Predicting forest structural diversity using satellite data

One of the core principles of sustainable forestry management is conservation and enrichment of biodiversity in forest ecosystems. Conservation and enhancement of wildlife habitats have been seen as one of the basic measures about biodiversity. To enhance the structural diversity in forest ecosystems has been considered to be the most practical way to accomplish this. It has been theoretically accepted that if the structural diversity increases in a forest ecosystem, many habitats hosting for many kind of the species are constituted. It is important that forest management applications should be performed to increase diversity of stand types and other land cover patches. It also needs to take silvicultural measurements to increase diversity at stand level. Appropriate silvicultural treatments can be practiced depending on the quantification of current structural diversity of stands. It is obvious that the inventory costs of measurements and observations based on field methods to determine structural diversity are quite expensive when we consider all of the stands in a forest management unit. Thus, efficient inventory methods should be developed to assess and map the structural diversity in the forest stands. In this respect, satellite data are very important information resources. Satellite data with different characteristics should be investigated to estimate the structural diversity in different forest ecosystems. In this thesis, it has been aimed at the modeling and mapping of the structural diversity in a forest area consisting of predominantly Anatolian black pine within the borders of Beyşehir in Konya, using the variables derived from SPOT-4, Quickbird2-PS and Quickbird2-MS satellite data. Diversity indexes were calculated with field measurements in 270 sample plots (including 30 trees) ranging in size from 100 m2 to 900 m2. In the next step, average brightness values and texture features of pixels corresponding to the sample plots were determined. The relations have been demonstrated for 270 sample plots (sub-plots) and 54 sample plots (main plots) with 1 ha size to evaluate a combination of five samples. Simple regression analysis and regression tree techniques were used to examination of the relationships between image variables and structural diversity. Simple regression analysis compared with regression tree technique was found more appropriate to determine stand diversity. More significant relations were found between the variables derived from Quickbird-2 satellite images and stand diversity (clark-evans index: R2=0,68, p<0.01; index of the distance between trees: R2=0,60, p<0.01; standard deviation index: R2=0,60, p<0.01; index of the average distance between trees R2=0,59, p<0.01; species richness: R2=0,51, p<0.01). The assessment was found more successful at main plot level (1 ha) (i.e. compared to the models having the highest significance level; R2 values increased 25% for Clark evans index, 39% for standard deviation index, 39% for index of the average distance between trees and 20% for species richness.). It was determined that both texture and brightness values of satellite data contributed to the estimating models. According to these results; it can be said that Quickbird-2 satellite data can be used to determine stand level diversity in the studied district or similar forest areas that do not require as much detail.