Feasibility of automating the determination of changes in forest areas using satellite images (Case Study: Central Alborz protected area)

Deforestation is one of the biggest global environmental challenges because forests are one of the most important components of the world's ecosystem and play a key role in adjusting the climate and regulating various biochemical cycles. The forests of northern Iran are considered one of the most valuable forest ecosystems in Iran, and they are among the rich forests in terms of the diversity of plant species. Therefore, detecting changes with the help of multi-temporal data in forest levels allows us to prevent further destruction by automatically identifying these changes. The main goal of this research is to identify the thresholds and apply them to the NDVI vegetation index images in MODIS sensors and automatic monitoring of forest areas.

This research was conducted in the Central Alborz protected area with an area of more than 398 thousand hectares and very rich vegetation with more than 1100 plant species. In this research, the MOD13Q1 product of the MODIS sensor and the normalized vegetation difference index (NDVI) with a spatial resolution of 250 meters and a temporal resolution of 16 days, as well as The MCD12C1 product images of the MODIS sensor with a spatial resolution of 5600 m and a temporal resolution of one year were used in three layers with the classification schemes of the International Geosphere-Biosphere Program (IGBP), University of Maryland (UMD) and the Modis Derived Scheme (LAI). The 16-day NDVI images were converted to monthly images with the maximum value algorithm and the PCA algorithm with 25 components was applied to them to eliminate errors and noise. Using the Google Earth system, 5 random polygons were selected on virgin forest areas, in order to estimate the value of the pixels and finally determine the thresholds. Vegetation homogeneity was checked using MCD12C1 images. Forest thresholds were determined by calculating the mean and standard deviation of the sample areas during 16 years. Finally, the changes were detected by applying a threshold on the images of 2001 and 2016 and applying Boolean operations with the bolean operators on them. Then, using the Google Earth system, these points were validated.

By applying PCA to the images, the amount of noise was reduced to the optimal level while maintaining the original information of the images. The results of investigating the homogeneity of the vegetation in 5 sample areas showed that they are homogeneous in three levels and are deciduous broadleaf forests The results obtained from calculating the standard deviation and monthly average of forest values indicate that the months of June, July, and August are most suitable for assessing changes in the forest cover within the study area. The threshold for these months was obtained as 8596.55, 8000 and 8497.51 respectively. By applying these thresholds to the images captured in these three months in 2001 and 2016, and then subtracting the images from each other, the pixels that experienced a decrease in forest value were identified. Finally, after performing the Boolean operation, 200 pixels with an approximate area of 1728.23 hectares were identified as changed points with a decrease in forest cover. Upon cheching, the overall accuracy for the determined points was found to be 88.5%. Hence, the application of these thresholds enables the automatic detection of changes in forest cover areas in an up-to-date and continuous manner.

In general, the Central Alborz protected area is one of the important forest centers of the country, and according to the results, we have lost a significant number of forests during the 16 years due to various reasons. Also, the results of this research indicate the importance of using automation methods in identifying forest cover changes by applying certain thresholds on the NDVI index images of the Modis sensor in remote sensing.