MNDWI is a modified version of NDWI, which has less noise than NDWI and can give proper output to identify the water content on the surface. The formula to derivate MNDWI follows below (Equation 1).

SAVI was first utilized by A.R Huete in the year 1988. Initially, the NDVI was used by most of the researchers to examine the condition of the vegetation over an area using satellite images, SAVI is a kind of modified version of NDVI which accounts for soil brightness in areas where vegetation cover is low or sparse. It helps minimize the influence of soil reflectance on vegetation monitoring.

The SAVI formula introduces a correction factor (L) to reduce soil noise, which is especially useful in arid or semi-arid regions. The formula to calculate SAVI is given below (Equation 2).

Where, L is the representation of the correlation factor ranging from 0 to 1, “0” is given to very high plant densities and “1” to very low plant densities. To detect the low-density vegetation in an urban area, SAVI will be more suitable than NDVI because SAVI can effectively work in an area with a vegetation cover of less than 15% (Xu 2008). However, NDVI will be only suitable for an area with a vegetation cover of more than 40%. In this study area, the L factor is taken as 0.3.

The NDBI efficiently emphasizes the built-up areas in remote sensing data sets. When combined with appropriate threshold selection, it can be utilized to identify impervious surfaces in urban environments. The NDBI values range between -1 to +1. The positive values in the NDBI represent the built-up area. Equation 3 was used to examine the NDBI.

IBI is used to extract the built-up more precisely using satellite images. It ranges between -1 to 1 in which zero to negative values represent the noise and other features in an area and the positive values represent the built-up. Equation 4 used to examine the IBI

In the present study, the SVM classification algorithm has been used to execute the LULC map of the study area in the ENVI platform. Landsat images have been utilized to map the LULC for 1994, 2004, 2014 and 2024. The satellite images were classified based on the training samples specified by the Region of Interest (ROI). The SVM calculated an optimal hyperplane for class separation based on training data, with the gamma parameter set to 0.01 to improve both efficiency and classification accuracy. ENVI software facilitated the classification, using a penalty parameter of 100 to minimize misclassification. A classification threshold of zero was set to ensure precise pixel assignments. The ROI was used to establish spectral signatures for five LULC categories: built-up, bare land, vegetation, forest, and water. The process included selecting training areas, generating ROIs, and classifying the images, with around 200 ground control points used for validation.

LULC changes were analyzed for the periods 1994−2004, 2004−2014, and 2014−2024 using post-classification comparison methods. A change detection matrix was used to track how each LULC class transitioned into others, with categories listed in rows and columns. Changes were quantified in both hectares and percentages, with percentage changes calculated using a specific formula (Equation 5).

In this a₂ refers to the area of the specified class in the current year; a₁ represents the area of the same class in the previous year.

MNDWI plays a crucial role in extracting water bodies in the study area and has been analyzed for the years 1994, 2004, 2014 and 2024 (Figure 3). The higher value MNDWI represents the areas that have water body, and vegetation. while the low values represent the built-up and bare soil areas. In 1994, approximately 54% of the area was classified as High to Very High in MNDWI, with only 5% falling under the Very Low category. By 2004, the Low to Moderate MNDWI class dominated the region, covering 59% of the area. In 2014, the Very Low to Low classes expanded significantly, occupying 59% of the study area. However, by 2024, the Very Low to Low classes decreased to 26%, indicating a positive trend in waterbody coverage. Spatially, the Very High MNDWI values were consistently observed in the southeastern and northern parts of the study area throughout all the years, though the extent of coverage varied. Conversely, lower MNDWI values were predominantly found in built-up areas, reflecting the impact of urbanization on waterbody distribution.

The extraction of vegetation can be enhanced by using SAVI instead of NDVI, particularly in areas with high soil influence (Figure 4). The lower SAVI values indicate bare soil, and rocky terrain & the higher value indicates urban with impervious surface domination. SAVI results indicate a decline in vegetation coverage over time. In 1994, the very high vegetation class was found in 86 sq. km (30%) of the study area and was distributed across the region. By 2004, this coverage decreased to 19% (56 sq. km.). The reduction continued, with the very high category covering only 27 sq. km (9%) by 2014, mainly along rivers and streams. However, in 2024, SAVI values showed a slight growth, with the very high class increasing to 14%. This improvement is likely due to increased soil moisture, particularly in the southern and central parts of the study area. The gradual reduction in vegetation coverage highlights the need to closely monitor land use and soil conditions to understand environmental changes.

The NDBI is used to extract built-up and concrete surfaces from satellite images. In the NDBI, built-up will show high value and vegetation, waterbodies and bare soil will show lower value (Figure 5). Analysis of NDBI values over the years reveals a gradual increase in the very high class of built-up areas. In 1994, approximately 14 sq. km (5%) of the taluk, mainly concentrated in Kumbakonam city and bare land near the river, exhibited very high NDBI values. By 2004, this area expanded to 22 sq. km (7%). In 2014, the very high NDBI class increased significantly to 45 sq. km (16%), primarily covering the central and southern parts of Kumbakonam Taluk. However, by 2024, the very high class decreased to 30 sq. km (10%), with its spatial distribution more concentrated in the central part, where Kumbakonam city is located.

The IBI is recognised for its effectiveness in extracting built-up areas and is considered more accurate than the NDBI. In this study, IBI values have shown a significant increase over recent decades (Figure 6). The very high IBI class covered 6 sq. km in 1994, expanded to 12 sq. km in 2004, increased further to 34 sq. km in 2014, and reached 51 sq. km in 2024. Compared to NDBI, IBI provided more precise results in identifying built-up areas. However, IBI does not perform well in clearly differentiating between bare land near rivers and built-up zones. Despite this limitation, the spatial analysis revealed a steady expansion of built-up areas over time, especially in the urban fringes of Kumbakonam. This trend highlights the ongoing urbanisation in the region, with growth concentrating around the city's periphery, indicating the importance of monitoring urban sprawl for future planning and land use management. The correlation shows the relation between the two variables, such as IBI and SAVI, with the help of a scatterplot (Figure 7) for 1994, 2004, 2014 and 2024. The result shows a negative correlation between IBI & SAVI. As the built-up increases, the vegetation decreases.

The temporal change of LULC of Kumbakonam was classified as per the NRSC level 1 classification, and the classes are built-up, agricultural land, vegetation, barren land and waterbody (Figure 8). In 1994, agricultural land was spread over 212 sq. km of the study area, and only 16 sq. km of land was barren, while waterbody was 14 sq. km and 18 sq. km of vegetation (Table 1). The build-up was spread in the centre and scattered in the southern part of the area, about 21 sq. km (Figure 9).

In 2004, the agricultural land decreased to 208 sq. km, while waterbody decreased to 10 sq. km and the vegetation drastically changed to 2 sq. km. Barren land was increased to 32 sq. km and build-up increased to 29 sq. km because some parts of the agricultural land changed into barren land & build-up. In 2014, the build-up was spread over about 38 sq. km and the barren land was extended by about 34 sq. km. Although agricultural land was reduced to 200 sq. km, vegetation was reduced to 1 sq. km and waterbody were reduced to 7 sq. km. In 2024, there was a rapid increase in the build-up to 83 sq. km and barren land was reduced to 3 sq. km. Whereas, the agricultural land, waterbody and vegetation were also reduced to 190 sq. km, 5 sq. km and 0.7 sq. km, respectively. In a LULC classification, accuracy assessment is a crucial aspect for analyzing the reliability of the generated image. In this study, a confusion matrix using ground truth ROIs was utilized to evaluate the accuracy. The overall classification accuracies for the years 1994, 2004, 2014 and 2024 were 83.64%, 84.27%, 91.17% and 94.62%, respectively, with Kappa Coefficients of 0.73, 0.74, 0.86 and 0.90.

From 1994 to 2024, there was a drastic increase in the build-up from 37.6% to 117.6 %, agricultural land, vegetation and waterbody decreased by -1.9 % to -5.2 %, -87.4% to -28.4 % and -26.5% to 32.8 % (Table 2). Whereas barren land was increased by 100% in 2004, but it was reduced in 2024 as-91.2%. Most of the agricultural land, vegetation and waterbody were changed into built-up areas from 1994 to 2024. The rapid increase in the build-up was mostly increased as a result of the population expansion and in 2021, the Kumbakonam town was upgraded to a municipal corporation.