Satellite based seasonal land use classification and change detection analysis of landsat-8 operational land imager

Abstract

This study investigated the Landsat-8 Operational Land Imager (OLI) to determine its suitability for change detection analysis and mapping applications due to its enhanced signal-to-noise ratio (SNR), spectral band configuration, technical superiority, improved system design and high radiometric resolution. Earlier Landsat series had a lower SNR, comparatively smaller radiometric resolution and limitations in spectral band configurations. Pre-classification methods e.g., image differencing; image rationing, vegetation indices and principal components used for change detection analysis do not indicate type of the change due to its limitations. Therefore, better technique of post classification change detection was used on OLI data in the study area which provided information of the type of the change i.e., from-to change detection. This paper evaluated the OLI support vector machines (SVM) classification suitability using data covering four different seasons (i.e., spring, autumn, winter, and summer) after pre-processing and atmospheric correction. After classification, the major change detection results of OLI SVM-classified data for the four seasons were compared to change detection results of six cases: (1) winter to spring; (2) winter to summer; (3) winter to autumn; (4) spring to summer; (5) spring to autumn; and (6) summer to autumn. Seasonal change in the shoreline resulted with the corresponding change in categories. The OLI data classifications were made by applying SVM classifier and due to its improved features, OLI data is found suitable for seasonal land cover classification and post classification change detection analysis. , compared to ALI and Landsat 7 [3][4][5]. The width of OLI band 5 is refined to exclude atmospheric absorption features at 825 nanometers (nm) and its panchromatic bandwidth is reduced to improve detection of vegetation vs. non-vegetation compared with previous Landsat series. In OLI, the new Cirrus band detects thin clouds more accurately compared to previous satellite-based sensors (i.e., ALI and Landsat-7) [6]. Landsat 8 provides regional monitoring through remote sensing with substantial improvements in data quality due to advances in its noise reduction and spectral coverage designs [7,8]. Studying different satellite sensors and their parameters (i.e., scanning systems, sensor characteristics, data systems, resolution, and so on) is important. However, studies examining the characteristics of Landsat-8 and its pre-and post-flight calibrations are rare in the literature [9][10][11][12][13][14][15].Change detection remains a challenging problem and is used different for classification and change detection analysis [16][17][18]. Different satellite based change detection methods have different advantages and disadvantages [19,20]. In some change detection methods, no training data is required [21] e.g., in image differences [22] and clustering based methods [23] provide only change vs no change. If training data is available, then post classification change detection analysis ...