Assessing the impact of terrain variables on LULC patterns using multinomial logistic regression in Churachandpur...

Terrain conditions significantly influence the spatial distribution of land use and land cover (LULC) patterns. This study investigates terrain- LULC relationship and the impact of terrain variables in LULC distribution in Churachandpur district. LULC maps for the years 2004, 2014, and 2024 were generated from Landsat imageries using supervised classification with the Random Forest algorithm in QGIS. Terrain variables including elevation, slope, aspect, and TWI were derived from SRTM DEM data. Terrain-LULC relationships is assessed with the help of multinomial logistic regression (MLR) in RStudio. The result shows high dependence of LULC on slope among all other terrain variables. The MLR model achieved an overall prediction accuracy of 50.93% with a Kappa coefficient of 0.3642, indicating moderate agreement. While agricultural land (AL) was predicted with high accuracy (90.36%), that of built-up areas (BU) was limited due to overlapping terrain characteristics with AL. Likewise, similarity between scrub forest (SF) and forest reduced prediction reliability. LULC change analysis show significant expansion of BU (91.70%) and loss of forest and AL between 2004 and 2024, driven largely by population growth and land-use pressure. Forest degradation and conversion to SF were especially pronounced, reflecting ecosystem stress. LULC classes exhibited spatial clustering based on the terrain conditions, with BU and AL concentrated on gentle slopes and mid-elevations. The study highlights a strong terrain-LULC relationship and demonstrates the value of remote sensing data, geospatial tools, and programming platforms in environmental research. These findings can support terrain-informed planning, conservation prioritization, and identification of ecologically sensitive zones in hilly regions.