Page 792 - NGTU_paper_withoutVideo

Modern Geomatics Technologies and Applications

Land use change detection and prediction using Similarity Weighted Instance-based
Learning, A Case Study: Tehran, Iran

1
2*
Ali Babaeian , Parham Pahlavani , Behnaz Bigdeli 3

1 GIS M.Sc. Student at School of Surveying and Geospatial Engineering, College of Engineering, University of
Tehran, Tehran, Iran
2 Assistant Professor at School of Surveying and Geospatial Engineering, College of Engineering, University of
Tehran, Tehran, Iran
3 Assistant Professor at School of Civil Engineering, Shahrood University of Technology, Shahrood, Iran
* pahlavani@ut.ac.ir

Abstract: The development of cities cannot be considered useful or harmful itself, but it will have irreparable
consequences if this development is unplanned and unbridled. Unplanned land use changes in cities not only disrupt
urban management but also cause damage to the environment. Therefore, modelling and predicting these changes can
play a significant role in urban management planning. In this study, a way to model and predict multiple land changes
has been provided. In this regard, a similarity weighted instance-based learning method was used. In this study, Landsat
satellite images were used in 2002, 2008 and 2014 to extract the land use map using the support vector machine (SVM)
classification method. Modelling was performed to reach the probability of change map, where pixels with higher
probability indicated that they belong to the intended land use class. The Multi Objective Land Allocation (MOLA)
method then identified potential areas for land use change for each land use class for 2020, using maps of the probability
of land use change from the changeable area between 2002 and 2008. Kappa coefficients are obtained for two algorithms.
Results showed the high capability of the proposed method used.

Keywords: Markov chain, Cellular automata, Land use change Prediction, Similarity Weighted Instance-based Learning

1. Introduction

Various factors such as migration to cities and increasing birth rates and socio-economic factors cause land use changes.
Population increase leads to the need for housing and the need for educational facilities and the need for recreational places, and
as a result, agricultural lands and forests become buildings and man-made artefacts. The extinction of plants and agricultural
lands not only damages the environment, but also causes economic losses. Construction, if not comprehensively planned and
managed, will cause traffic and overcrowding in infrastructure-free areas. In this regard, a vision of the future, provided by
modelling would be needed for experts and politicians to plan for the future.
Mozumder et al. [1] compared logistic regression and MLP methods to evaluate their applicability for built-up transitions. They
simulated land use changes for the 1989-2001 period to produce transition potential maps for 2011 and validated their results
with multi-regression validation method. Results indicate that MLP method predicted changed areas with more accuracy than
logistic regression method.
Shrestha et al.[2] employed SimWeight model to calculate land use change maps and evaluated these changes on streamflow
and sediment in a basin. They concluded that land use transition potential modelling can result in spatial variations of change.
Therefore, land use demand uncertainty causes the highest streamflow and sediment load changes.
Bununu [3] simulated urban expansion in Nigeria. They used SimWeight to calculate transition potential maps. Satisfactory
outcomes from relative operating characteristic and kappa index of agreement demonstrated ability of method to model transition
potential.
Due to recent advances in image-based GIS, it is possible to use a variety of digital data and satellite images,
simultaneously. By conducting this research, it is possible to use GIS such as data collection, storage, management, simulation,
analysis, and display of complex spatial information to evaluate various and effective aspects of urban development at the lowest
cost and with appropriate accuracy. On the other hand, having enough information about the future of urban dispersal in the

1