The assessment of habitat quality is instrumental in preserving regional species diversity and ecosystem health, thereby forming the theoretical foundation for sustainable urban development. While the Invest model is a commonly employed tool for habitat quality evaluation, it fails to consider the terrain. This study, centered on Haitan Island, introduces the terrain diversity index to rectify the Invest model’s lack of terrain evaluation. The terrain diversity index, encompassing indices for terrain slope, undulation, and humidity, combined with the Invest model, was applied for a comprehensive assessment of the study area’s habitat quality. Furthermore, the distribution characteristics of habitat quality on Haitan Island, China, were examined using Moran’s I and LISA indices. The research indicates that forest land is the primary land cover type on Haitan Island, with blue-green space comprising forests, farmland, water bodies, and grassland, making up 66.8% of the island’s area, thus implying a positive overall ecological base. Habitat quality distribution within the study area displays spatial heterogeneity, with regions of superior habitat quality primarily found in the northeast areas such as Junshan. Compared to the standalone Invest model, the combined method considering terrain and vegetation cover types yields a more sensitive impact on habitat quality evaluation and improves the precision of identifying superior habitat quality by 56.7%. Spatial autocorrelation analysis revealed that the comprehensive habitat quality index in the study area exhibited clustered distribution. Hotspots were mainly identified in areas like Junshan and the western mangrove wetland, regions with a high concentration of habitat quality values, while low-value clusters were mostly found in the central city and southwestern plains. This study offers a novel methodology for habitat quality evaluation, compensating for the traditional Invest model’s neglect of terrain factors, and enriching the research on island habitat quality. It can provide fresh approaches and references for future habitat-related studies.

As climate change intensifies and urbanization continues, Fuzhou faces growing public health risks from the urban heat island effect. This study develops a Heat Risk Index (HRI) by integrating the Local Climate Zone (LCZ) classification with the Crichton Risk Triangle framework. Hazard was characterized using quality-controlled Landsat LST and the proportion of valid MYD11A2 composites exceeding a citywide P90 LST threshold. Exposure was represented by normalized population density; ln(P + 1) was used only for cartographic classification. Vulnerability incorporated older population, nighttime light, NDVI, and MNDWI. The continuous HRI was calculated as H × E × V, and ln(HRI) was used only for Jenks five-class map presentation. The results show clear spatial differences among LCZ types and counties, with higher risks concentrated in densely built and populated areas of the five urban districts, Changle, and localized county centers. LCZ differences in LST and HRI were statistically significant (both p < 0.001), and HRI showed significant positive spatial autocorrelation. These findings provide a basis for LCZ-specific heat-risk management while acknowledging uncertainty associated with data-year mismatch, spatial resampling, remotely sensed surface temperature, and the absence of independent health-outcome validation.

We evaluated 30 blue and 31 green spaces in Fuzhou using a July land surface temperature (LST) composite from four quality-controlled Landsat 8 scenes acquired during 2020–2022. A cubic temperature–distance procedure produced positive cooling estimates for all blue spaces and 30 green spaces. Modeled carbon-equivalent potential (MCEP) was calculated as an LST-based heat-energy equivalent under reference parameters, and MCEPA as MCEP per hectare. MCEP totaled 507.07 t CO₂. Median MCEP was 5.32 t for blue spaces and 2.42 t for green spaces (p = 0.284). Unadjusted MCEPA was higher in blue spaces (median 1.38 versus 0.90 t CO₂ ha⁻¹; p = 0.035; Cliff’s δ = 0.318), but the contrast was not significant after Holm correction (adjusted p = 0.211) and may partly reflect differences in patch-area distributions. MCEP increased with patch area, with allometric exponents of 0.606 and 0.865 for blue and green spaces, respectively. Because area is an integration dimension in MCEP, these are scale-dependent relationships. Bayesian Gamma models indicated positive area associations, whereas credible intervals for surrounding-ring NDVI and landscape shape index included zero. Spatial allocation reached 220 of 995 communities. The results describe surface cooling and conditional carbon-equivalent potential, not measured electricity savings or avoided emissions.