Urban Growth Modeling Using Cellular Automata with Multi-Temporal Remote Sensing Images Calibrated by the Artificial Bee Colony Optimization Algorithm

Fereydoun Naghibi,Bryan Pijanowski,Mahmoud Delavar

doi:10.3390/s16122122

Fereydoun Naghibi, Bryan Pijanowski + Show 1 more

Open Access

https://doi.org/10.3390/s16122122

Copy DOI

Journal: Sensors	Publication Date: Dec 14, 2016
Citations: 29	License type: CC BY 4.0

Affiliation: Purdue University West Lafayette, University of Tehran

Abstract

Cellular Automata (CA) is one of the most common techniques used to simulate the urbanization process. CA-based urban models use transition rules to deliver spatial patterns of urban growth and urban dynamics over time. Determining the optimum transition rules of the CA is a critical step because of the heterogeneity and nonlinearities existing among urban growth driving forces. Recently, new CA models integrated with optimization methods based on swarm intelligence algorithms were proposed to overcome this drawback. The Artificial Bee Colony (ABC) algorithm is an advanced meta-heuristic swarm intelligence-based algorithm. Here, we propose a novel CA-based urban change model that uses the ABC algorithm to extract optimum transition rules. We applied the proposed ABC-CA model to simulate future urban growth in Urmia (Iran) with multi-temporal Landsat images from 1997, 2006 and 2015. Validation of the simulation results was made through statistical methods such as overall accuracy, the figure of merit and total operating characteristics (TOC). Additionally, we calibrated the CA model by ant colony optimization (ACO) to assess the performance of our proposed model versus similar swarm intelligence algorithm methods. We showed that the overall accuracy and the figure of merit of the ABC-CA model are 90.1% and 51.7%, which are 2.9% and 8.8% higher than those of the ACO-CA model, respectively. Moreover, the allocation disagreement of the simulation results for the ABC-CA model is 9.9%, which is 2.9% less than that of the ACO-CA model. Finally, the ABC-CA model also outperforms the ACO-CA model with fewer quantity and allocation errors and slightly more hits.

Highlights

In recent decades, rapid urbanization has led to many negative impacts on the environment, such as the loss and fragmentation agricultural lands and of natural areas that support wildlife
To model this process several urban growth models have been presented such as Markov chain models [19], spatial logistic regression [20], multi-criteria evaluation [19], cellular automata (CA) [16,21,22,23], agent-based models [17,24] and machine learning and artificial intelligence (AI)
Of the simulated the actual modelland-use prediction capability, themade predicted map isofverified by the reference of similarities

Summary

Introduction

Rapid urbanization has led to many negative impacts on the environment, such as the loss and fragmentation agricultural lands and of natural areas that support wildlife To avoid these impacts, anticipatory planning has to be considered [1]. Urban growth models have been proposed to use the capabilities of a new generation of spatial analysis tools within the geospatial information systems (GIS) framework. They investigate urban regions using various multi-temporal datasets such as remote sensing images to detect changes over the time [2,3,4,5,6,7]. To model this process several urban growth models have been presented such as Markov chain models [19], spatial logistic regression [20], multi-criteria evaluation [19], cellular automata (CA) [16,21,22,23], agent-based models [17,24] and machine learning and artificial intelligence (AI)

Results

Discussion

Conclusion