Compiling a Preliminary Regional Geological Map Using Sentinel-۲ Satellite Imagery and the Random Forest Algorithm in the East of Iran

This study employs Sentinel-۲ satellite images along with the random forest algorithm to create a regional geological map. For this purpose, the independent variables consist of the images for ۱۰ Sentinel-۲ bands of the Khosuf-I region, while the class labels consist of a geological map of Khosuf-I divided into three and fifteen rock units. The classification accuracy of the resulting model is ۹۰.۹۷ and ۸۴.۸۵% for the three-class training and testing data, and ۹۴.۷۶ and ۶۳.۹۲% for the fifteen-class training and testing data, respectively. These models are then applied to the Sentinel-۲ satellite images' data of the Birjand-IV region to prepare two preliminary geological maps. The Birjand-IV region's three-class geology map reveals that igneous rocks are present in the northern and southern regions, while sedimentary rocks occupy the middle section and metamorphic rocks are found within the region's igneous masses. Similarly, the fifteen-class geology map of Birjand-IV indicates that andesite, dacite, intermediate tuff rock units, and metamorphic rocks characterize the northern region. Conversely, the southern part of the region is mainly composed of ophiolite, flysch sediments, basaltic and ultra-basic volcanic rocks, and limestone and shale interlayers. Field studies in three areas confirm the accuracy of the preliminary geology maps.

کلیدواژه ها:

Geological map ، Random Forest Algorithm ، SMOTE algorithm ، Image Processing ، Sentinel- ۲ image processing