3

I noticed something strange in my NDVI time series. Here I plotted NDVI for some geometries of interest in European countries between 1985 and 2020. NDVI within European Geometries

You will see there is a noticeable jump in NDVI between 2011 and 2013. This corresponds to the switch between Landsat5 and Landsat8. From year to year within Landsat5 years, the NDVI values usually only change by 1%-3%. However, between 2011 and 2013 all countries have NDVI increases of 5%+.

I generated by NDVI rasters from Google Earth Engine. I have checked over my scripts and do not see any errors. What could be causing this?

Here are links to my GEE scripts

Landsat5 NDVI: https://code.earthengine.google.com/2af4a03d52cf2032639e54422de28fa6 (lines 67-68 define the Landsat5 bands)

Landsat8 NDVI: https://code.earthengine.google.com/e5fa282d87134e1a454265d8e4c42862 (lines 60-61 define the Landsat8 bands)

1
  • The GEE code that korndog linked in his comment includes the coefficients for transforming ETM+ to the spectral response of OLI. I was interested in transforming OLI to ETM+. I used the OLS regression model in the Roy et al paper. Here is the code with the OLI to ETM+ coefficients: ` var coefficients = { itcps: ee.Image.constant([0.0183, 0.0123, 0.0123, 0.0448, 0.0306, 0.0116, 0]) .multiply(10000), slopes: ee.Image.constant([0.8850, 0.9317, 0.9372, 0.8339, 0.8639, 0.9165, 1]) }; `
    – Tris
    Oct 18, 2021 at 21:58

2 Answers 2

4

This is a "real" pattern that has been noted by many workers (see below for two good examples) and attributed with the different spectral responses of TM/ETM+ and OLI, though other factors could also play a role, e.g. LEDAPS versus LaSRC to derive surface reflectance products (if that's what you're using), whiskbroom versus pushbroom sensor.

Justin Braaten has a good GEE community tutorial on "harmonizing" TM/ETM+ to OLI using the coefficients derived by Roy et al. (2016): https://developers.google.com/earth-engine/tutorials/community/landsat-etm-to-oli-harmonization. Depending on how much fuss you want to make about this, you could also look into deriving a corrected dataset for your own study area.

EDIT: TM and ETM+ are often treated as interchangable, but note that this is a simplification.

Holden, C. E., & Woodcock, C. E. (2016). An analysis of Landsat 7 and Landsat 8 underflight data and the implications for time series investigations. Remote Sensing of Environment, 185, 16-36.

Roy, D. P., Kovalskyy, V., Zhang, H. K., Vermote, E. F., Yan, L., Kumar, S. S., & Egorov, A. (2016). Characterization of Landsat-7 to Landsat-8 reflective wavelength and normalized difference vegetation index continuity. Remote sensing of Environment, 185, 57-70.

5
  • Awesome! Thanks for sharing this. I'd like to hear your thoughts on transforming OLI to ETM+, rather than the other way around. Would it be advisable to use the Roy et al. coefficients to do this?
    – Tris
    Sep 22, 2021 at 18:46
  • 2
    Yes, no issue with that as long as you use the gains/offsets in their paper that were derived to predict ETM+ (e.g. Table 2 row 3, 6, ...). OLI is the better sensor and the new standard for Landsat's future, but there are many scenarios where it would be desirable to use ETM+ as the standard.
    – korndog
    Sep 22, 2021 at 19:38
  • Also, note the edit in my amswer on the interchangability of TM and ETM+.
    – korndog
    Sep 22, 2021 at 19:50
  • Follow-up question: there looks to be another - but less noticeable - jump down in greenness from 2001 and 2003. Could anything be causing this? @korndog
    – Tris
    Sep 29, 2021 at 19:51
  • 1
    @Tris nothing that I know of, though western Europe did experience extreme drought in 2003
    – korndog
    Sep 30, 2021 at 22:35
0

With Collection 2 there is no need for any cross-sensor harmonization. More info is to be find in this link: https://developers.google.com/earth-engine/faq#is_cross-sensor_landsat_surface_reflectance_harmonization_needed

2
  • 2
    Your answer could be improved with additional supporting information. Please edit to add further details, such as citations or documentation, so that others can confirm that your answer is correct. You can find more information on how to write good answers in the help center.
    – Community Bot
    Feb 26, 2023 at 17:50
  • Link-only answers are vulnerable to link rot, and are likely to be deleted by automated review.
    – Vince
    Feb 26, 2023 at 18:03

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.