Scenario-based land cover change modeling and its implications for landscape pattern analysis in the Neka Watershed, Iran

AuthorsJoorabian Shooshtari, S., Gholamalifard, M.
JournalRemote Sensing Application: Society and Environment
Paper TypeFull Paper
Published At2015-07-15
Journal GradeISI (WOS)
Journal TypeElectronic
Journal CountryNetherlands

Abstract

Land cover changes and urbanization cause destruction of natural habitats and threaten biodiversity. Land cover modeling is one of the most important procedures for evaluating this trend. This study was performed with the objective of comparing multi-layer perceptron (MLP) artificial neural network with logistic regression (LR) in predicting land cover change and quantifying future landscape change using landscape metrics in the Neka River Basin, a small part of the eastern Hyrcanian forest, in northern Iran. For this purpose, first, change analysis was carried out using satellite imagery, from 1987 to 2011. Then, transition potential modeling was conducted using MLP and LR in 5 different scenarios. A Relative Operating Characteristic (ROC) analysis was carried out to detect the degree of correlation between variables and transitions in LR. In addition, the accuracy rate for assessing the transition potential modeling using MLP was employed. Land cover change prediction was conducted using prediction for 2011 and 2017. The accuracy assessment model was determined by comparing the actual land cover map of 2011 with the predicted land cover map of 2011. Landscape indices for 1987, 2001, 2006, 2011, and 2017 were calculated and analyzed using Fragstats to determine the impact of land cover change on landscape fragmentation. The result showed that during 1987–2001, agriculture was the main contributor to the increased built-up area. The most important transition was the conversion of agriculture to orchard and residential, between 2001 and 2006. Forest regenerated from orchard and agricultural lands, between 2006 and 2011. The maximum and minimum amounts of Cramer's V were obtained for the empirical likelihood to change variable and distance from the river. Overall kappa obtained in the best scenario based on LR was 88%. Furthermore, the prediction results showed that major deforestation will occur in surrounding forest areas and most residential development is in the outskirts of the town of Neka. Increased fragmentation in the landscape will continue in 2017, more shape complexity will be observed, and habitats of the Neka Basin will become more diverse and abundant. The results of this study provide useful information for Reducing Emission from Deforestation and Forest Degradation (REDD) project.