Fuzzy Majority Voting-Based Fusion of Markovian Probability for Improved Land Cover Change Prediction: A Case Study of Delta, Egypt.

Abstract

This research predicts the future changes of Delta, Egypt using Landsat satellite images of 2002, 2010 and 2015 through the fusion of Markovian probability maps for improved land use/cover changes (LUCC). First, each of the three satellite images was radiometrically and atmospherically corrected. After that, two different classification algorithms were used to prepare the base maps for: 2002; 2010; and 2015, with three major classes: land areas; cultivated areas; and water bodies. The classifiers used include: Self-Organizing Map (SOM); and Classification Trees (CTs). A total of 6 uncorrelated feature attributes have been incorporated in the classification in order to mitigate the impacts of classification errors on change detection. The classified images of 2002 and 2010 were then used to predict the 2015 land use with a Markov Chain Model. As a result, two Markovian probability images were obtained, one for the SOM-based model and the other for the CTs one. The Fuzzy Majority Voting (FMV) was then applied for combining the two Markovian probability images. This would lead to an enhanced prediction of 2015 land use. At the end, the final predicted image for 2015 land use was validated with the 2015 classified image. Two stages of validation procedures were applied in this research: 1) the Kappa index of agreement (KIA) was used to validate the overall performance of the prediction process with the most accurate prediction of 0.8871 being achieved; 2) the components of agreement and disagreement were used to gain a detailed idea about the performance of the prediction process with agreement and disagreement of 82.13% and 17.87%, respectively, being achieved. On the other hand, the Markov model extrapolates that coastal area decreased from 35.80% to 35.30 percent of the total study area during 2002–2015. Finally and as compared with 2015, the prediction of 2050 land use shows 1.11% increase of water body.