Creation of land use and land cover maps of the Khachmaz-Shabran region of Azerbaijan using machine learning methods

Abstract

his study aims to produce Land Use/ Land Cover (LU/LC) maps of the Azerbaijan KhachmazShabran region using machine learning (ML) methods and remotely sensed data. We used two
common ML classification algorithms, Random Forest (RF) and Support Vector Machine
(SVM). We generated two different LU/LC maps using an Azersky satellite image with a spatial
resolution of 1.5 and a Sentinel-2 image with 20 m spectral bands. Both images were acquired
on 1 August 2020. We implemented LU/LC class definitions of Level 2 of the CORINE
nomenclature. After the classification step, an error matrix was created using the same
reference points for Azersky and Sentinel-2. We compared the results of two classifications to
determine the better-performing approach in obtaining the region's LU/LC maps.