بررسي قابليت انتقال كاربري اراضي و پوشش زمين با استفاده از روش‌هاي رويه يادگيري بر مبناي نمونه وزني مشابهت، رگرسيون لجستيك و ژئومد (مطالعه موردي: حوزه بسطام شهرستان سلسله)

Investigating the transmission potential of land use and land cover using Similarity Weighted Instance based Learning, Logistic regression and Geomod methods (Case study: Bastam basin, Selseleh city)

زﻣﯿﻨﻪ و ﻫﺪف: ارزﯾﺎﺑﯽ و ﺑﺮآورد ﻗﺎﺑﻠﯿﺖ اﻧﺘﻘﺎل ﺑﺎ ﺻﺤﺖ ﺑﺎﻻ، ﯾﮏ ﮔﺎم ﻣﻬﻢ در روﻧﺪ ﻣﺪلﺳﺎزي و ﭘﯿﺶﺑﯿﻨﯽ ﺗﻐﯿﯿﺮات ﮐﺎرﺑﺮي اراﺿﯽ و ﭘﻮﺷﺶ زﻣﯿﻦ اﺳﺖ. ﻫﺪف از اﯾﻦ ﭘﮋوﻫﺶ، ﺑﺮرﺳﯽ ﻗﺎﺑﻠﯿﺖ ﺗﻐﯿﯿﺮات ﮐﺎرﺑﺮي اراﺿﯽ و ﭘﻮﺷﺶ زﻣﯿﻦ ﺑﺎ اﺳﺘﻔﺎده از روشﻫﺎي روﯾﻪ ﯾﺎدﮔﯿﺮي ﺑﺮ ﻣﺒﻨﺎي ﻧﻤﻮﻧﻪ وزﻧﯽ ﻣﺸﺎﺑﻬﺖ، رﮔﺮﺳﯿﻮن ﻟﺠﺴﺘﯿﮏ و ژﺋﻮﻣﺪ اﺳﺖ. روش ﺑﺮرﺳﯽ: ﻧﻘﺸﻪﻫﺎي ﮐﺎرﺑﺮي اراﺿﯽ و ﭘﻮﺷﺶ زﻣﯿﻦ ﻣﺮﺑﻮط ﺑﻪ ﯾﮏ دوره زﻣﺎﻧﯽ 30 ﺳﺎﻟﻪ )1364 ﺗﺎ1394( ﺑﺎ اﺳﺘﻔﺎده از ﺗﺼﺎوﯾﺮ ﻫﺎي ﻟﻨﺪﺳﺖ 5 و 8 ﺗﻬﯿﻪ ﺷﺪ. ﻣﺪلﺳﺎزي ﻗﺎﺑﻠﯿﺖ اﻧﺘﻘﺎل ﮐﺎرﺑﺮي اراﺿﯽ و ﭘﻮﺷﺶ زﻣﯿﻦ ﺑﺎ اﺳﺘﻔﺎده از روشﻫﺎي روﯾﻪ ﯾﺎدﮔﯿﺮي ﺑﺮ ﻣﺒﻨﺎي ﻧﻤﻮﻧﻪ زﻧﯽ ﻣﺸﺎﺑﻬﺖ، رﮔﺮﺳﯿﻮن ﻟﺠﺴﺘﯿﮏ و ژﺋﻮﻣﺪ و ﻣﺘﻐﯿﺮﻫﺎي ﺗﺎﺛﯿﺮﮔﺬار در روﻧﺪ ﺗﻐﯿﯿﺮات ﺻﻮرت ﮔﺮﻓﺖ. ﻣﯿﺰان ﺻﺤﺖ ﻧﺘﺎﯾﺞ ﺑﻪ دﺳﺖ آﻣﺪه از ﻣﺪلﻫﺎ ﺑﺎ اﺳﺘﻔﺎده از ﻧﻘﺸﻪ واﻗﻌﯿﺖ زﻣﯿﻨﯽ ﺗﻌﯿﯿﻦ ﺷﺪ. ﻣﺮاﺣﻞ اﺟﺮاﯾﯽ اﯾﻦ ﭘﮋوﻫﺶ در ﺑﺎزه زﻣﺎﻧﯽ ﺳﺎلﻫﺎي 1395 ﺗﺎ 1396 اﻧﺠﺎم ﺷﺪ. ﯾﺎﻓﺘﻪﻫﺎ: ﻣﯿﺰان ﺿﺮﯾﺐ ﮐﺎﭘﺎ ﺑﺮاي روشﻫﺎي روﯾﻪ ﯾﺎدﮔﯿﺮي ﺑﺮ ﻣﺒﻨﺎي ﻧﻤﻮﻧﻪ وزﻧﯽ ﻣﺸﺎﺑﻬﺖ، رﮔﺮﺳﯿﻮن ﻟﺠﺴﺘﯿﮏ و ژﺋﻮﻣﺪ ﺑﻪ ﺗﺮﺗﯿﺐ 0/84، 0/76 و 0/67 ﻣﺤﺎﺳﺒﻪ ﺷﺪ. ﺑﺮرﺳﯽ ﻧﻘﺸﻪﻫﺎي ﭘﯿﺶﺑﯿﻨﯽ ﺷﺪه ﺑﺮاي ﺳﺎل 1409 ﺑﺎ اﺳﺘﻔﺎده از روش روﯾﻪ ﯾﺎدﮔﯿﺮي ﺑﺮ ﻣﺒﻨﺎي ﻧﻤﻮﻧﻪ وزﻧﯽ ﻣﺸﺎﺑﻬﺖ و زﻧﺠﯿﺮه ﻣﺎرﮐﻮف ﻧﺸﺎن داد ﮐﻪ ﻣﺴﺎﺣﺖ ﻣﻨﺎﻃﻖ ﻣﺴﮑﻮﻧﯽ، ﺑﺎﻏﺎت و اراﺿﯽ ﮐﺸﺎورزي روﻧﺪ اﻓﺰاﯾﺸﯽ و ﻣﺴﺎﺣﺖ اراﺿﯽ ﺑﺎﯾﺮ، ﺟﻨﮕﻞﻫﺎ، ﻣﺮاﺗﻊ و ﻣﻨﺎﺑﻊ آﺑﯽ روﻧﺪ ﮐﺎﻫﺸﯽ ﺧﻮاﻫﻨﺪ داﺷﺖ. ﺑﺤﺚ و ﻧﺘﯿﺠﻪﮔﯿﺮي: در ﻧﻬﺎﯾﺖ ﻧﺘﺎﯾﺞ ﺣﺎﮐﯽ از دﻗﺖ ﻧﺴﺒﺘﺎً ﺑﺎﻻي ﺳﻪ روش در ﺑﺮآورد ﻗﺎﺑﻠﯿﺖ ﺗﻐﯿﯿﺮات ﮐﺎرﺑﺮي اراﺿﯽ و ﭘﻮﺷﺶ زﻣﯿﻦ اﺳﺖ ﭘﺎﻣﺎ ﺑﺎ ﺗﻮﺟﻪ ﺑﻪ ﺿﺮاﯾﺐ ﮐﺎﭘﺎي ﺑﻪ دﺳﺖ آﻣﺪه، دﻗﺖ روش روﯾﻪ ﯾﺎدﮔﯿﺮي ﺑﺮ ﻣﺒﻨﺎي ﻧﻤﻮﻧﻪ وزﻧﯽ ﻣﺸﺎﺑﻬﺖ ﺑﯿﺸﺘﺮ از دو روش دﯾﮕﺮ ﺑﻮده اﺳﺖ.

Background and Objective: Assessing and estimating the high-accuracy transmission potential is an important step in the process of land use and land cover changes modeling and predicting. The aim of this study is to investigate the transmission potential of land use and land cover changes using Similarity Weighted Instance based Learning, Logistic regression and Geomod methods. Method: The land use and land cover maps for a 30-year period (1985-2015) were prepared using Landsat 5 and 8 satellite imagery. Land use and land cover transmission potential modeling was done using Similarity Weighted Instance based Learning, Logistic regression and Geomod methods and effective variables in the process of change. The accuracy of the results obtained from the models was determined by comparing with ground reality map for mentioned year. Findings: The Kappa coefficient of Similarity Weighted Instance based Learning, Logistic regression and Geomod were 0.84, 0.76 and 0.67, respectively. The investigating predicted maps for 2030 prepared by Similarity Weighted Instance based Learning and Markov chain showed that the area of residential areas, gardens and agricultural lands is increasing and the area of bare land, forests, pastures and water resources will have a decrease trend. Discussion and Conclusion: Finally, the results indicate a relatively high accuracy of three methods in estimating the transmission potential for land use and land cover changes, but according to the kappa coefficients, the accuracy of Similarity Weighted Instance based Learning method more than the other two methods.