Posted on June 27, 2013 by Howard Oh
After a late night discussion with many REU students and grad students (especially Emily!) I had a new mission. Everyone helped me brainstorm throughout the whole day and Emily helped me focus on another area that may possibly have different vegetation growing on opposite sides of the area we call Narnia. I was still continuing my idea of trying to classify some native plants versus non-native plants but with a new focus on trying to correlate the vegetation with geology. With this new focus I headed back out to the field with Paul and got a phantom going. It was raining for most of the morning and I was concerned that I was not going to be able to fly the quadcopter and get the imagery I needed, again. Luckily the sky cleared up and we had a window in between rain clouds to get about 6 good flights in with good ground cover. Back at the barn, I stitched the imagery and saw that it had good coverage. The problems I faced were when I downloaded my XY points from a handheld GPS. Because of the dense canopy cover, some of my points were off my 10 meters. The tree I thought I may classify separately was the Banyan tree because of its distinct canopy cover. However, when I looked at the imagery it looked very similar to the Koa Haole and in some parts of the image, it was difficult to differentiate between plant species, especially when my ground truth points were not accurate. Even eCognition had a difficult time correctly segmenting out the separate tree canopies.
Back at the barn, Dr. Wechsler gave me more insight on the aquifers and DEM data and how they might be a factor to the vegetation in the area. She also helped me get in touch with Dr. Becker and he also enlightened me that the soils were more of an important focus and had a stronger correlation to the vegetation. There was also the idea of trying to map out areas with different salinity and see if some plants were more dominant than others in either fresh, salt or brackish water. Unfortunately, I was unable to find usable imagery in that area to get down to the species level classification for that type of analysis.