Evaluating the Pattern of Forest Cover Changes Using Fuzzy Object-Oriented Techniques (Case Study: Kaleybar County)

Forest monitoring requires an automatic systems to analyze large-scale remote sensing data. Nowadays with development of remote sensing technology, a large amount of spatial data is available. Satellite data is the fastest and cheapest method for researchers to provide land cover mapping. Considering the wide variation of forest usage and destruction in recent years, producing the map of the forest area and examining the process of its changes in certain time periods is necessary. In this research, we tried to investigate land cover changes in Kaleybar county during a 27-year period, especially forest cover using fuzzy object-oriented techniques using Landsat imagery. For this purpose, the TM Landsat 5 (1990), the ETM Landsat 7 (2000 and 2010) and the OLI Landsat 8 image for 2017 were used. In this study, ESP algorithm was used to optimize the scale in order to improve the image segmentation results and subsequently increase the accuracy of classification results. The result shows that there has been a decline in lands with forest cover and 1st grade pastures in Kaleybar county during the period from 1990 to 2017. However, we see the increment in 2nd grade pastures, arid and residential lands, which indicates the general trend of destruction in the region through the replacement of 1st grade pastures and forest lands by other uses such as 2nd grade pasture and arid and residential areas. In 27 years, the forest lands have fallen by 5.1 percent that is equivalent to 107 square kilometers. There are many factors affecting forest cover changes in the region, including the increase of habitation centers, deforestation and conversion of forest to farmlands.
Extended Abstract
1- Forest monitoring requires an automatic systems to analyze large-scale remote sensing data. Nowadays with the development of remote sensing technology, a large amount of spatial data is available. Utilization of the satellite data provides the fastest and cheapest method for researchers to prepare land cover mapping. In order to obtain a proper planning and management for natural resources, especially forests, accurate and timely information maps are required. Considering the wide variation of forest usage and destruction in recent years, it is necessary to prepare maps of the forest areas and to examine the changes occurred in them during certain time periods. The Arasbaran forest habitats, which were covering a large area in the past, are nowadays limited to the small parts of the Kaleybar, Khodafarin, Ahar and Jolfa counties in East Azarbaijan province with the total area of 140,000 hectares. Vegetation of Kaleybar county is very rich and important compared with other parts of the East Azarbaijan province and encompasses large forests with a variety of rare trees and natural grasslands, though it has suffered many changes in recent years. Therefore, this research tries to use object-oriented method, especially fuzzy object-oriented to increase the accuracy of Landsat images classification and the land use change trend, especially the forest cover of Kaleybar, for the period of 27 years from 1990 to 2017, while high spatial resolution satellite images are not available.
2- The Kaleybar county has an area of about 2112 Km2 and covers 3.2 percent of East Azarbaijan province in its northeast. The purpose of this study is to evaluate the changes in the Kaleybar county with an emphasis on forest lands. For this aim, Landsat satellite images of TM Landsat 5, ETM Landsat 7 and OLI Landsat 8 with range of 1868 to 3300 from were processed with the eCognition Developer software for the period of 1990 to 2017 with 10 year time series. The fuzzy object oriented approach was used to extract the land cover of different vegetation indices, as well as homogeneous texture data, shape, compression and brightness. The results were then calculated and finalized in ArcGIS software after accurate evaluation.
3- In this research, images in 200 scales sorted consecutively from 1 to 200 were segmented using low to high multi-functional hierarchical segmentation approach with shape coefficient of 0.4 and compression coefficient of 0.5 in order to construct LV graphs and the appropriate scales for image segmentation were determined using the plotted graphs. By predicting the appropriate scale for creating image units using the algorithm (ESP), the scale of 15 with coefficients of shape and compression 0.3 and 0.5 respectively was scaled as the appropriate scale for extraction of Landsat 5 and 7 satellite images, and the scale of 130 with shape coefficient of 0.4 and compression coefficient of 0.5 was chosen as the appropriate scale for Landsat 8 satellite OLI images. Classifying the selected images using the fuzzy object-oriented method, the land cover changes were calculated and mapped in the Arc GIS software. For the 27-year period, the largest changes have occurred in the forest and inferior lands. The difference was that forest lands have declined with negative gradient, and consequently, inferior lands have increased with positive gradient. In 27 years, forest land areas have decreased by 5.1 percent, equivalent to 107 Km2.
4- The purpose of this study was to detect land cover changes in Kaleybar county during 27-year period. For this, remote sensing satellite imagery was used and after preparing the land cover map for all four time periods, the area of six classes of land cover was obtained and the land cover changes map was extracted. To better comparison of the changes occurred in these four periods, these changes were quantitatively calculated. Results show that during the period from 1990 to 2017, there has been a decline in lands with forests and 1st grade pastures. On the other hand, it is seen that 2nd grade pastures, Bayer lands and residential have increased, which indicates the general trend of destruction in the region through the replacement of 1st grade pastures and forest lands by other uses such as 2nd grade pasture and arid and residential areas. During this 27 year period, the forest lands have decreased by 5.1 percent that is as equal as 107 km2. In this research, classification results with >90% accuracy for each of the four image periods of imaging indicate the ability of the fuzzy object-oriented method in land cover studies. The method applied in this study can determine the land cover changes over time, and also determine the land degradation trend quantitatively and accurately.