چکیده این تحقیق با هدف پیشبینی پراکنش مکانی تخریب جنگلهای شمال استان ایلام و به منظور ردیابی عوامل مؤثر بر تخریب جنگل صورت گرفت. در این بررسی تأثیر هفت فاکتور فاصله از جاده و مناطق مسکونی، شاخص قطعهبندی جنگل، جهت جغرافیایی، ارتفاع از سطح دریا و شیب و همچنین فاصله از مرز جنگل و غیرجنگل بر روی تخریب جنگل، مورد مطالعه قرار گرفت. در این مطالعه، برای بررسی تغییرات جنگل، دادههای سنجنده MSS مربوط به سال 1355 و سنجنده TM مربوط به سال 1386 مورد پردازش و طبقهبندی قرار گرفتند. تصاویر مورد بررسی به دو کلاسه جنگل و غیرجنگل طبقهبندی شدند و به منظور بررسی عوامل تخریب، نقشهی تخریب جنگل با متغیّرهای مکانی فیزیوگرافی و انسانی وارد مدل شد. برای مدلسازی و برآورد پراکنش مکانی تخریب جنگلهای منطقهی مورد مطالعه از روش آماری رگرسیون لجستیک استفاده شد. نتایج نشان میدهد که در طول 31 سال حدود 19294 هکتار از سطح جنگلهای منطقه کاهش یافته است. با توجه به نتایج مدلسازی مشخص شد که در قطعات جنگلی گسسته و در مناطق نزدیک به مرز جنگل و غیرجنگل تخریب بیشتری صورت گرفته است. همچنین متغیّرهای شیب، فاصله از مراکز جمعیّتی و جاده با مقدار تخریب رابطهی عکس دارند و با افزایش ارتفاع از سطح دریا در این منطقه مقدار تخریب کاهش مییابد. در نهایت، یک مدل مکانی ساده که توانایی پیشبینی پراکنش مکانی تخریب جنگل را با استفاده از رگرسیون لجستیک دارد، ارائه شد. کلیدواژهها: مدلسازی تخریب، سنجش از دور، رگرسیون لجستیک، GIS، جنگلهای شمال ایلام.
Geography and Development
10nd Year - No. 29 - Winter 2013
Received : 25/5/2011 Accepted : 17/7/2012
PP : 10 - 13
Modeling Deforestation Using Logistic Regression, GIS and RS
Case study: Northern Forests of the Ilam Province
Dr. Saleh Arekhi
Assistant Profesor of Geography
University of Golestan Aliakbar Jafarzadeh
M.Sc of Forestry
University of Sari
M.Sc of Watershed Management
University of Tarbiat Modarres Noor
Introduction Land use and land cover are not static and are frequently subject to change due to human activities, also, with attention to the trend of deforestation in recent years, is very important to estimate the degradation in the different time periods. Access to information related to the past and realizing the changes are necessary for solving deforestation problem. Identify and detect these changes can help planners and managers to identify effective factors in land use and land cover changes and have to be useful and effective programming for their control. Iran west forests is important in terms of area, environmental issues, soil and water resources conservation which during the past decades due to social and economic factors, lack of comprehensive management and etc lost its production capacity and this trend threatens the future of the region forests. As a result, planning and management of these forests are associated with many problems which lack the necessary studies in this area will contribute to the above issue. To explore and evaluate these changes, using remote sensing and GIS techniques and tools can have a considerable effect to generate spatial information and having analytical capabilities. In the present study, using the information resulting from comparing satellite images of two different periods, identify changes trend in the studied forests and then with applying GIS analyzes to identify effective factors in these changes can measure and the results presented in thematic maps and finally suggested degradation probability model for the study area.
Research Methodology In this research, remote sensing data include images of MSS (1976) and TM and also 1:50.000 topographic maps and 1:20.000 aerial photos of region were used. In order to classifying and preparing forest map related to the study years, training samples are prepared with help of ground operations using GPS device and also using processing and satellite images enhancement in environment of IDRISI software and finally, training sites of forest and non-forest areas were prepared. Also, the spectral response of different bands for each of training site classes were drawn and interpreted. After geo-referencing satellite images using ground control points, map of forest extent related to years of 1976 and 2007 through classification on original and processed images of MSS and TM was prepared. After preparing maps of forest extent related to years of 1976 and 2007, each of the mentioned maps were classified into two categories of forest and non-forest and after ensuring the accuracy of the maps produced in GIS environment for preparing changes map in forest area, both maps at the beginning and end of the studied period were crossed with each other. The most important factors in the forest degradation trend are the natural and human factors .In this research, map of slope, aspect, elevation classes as natural factors affecting changes and distance from residential areas, roads, distance from the edge of the forest and forest fragmentation index as human factors were considered in the event of changes. In this direction, map of above factors using Arcview and Idrisi software was prepared in GIS environment and was used to analyze. For preparing maps of classes of slope, aspect, elevation, digital elevation model of the study area should be prepared.
Modeling Deforestation Using Logistic Regression, GIS and RS …
Discussion and Results Accuracy assessment results of classified maps are presented in the related table. With consideration to high amounts of 83% of overall accuracy, these maps can be used for preparing degradation map. After processing the satellite images, amount map and spatial distribution of forest areas in study area was prepared in years of 1976 and 2007 with use of MSS and TM satellite images. With overlay the classification results of two periods, occurred changes map was obtained. From this map, we can extract amount of degraded forest and location of these changes. Results obtained of comparison of two maps related to beginning and end of the time period (1976 and 1388) shows that during this period, 19294 hectare of forest regions area has been decreased. In logistic regression, chi-square statistic (-2Log Likelihood) or (-2LL) is widely used. When a model has a poor fit, having large amount and when a model has a good match, its value is small. In this study, chi-square value was 117.309. Coefficients of the variables in the regression equation are very important and logit equation coefficients list. 1 equation shows statistical model of degradation probability prediction obtained of logistic modeling with seven variables: slope, aspect, elevation, distance from roads and population centers, distance from the edge of the forest and forest segmentation index. Pseudo R2 for the model is 0.1608 and the ROC regression coefficient amount was equal to the amount of 0.7678. Based on this equation, the spatial distribution map of the study area forests was obtained. (1) Forest degradation probability = -4.2303+0.007207(aspect)- 0.020488(distance from villages) - 0.001495(elevation from sea level) – 0.00116 (slope) -0.001908 (distance from road) – 0.087687 (distance from the forest edge) + 0.14544 (forest fragmentation pattern)
Conclusion This study has been taken place with objective of estimating the spatial distribution of Zagros forests and in order to detect effective factors on forest degradation. In this study, the effect of seven factors, distance from roads and residential areas, forest fragmentation index, aspect, elevation levels, slope and distance from forest and non-forest edge on forest degradation rate was studied In this study, to investigate changes in forest, MSS sensor related to year of 1976 and TM sensor data related to year 2007 were processed and classified. The Studied images were classified in to two classes of forest and non-forest and in order to study the degradation factors, map of forest degradation with spatial variables of physiographic and human were entered into the model. For modeling and estimating the spatial distribution of studied forests degradation, of logistic regression statistical method was used. Output of regression logistic with Pseudo R2 equal to 0.1608 and ROC of 0.76 represents a relative agreement with the actual degradation and ability to fit the model to estimate changes in forest area. Also, with consideration to negative coefficients related to distance from residential areas and roads can say with decreasing the distance then these factors, urban expansion and man-made regions development, road construction and increase in population plays role in forest degradation of the study area.
1.Amini M.R, Shataee S, Moaieri M.H, Ghazanfari H (2009). Deforestation modeling and investigation on related physiographic and human factors using satellite images and GIS (Case study: Armerdeh forests of Baneh), Iranian Journal of Forest and Poplar Research. 16 (3). 2.Bagheri R, Shataee S (2010). Modeling loss of forest extent using the logistic regression, Case Study: Ghehal Ghay watershed of the Golestan province, Iran Journal of Forest, Iran forestry society, second year, 2 (3). 3.Braimoh A, Vlek P.L.G (2003). Modeling land use change in Northern Ghana, Conference, technological and institutional innovations for sustainable rural development, Gottingen, Germany. 21. 4.Clark W.A, Hosking P.L (1986). Statistical Methods for Geographers (Chapter 13), New York: John Wiley & Sons. 5.Eastman J. R (2002). Idrisi for windows user’s guide ver.32, Clark labs for cartography technology and geographic analysis, (Clark University). 6.Giriraj A, Ullah M.I, Murthy M.R, Beierkuhnlein C (2008). Modelling Spatial and Temporal Forest Cover Change Patterns (1973-2020). A Case Study from South Western Ghats (India). Sensors, 8. 7.Gomez-Mendoza L, Vega-Pen A.E, Ramirez M.I, Palacio-Prieto, J.L, Galicia L (2006). Projecting land-use change processes in the Sierra Norte of Oaxaca, Mexico. Applied Geography, 26. 8.
Modeling Deforestation Using Logistic Regression, GIS and RS …
Gruenberg W.D, Curtin P, Shaw W (2000). Deforestation Risk for the Maya Biosphere Reserve, Guatemala. School of Renewable Natural Resources, The University of Arizona, Tucson, Arizona, USA.
9.Jat M.K, Khare P.K, Khare D (2008). Monitoring and modelling of urban sprawl using remote sensing and GIS techniques. International Journal of Applied Earth Observation and Geoinformation, 10. 10.Mas J.F, Puig H, Palacio J.L, Sosa-Lopel A (2004). Modeling Deforestation using GIS and Artificial Neural Networks. Environmental Modeling & Software, 19. 11.Matheron G (1970). La the´orie des variables ge´ne´ralise´es et ses applications. Les Cahiers du Centre de Morphologie Mathe´matiques de Fontainebleau, 1 (2). 12.Matthew L, Robert J, Smith R.J, Nigel L.W (2004). Mapping and predicting deforestation patterns in the lowlands of Sumatra. Biodiversity and Conservation, 13. 13.Mertens B, Lambin E. F (1999). Modelling land cover dynamics: integration of fine-scale land cover data with landscape attributes. International Journal of Applied Earth Observation and Geoinformation, 1. 14.Mesghari S (2002). Investigating Changes in forest areas using GIS and remote sensing, Tehran: Researcgh Project, Technical College, Khajeh Nasir Toosi University. 15.Miriam S.W, Taylor V.S (2010). Modeling social and land-use/land-cover change data to assess drivers of smallholder deforestation in Belize. Applied Geography 30. 16.16. Njarluo S (2005). Investigating Changes in forest area using aerial photographs, topographic maps and IRS and ETM+ images, M.Sc thesis of Forestry,Gorgan University of Agriculture and natural resources. 17.Pierre Bavqa M (2004). Investigating Changes of forest extent in relationship with topographic factors and man-made areas, Case Study: Eastern Forests of the Gilan province, M.Sc thesis of forestry, Tehran University. 18.Pontius R.G, Schneider L (2001). Land-use change model validation by a ROC method for the Ipswich watershed, Massachusetts, USA. Agriculture, Ecosystems and Environment. 85(1-3). 19.Rafieian O, Darvishsefat A.A, Namiranian M (2006). Determining changes extent of the Iran northern forests between the years of 1993 to 2001 Using ETM+ Images, Journal of Agriculture and Natural resources sciences and technology, 10 (3). 20.Ranjbar A (2002). Investigating and estimating forests degradation trend using GIS and remote sensing data, M.Sc thesis of remote sensing, Khajeh Nasir Toosi University. 21.Scheer L, Sitko R (2007). Assessment of some forest characteristics employing ikonos satellite data. Journal of Forest Science, 53. 22.Shataee S, Hosseinali Zadeh M, Ayobi S (2007). Investigating capability of ETM+ spectral data in estimating the amount of organic matter in the surface soil, Rangeland Journal, first year, 1.