Estimating Canopy Cover in Pinyon-Juniper Woodlands Using ADAR Imagery

Kristin Vanderbilt

Please click on pictures for a detailed view.....

 

Background:

  • Net primary productivity (NPP), the net flux of carbon into plants per unit time, is a key ecosystem variable measured at the Sevilleta Long Term Ecological Research (LTER) site.
  • To facilitate calculation of NPP for pinyon-juniper woodlands, which consist of patches of trees interspersed with areas of grass or bare soil (Figure 1), an estimate of the canopy cover of trees in this community is needed.

Results:

NDVI:

NDVI was calculated from the ADAR image as

NDVI = (Band 4 - Band 3)/(Band 4 + Band 3). NDVI values of 5 and above were found to represent tree cover by mapping tree location to NDVI of specific pixels (Figure 5). Of a total 9741 pixels, 7713 (79%) represented tree cover.

Figure 5. NDVI of study plot.

Figure 1: Pinyon-Juniper woodland at the Sevilleta LTER

 

Objective:

  • To evaluate how well ADAR imagery can be used to estimate tree cover in pinyon-juniper woodland so that
    • the percent of the landscape that is grassland is known, and

    • an estimate of pinyon and juniper NPP may be made based on canopy cover
Methods:

Based on a 100 m X 100 m plot with mapped tree locations in pinyon-juniper woodland (Figure 2) and an ADAR image with 1 m resolution for the same area (Figure 3), I estimated tree cover from:

  • NDVI
  • Unsupervised land cover classification
  • Supervised land cover classification
The study plot is located in the eastern part of the Sevilleta National Wildlife Refuge in central New Mexico (Figure 4).

UNSUPERVISED LAND COVER CLASSIFICATION:

Four landcover types were defined based on the results of the unsupervised clustering algorithm used by GRID (Figure 6). This method indicated that 26% or the plot was covered by trees.

Figure 6. Unsupervised classification of ADAR image in to four cover classes.

 

Figure 2: Species locations in 100 X 100 m plot superimposed on ADAR image.

Figure 3: Study plot delimited on ADAR image.

SUPERVISED LAND COVER CLASSIFICATION:

A training set for four classes (trees, grass, mixed (grass/rock/soil), and shade) was defined from which Grid developed a spectral signature file. This file was used to categorize all pixels into the cover class it most closely resembled. This method indicated that 76% of the plot was covered by trees (Figure 7).

Figure 7. Results of supervised classification of ADAR image into four cover classes.

 

 

Figure 4. Vegetation map of the Sevilleta National Wildlife Refuge, NM

Conclusion:

Tree cover on this plot is estimated visually to be between 40 and 50%. None of these methods, as applied, is a good mechanism for estimating tree cover in pinyon-juniper woodland.