I am analyzing the data of Landsat 8 on the evolution of the vegetation indices through NDVI & WDVI and it seems that the values obtained (either raw images that correct TOA & BOA) are very different from those of Lansat 5/7 ETM + and MODIS too, in particular, these values appear to be much lower. The problem is in the different wavelength range of the NIR band of Landsat 8? Some of you have already done the analysis and had the same feeling?
I asked the USGS team and recevied the answer below:
Hello,
Thank you for your message. Yes, the narrow wavelengths of Landsat 8 will cause NDVI differences between L8 and L7. The wavelengths for all Landsat sensors can be found on http://landsat.usgs.gov/band_designations_landsat_satellites.php; a graphical comparison can be seen on http://landsat.usgs.gov/ldcm_vs_previous.php.
We have inquired scientists to provide details on using L8 NDVI with L7 NDVI; at the current time, there is nothing published to explain how the differences of NDVI between the two sensors can be adjusted to become closer in values; at this time we recommend users calculating NDVI for L8 simply be aware that there will be differences.
I'm sorry there is nothing more I can provide to you at this time.
Best regards,
Linda
landsat.usgs.gov
You may generate some indices using landsat image
http://www.slideshare.net/kabiruddin/introduce-variable
Ciao Pasquale,
indeed, there seems to be a tricky aspect in comparing band-wise surface reflection and vegetation indices between the older ETM/TM-sensor and the Landsat 8 OLI. As far as I see, many seem to struggle with that issue as the number of publication on that subject keeps fairly low. I might believe that you consulted already the paper of Li et al (2013) (linked) in Remote Sensing, where they found a very high correlation (R2>0.9) in the indices, especially in the upper index-ranges. I must admit, that I could not confirm such good correlations when comparing landsat scenes of arid environments in Madagascar. By using 10000 random locations on LEDAPS-corrected Landsat 7 images and DOS-corrected Landsat 8 scenes, I could establish a correlation in NDVI of R2=0,84. Not bad, but of course still not really comparable. Interestingly enough, a comparison of the individual bands resulted in much lower correlation factors between 0.6 - 0.7, therefore I'd look at the "strong" index-correlation very carefully...
I wonder if there has already been any studies on the L8-NDVI and primary production correlation so far; I haven't found any yet.
I don't know in what environment you are working and what your findings are so far. If you need more information about my study, please don't hesitate to ask.
Cheers, Pascal
http://www.mdpi.com/2072-4292/6/1/310/pdf
Ciao Pascal, many thanks for your comments. I discover the paper you mention after the publication of my question about the different value of NDVI for ETM/TM-sensor and the Landsat 8 OLI. I'm re-proccessing the data I used and the first results seem to contradict my first impression. When I'll have more consistent data I will write to you again.
Cheers, Pasquale
Hello and thank you very much for your informative answer Pascal!
I think i am facing similar problems. I am comparing NDVI and SAVI time series from Landsat 5 TM (1985-1995) and Landsat 8 as well as Sentinel-2. The test sites are 284 vegetation plots in high mountain areas so most of them have a very low plant cover. The absolute values vary a bit with the NDVI and a lot with the SAVI. Also the phenological seasons i get from Landsat 5 start way later and end far sooner than the ones from Landsat 8 and Sentinel-2. My thought was that the lower radimetric resolution of LS5 fails to detect the lowest differences of the vegetation reflection at start and end of season.
I atached four graphs that include 6 jears from each timeseries of all plots together. All data was converted to TOA-Reflectance.
Could the difference in the SAVI also be caused by the narrower NIR-Band of Landsat8?
Thank you for your answers!
Cheers, Matthias
- Badges
- Science topic
- Similar topics
- Geoscience
- Remote Sensing