<scp>MNDBI</scp> : A Novel Spectral Index Integrated With <scp>AI</scp> for Enhanced Urban Feature Extraction, Growth Analysis, and Future Prediction

ABSTRACT Urban growth and its mapping have been a major focus of research using remote sensing data since the invention of remote sensing. The multispectral images contain enormous data about different land cover objects, and various techniques have been used to extract them. This work leverages the power of artificial intelligence by employing a spectral index‐based approach, enhanced with machine learning for classification, to highlight urban features from multispectral images. The NDBI, which uses SWIR and NIR bands, is modified, and a new index named Modified Normalized Difference Built‐up Index (MNDBI) is proposed with three bands viz., SWIR1, SWIR2, and NIR. MNDBI outperforms the results of existing indices with 97% accuracy, whereas NDBI using SWIR1 and SWIR2 shows 83.89% and 88.97% accuracy, respectively. The MNDBI‐SVM approach is applied to Landsat‐5, Landsat‐7, and Landsat‐8 images from 1985 to 2020 in Las Vegas to analyze urban growth. Furthermore, the predictive capabilities of artificial intelligence are utilized through an Autoregressive Integrated Moving Average (ARIMA) model to forecast future urban development. This work has two objectives: (1) accurately extracting urban areas using the MNDBI and SVM classification and (2) predicting urban growth using the ARIMA model. This combined approach allows for estimating a critical saturation point for urban expansion.