I have a problem with a phenological study. The study area is in temperate zone of Asia. I have been doing phenological research using MOD VI product. I wanted to extract phenological events (onset/offset date, length etc) from NDVI and EVI, based on Greatest increase NDVI value method. After calculating the NDVI ratio, also I built a graph by result of NDVI ratio. But during determination of the average onset date of vegetation green-up and dormancy based on max and min value of NDVI ratio, some unreliable thing get me into trouble. For example: in forest area, max value of NDVI ratio occurred in middle of July corresponding threshold is 0.4 while in grassland area, onset of green-up occurred in middle of May corresponding threshold of 0.09. When I saw the graph, onset of green up in forest already started in beginning of May. So that's why I said it is unreliable.

Therefore, I wish to ask:

  1. Did you have a such kind of problem If you did phenological research using Greatest increase NDVI approach?
  2. If so, how did you solve this problem to determine onset date of green-up and dormancy.
  • You will have describe your methods better, what is; MOD VI, Greatest Increase NDVI? Have you done atmospheric correction? Do you have any water/lakes in the image and what values are being returned for them? – BWill Jun 8 '11 at 2:09

Hard to say without looking at the data, but there are some issues that you need to take into account. First, there's angular variation not corrected by the MxD13 data, and its effect can be different for grasses (continuous canopies if you want) and forests (discontinuous). You can use the MODIS NBAR product (MCD43?) to use nadir-reflectance. This product works quite well if you are not hampered by cloud.

Then there's a problem with any approach that takes a VI as a measure of phenology. Due to uncertainty in surface reflectance, angular effects (see above) and the influence of soil in low LAI scenarios, NDVI is very noisy. Typically, fitting a predefined shape improves things a bit. Typically, (double) sigmoid or logistic functions are used for onset of greennes and senescence.

Alos, bear in mind that these suggestions are basically a run down of the MODIS land cover dynamics product (MCD12). There are many papers that use/validate this product, and you should be aware of them. In particular, I would point out the following papers:

  1. "Land surface phenology from MODIS: Characterization of the Collection 5 global land cover dynamics product", Ganguly, S., Friedl, M.A., Tan, B., Zhang, X. and M. Verma, Remote Sensing of Environment, 2010, 114, pp. 1805-1816.

  2. "Global vegetation phenology from Moderate Resolution Imaging Spectroradiometer (MODIS): Evaluation of global patterns and comparison with in situ measurements", Xiaoyang Zhang, Mark A. Friedl, Crystal B. Schaaf, Journal of Geophysical Research, Vol. 111, G04017, doi:10.1029/2006JG000217, 2006

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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