Multispectral optical imagery from Sentinel-2, featuring 13 bands with spatial resolution as detailed in Table 1, was utilized in this study. Specifically, Band 2 (blue), Band 3 (green), Band 4 (red), and Band 8 (Near Infrared) were selected for Land Cover and Land Use (LCLU) and Normalized Difference Vegetation Index (NDVI) calculations. The imagery, acquired on January 23, 2022, was sourced from the USGS Earth Explorer.

For LCLU analysis, we applied unsupervised classification techniques in digital image processing (DIP), following methodologies established by Karuppaiah et al. (2021) 21, Tamilenthi et al. (2011) 12, 13, and Jensen (1989) 22. These methods have been instrumental in studying various land cover features and resource applications, as documented by Navalgund et al. (1996) 23.

Additionally, Shuttle Radar Topographic Mission (SRTM) data was employed to create slope and Digital Elevation Model maps (Figure 3). The topographic map, at a 1:50,000 scale, illustrated primary geographic features, including contours.

High-resolution images, while offering exceptional detail, come with higher costs and limited photographic coverage. In contrast, multispectral medium resolution images, which are freely available, are more suitable for larger landscapes despite their limited temporal acquisition and potential cloud cover interference. Satellite imagery, therefore, proves highly effective for viewing expansive areas.

Sirumalai hill is strategically connected to the districts of Dindigul and Madurai, with transportation primarily via a roadway featuring 20 hairpin bends. Regular buses and taxis operate from both cities. Figure 1 illustrates the study area's location map, derived from Survey of India (SoI) toposheets, detailing the reserved forest boundary, forest spread, roads, benchmarks, pillars, painted standing trees, and poles marking internal boundaries. Figure 2 presents an aerial view of Sirumalai hill using Sentinel-2 imagery, highlighting the topographic expression of the reserved forest boundary.

Sentinel-2 satellite imagery, collected at processing level 1C for January 2022, contains pixel values in surface reflectance. This imagery necessitated atmospheric corrections and digital image processing techniques to accurately extract features from each pixel. This processing is crucial for deriving meaningful data about land cover and land use, ensuring precise analysis and interpretation of the forest landscape.

The use of unsupervised classification techniques, particularly through the recording method, proved essential for accurately reclassifying mixed pixels. This approach allowed for a thorough examination of the study area, revealing that 79.31% is densely forested, while the remaining portion is comprised of various other features.

Figure 3

Within the forest area, open spaces are devoid of trees or plantations primarily due to timber extraction and landslide-prone slopes 24, 25, 26. Additionally, some regions have experienced vegetation removal. Open forest areas, now dominated by shrubs and tall grasses, cover 58.25 sq km, accounting for 20.04% of the total study area. Approximately 1.3 sq km (0.45%) has been converted to agricultural land. Furthermore, 0.6 sq km (0.21%) of forested land has been transformed into human settlements. However, the potential for rapid settlement expansion is limited due to varying sub-soil surface roughness and other factors such as slope and geology.

In the forested area, open spaces without any trees or plantations are evident due to timber collection and landslides on sloped regions 24, 25, 26.

Additionally, some areas have been cleared of vegetation. Within these open forests, shrubs and tall grasses have grown, covering an area of 58.25 sq km, which constitutes 20.04% of the total study area. Moreover, approximately 1.3 sq km (0.45%) of the forest has been converted into agricultural land. Human settlements have expanded into 0.6 sq km (0.21%) of forested land. However, the potential for rapid settlement growth is limited due to the variable sub-soil surface roughness, influenced by factors such as slope and geology.

Figure 4

According to the Census of India 2011, Sirumalai village has a total population of 5,041 (Indian Village Directory). The housing structures in the village are diverse, with 60% comprising huts and asbestos-roofed dwellings, while the remaining are tiled-roof and a few congregate houses. The huts and asbestos-roofed residences are predominantly scattered due to the challenging hill slopes, whereas the tiled and congregate houses follow a linear pattern.

The study identified two primary settlement patterns: linear and scattered, as depicted in Figure 4. The linear settlements consist of approximately 500 houses, which are situated along the roadsides at the top of the hill in flat regions with slopes ranging from 0 to 8 degrees. In contrast, the scattered settlements are found in the densely forested middle areas, highlighting the influence of the topography on settlement distribution.

Google Earth provides high-resolution satellite imagery that is invaluable for verifying data obtained from low-accuracy optical remote sensing. For this study, land use features were extracted from Google Earth Pro in vector file format (.kml), which was instrumental in identifying settlement patterns and distinguishing between forest, open, and agricultural lands. The classified images of the study area were cross verified with coordinates on Google Earth to examine settlement patterns and other features accurately.

Slope is a critical factor influencing settlement growth, especially in hilly regions. In the study area, the slope degree ranges from 0° to 68.94°. Table 3 categorizes the land cover and land use types according to their slope ranges. Notably, the slope range of 0° to 8.65° covers an area of 71 sq km, which is generally suitable for settlement growth. However, in the study area, only 0.6 sq km within this slope range is classified as settlement by Sentinel-2 data. The remaining 70.4 sq km in this slope range is predominantly covered by vegetation.

Figure 5

Figure 6

Table 3

S.No	Slope Range in degree	Slope area in sqkm	LCLU category
1	0 - 8.65	71	Settlement
2	8.65 - 16.22	73	Agriculture
3	16.22 -23.79	84	Dense forest
4	23.79 - 32.71	66	Sparse vegetation
5	32.71 - 68.94

The NDVI (Normalized Difference Vegetation Index) is a widely used metric for assessing vegetation health, derived from optical remote sensing by combining red and near-infrared (NIR) bands. Previous studies have demonstrated that NDVI can provide valuable insights into vegetation dynamics (Mather, 1999; Foody et al., 2001; Li X. et al., 2007 27). For this study, we utilized band-4 (Red) and band-8 (NIR) for NDVI calculations.

resents the band values used, and the NDVI was computed using the specified formula after performing geometric correction on the raw imagery. The spectral reflectance values for the pixels ranged from -0.02 to 0.82. Typically, NDVI values range from -1 (indicating no vegetation) to 1 (indicating dense vegetation). The variation observed in NDVI values across different elevation zones provides insights into vegetation distribution and health in relation to elevation changes.

NDVI calculated from the following formula:

Figure 7

A fishnet grid was established for raster data extraction with cell dimensions of 1000 x 1000 meters, comprising 21 rows and 31 columns. This setup was utilized to analyze the correlation between Land Cover/Land Use (LCLU), Normalized Difference Vegetation Index (NDVI), and slope in the context of a multiple linear regression (MLR) model. NDVI values were interpreted as follows: values below 0 represent non-vegetated areas such as rocky surfaces or soil reflections, negative values indicate built-up areas or water bodies, values between 0.5 and 0.6 correspond to grassland and sparsely vegetated areas, and values above 0.6 signify densely vegetated areas with healthy vegetation.

For scatter plots, ensure that the x-axis and y-axis are labeled correctly as "Slope" and "LCLU," respectively. Each data point should be plotted with a trend line to visualize the best fit among the data sets. The scatter plot should illustrate the dispersion of points both above and below the trend line, indicating variability in the data.

The MLR model showed a significant correlation between settlement and slope. A total of 294 samples were analyzed, incorporating LCLU, slope, and NDVI. The regression statistics revealed a multiple R value of 0.4, with both R² and adjusted R² values at 0.2, and a standard error of 0.3. The intercept coefficient was 3.03, while NDVI and slope coefficients were -2.7 and -0.007, respectively, suggesting that lower NDVI and slope values are associated with increased settlement activity. The model's significance is underscored by an F-value of 42.66, indicating strong statistical significance.

Figure 8 illustrates the scatter plot showing the relationship between slope and LCLU. The trend line demonstrates a positive correlation, indicating that a reduction in slope from 18 to 16 degrees is associated with an increase in agricultural land, while slopes below 16 degrees are more favorable for settlement construction.

Figure 8

Understanding the distribution of forested land, agricultural areas, settlements, and water bodies is crucial for local tribes and communities. This study highlights that topography, particularly slope, significantly impacts vegetation indices, as evidenced by variations in NDVI. The analysis of land cover and land use (LCLU) in conjunction with slope data reveals that slope plays a pivotal role in influencing settlement patterns. Settlements tend to be concentrated in areas with slopes between 8 to 16 degrees, while regions with slopes less than 8 degrees are more likely to feature linear settlement patterns.

The research indicates a strong correlation between slope and settlement growth, with built-up areas decreasing as slope increases. Currently, there remains a potential area of 70.04 sq km within favorable slope ranges that could be utilized for development purposes. Effective forest landscape management could enhance economic benefits for tribes, improve agricultural development, and elevate living standards.

Future recommendations include investigating plant species variations based on elevation and slope, identifying landslide-prone zones to prevent construction in vulnerable areas, and utilizing high-resolution imagery and GPS for better planning of agricultural and residential infrastructure.