Study Area

The study area is two canyons in New Mexico within the Sandia Mountains east of Albuquerque. The North Canyon is Pino Canyon and the South Canyon is Oso Canyon. The purpose of the study is to use NDVI to determine areas that have been affected by drought related insect infestation and to compare the health of the two canyons. High resolution imagery of the area was downloaded from http://www.bernco.gov/download_Orthoimagery/, aster data was downloaded from glovis, landsat data was downloaded from earth explorer and the study area shapefiles along with project instruction were provided by James Norton with the Centre of Geographic Sciences. The data is projected in StatePlane NAD 1983, HARN (US Feet), New Mexico Central, FIPS 3002.

Methods

After acquiring the necessary datasets, they were loaded into arcmap and the appropriate data frame coordinate system was selected.

The first step was to combine the aster bands into one composite image using Raster Processing — Composite Bands tool.

After the composite was created the raster was clipped to the study area using Raster Processing — Clip Tool

Before analyzing the aster imagery using the NDVI indices, control areas need to be outlined. This means that we need to analyze the aerial photography and create polygons around zones of dead trees and living trees. A shapefile called trees was created and a column called “status” was added. As control area polygons were drawn there were designated as “live” or “dead”.

The composite aster image was then classified using NDVI Indices through the image analysis window. (Window — Image Analysis)

After the results were computed the data was compared to the original control study sites as shown to the right.

Discussion

This method was not as successful as hoped because of the affect of shadows, cloud, bare ground and soil moisture. Dead trees are fairly colourless and therefore are likely to be classified with bare soil. Variable tree density also affects the NDVI as bare ground has a similar response to dead trees. This can give a low NDVI even when the area looks green. It could potentiall be useful to classify the area when there is snow cover after windy weather.

The brightest area in the image are mostly valleys that follow river ways. This has a higher NDVI likely due to an increase in tree density in lower lying areas. Additional soil moisture could play a role as could north facing shadows which reduce reflectance.

  • ArcMap : Classifying tree damage using aerial photography, aster data and NDVI Indices
  • ArcMap : Classifying tree damage using aerial photography, aster data and NDVI Indices
ArcMap : Classifying tree damage using aerial photography, aster data and NDVI Indices
ArcMap : Classifying tree damage using aerial photography, aster data and NDVI Indices

If you look at the image to the left, it is noticeable that the shadowed areas have higher NDVI values. The dead area is the darkest spot in the image. The live tree zone is in the middle of the grayscale for the image.

 

Next to compare the overall statistics of the two images the extract by mask tool which extracts raster data based on vector layers, was used. Statistics were extracted from the grayscale NDVI classified file individually using the Osa and Pino ROI vector files. This can be done visually (Pino appears darker which would suggest it has lower tree health)

The results showed that Oso had a mean of 122.5 and Pino had a mean of 116.5. This shows that Oso does contain more healthy trees than Pino Canyon (or at least more trees!).

 

Conclusion

Overall the multispectral imagery did well at determining tree density but the NDVI analysis was not quite accurate enough to distinguish bareground from areas with dead trees.

For fun the area was looked at in arcglobe with the gps points plotted on a hike track within the Pino Canyon.  The results are shown below.