Professional Documents
Culture Documents
Lab 19: Vegetation Indices: Introducing The NDVI 11/04/10
Lab 19: Vegetation Indices: Introducing The NDVI 11/04/10
Fall, 2010
11/04/10
The data you will use for this lab is Landsat Thematic Mapper imagery of the Del Norte are in southern Colorado. The
purpose of this lab is twofold. First you will be introduced to the difference in values of the Normalized Difference Vegetation Index (NDVI) that can result from using data that has not been calibrated (raw data). A very common mistake
made in computing the NDVI is to use the raw digital numbers recorded by the sensor. These values are not calibrated
for sensor gain and offset and the amount of radiation incident on the target (affected by solar zenith angle, distance of
the Sun from the Earth and topographic and atmospheric effects). For this lab, you have been given data that that not
calibrated (L5034034_03420050522.img) and that have been corrected for the effects of terrain, atmosphere, sensor
position and illumination conditions (L5034034_03420050522r.Img). For the second part of the lab, you will use the NDVI image computed from the reflectance data (NDVI_ref) to classify the image into 3 land cover classes.
1.
2.
3.
Open ArcMap and choose to Start using ArcMap with A New empty Map
4.
Add the two image files and the point shapefile to the map and change the band combinations so that Band 4 is
displayed with red, Band 3 is displayed with green and band 2 is displayed with blue.
5.
Switch between the two images and make a mental note of some of the differences between them.
6.
Now add the following individual bands from the image files:
L5034034_20050522.img - bands 3 and 4
L5034034_20050522r.img - bands 3 and 4
183
Fall, 2010
What is the difference between an NDVI calculated with a raw image and a reflectance image?
1.
Open the Spatial Analyst Toolbar in ArcMap. From the View menu in ArcMap, select Toolbars, then Spatial
Analyst.
2.
Set your working directory in Spatial Analyst to the same directory where you stored the data for this lab. Choose
Options from the Spatial Analyst tool bar then enter your working directory in the field called working directory:
3.
Open the Raster Calculator (Spatial Analyst tool bar) and enter the following expression to calculate the NDVI from
the reflectance data (its quite tricky to get this right!)
NDVI_ref = float([l5034034_03420050522r.img - Band_4] - [l5034034_03420050522r.img - Band_3]) /
float([l5034034_03420050522r.img - Band_4] + [l5034034_03420050522r.img - Band_3])
4.
Including the float command is very important. The data in the images are integer. If you do not instruct the software
to treat the data as Float, then your calculation will result in an image that shows only 0s and 1s
When you have run these calculations correctly, you should have 2 NDVI images stored in your working directory:
NDVI_ref - calculated from the reflectance imagery
NDVI_raw calculated from the uncalibrated imagery.
It is difficult to figure out what the difference is between the NDVIs just by looking as these images. For this part of the
lab, you should address the question:
To answer this question, you will need to (i) calculate the difference between the NDVI images and then (ii) classify the
results into negative difference (1), insignificant difference (2) and positive difference (3)
184
Fall, 2010
185
Fall, 2010
Using the NDVI to classify land cover in the Del Norte Basin
The Del Norte is one of the most productive snowsheds in the Upper Rio Grande (URG) basin. Around 50% of the surface
water supply in New Mexico comes from melting snowpack in the Del Norte and other snowsheds in the URG. We can
use remotely sensed imagery to help us determine snowpack extent, but mapping snow cover is complicated by forest
cover. It is difficult to map snow cover under trees. In this part of the lab you will examine the values of the NDVI for different land covers and you will determine how you might use the NDVI to classify the image into 3 classes:
Snow (1)
Trees and Snow (2)
Vegetation (whether it is trees, meadow, riparian, agricultural) (3)
1.
2.
Examine the NDVI values for the different land cover classes. (These NDVI values were extracted from the
NDVI_ref layer.) Note: you should group the different types of vegetative land cover together to make a generic
vegetation class.
3.
Choose thresholds in the NDVI values that distinguish the three classes: Snow (1), Trees and Snow (2), Vegetation
(3)
4.
5.
Reclassify NDVI_ref to the 3 classes: Snow (1), Trees and Snow (2), Vegetation (3)
The image below is an example of how to set-up the reclassification - but DO NOT USE THE THRESHOLD VALUES given in
this example. You will lose 5 points if you copy the thresholds that I have used - they are deliberately incorrect! You
must determine the thresholds for yourself