Rapid urban expansion around university campuses, often associated with increasing demand for student housing and supporting services, can alter local thermal environments. Although Landsat-based indices are widely used to study urban thermal environments, campus-driven densification in humid tropical peri-urban settings remains rarely quantified. This study examines multi-temporal (2007–2025) thermal dynamics in a peri-urban area surrounding a major university campus in humid tropical Indonesia. Landsat imageries and 2025 building footprint data were integrated to classify the landscape into very low-, low-, and medium building-density zones. Land Surface Temperature (LST) and the Urban Thermal Field Variance Index (UTFVI) were derived, then Pearson’s correlation and Principal Component Analysis (PCA) were applied to capture the dominant covariance structure among LST and key biophysical indices (NDVI, NDWI, and NDBI). Results reveal a clear density-dependent thermal gradient: increasing building density corresponds with declining vegetation cover, expanding built-up surfaces, and rising surface temperatures. Medium-density settlements consistently experience the highest thermal stress, while very low-density zones retain comparatively favorable conditions. NDVI shows strong negative correlations with LST (r = − 0.51 to − 0.73), whereas NDBI is strongly positive (r = 0.60–0.76). NDWI also exhibits positive correlations (r = 0.43–0.72), reflecting vegetation moisture signals within the humid tropical environment rather than a direct cooling effect. Statistical analysis results indicate the thermal environmental transformation associated with campus-driven development. These findings highlight the environmental implications of peri-urban settlement densification around expanding university campuses and emphasize the importance of integrating vegetation conservation and climate-sensitive planning in rapidly developing institutional landscapes.
Rapid urbanization and declining ecological infrastructure have significantly intensified Urban Heat Island (UHI) effects and thermal vulnerability in metropolitan cities. This study presents a geospatial framework, “Built-up Optimization”, for microclimate resilience assessment and adaptive urban design in Delhi using remote sensing and patch-based spatial analysis. The framework integrates Land Use Land Cover (LULC) classification, Land Surface Temperature (LST) estimation, green-to-built-up ratio (GBR) analysis, and quartile-based hotspot prioritization techniques to identify thermally vulnerable urban zones. The LULC analysis revealed substantial landscape transformation, where 35.37
Land use/land cover (LU/LC) classification underpins environmental monitoring and land management. This study compares AdaBoost, CatBoost, LightGBM, and XGBoost for LU/LC classification of a high-resolution Pléiades image of Hopa District, Artvin Province, Türkiye. Four datasets were built incrementally: the original red, green, blue and near-infrared (RGB–NIR) bands; plus a Digital Elevation Model (DEM); plus the Normalized Difference Vegetation Index (NDVI); and plus three principal components (PCA1–PCA3) of Gabor-filtered imagery. On the four-band baseline all four algorithms performed almost identically (accuracy 0.768–0.780; Cohen’s kappa 0.628–0.648), indicating a ceiling set by the spectral data rather than by the learner. DEM produced the largest single gain (kappa +0.110 to +0.154); NDVI added almost nothing (+ 0.001 to +0.007), being a transformation of bands already present; the Gabor-derived components added a further +0.047 to +0.080, two thirds of it from PCA1 alone, and separated the algorithms most clearly. XGBoost performed best on the nine-band dataset (accuracy, weighted precision, weighted recall and weighted F1-score all 0.925; kappa 0.883), raising kappa from 0.648 to 0.883 relative to the baseline. CatBoost and LightGBM were competitive; AdaBoost was weakest throughout. Integrating spectral, topographic, and textural features therefore substantially improves LU/LC classification in this setting.
Groundwater in the crystalline terrain of Manachanallur Taluk, Tamil Nadu, is stored in a thin weathered–fissured profile, and no zonation combining surface and subsurface evidence exists for the taluk. Groundwater potential zones were delineated over 378.97 km2 from 17 terrain, structural, surface-cover, climatic, anthropogenic, geological and hydro-geophysical layers on a 30 m grid (EPSG:32,644), derived from the Copernicus GLO-30 elevation model, Landsat 8/9 and Sentinel-2 imagery of 2024 and ten vertical electrical soundings (VES) inverted with a four-layer model. Nineteen candidate layers were screened with Spearman correlation, Cramér's V, correlation ratio and generalised variance inflation factor, and the retained layers were weighted by Shannon entropy. Lineament density (0.126), Topographic Wetness Index (0.108) and aquifer longitudinal resistivity (0.107) received the largest weights. The High and Very High classes cover 145.8 km2 (38.5
In the Democratic People’s Republic of Korea (DPRK), long-term drought monitoring has only used limited indices such as the vegetation condition index (VCI). Vegetation health index (VHI) is a remote sensing drought monitoring index that has been widely used worldwide for the past 20 years. However, in most studies, the contributions of the VCI and the temperature condition index (TCI) to VHI were assumed to be equal. To overcome this shortcoming, we determined the optimal contributions of VCI and TCI to VHI using the correlation between standard precipitation evapotranspiration index (SPEI) and VHI, and obtained an improved VHI. Based on MODIS remote sensing data, the spatio-temporal variability of pixel-wise VHI at three months (April–June) time scales in the DPRK from 2001 to 2020 was analyzed using Theil-Sen slope estimator, Mann-Kendall test, and the future trends were predicted using the Hurst index. The proportion of regions dominated by VCI is 33
