Question
Asked 10th Jul, 2014

How does one calculate the percent vegetation cover using NDVI method?

I calculated the percent vegetation cover in a given study area using NDVI method and just to want make sure that my method was correct and the result is reliable. I just did this through simply calculation of NDVI and then I reclassified my map (I used +0.2 as the threshold to separate vegetation and non-vegetation) and finally calculated the proportion of the green to whole area. However, someone recommended me using Green Vegetation Fraction to produce the vegetation map and calculation the percent vegetation cover. Any ideas?

Most recent answer

2nd Feb, 2019
Chahrazed Salhi
Ecole Nationale Supérieure d'Hydraulique
tell me please, how can i calculate NDVIsoil ?

Popular Answers (1)

23rd Jul, 2014
José M.C. Pereira
University of Lisbon
The NDVI is very sensitive to soil colour, soil moisture, shading effects caused by solar and viewing angle geometry, so it would be wise to start by making sure the NDVI is appropriate for your study conditions. There are dozens of alternative vegetation indexes. Also, the value of the NDVI, as of other vegetation indexes, is affected by other factors than vegetation cover: leaf area index, phenology and physiological status of the vegetation, vertical structure of the plant community (e.g. open grassland vs. multi-layered forest). Thresholding the NDVI at 0.2 and considering that anything above that is vegetation, and equivalent in terms of projected cover over the entire range of above-threshold values is a very crude approach. However, it may be enough for your purposes, but that is impossible for me to assess, given the lack of information on your study conditions.
9 Recommendations

All Answers (12)

10th Jul, 2014
Carmelo Cammalleri
European Commission
Your method assumes that a pixel is either fully vegetated (> 0.2) or not vegetated (< 0.2). However, even for vegetated pixels you surely have different coverage. Consider that a fully vegetated pixel should have NDVI > 0.8 (circa).
You should compute for each pixel a fraction vegetation coverage (following on of the relationships proposed in the literature) and then average these values on the whole area.
See the attached publication for one of the possible relationships (eq. 1) to use.
 
 
3 Recommendations
16th Jul, 2014
Miguel Armando López-Beltrán
Universidad Autónoma de Sinaloa
I agree that you use green vegetation fraction.
2 Recommendations
17th Jul, 2014
Mohammad Ali Nezammahalleh
University of Tehran
you can use some other image data source. it can help you to find more accuracy
2 Recommendations
17th Jul, 2014
Asmat Ali
PMAS - Arid Agriculture University
I agree with @Mohammad Ali Nezammahalleh. In this way, you would be able to do comparison as well.
2 Recommendations
23rd Jul, 2014
José M.C. Pereira
University of Lisbon
The NDVI is very sensitive to soil colour, soil moisture, shading effects caused by solar and viewing angle geometry, so it would be wise to start by making sure the NDVI is appropriate for your study conditions. There are dozens of alternative vegetation indexes. Also, the value of the NDVI, as of other vegetation indexes, is affected by other factors than vegetation cover: leaf area index, phenology and physiological status of the vegetation, vertical structure of the plant community (e.g. open grassland vs. multi-layered forest). Thresholding the NDVI at 0.2 and considering that anything above that is vegetation, and equivalent in terms of projected cover over the entire range of above-threshold values is a very crude approach. However, it may be enough for your purposes, but that is impossible for me to assess, given the lack of information on your study conditions.
9 Recommendations
22nd Sep, 2014
Jose Gonzalez de Tanago
Wageningen University & Research
Maybe that comes too late. Which kind of vegetation do you want to derive the percent of cover? and which NDVI method are you refering to? could you provide a reference for the NDVI method you mention?
If you want to derive pecentage of TREE cover and you have access to very high resoultion imagery (at least in google earth) you could derive training areas with calculated percentage tree cover (digitazing the canopy cover) and use a Random Forest regresion model to derive the percentage of tree cover for all your pixels. This method was used to derive the global percentage of tree cover "VCF" product either MODIS or Landsat based. The method is described in the references attached by Hansen et al.
Another  alternative (I think would probably be the most accurate approach) is calculate the fraction cover by (GREEN) vegetation through Spectral Unmixing. It will require you: -1st know which are the main "components" that cover your pixels( normally: soil, green vegetation, non-green vegetation (bark, etc), water?, and SHADE!), 2nd build  your spectral library for enmembers of these components (you need training areas of pure pixels representings the spectral signature of those), 3rd calculate each fraction cover (could be implemented in many softwares, ENVI, or open source softwares are available). There are many references following spectral (un)mixing approach to map vegetation cover. If you have only multispectral data (e.g. Landsat) I found useful the reference by Rogan et al. (see ref. attached)
3 Recommendations
14th Jan, 2015
Trevor Moffiet
University of Newcastle
Values of NDVI (or any other spectral vegetation index) can't be directly related to % vegetation cover for a scene of interest unless there has been some calibration study performed between the spectral index values and ground truth measurements of % foliage cover.  In other words, you need to predict the level of foliage cover using a calibration model (prediction model) that is suitable for the scene of interest. Note that NDVI 'saturates' at high cover and is not useful for distinguishing between different levels of foliage cover at high levels of foliage cover. You will find some articles on my ResearchGate publications page that relate to vegetation indices and the estimation of % foliage cover.
9 Recommendations
19th Jan, 2016
Kirubakaran Muniraj
I agree with @Jose Gonzalez de Tanago Menaca (Alternative method)
20th Sep, 2016
Jahangir Ali
National University of Sciences and Technology
Kindly tell me that I  want to know how much is the vegetation area and how much is the barren land in my study area.  How can I do that. Regards
19th May, 2017
Jahangir Khan
Bahria University Karachi Campus
Could anyone please share, the processing (may be in Arcmap or Erdas) for computing NDVI for Soil and NDVI for Vegetation covers. I need to work with Fractional Vegetation Cover (FVC)
1 Recommendation
20th Nov, 2017
Linyuan Li
Beijing Normal University
@Jahangir Khan
Hello, we just published a paper about estimation of FVC based on a simple physical method.
If you have any question about it, please let me know.
Can you help by adding an answer?

Similar questions and discussions

Related Publications

Got a technical question?
Get high-quality answers from experts.