In this paper, while comparison of different machine learning techniques for classifying of multi-temporal remotely sensed images of Karbala metropolis, accurate information of land use and land covers (LULC) of this city was prepared and the occurred changes between 2002 and 2021 were determined based on post-classification comparison (PCC) method. To improve the classification results, spectral bands have been pan-sharped with panchromatic band using Gram-Schmidet method and textural features have been extracted using the GLCM matrix. The results show that the support vector machine (SVM) and maximum likelihood classification (MLC) methods provided the most satisfactory results, and SVM is the most accurate method due to its non-parametric nature. The overall accuracy (O.A) of the classification for 2002 and 2021 images with SVM were 89.76% and 88.05%, respectively, and for MLC method in the best case were, 85.94% and 85.36%, respectively, which were approximately 19% more than the Minimum Distance (MD) classification. The pan-sharpening led to an increase of 1.5% and 3%, of the O.A for 2002 and 2021 images respectively, and the use of textural features was useful in separating the urban area from non-urban areas. Comparison of the multi-temporal LULC maps of Karbala shows that between 2002 and 2021, approximately 33 hectares of vegetation covers and 25 hectares of agricultural lands were changed to urban areas, and in general, the area of Karbala has tripled. The significant growth of the metropolis of Karbala and the degradation of urban green space are important issues that should be seriously considered by the relevant city managers and planners.