Performance Evaluation of Supervised Classifiers for Land Use and Land Cover Mapping Using Sentinel-2 MSI Image

Ritesh Kumar1, Akshay Rai2, Varun Narayan Mishra3*, Pranaya Diwate3 and V. S. Arya1

1Haryana Space Applications Centre (HARSAC), CCS HAU Campus, Hisar-125004 (HR), India

2Leads Connect Services Pvt. Ltd., Noida Sector-16 (UP), India

3Centre for Climate Change and Water Research, Suresh Gyan Vihar University, Jaipur-302017 (RJ), India

(*Corresponding author, Email:


Land use and land cover (LULC) classification using remote sensing data has received substantial interest, as it is one of the most critical parameters for various environmental and regional planning applications. This study aims to examine the performance of different classification algorithm viz., maximum likelihood classifier (MLC), support vector machine (SVM) and random forest (RF) on Sentinel-2 MSI image. The highest overall classification accuracy was found using RF (88.73%), followed by SVM (85.35%) and MLC (83.10%). Moreover, an investigation was also carried out on the tree cover and its density in the study area using the best result providing RF classifier. The assessment of tree cover and its density showed an overall accuracy of 92.18%. The results show that the dense tree cover area is 71.40 km2, moderate tree cover is 81.87 km2 and open tree cover is 26.33 km2 in the study area. The work suggests that the RF classification algorithm has significantly improved the LULC and tree cover mapping in terms of overall classification accuracy. This study has proven the applicability of the remote sensing dataset for LULC derivation and assessment of green tree cover in any area with minimum field survey.

Keywords: LULC, SVM, RF, Accuracy, Sentinel-2 MSI

× How can I help you?