Geospatial Land Use/Land Cover (LULC) classification of the upper Wainganga Basin, Madhya Pradesh

Introduction

Land use applications encompass baseline mapping as well as continuous monitoring. While land use is usually derived from land cover, the two concepts are closely connected and frequently used interchangeably (Chaudhary et al., 2018) [5] . The primary goal of image classification is to automatically assign each pixel in an image to specific land cover classes or categories. Typically, multispectral data are utilized for this process, where the spectral characteristics of each pixel serve as the numerical foundation for classification (Jensen, 1996) [14] . LULC classification is the process of assigning specific land cover categories to individual pixels and organising them into distinct classes. Examples of such categories include water bodies, urban areas, forests, agricultural lands, grasslands, built-up regions, mountains, and highlands (Anderson, 1976) [1] . The primary objective of image classification is to automatically categorise all pixels in an image into specific land cover classes or themes. In essence, different surface features exhibit unique combinations of spectral reflectance characteristics based on their inherent properties (Hamad, 2020) [11] . LULC maps serve a crucial role in the planning, management, and monitoring of programs across local, regional, and national scales (Hamad, 2020) [11] . Land use refers to the human utilisation of land for various purposes such as agriculture, forestry, settlement, and industry, whereas land cover denotes the natural or artificial physical material at the surface of the earth, including vegetation, water bodies, bare soil, and built-up areas (Anderson et al., 1976) [1] . The Upper Wainganga Basin, a significant sub-basin of the Godavari River system, is characterised by its diverse physiographic, geological, and climatic conditions. The basin serves as an important source of water for agriculture, domestic use, and industry, while also supporting rich biodiversity and forest ecosystems. Remote sensing (RS) and geographic information systems (GIS) have emerged as powerful tools for analysing spatial and temporal patterns of land use and land cover. The advent of satellite-based earth observation systems provides synoptic, repetitive, and multi-spectral data that can be used to detect changes in land surface conditions with high accuracy (Jensen, 2007) [15] . Through the integration of RS and GIS, it becomes possible to analyse and visualise spatial phenomena, identify trends, and model relationships among various biophysical parameters. Moreover, RS and GIS techniques are cost-effective and time-efficient compared to conventional ground-based surveys (Lu et al., 2011) [22] . The advent of high spatial resolution satellite imagery and more advanced image processing and GIS technologies has resulted in a switch to more routine and consistent monitoring and modeling of land use/land cover patterns. Geographic Information Systems (GIS) provide powerful tools for analyzing land use issues, as well as for land use planning and modeling, as noted by several researchers. Studies by Zsuzsanna et al. (2005) [33] , Mensah et al. (2018), and Harshada et al. (2012) [12] examined land use and land cover changes using both unsupervised and supervised classification methods. Remote sensing has been widely used in updating land use/cover mapping has become one of the most important applications of remote sensing (Lo and Choi, 2004).

Materials and Methods

Study Area

The study was conducted in the Upper Wainganga sub-basin. The Wainganga River is a major tributary of the Godavari River, flowing through the states of Madhya Pradesh and Maharashtra. The total Wainganga basin area is 49949.48 km 2 . while the latitude extension is 19°30'N to 22°30 N' and the longitude extension is 79°00'E to 80°30 E'. The total length of the Wainganga River is 638.91 km, of which 270.2 km lies in Madhya Pradesh. This study focuses on the Upper Wainganga Basin, covering an area of 8,425.72 km 2 . The sub-basin is located across the districts of Balaghat, Seoni, Chhindwara, and Mandla in Madhya Pradesh. Geographically, the study area extends between latitudes 21°40'N to 22°45'N and longitudes 79°00'E to 80°30'E. The average annual rainfall in Seoni, Balaghat, and Chhindwara ranges between 1400 mm to 1600 mm. Maximum of the rainfall is received during the South-West monsoon from June to October. In the winter, the minimum temperature is 7 °C and the maximum temperature is 13 °C. In summer, the maximum temperature ranges from 39 °C to 47°.

Database preparation

Spatial data used in the study included satellite imagery obtained from the NRSC Bhuvan portal (https://bhoonidhi.nrsc.gov.in). The study utilised satellite imagery obtained from the Sentinel-2B mission for the period of November-December 2024. The data, featuring a spatial resolution of 10 meters, was sourced from the Bhoonidhi portal of the National Remote Sensing Centre (NRSC). ERDAS Imagine was employed for spatial data processing and the generation of thematic maps, enabling efficient handling and analysis of geospatial datasets. QGIS 3.38.0, an open-source GIS platform, was also used for spatial data analysis and thematic mapping to complement ArcGIS outputs. Additionally, Microsoft Excel was utilised for data preprocessing, performing statistical analyses, and Accuracy assessment, ensuring accurate evaluation and interpretation of the results. The 2024 LULC classification of the Upper Wainganga River Basin was analysed using Remote Sensing (RS), GIS, and multitemporal Sentinel-2B satellite images downloaded from NRSC, Images were mosaicked in ERDAS Imagine 2014 and QGIS 3.38.0, and the study area was subset for further analysis. Unsupervised classification with different band combinations and the maximum likelihood method was applied, resulting in five land cover classes. Field surveys were conducted to validate the classification accuracy.

Remote-sensing classification process

Remote-sensing classification is a complex process and requires consideration of many factors. The major steps of image classification may include the determination of a suitable classification system, the selection of training samples, image pre-processing, and feature extraction, the selection of suitable classification approaches, post-classification processing, and accuracy assessment. The user's need, scale of the study area, economic condition, and analyst's skills are important factors influencing the selection of remotely sensed data, the design of the classification procedure, and the quality of the classification results (D. Lu and Q. Weng, 2007) [22] . The classifiers use algorithms to analyse unknown pixels (referred to as false positives) in grayscale images and group them into various land use classes based on standard clustering methods that consider the proximity of similar image values. 1. Choose the number of classes (land use) 2. Group pixels based on spectral similarity (e.g., water, forest, urban, barren land) 3. Assign descriptive names to each cluster 4. Evaluate the classification outcomes Unsupervised classification groups spectral image pixels based on similarity, relying on spectral distance, particularly the Local Projection Neighbourhood (LPN), with minimal user input. Users may specify the number of classes, which are later assigned physical meanings, though land cover information is not predefined (Liu, 2008) [21] . Clustering algorithms, such as Hierarchical, Centroid-Based, Fuzzy, and Supervised Clustering, organise pixels into spectral classes. Reliable LULC classifications are often provided by institutions like the Survey of India and the Remote Sensing and Geological Survey agencies.

Accuracy Assessment

Accuracy assessment is essential for evaluating and quantifying a LULC map by comparing classified data with reference data. An error matrix, along with user's and producer's accuracy, is commonly used to measure classification performance (Congalton et al., 1991) [6] . PA measures the probability that a reference pixel is correctly classified, while UA indicates the likelihood that a pixel classified as a certain category actually represents that category in reality (Congalton, 1991; Foody, 2002) [6,10] . The Kappa coefficient (K) quantifies agreement beyond chance and was computed to assess the reliability of the classified map. 2. In the Upper Wainganga basin, five land use and land cover types were chosen: Water body, Forest, Agriculture/Other Vegetation, Fallow/Open/Barren, and Built-up area.

Land Use/Land Cover Status

Figure 3 illustrates the spatial distribution pattern of land use and land cover across the Upper Wainganga Basin for the year 2024. The analysis indicates that, in 2024, approximately 1.36% (114.78 km²) of the basin area was occupied by water bodies, 45.93% (3870.37 km²) was covered by forest land, 32.01% (2697.18 km²) was utilized for agriculture or other vegetation types, 15.81% (1332.37 km²) comprised fallow, open, or barren lands, and 4.87% (411.04 km²) was identified as built-up or urban area.

Accuracy Assessment of LULC Classification

In this study, the accuracy of the 2024 classified map was evaluated using 500 reference points distributed across the major land cover classes. The results were summarised through a confusion matrix/Error matrix (Table 3), which illustrates the agreement and disagreement between the reference data (rows) and the classified data (columns). The classification accuracy was evaluated using the Kappa coefficient, which showed a substantial agreement (0.6099) for 2024 (Landis & Koch, 1977) [20] . Water body (0.792), built-up (0.732), and forest (0.682) had the highest accuracy, while agriculture/other vegetation (0.499) and fallow/open/barren (0.516) were lower (Table 5). Overall, the 2024 classification was robust, though distinguishing agriculture from fallow areas remained challenging due to spectral similarity in mediumresolution imagery.

References

1. Anderson JR, Hardy EE, Roach JT, Witmer RE. A land use and land cover classification system for use with remote sensor data. U.S. Geological Survey Professional Paper 964. 1976.
2. Appiah Mensah A, Akoto Sarfo D, Partey S. Assessment of vegetation dynamics using remote sensing and GIS: a case of Bosomtwe Range Forest Reserve, Ghana. Egypt J Remote Sens Space Sci. 2018;22:145-154.
3. Chakraborty D, Saha S, Sharma A. Land use/land cover change and its impact on surface runoff: a case study of Silabati watershed, West Bengal, India. Appl Geogr. 2016;68:48-57.
4. Chaudhary BS, Saroha GP, Yadav M. Human induced land use/land cover changes in northern part of Guragon District, Haryana, India. J Hum Ecol. 2018;23:243-252.
5. Congalton RG. A review of assessing the accuracy of classifications of remotely sensed data. Remote Sens Environ. 1991;37(1):35-46.
6. Falkenmark M, Rockström J. The new blue and green water paradigm: breaking new ground for water resources planning and management. J Water Resour Plan Manag. 2006;132(3):129-132.
7. Feldman AD. Hydrologic Modeling System HEC-HMS: technical reference manual. U.S. Army Corps of Engineers, Hydrologic Engineering Center; 2000.
8. Fohrer N, Haverkamp S, Eckhardt K, Frede HG. Hydrologic response to land use changes on the catchment scale. Phys Chem Earth Part B Hydrol Oceans Atmos. 2001.
9. Foody GM. Status of land cover classification accuracy assessment. Remote Sens Environ. 2001;80(1):185-201.
10. Hamad R. An assessment of artificial neural networks, support vector machines and decision trees for land cover classification using Sentinel-2A data. Sciences. 2020;8(6):459-464.
11. Harshada K, Ingle S. Land use land cover classification and change detection using high resolution temporal satellite data. J Environ. 2012;1:146-152.
12. Jain SK, Kumar S, Varghese J. Site suitability analysis for water conservation structures using remote sensing and GIS in a semi-arid region of India. J Earth Syst Sci. 2015;124(5):1031-1045.
13. Jensen JR. Introductory digital image processing: a remote sensing perspective. 2nd ed. Upper Saddle River, NJ: Prentice Hall; 1996.
14. Jensen JR. Remote sensing of the environment: an earth resource perspective. 2nd ed. Upper Saddle River, NJ: Pearson Prentice Hall; 2007.
15. Kalpana YB, Nandhagopal SM. LULC image classifications using K-Means clustering and KNN algorithm. Dynamic Syst Appl. 2021;30.
16. Koneti S, Sunkara SL, Roy PS. Hydrological modeling with respect to impact of land-use and land-cover change on the runoff dynamics in Godavari River Basin using the HECHMS model. ISPRS Int J Geo-Inf. 2018;7(6):206.
17. Kumar M, Roy PS, Joshi PK. Forest fragmentation and water resources management in central India. Environ Monit Assess. 2018;190(5):298.
18. Lambin EF, Geist H, Lepers E. Dynamics of land-use and land-cover change in tropical regions. Annu Rev Environ Resour. 2003;28:205-241.
19. Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics. 1977;33(1):159174.
20. Liu J, Gao M. An unsupervised classification scheme using PDDP method for network intrusion detection. In: IEEE Second International Symposium on Intelligent Information Technology Application. 2008. p. 658-662.
21. Lu D, Weng Q. A survey of image classification methods and techniques for improving classification performance. Int J Remote Sens. 2007;28(5):823-870.
22. Malczewski J. GIS-based multicriteria decision analysis: a survey of the literature. Int J Geogr Inf Sci. 2006;20(7):703726.
23. Munsi M, Malaviya S, Oinam G, Joshi PK. A landscape approach for mapping forest cover change using multitemporal satellite data. Asian J Geoinformatics. 2010;10(1):1-8.
24. Rajendrakumar S, Parvati VK. Automation of irrigation system through embedded computing technology. In: Proceedings of the 3rd International Conference on Cryptography, Security and Privacy. 2019. p. 289-293.
25. Rawat JS, Kumar M. Monitoring land use/cover change using remote sensing and GIS techniques: a case study of Hawalbagh block, district Almora, Uttarakhand, India. Egypt J Remote Sens Space Sci. 2015;18(1):77-84.
26. Roy PS, Meiyappan P, Joshi PK, Kale MP, Srivastava VK, Srivastava SK, et al. Development of decadal (1985-19952005) land use and land cover database for India using remote sensing data. Remote Sens. 2015;7(3):2401-2430.
27. Singh A. Digital change detection techniques using remotely-sensed data. Int J Remote Sens. 1989;10(6):9891003.
28. Singh P, Sinha VSP, Kumar P. Land use and land cover change detection through remote sensing and GIS: a case study of Ranchi district, Jharkhand, India. Int J Adv Remote Sens GIS. 2018;7(1):2591-2600.
29. Yokoyama K, Kurosu T, Shi, Fujita M. Land use classification with textural analysis and the aggregation technique using multi-temporal JERS-1 L-band SAR images. Int J Remote Sens. 2001;22(4):595-613.
30. Zsuzsanna D, Bartholy J, Pongracz R, Barcza Z. Analysis of land-use/land-cover change in the Carpathian region based on remote sensing techniques. Phys Chem Earth. 2005;30:109-115.