Abstract Accurate detection and prediction of land-use and land-cover (LULC) dynamics are crucial for sustainable environmental planning and resource management, especially in ecologically sensitive regions such as the Nilgiris district in the Western Ghats, particularly in Tamil Nadu. This study investigates the spatial and temporal patterns of LULC changes in Gudalur Taluk over a 40-year period (1995–2035) using multi-temporal satellite imagery and machine learning (ML) techniques. Landsat imagery from the years 1995, 2005, 2015, and 2025 was used for classification, while LULC for 2035 was predicted using the random forest (RF) classification model, known for its robustness and high classification accuracy in complex landscapes. Seven major LULC classes – water body, built-up land, forest blank, evergreen forest, deciduous forest, agricultural land, and barren land – were analysed. The classified outputs indicate a notable decline in agricultural land (from 8.90% in 1995 to 4.83% in 2035) and evergreen forest (from 20.43% to 13.83%), largely attributed to the expansion of built-up land and forest blanks. Built-up areas increased significantly, rising from 1.73% in 1995 to 6.82% in 2035, reflecting urbanisation and developmental pressures. Simultaneously, forest blank areas increased from 32.73% to 44.77%, indicating some removal or shift in the forested landscapes. Water bodies and deciduous forests also showed continuous decreases over time. The spatial trends seem to show some similarity in transition from forest cover and agricultural landscapes to mixed and degraded landscapes while all influenced by anthropogenic changes and urbanization impact. The RF model presented a non-linear pattern while capturing the spatial heterogeneity for the region, which supports using a RF model for future LULC modelling of hilly areas. The results provide a valuable baseline, for regional planners as well as conservation agencies, to devise plans for effective land management practices to reduce negative impacts to the environment. Research highlights <UnorderedList Mark="Bullet"> <ItemContent> RF-based remote sensing mapped and predicted LULC change (1995–2035). </ItemContent> <ItemContent> Evergreen forest and agriculture declined; built-up areas expanded. </ItemContent> <ItemContent> Random Forest captured non-linear LULC transitions in hilly terrain. </ItemContent> </UnorderedList>

Long-term land use and land cover change detection and prediction using remote sensing and ML approaches (1995–2035) for Gudalur Taluk, Tamil Nadu

期刊论文
发表日期：

2026年6月25日

查看全文

J Thavaseelan,
J Jegan

摘要

Abstract

Accurate detection and prediction of land-use and land-cover (LULC) dynamics are crucial for sustainable environmental planning and resource management, especially in ecologically sensitive regions such as the Nilgiris district in the Western Ghats, particularly in Tamil Nadu. This study investigates the spatial and temporal patterns of LULC changes in Gudalur Taluk over a 40-year period (1995–2035) using multi-temporal satellite imagery and machine learning (ML) techniques. Landsat imagery from the years 1995, 2005, 2015, and 2025 was used for classification, while LULC for 2035 was predicted using the random forest (RF) classification model, known for its robustness and high classification accuracy in complex landscapes. Seven major LULC classes – water body, built-up land, forest blank, evergreen forest, deciduous forest, agricultural land, and barren land – were analysed. The classified outputs indicate a notable decline in agricultural land (from 8.90% in 1995 to 4.83% in 2035) and evergreen forest (from 20.43% to 13.83%), largely attributed to the expansion of built-up land and forest blanks. Built-up areas increased significantly, rising from 1.73% in 1995 to 6.82% in 2035, reflecting urbanisation and developmental pressures. Simultaneously, forest blank areas increased from 32.73% to 44.77%, indicating some removal or shift in the forested landscapes. Water bodies and deciduous forests also showed continuous decreases over time. The spatial trends seem to show some similarity in transition from forest cover and agricultural landscapes to mixed and degraded landscapes while all influenced by anthropogenic changes and urbanization impact. The RF model presented a non-linear pattern while capturing the spatial heterogeneity for the region, which supports using a RF model for future LULC modelling of hilly areas. The results provide a valuable baseline, for regional planners as well as conservation agencies, to devise plans for effective land management practices to reduce negative impacts to the environment.

Research highlights

RF-based remote sensing mapped and predicted LULC change (1995–2035).

Evergreen forest and agriculture declined; built-up areas expanded.

Random Forest captured non-linear LULC transitions in hilly terrain.