The development of spectral indices for detecting built-up land areas and their relationship with land-surface temperature

ABSTRACT

Different spectral indices were provided for the classification of built-up lands using satellite images, but the importance of the blue band in the classification, fusion, and image enhancement prior to the calculation of spectral indices has received less attention so far. This study introduces a new index based on blue band and run it along with six other spectral indices. The method of study is as follows, after the selection of Landsat 7 and Landsat 8 images, the initial pre-processing was done on the images and in the second stage, the enhancement and the fusion were performed. In the third stage, the built-up lands were classified using spectral indices. Otsu’s thresholding method was used on all indices to separate built-up and non-built-up lands. To assess the accuracy, 3500 reference points were used. The results showed that when fusion and image enhancement were performed, overall accuracy (OA) increased by 3% to 6.71% for Landsat 7 images, and 2.14% to 6.71% for Landsat 8 images. The results also showed that after fusion and image enhancement, the visible blue band and first shortwave infrared index (VbSWIR1-BI) with an OA of 92.88% (Landsat 7) and 91.68% (Landsat 8) indicate the highest OA. The percentage of built-up lands changes according to spectral indices ranges from 6.38% to 25 .76% and the largest and lowest amount of built-up land on the basis of the VbSWIR1-BI index belonging to districts 12 and 22, respectively. The results of the spectral indicator show dispread growth during the 14 year period for Tehran. Finally, a significant relationship was observed between surface temperature and built-up land classified by spectral indices.