Mega-trees play a disproportionate role in forest ecosystems, concentrating large shares of biomass and carbon and sustaining high biodiversity, thereby influencing forest productivity, dynamics and ecosystem services. However, their growth and persistence are constrained by abiotic conditions, particularly in highly fragmented and human-modified landscapes such as the Brazilian Atlantic Forest. This study aimed to map and quantify emergent trees and mega-trees in a fragmented Atlantic Forest landscape in Pernambuco (Northeast Brazil) and to model the habitats that favour their occurrence and long-term survival. We applied machine-learning habitat modelling (MaxEnt) using terrain-derived predictors from a Digital Terrain Model (DTM) at multiple spatial scales (10, 30, 50 and 100 m). Across resolutions, terrain variables were consistently associated with suitable habitats, likely acting as spatial proxies for several non-exclusive processes (water redistribution, nutrient availability, solar exposure, wind shelter, microclimatic buffering and land-use filtering). The Topographic Position Index (TPI) was one of the most influential predictors, suggesting that concave landforms may represent candidate sheltered microsites for mega-tree persistence. Because terrain integrates several co-varying processes, we interpret this as an association rather than evidence of a single causal mechanism. The 10 m model achieved the highest performance (AUC = 0.906), while intermediate resolutions (30 and 50 m; AUC = 0.867 and 0.853) better represented habitat conditions surrounding mega-trees, reducing omission errors seen at 100 m. Overall, terrain-mediated refugia emerge as conservation-relevant targets, providing a spatial framework to support management and restoration planning for mega-trees in fragmented tropical forests.
Urban structure and functional layout create complex population activity hotspot (PAH) with polycentricity, spatiotemporal dynamics, and indistinct boundaries. Current research often overlooks spatial heterogeneity and assumes linear correlations, limiting both the accurate characterization of PAH and the exploration of nonlinear relationships between PAHs and urban features. Therefore, we propose an adaptive region search method with dual-consideration of polycentricity and activity intensity gradient (DPAIG) for accurately identifying high-density population activity zones and their dynamic perimeters. Subsequently, a 2D Gaussian fitting model (2DGFM) quantifies PAH morphology with concise parameters. Using interpretable machine learning, we analyze nonlinear relationships between PAH morphology and urban features. A case study in Wuhan demonstrates that the boundaries identified by DPAIG closely align with geographic boundaries. The Gradient Volatility Index (GVI) is 0.586, a 7.57% reduction relative to the second-best method, with smoother gradient transitions. The 2DGFM offers a high-precision parametric representation of PAH morphology (R2 = 0.939). Further analysis identifies building-related and natural terrain features as having the highest predictive contribution to the extent and orientation of PAHs, respectively, showing “inverted U-shaped” and “W-shaped” relationships with notable threshold effects. This study presents a quantitative framework for characterizing dynamic PAHs and examining their nonlinear relationships with related urban factors, providing valuable insights for emergency management and urban sustainability.
As weather extremes become more frequent and persistent, cities need to understand how weather conditions are associated with everyday spending. Yet, existing research rarely provides a comprehensive picture of these relationships across spending amount, transaction frequency, sectoral composition, and cross-city variation. By linking nationwide credit card transactions from South Korea covering 2019 to 2020 with city-level meteorological data, this study examines associations of apparent temperature, precipitation, and extreme-heat persistence with everyday spending. Using 20–25°C as the reference range, the results show that days with apparent temperatures of 33°C or higher were associated with an estimated 3.44% increase in overall transactions and an estimated 2.30% decrease in overall expenditure. Sectoral patterns were uneven, with higher transactions and expenditure in several retail, restaurant, and recreation categories but lower transactions and expenditure in Home and Internet Shopping. Beyond heat on a given day, a 10-mm increase in precipitation was associated with 1.05% fewer transactions and 0.94% lower expenditure. Meanwhile, each additional extreme-heat day during the preceding three days was associated with declines of 0.34% in transaction frequency and 1.60% in expenditure. A cross-city comparison further suggests that climatic context, coastal location, age structure, and population density were associated with heterogeneity in spending responses to weather. As global warming increases heat exposure, sectors and cities may face different economic gains and losses. Heat-adaptation strategies should therefore account for daily service demand, vulnerable populations, and local weather sensitivity.
