Land Cover prediction with Urban Sprawl Indices: Leveraging Machine Learning and Stochastic Methods

Urban sprawl, an inherent aspect of contemporary urbanization, necessitates precise prediction and management strategies to mitigate its adverse ramifications on urban landscapes. This study delves into the urban dynamics of Tehran, employing sophisticated machine learning methodologies and satellite imagery to scrutinize patterns of urban expansion. By leveraging the XGBoost algorithm alongside Landsat satellite images from 2011 and 2016, the research illuminates the shifting land cover configurations within the city. In the preparation of the 2021 land cover map, numerous urban sprawl measurement indices, including Elevation, Vertical Density, Gross Population Density, Net Population Density, Fractal Dimension, Immigration Rate, and Land Use Mix, were harnessed to train the Random Forest algorithm. This integrated approach, coupled with a Markov chain, yielded a precision of 0.8838, affirming its efficacy through comparison with this year's reference map. This precision underscores the superiority of this method over the CA-Markov alternative, which achieved a precision of 0.8658. By scrutinizing these methodologies and elucidating Tehran's urban sprawl dynamics, this study offers valuable insights to the academic discourse on urban management, facilitating the pursuit of sustainable development in Tehran and analogous urban contexts.

کلیدواژه های Land Cover prediction with Urban Sprawl Indices: Leveraging Machine Learning and Stochastic Methods:

Urban sprawl ، GIS ، Remote sensing ، Machine learning ، CA-Markov