Urban rat infestations represent a persistent public health challenge, yet municipal pest management programs often operate reactively and with limited resources. This study applies a suite of spatiotemporal statistical approaches (geographically weighted regression, spatial lag modeling, Random Forest, and Extreme Gradient Boosting) to predict the spatial distribution of rat sightings across Chicago, Illinois, over a fifteen-year observation window (2011–2025). Using 311 service request data linked to a hexagonal grid (150 m cells) and a set of environmental, demographic, and property-value predictors, we assess the individual and ensemble performance of these models on held-out 2025 rat-sighting data. Our comparative analysis demonstrates that the machine learning frameworks and the spatial lag model generally outperformed geographically weighted regression. Specifically, Random Forest, the spatial lag model, and XGBoost each explained approximately 72
Urban trees moderate city climates, yet how their leaf-area density relates to projected future climates is poorly constrained globally. We fitted a Random Forest model to monthly temperature and precipitation normals and the multi-year mean leaf area index (LAI) derived from a satellite-informed machine-learning urban tree product across 21,590 urban cells, then projected end-of-century change under six Shared Socioeconomic Pathways. Leave-one-continent-out validation gave R^2 = 0.846 (95 0.813 – 0.873 from spatial-block resampling), but that spatial skill does not transfer to time: driven by observed reanalysis anomalies over exactly the satellite record, the same relationship reproduces neither the magnitude nor the pattern of observed change ( ρ = +0.087 ). The response variable is itself a machine-learning product built from overlapping climate predictors, so the spatial fit is not independent evidence of that relationship either. Under modified Mann–Kendall tests with false-discovery-rate control, 19.8
Urban forests have gained increasing attention as nature-based solutions, yet quantitative information on aboveground biomass and carbon storage in small urban forest remnants remains limited, particularly in small municipalities. This study aimed to quantify aboveground biomass, carbon stocks, and their illustrative economic valuation in an urban Tiny Forest located in southeastern Brazil. A complete tree inventory was conducted across a 2.76-ha area, including all living trees with a diameter at breast height (DBH) ≥ 5 cm. Aboveground biomass was estimated using an allometric equation developed for urban trees, carbon stocks were calculated using a conversion factor of 0.47, and an illustrative economic valuation was derived from a reference price in the voluntary carbon market. The inventory recorded 539 trees belonging to 68 species and 24 botanical families, with native species accounting for 79.0
Urban expansion is changing bird assemblages across sub-Saharan Africa, where the sorting of species along urban–rural gradients is understudied. We present a multi-scale analysis of bird assemblages at four locations in Minna, central Nigeria, combining trait-environment modelling, ordination, beta-diversity analysis and machine learning. We recorded 115 species (15 orders, 50 families) in 71 single-visit, 10-minute, fixed-radius point counts (April to July 2021). Because locations were sampled at different times, location and season are partly confounded. Bird detection was weakly predicted (AUC = 0.676): generalists and frugivorous birds were more likely to be detected at multiple stations compared to specialists. Abundance where detected was weakly predicted (R² = 0.133) and higher among seed-eating species. Trait-environment interactions did not improve prediction, indicating trait-associated detection rather than trait-mediated environmental filtering. No community metric was predictable at a withheld location, and no feeding-guild model had predictive skill. Composition differed among locations (PERMANOVA pseudo-F = 1.68, R² = 0.070, p = 0.001), although dispersion also differed and the three locations surveyed within nine days did not differ significantly. Beta-diversity was mostly explained by replacement (0.65) compared to nestedness (0.07). Shannon diversity and functional dispersion were lowest in Gidan Kwano, whereas richness did not differ, even after rarefaction. Composition was associated with tree density, tree size, and elevation, yet not independent from geographic distance. Turnover-driven replacement suggests that networks of habitat patches would provide better protection of regional diversity, which should be confirmed through repeated surveys at same seasons.
Wildlife roadkill represents a major direct threat to many species, affecting population dynamics, and in some cases driving species toward extinction. Although roadkill has been widely documented in Colombia, studies in the Caribbean Region, which encompasses key vulnerable ecosystems such as the Tropical Dry Forest (TDF), remain scarce. Here, we assessed whether road characteristics influence the taxonomic composition of vertebrate roadkill on peri-urban roads in the department of Córdoba. We conducted systematic bicycle surveys along three inter-municipal roads over a three-month period during the dry-to-wet transition season, recording species identity and local road characteristics, including artificial lighting near houses, fences, water bodies, traffic signals, and wildlife crossings. We recorded 317 roadkill events, most of which were identified to species (217), with reptiles representing the largest proportion of records, followed by amphibians, mammals, and birds. The taxonomic composition of roadkill records was associated with local road characteristics. Mammal records were more likely on illuminated road sections and in the absence of fences, but less likely near water bodies. Bird records were less likely in the absence of fences and in the presence of wildlife crossings. Amphibian records were more likely in the presence of pole fences and less likely on illuminated road sections, whereas reptile records were more likely near water bodies. These findings indicate that local road characteristics shape the relative representation of vertebrate groups among roadkill records in peri-urban TDF landscapes. Mitigation measures are necessary to reduce mortality and preserve ecosystem functionality.
