Determination of impervious area of Saroor Nagar Watershed of Telangana using spectral indices, MLC, and machine learning (SVM) techniques

Urbanization affects the local wind and water cycle by changing the natural surface and atmospheric conditions, which further changes the local climate and climate system. Assessment of built-up-area changes in a rapidly growing urban area within a short time is a prime factor for administrators for better environmental assessment and sustainable development of urban areas. Traditional survey approaches, on the other hand, are unable to meet the demand for rapid urban land management development, and there is a pressing need to develop new methods to address the limitations of existing ones. This study compares various urban spectral indices to other existing approaches in order to determine which index provides a better representation of the impervious area in the urban watershed. Landsat 8 OLI (Operational Land Imager) satellite images acquired on 15 March 2014 and 31 March 2020 are used in the present study. Indices, namely Modified Built-up Index (MBUI), SwiRed Index (SwiRed), and Enhanced Built-up and Bareness Index (EBBI), were utilized to extract impervious areas. Thresholding of indices is done by comparing them with 1000 reference points taken from Google Earth imagery of the respective years. The accuracy of the urban indices is assessed by comparing the results with high-resolution Google Earth Satellite Images. The impervious area is extracted from spectral indices and other remote sensing techniques such as maximum likelihood classification and support vector machine classification techniques. The overall accuracy of SVM, MLC, MBUI, EBBI, and SwiRed for the 2014 dataset was found to be 95.1%, 90.8%, 83.9%, 78.9%, and 87.3%, respectively, and the overall accuracy of SVM, MLC, MBUI, EBBI, and SwiRed was found to be 96%, 89.2%, 89.1%, 85.5%, and 92.6%, respectively. Impervious areas in the heterogeneous urban environment can be monitored in a better way and within lesser time by using spectral indices generated using Landsat 8 OLI (Operational Land Imager) satellite data.