The use of image analysis of leaf samples to determine vegetation health at Lake Merritt, Oakland, CA

Johann Curry, David Galvez, Ayesha Iqbal, Lisa Mar Asha Ngilbus, and Jenny Phan

The primary objective of our investigation was to use a newly developed image analysis technique to examine the overall health of vegetation located along the edge of the Lake Merritt Park and National Wildlife Refuge in downtown Oakland, California. Adapting techniques typically used in remote sensing we developed a strategy to calculate the amount of damage area in each individual leaf. Relative health of trees was determined by collecting leaf samples from various trees in designated areas, recording digital images of these samples, and then utilizing image analysis software to determine the percentage of apparent damage on a given sample. Then we calculated the average damage area of our samples of leaves. To infer the health of the entire tree as well as the entire designated area. We decided to compare three different areas of the park surrounding the lake based on proximity to the lake, nearby streets, and surrounding vegetation. We made the assumption that areas with heightened activity and traffic flow would produce the most damage of vegetation due to pollution. Thus, designation of areas was based on general observations of the vicinity. The first area, which was located along the north east edge of the lake, had minimum damage � averaging only about six percent per tree. This area was characterized by its close proximity to the lake. The second area, situated in Lakeside Park, was distinctive due to its dense stand of trees and minimum human activity. Nevertheless this location showed the greatest amount of leaf damage � about nine percent. The last area was located along the west edge of the lake, along Lakeside Drive. This area, adjacent to heavy traffic, averaged the least amount of damage � less than one percent. This is only our first attempt to try to correlate the % damage area on a sample of leaves to the quality of the environment. We had to develop several procedural strategies to allow us to collect reliable data. The first obstacle was the variation of tree species within and between the selected area of the park. This prevented us from choosing a particular species of tree as indicator. Despite the limitation, we believe that with additional refinement this methodology could provide a new qualitative strategy to measure the general health of the environment around us.

Please download the full PDF for map, graphs & figures, and cited works.