Land use change analysis and prediction of urban growth using multi-layer perceptron neural network Markov chain model in Faridabad- A data-scarce region of Northwestern India

Abstract

Present research aims to examine the transformations of land use and land cover (LULC) within the Faridabad district, India, using high-resolution remotely-sensed images. LULC change analysis over the years 2007–2022 revealed a significant decline in agricultural land from 65.4% of the total area in 2007 to 53.9% in 2022. Conversely, considerable increases have been observed in urban built-up areas (from 58.2% in 2007 to 93.3% in 2022), industrial areas (from 13.7% to 26.9%). Vegetation coverage decreased from 18.9% in 2007 to 12.7% in 2022 after primarily alleviating in 2017 due to green initiatives. Further, the LULC maps of 2007 and 2012 were used to predict the LULC of 2017 using Multi-Layer Perceptron Neural Network (MLPNN)-integrated Markov Chain Model (MCM). Subsequently, predicted LULC of 2017 were compared with observed LULC of 2017 to validate the model. Additionally, the integrated model has been applied to predict and validate LULC of 2022. Validation results produced R² values and K statistics >0.8 for both 2017 and 2022 confirming the efficacy of the model. Finally, future LULC scenario has been predicted for 2027. Comparison of predicted LULC for 2027 with observed LULC of 2022 revealed that built-up would increase by 3.8% (built-up 149.3 km² in 2022 and 154.9 km² in 2027). Vegetation would decrease by 3.1% (12.7 km² in 2022 and 12.3 km² in 2027). From the present findings, it is recommended that a continuous monitoring is required to analyse the efficacy of implemented measures and adapt strategies as necessary.