مدل‌سازی تخریب جنگل‌های زاگرس با استفاده از رگرسیون لجستیک (مطالعه موردی: جنگل‌های چرداول استان ایلام)

از آنجا که تغییر کاربری اراضی و تخریب جنگل‌ها، نشان‌دهندۀ ارتباط مستقیم و متقابل انسان و محیط‌زیست طبیعی آن است، درک بهتر فرایندهای اجتماعی و بیوفیزیکی که ایجادکننده تغییرات و تخریب اراضی هستند، می‌تواند نقش مهمّی در سیاست‌گذاری و اجرای اقدامات پیشگیرانه و تصمیم‌ها داشته باشد. به‌منظور بررسی روند تخریب پوشش جنگلی شهرستان چرداول در استان ایلام، از تصاویر سنجنده‌های پیمایش‌گر چندطیفی و تصویربردار عملیاتی زمین ماهواره لندست مربوط به سال‌های 1366 و 1393 و روش مدل‌سازی رگرسیون لجستیک استفاده شد. برای بررسی عوامل تخریب، نقشه تغییر پوشش جنگلی با متغیّرهای فیزیوگرافی (شیب، جهت و ارتفاع) و انسانی (فاصله از جاده و فاصله از مناطق مسکونی) وارد مدل رگرسیون لجستیک شد. نتایج پژوهش نشان داد طی 27 سال، حدود 10332.82 هکتار از جنگل‌های شهرستان چرداولتخریب شده است که نشان از کاهش سالانه 382.67 هکتار از سطح جنگل‌های منطقه دارد؛ همچنین، نتایج مدل‌سازی نشان داد که متغیّر جهت دامنه با دارا بودن بیشترین ضریب تاثیر (0.7267)، شاید مهم‌ترین عامل بیوفیزیکی تاثیرگذار بر تخریب جنگل در منطقه مورد مطالعه بوده است؛ پس از آن، به‌ترتیب متغیّرهای شیب و ارتفاع از سطح دریا در تخریب احتمالی جنگل تاثیرگذار بودند. متغیّرهای فاصله از روستا و فاصله از جادّه هم رابطه معکوس با مقدار تخریب در منطقه مورد مطالعه دارند. ارزیابی مدل رگرسیونی برازش‌داده‌شده با شاخص‌های ویژگی عملیاتی نسبی (معادل 0.8493) وضریب تشخیص کاذب (معادل 0.2248) هم بیانگر قابلیت بالای مدل به‌منظور توصیف تغییرات و تعیین مناطق مستعد تغییر است.با توجّه به سرعت تخریب سالانه جنگل در این منطقه که بیشتر از متوسّط جهانی است، در صورت عدم برنامه‌ریزی پیشگیرانه توسّط برنامه‌ریزان استانی و کشوری، شاید در آینده‌ای نه‌چندان دور شاهد پدیده بیابان‌زایی در شهرستان چرداول باشیم.

Modelling the Trend of Zagros Forest Degradation using Logistic Regression (Case study: Chardavol Forest of Ilam province)

Since land use change and forest degradation represent a direct and interrelated relationship between human and their natural environment, understanding the social and biophysical processes that cause land use change and degradation can play an important role in policing and implementing preventive measures and decisions. In order to investigate the forest cover degradation trends of Chardavol county in Ilam province, satellite images of MMS and OLI Landsat sensors for the years 1987 and 2014 and regression logistic modeling methods were used. To investigate the causes of degradation, forest cover changes map and physiographic (slope, aspect, altitude) and human (distance to road and distance to residential areas) variables were integrated into regression logistic model. The results of study showed that about 10332.82 ha of forest cover has been reduced in Holeilan division of Chardavol County during 27 years. This amount of forest cover reduction includes 382.68 ha annually. In addition, the results of modelling showed that aspect variable with the highest coefficient (0.7267) is probably the most biophysical factor affecting on deforestation in the study area. After that, slope and altitude variables probably affected deforestation, respectively. Distance to villages and road variables in study area are both inversely related to the amount of forest degradation. Assessment of regression model fitted with ROC (0.8493) and PseudoR2 (0.2248) indices indicated the ability of the model to describe the changes and to identify the areas prone to change. According to the annual rate of deforestation in the area, which is more than global average, in the absence of proactive planning by provincial and national planners, perhaps, we will see the desertification phenomenon in Chardavoul County in the near future. Extended Abstract 1Introduction Zagros forests have long been the habitat of the inhabitants and nomads in these areas and have been exposed to many damage. The issue of degradation and reduction of Zagros forests has emerged as one of the crises in recent years. Destruction and land use /land cove changes represent a variety of social and environmental factors. Since land use change and forest degradation represent a direct and interrelated relationship between human and their natural environment, understanding the social and biophysical processes that cause land use change and degradation can play an important role in policing and implementing preventive measures and decisions. 2Materials and Methods Satellite images of MMS and OLI Landsat sensors for the years 1987 and 2014 and regression logistic modeling methods were used in order to investigate the forest cover degradation trends of Chardavol County in Ilam province. Image method with 46 ground control points was used to do the geometry correction of the images of 2014. In order to do the geometry correction of images of 1987, after correction of the image of 2014, the image to image method with 42 ground control points was used. The supervised classification method of support vector machines was used to classify the satellite images of the respective years. Conducting investigations in the studied area as well as reviewing past research in similar areas to identify the important factors of forest degradation in the region, it turned out that five factors (elevation to sea level, aspect, slope, distance to village, and distance to road) are more important and effective in destroying forests in the region. Then, forest cover changes map and physiographic (slope, aspect, altitude) and human (distance to road and distance to residential areas) variables were integrated into regression logistic model. 3Results and Discussion The results of the supervised classification in the studied area were compared and statistically analyzed for classification accuracy using general and Kappa reliability coefficients, as the images of years 1997 and 2014 had a total accuracy of 86.11 and 86.39%, respectively. The results of study showed that about 10332.82 ha of forest cover has been reduced in Chardavol County during 27 years which indicates an annual decline of 382.68 hectares or 0.37 percent of the initial level of forests in the area. This amount of forest cover reduction includes 382.68 ha annually. In addition, the results of modelling showed that aspect variable with the highest coefficient (0.7267) is probably the most biophysical factor affecting deforestation in study area. After that, slope and altitude variables probably affected deforestation, respectively. Distance to villages and road variables in study area are both inversely related to the amount of forest degradation. That means the greater the distance from the road and the village, the less the decline in forest cover or forest destruction is. The results of degradation in different aspects indicate that the flat areas in the area have the highest damages (4292.96 ha) and the west aspects had the least amount of destruction (278.42 ha). Assessment of regression model fitted with ROC (0.8493) and PseudoR2 (0.2248) indices indicated the ability of the model to describe the changes and to identify the areas prone to change. To further improve the model, the more variables such as socioeconomic data (population, average income level and welfare of residential areas, number of livestock in residential areas) climate data and so on can be included in the prediction model. 4Conclusion Considering the comparison of the results of this study and the factors affecting the process of forest degradation in this region, further studies needed to investigate the factors affecting forest degradation in each region. Because the factors affecting forest degradation are often specific to each region and are different from other areas, even if the factors are common, their degree of importance will vary from region to area. According to the annual rate of deforestation in the area, which is more than global average, in the absence of proactive planning by provincial and national planners, perhaps, we will see the desertification phenomenon in Chardavoul county in the near future. Since spatial modeling is a good tool for better understanding of the causes of land use/land cover changes, it is hoped that the results of this research will be considered in future planning that is relevant to landuse /cover change.