier 1 or Tier 2) of the Landsat data. Attention should also be paid regarding the presence of optically thick clouds and high altitude cirrus clouds in the Landsat imagery.
It would be ideal if you use SR Level 2 data that was derived from T1, L1T and cloud free Landsat 8 and Landsat 5 TM scenes for the vegetation biomass evolution analysis. But as you rightly pointed out, even atmospheric correction may not be sufficient for comparing Landsat 8 and Landsat 5 TM, obviously, due to the fact that the spectral and radiometric resolutions are different in both these sensors. In addition, Landsat 8 has higher Signal to Noise ratio. Therefore, you would have to come up with some spectral band transformation parameters that would compensate for the differences. There is a similar discussion on RG that contains links to relevant articles that should be helpful for your study. So check them out.
- https://lta.cr.usgs.gov/DD/landsat_dictionary.html#data_type_l1
- https://lta.cr.usgs.gov/DD/landsat_dictionary.html#collection_category
- https://www.researchgate.net/post/Is_there_any_aggregated_Spectral_Band_Adjustment_Factor_SBAF_for_Landsat_5_and_8
Thank you very much for your answers ! !
I already got the Level 2 Surface Reflectances products, and took the cloud-free ones.
In the recommended articles, I found this pretty usefull:
VILLAESCUSA-NADAL, Jose Luis, FRANCH, Belen, ROGER, Jean-Claude, et al. Spectral Adjustment Model's Analysis and Application to Remote Sensing Data. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2019.
And the associated parameters to adjust the bands (various sensors) :
http://salsa.umd.edu/pages/resources.php
Thank you again,
- Badges
- Science topic
- Similar topics
- Methodology
- Sensors