Jump to content

 
Photo

Determining "acceptable" NDVI number

- - - - -

  • Please log in to reply
6 replies to this topic

#1
htomita

htomita

    Newbie

  • New Member
  • Pip
  • 5 posts
  • United States

Hello all,

Very new to the GIS field here, and one of the first topics I'm researching as an intern is the vegetation range for an NDVI layer. I read on Wikipedia that 0.3-0.8 is acceptable, but there is no source, and some of the linked sources that look intriguing are behind paywalls. I used 0.38 for the NDVI, and >8m for the height to determine a tree canopy layer, and it seems to be accurate. However, I would like to read more about NDVI to understand the methodologies that are commonly used when creating a NDVI, especially when they pertain to specific types of vegetation layers. Was my ad hoc procedure acceptable, and where can I find academic or technical papers to learn from their methodology. Textbooks are welcome too, provided they are available online. If you think it is a very useful GIS textbook in general, please do recommend, and I will seek it out at the libraries around here.

tl;dr:
a) is 0.38 an "acceptable" NDVI
B) you can recommend good online resources (e.g., textbooks, academic or technical paper troves NOT behind paywalls, etc.) on NDVI or GIS in general.

edit: also, if this is in the wrong section, I apologize. Please let me know where I should move it to.
Much obliged, Hiroshi

Edited by htomita, 12 May 2011 - 12:38 PM.


#2
SaultDon

SaultDon

    Master Contributor

  • Validated Member
  • PipPipPipPip
  • 153 posts
  • Gender:Male
  • Location:BC
  • Interests:Environmental/Wildlife GIS, Aboriginal Land Use and Occupancy
  • Canada

http://abstracts.ran...egetation_index

and

http://www.fsnau.org..._Index_NDVI.pdf (from wiki)


I like how they always state Value X 'commonly' represents Y.


There are tonnes of references listed in those two links that may point to something interesting.
Look at the first link in particular, they reference actual Case Studies where NDVI was used.


One of the earliest applications of NDVI was in:
Rouse, J. W., R. H. Haas, J. A. Schell, and D. W. Deering (1973). Monitoring vegetation
systems in the Great Plains with ERTS, Third ERTS Symposium, NASA SP-351 I, 309-
317.


If you can get your hands on that, it might be helpful.

#3
htomita

htomita

    Newbie

  • New Member
  • Pip
  • 5 posts
  • United States

http://abstracts.ran...egetation_index

and

http://www.fsnau.org..._Index_NDVI.pdf (from wiki)


I like how they always state Value X 'commonly' represents Y.


There are tonnes of references listed in those two links that may point to something interesting.
Look at the first link in particular, they reference actual Case Studies where NDVI was used.


One of the earliest applications of NDVI was in:
Rouse, J. W., R. H. Haas, J. A. Schell, and D. W. Deering (1973). Monitoring vegetation
systems in the Great Plains with ERTS, Third ERTS Symposium, NASA SP-351 I, 309-
317.


If you can get your hands on that, it might be helpful.



Hi SaultDon,

Thanks for the speedy response. Sorry I didn't get back to you until now. I thought I signed up for email notifications of responses, but I didn't get one, so I need to double check that.

So I went through the pdf you pulled f/ Wiki, and it was a useful overview explanation of NDVI, but I wasn't able to make sense of the last sentence.

"Theoretically, NDVI values are represented as a ratio ranging in value from -1 to 1 but in practice extreme negative values represent water, values around zero represent bare soil and values over 6 represent dense green vegetation."


Did the author mean 0.6, instead of 6? That was a tad confusing; I just assumed it meant 0.6.. I have only been reading that NDVI goes from -1 to +1, and anything under zero "do not have any ecological meaning" (quote f/ other article you linked). And if >0.6 means "dense green vegetation", then does my NDVI filter of 0.38 not count?

So the link from Rangeland Methods had some great links at the very bottom under Additional Information. This site from NASA provided the best explanation of NDVI yet. The quote on the right, "If there is much more light in NIR waves than in visible wavelengths, then the vegetation in that pixel is likely to be dense", in concert with the nice diagram of NIR and Visible light bouncing off a green and withered tree made a lot of sense.

This FAQ on Vegetation in Remote Sensing is written by a Caltech prof in '94, but the information is extensively thorough and concise.
So that FAQ alone practically answered all my questions. The "Problems" section near the bottom (27-33) proved especially helpful. What I gathered was...

27) How well do these vegetation indices work in areas with low
vegetation cover?
A) Generally, very badly. (more details on the website)

31) How low a plant cover is too low for these indices?
A) These are rules of thumb, your mileage may vary:
RVI, NDVI, IPVI = 30%
SAVI, MSAVI1, MSAVI2 = 15%
DVI = 30%
PVI, WDVI, GVI = 15%

SUMMARY: In order of preference for each type of sensor:
TM or MSS (or any broad-band sensor)
1: NDVI (or IPVI)
2: PVI
3: SAVI (top of list for low vegetation)
4: MSAVI2
Since this is my first foray into remote sensing, I can't verify the accuracy of FAQ on Vegetation in Remote Sensing. Are there any remote sensors out there who would be will to take a quick gander and tell if this is all up to date methodology? Particularly the bit on SAVI (#16).

I'm going to see if I can run a SAVI (Soil Adjusted Vegetation Index) on the LandSat data. FAQ on Vegetation in Remote Sensing contains the equations and a lot of good jazz. It recommends NDVI for denser vegetated areas, while SAVI is better for dealing with "soil noise".

I read back to my original post, and I neglected to mention that I'm studying vegetation in an urban area (Pasadena, CA), so these indices might not be as well suited? Yikes. Again, any remote sensors, please, please chime in.

This is the where the data is from, the LAR-IAC program run in LA County to gather various remotely sensed data. This page details how it has been used to get NDVI indices, but used a tolerance of >0.1, while we went with >0.38 because we're looking at the tree canopy. My bad for not mentioning the urban environment. I didn't take that into account until I started reading the FAQ and how it focuses primarily on remote regions w/ interference f/ roads and glare f/ buildings.

Anyway, this FAQ gave me a good understanding of various vegetation indexes and which to use. Next step is to determine which is preferable for urban environments like Pasadena. I think I'll call the Mark Greninger in LA County tomorrow. Thanks for your help thus far!

Edited by htomita, 13 May 2011 - 12:03 AM.


#4
SaultDon

SaultDon

    Master Contributor

  • Validated Member
  • PipPipPipPip
  • 153 posts
  • Gender:Male
  • Location:BC
  • Interests:Environmental/Wildlife GIS, Aboriginal Land Use and Occupancy
  • Canada

That 6 also caught my eye too, so not sure what they mean.

But like in the MODIS NDVI (PDF 44kb), they use a range of values from 0-255 (a stretched -0.25 - 1 of the NDVI values).

"NDVI has been scaled using the formula (NDVI * 200)+50, this yeilds values from
0-250 corresponding to NDVI range of -0.25 - 1
o No Data value: 253
• Data format for the Band files is geotiff with 16 bit Signed integer
o No Data value: -200
o Fill value: -28672 "



I found a PDF for that Song et al project which has all sorts of information, like options for the atmospheric corrections.
http://www.unc.edu/c.../Song01_RSE.pdf


I think this would turn into an excellent discussion on gis.stackexchange.com (add that to your list of must have GIS sites). Cross post your original question (or revised) and share that link here so we can follow it.

#5
htomita

htomita

    Newbie

  • New Member
  • Pip
  • 5 posts
  • United States

SaultDon,
I glanced through the Song et. al., but it seems to focus on comparing new and old methods of atmospheric correction, with the conclusion that new methods aren't that amazing after all. Do you feel that I could use some of those correction techniques for urban areas?

I took your advice and cross posted this to gis.stackexchange. If you like, please follow the post there.

#6
SaultDon

SaultDon

    Master Contributor

  • Validated Member
  • PipPipPipPip
  • 153 posts
  • Gender:Male
  • Location:BC
  • Interests:Environmental/Wildlife GIS, Aboriginal Land Use and Occupancy
  • Canada

I've only ever done it in an environmental context, this is often in rural areas. That's why I think the stackexchange may help; a person with an extensive RS background may be able to give a thorough response. Or you might even get help with phrasing the question in general.

I was looking at those references because they seem to derive an 'acceptable' NDVI value from an overall inspection of the average reflectance of the covertypes. Its that line graph in one of those links where they look at grasslands as an example. They're lucky and had a spectrometre.

Below is the link to that image.
http://abstracts.ran...al_response.jpg
A high NDVI value will correlate with a high difference in the two bands, red and near IR. You can see this with Montane Shrubs.
A low NDVI value would indicate, like the exposed rock, low to nil vegetation. Trying to compensate for the Senescent Vegetation (PDF) is always fun.

Are you able to do any field truthing to get some points from maybe a GPS unit to validate? Get some points for your 'desirable' NDVI, then get some points for the undesirables... Run some NDVI tests and overlay those points.

Do you have a summary of the covertypes you would expect to find in your study area?
Try to mask out those undesirable covertypes that you may think are skewing your NDVI values. One thing I often do is mask out some of those areas based on a subset of a supervised classification.


So overall, the above may explain why you are getting these 'desirable' low NDVI values, because your area contains an urban setting which can give a low reflectance.

#7
htomita

htomita

    Newbie

  • New Member
  • Pip
  • 5 posts
  • United States

Oh I see now, so when I first glanced at that graph, I couldn't read it and determine where the NDVI number was since it isn't explicitly labeled. Now I see that the steeper the slope, the greater the difference between the red and Near IR wavelengths, which in turn equates to a higher NDVI.

I might be able to do some field truthing. I will ask when I get into the office on Monday.

Summary of cover types: Actually just one type...a general tree canopy layer. :P So we decided that anything over 8ft and was bushy green by NDVI standards would be included in the layer. I was asked to research this NDVI process and see if we need to do more adjustments for urban areas.

I didn't think about masking, I will also give that a try on Monday if there turn out to be some tricky areas.

Thank you Donovan for all the leads and help. From this, I have more trials to run the data through, and see which methods works the best with our urban area.




0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users

-->