Monitoring and Prediction of Urban Growth Using Multitemporal Images and GIS Techniques (A Case Study of Bojnourd City)

Document Type : Research Paper

Author

Abstract

Introduction:

Rapid population growth and human activities have resulted in unsustainable exploitation of natural resources. Studying land changes, change detection and prediction are essential for analyzing developmental consequences over time and also for decision-making and implementing appropriate policy responses relating to land uses.
Land-Use Change modeling have become a top research topic in many scientific field recently. There are many approaches and models in use to perform spatially simulations, but among many land use modeling tools, LCM offers many advantages.
This modeler evaluates land changes between two different times, calculates the changes trend, gain and loses, Persistence and displays the results with graphs and maps. There are three approaches to produce the probability map: logistic regression, multilayer perceptron (MLP) and a similarity-weighted (SimWeight).
Multi-layer perceptron (MLP) creates a transition potential map for each group of transitions in order to allocate the simulated transitions. It use a Markov matrix to extrapolate the quantity of each transition and persistence.
Markov matrix generally obtains through the comparison of the land use maps from two dates. Markov chain projection provides the model with the estimated areas of each land use category for future dates and the amount of change for each transition.
LCM needs explanatory variables to improve the understanding of the causes, locations, and trends of land use changes. These variables are selected when they exhibit relatively high Cramer coefficient values for land cover categories. Cramer’s Coefficient indicates the degree to which each explanatory variable is associated with the distribution of land cover categories.
This study demonstrated that human disturbance include road and urban distance were the key factors in determining transition model in land change prediction.
We predicted land use changes in Bojnourd city using multi-temporal remote sensing data and a multi-layer perceptron (MLP) neural network with a Markov chain model. Remote sensing techniques applied to classify satellite imagery for 2001, 2008 and 2015.
Several studies have developed different modeling methods to simulate the pattern and consequences of land use changes. Arekhi (2011) modeled deforestation using logistic regression, GIS and RS in the Iran's west forests. In this research the effects of seven factors, distance from roads and residential areas, forest fragmentation index, distance from forest edge, aspect, elevation and slope was studied. The results indicated that with decreasing the distance from residential areas and roads, forest degradation will be more and the most of the deforestation occurred in the fragmented forest cover.
In other reaserch Joorabian Shooshtari and Gholamalifard (2015) explored changes in landscape pattern in northern Iran's Neka Basin for 1987, 2001, 2006, 2011, and 2017. Their studies revealed that during 1987–2001, agriculture was the main contributor to the increased built-up area, between 2001 and 2006 agriculture converted to orchard and residential, and between 2006 and 2011 forest regenerated from orchard and agricultural lands.
Numerous studies have assessed urban growth with different modeling methods around the world. These studies though mapped and focused on determining whether a change has occurred and how the change has evolved over time.
Wilson & Weng (2011), studied impacts of urban land use and climate changes on surface water quality within Des Plaines River watershed, Illinois. Low density residential growth, normal urban growth, and commercial growth are three future scenarios in this study that specified with Land Change Modeler (LCM).

Materials and methods:

Land Change Models can be very useful tools for environmental and urban growth research concerning about land use change. LCM was used in this research to predict the land use map in 2015 using the following procedure: Change analysis and choice of explanatory variables, Transition potential modelling, Change prediction map and Model assessment.
A total area of 14438/03 hectares of Bojnourd city was taken as study area which has potential for expansion. IDRISI Andes was used to determine land change using three different land-use maps from 2001, 2008 and 2015. In This Study, A series of satellite images of Landsat Enhanced Thematic Mapper Plus (ETM+) and Operational Land Imager (OLI) data (2001, 2008, and 2015) respectively were used to produce classified land use/cover map. It's necessary to assess the satellite data for their image quality.
Maximum likelihood classification were used to derive 3 land use categories in the study area. This way is based on the probability density function that is associated with a particular training site signature.
Accurate assessment land use maps, using ground control points, visual interpretation and Google earth were controlled. The classification accuracy and kappa coefficients was evaluated for land use maps.
The Land Change Modeler module in IDRISI software was utilized for land change detection and change trend analysis. For change analysis and prediction, first of all Land use maps for 2001 with 2008, 2008 with 2015, and 2001 with 2015 were used for analysis and detection of changes. Net change, Gains and losses, persistence and other modules were used to evaluate map transition potential.
The study used several variables including distance from road, distance from settlement, distance human disturbance, distance from vegetation edge, slope and qualitative variables.
Two land cover maps of two different times (2001 and 2008) were applied to predict potentially transition in the future. LCM available as IDRISI and ArcGIS extension) is a useful tool for the assessment and projection of land cover changes. Different modules are available to do this like cross tab module, gains and losses and etc.

Results:

The land-use maps were produced by supervised maximum likelihood classification and 3 classes (settlement areas, vegetation and burren land) were considered. Kappa coefficients obtained in this study was above 80%.
In this study, 3 major land use categories identified and mapped after field surveys, literature reviews and visual interpretation. Neural network training was carried out with the default setting (learning rate from 0.01 to 0.001, momentum 0.5, number of hidden nodes calculated as the average between numbers of input and output nodes, 10,000 iterations).
4 sub-models were identified which included burrenland to settlement areas, vegetation to settlement areas, vegetation to burrenland and burrenland to vegetation. Land use map of 2015 was predicted by using changes that occurred during the years 2001 and 2008.
The analysis of changes shows the expansion in settlement areas (1203 ha), burgenland (737 ha), vegetation area (554 ha) during the years 2001– 2015. The land use change analysis for the next period (2008-2015), indicates that the area of settlement areas has increased. The transition from Burgenland and vegetation to settlement areas was 980 ha.



Discussion and conclusion:

The objective of this research is to evaluate LCM as a land use model, focusing on its predictive power for the assessment of transition potential.
This study used Landsat ETM+ imageries of 2001 and 2008, and OLI/TIRS of 2015 to identify, classify, Assess and interpret changes in a city area. The land cover categories and their changes for these years were generated and analyzed in the Idrisi environment.
Land use scenarios for 14 years from 2001 to 2015 was performed using Markov analysis in LCM module in IDRISI software.
Land Change Models can be very useful tools for environmental and urban growth research concerning about land use change. Rapid urban growth in last decades is a big problem which prompt concerns about environmental issues over the accompanied environmental issues and the degradation of economical sustainability.
Land use maps are very vital for decision makers and environmental management purposes to evaluate land changes and the Causes of land degradation. LCM provides comparable and understandable maps and graph to demonstrate natural and environmental conditions. Monitoring land changes can provide valuable information for regional management and planning, but it's not enough. Prospective simulation supports decision-making for urban planner and environmental management. LCM, provides great advantages such as better monitoring changes, describe change trend, quantifying changes and also it can answer to questions with different scenarios like "what would happen if ".

Keywords

Main Subjects