Simulation of land use land cover change in Melbourne metropolitan area from 2014 to 2030: using multilayer perceptron neural networks and Markov chain model

ABSTRACT

The explosive growth of the Melbourne metropolitan area (MMA) population in the last two decades and the spatial expansion of the city to the periphery have necessitated the analysis of land use land cover (LULC) changes. To this end, Landsat 8 Operational Land Imager (OLI) and Multilayer Perceptron (MLP) neural networks, Markov chain model, GIS and TerrSet software package were used. Initially, nine driving variables affecting the future development of the city were identified. Then, using the maximum probability estimation model, land uses were classified into six categories in the MMA with an area of 8819.6 km² for 2014, 2017 and 2020. During 2014–2020, two land uses have experienced positive changes (residential and cultivated land uses with 16.48% and 11.56% respectively) and four of them (forest cover with −5.74%, grass and green space with −9.72%, barren lands with −11.14% and water body with −1.45%) have experienced a decrease in area. To validate the prediction of land use changes until 2030, the kappa coefficient (0.617) was used. The result of using Cramer's V statistics (above 0.15) to measure the effect of driving variables on land use forecast was acceptable. LULC change was predicted by MLP and Markov chain model for 2030. The results showed that two land uses (residential and industrial land use with 26.3% and cultivated land with 6.99%) will have positive change and four of them (forest coverage with −17.11%, barren land with −9 29%, grass and green space with −5.33% and water bodies with −1.38%) should have negative growth. Spatial changes of land use will occur mostly in the northern and north-western regions on barren land, and on the east and southeast on forest and green spaces. Findings are useful for planning authorities in MMA where it is highly affected by population growth and LULC change.

KEYWORDS:

• Driver variables
• forecasting
• Markov chain
• Melbourne
• multilayer perceptron
• spatio-temporal change

Acknowledgments

I would like to thank the honourable anonymous reviewers who carefully read the article and made valuable suggestions and constructive feedback that helped to improve the quality of the article.

Disclosure statement

No potential conflict of interest was reported by the author(s).