Classification of ASTER Data by Neural Network to Mapping Alterations Related to Copper and Iron Mineralization in Birjand

The studied area located in eastern Iran shows a high potential for various mineralizations, especially copper due to its tectonic activity. Remote sensing data can effectively distinguish these areas because of the sparse vegetation. Therefore, in this study, the ASTER (Advanced Spaceborne Thermal Emission and Reflection Radiometer) multi-spectral data was used to recognize argillic, sericite, propylitic, and iron oxide alterations associated with copper mineralization. For this purpose, two categories (porphyry copper-iron and advanced argillic-iron) related alterations were considered to perform the classification of a 2617 square kilometer area using a neural network classification algorithm. To evaluate the accuracy of the classifier, the confusion matrix was computed, which provides overall accuracy and the kappa coefficient factors for assessing classification accuracy. As a result, 64.17% and 83.5% of overall accuracy, and 0.602 and 0.807 of the kappa coefficient were achieved for the advanced argillic alterations and porphyry copper categories, respectively. Ultimately, the validation of the classifications was carried out using the normalized score (NS) equation, employing quantitative criteria. Notably, the advanced argillic class emerged with the top normalized score of 2.25 out of 4, signifying a 56% alignment with the geological characteristics of the region. Consequently, this outcome has led to the identification of favorable areas in the central and northeastern parts of the studied area.