Similarity of land cover, through digital classification of satellite images, and its field verification in the city of Celendin, Cajamarca, 2022

Abstract

The research aimed to calculate the percentage of similarity between land cover identified in the field through sampling points and the corresponding categories in the raster map derived from unsupervised classification, using a Landsat 8 satellite image of the area of interest. A sample exceeding thirty-five sampling points was utilized, providing a confidence level of ninety percent and a ten percent margin of error. Images of the area of interest were downloaded from the United States Geological Survey (USGS), and a combination of the first seven bands was performed. Subsequently, this composition was merged with band eight to enhance spatial resolution, achieving a resolution of fifteen meters. Radiance and reflectance corrections were then applied. With the corrected images, band combinations 543 for vegetation, 764 for forest, and the Bare Soil Index (BSI) were conducted. These data were represented on a raster map, and sampling points were superimposed on each category to determine the number of points coinciding both in the field and on the raster map. The results revealed a 93.5% match in the vegetation category, 95% in bare soil, and 91% in the forest category.