The evolution of social-ecological system interactions and their impact on the urban thermal environment

Urbanization and intensifying heat waves are driving deterioration of the urban thermal environment (UTE), with significant implications for human health, energy consumption, and air pollution. The UTE is shaped by complex interactions within social-ecological systems (SESs), where urbanization alters surface structure and energy flows. While remote sensing has clarified spatial and seasonal patterns of UTE (e.g., stronger daytime than nighttime UHI; summertime peaks), interannual evolution and its linkage to stages of urban development remain insufficiently resolved and spatially heterogeneous. Notably, rapidly urbanizing regions show stronger UHI enhancement than highly urbanized areas, suggesting multiple steady states and a potential Environmental Kuznets Curve (EKC)-like inverted U trajectory for UTE as cities develop. Prior studies have often examined isolated drivers (social or ecological) due to a paucity of time-series data, limiting system-level understanding of feedback pathways critical for integrated mitigation. This study proposes and operationalizes a SES framework linking social (e.g., population, imperviousness, economic activity) and ecological (e.g., vegetation quality, ecological land) factors to surface energy balance processes that determine UTE. We evaluate 136 Chinese urban areas (2000–2021) to: (1) characterize spatio-temporal UTE evolution and identify nonlinear phases; (2) quantify single and interactive effects of social and ecological drivers; and (3) construct a social-ecological status index (SESI) that integrates SES feedback pathways to identify thresholds and implications for win-win heat mitigation strategies.

The study builds on extensive work documenting spatial and temporal variability of UTE and SUHI using remote sensing, showing stronger daytime SUHI and summertime anomalies, but with unresolved interannual dynamics and strong spatial heterogeneity. Prior global and regional analyses indicate that UHI trends are often stronger in rapidly urbanizing regions (Asia, Africa) than in mature urban systems (Europe, North America), implying developmental-stage dependencies. Research has linked UTE to social drivers (population density, impervious surfaces) and ecological factors (blue-green infrastructure), often substituting space-for-time due to scarce time series. Some studies suggest an EKC-like relationship between environmental impacts and development, raising the possibility that UTE follows an inverted-U over urban development. The literature also highlights nature-based solutions (green/blue spaces, connectivity, urban forests) as effective mitigators, while suggesting that population density may exert indirect rather than direct effects on UTE via interactions with other factors. Gaps include integrated, time-series-based assessments of SES interactions, quantification of critical feedback pathways, and thresholds where SES development shifts UTE trajectories from intensification to mitigation.

Study area: 136 urban areas (urban population >1 million; accounting for polycentric cities) across China, spanning diverse climate zones and rapid urbanization contexts, 2000–2021. Datasets: MODIS Terra MOD11A2 land surface temperature (LST, 1 km, 8-day composite) and MOD13A2 enhanced vegetation index (EVI, 1 km, 16-day composite); integrated nighttime lights (DMSP-OLS and SNPP-VIIRS) converted to Vegetation Adjusted Nighttime Light Index (NLVA, 0–1); WorldPop population density; China Land Cover Dataset (30 m) classified into nine types (overall accuracy 79.31%) for annual impervious and ecological land area estimation; administrative and urban boundaries from public sources. UTE quantification: ΔT = Tu − Tr, where Tu is LST over the urban area, Tr over a reference area selected within 10–30 km buffers as cropland/forest with EVI > 0.4 to ensure consistent land type and minimize climate context differences. Focus is summer daytime (June–August) annually from 2000–2021. Trend detection: Ordinary least squares (OLS) linear regression of ΔT vs. year for general trends and significance. Piecewise linear regression detects inflection points (≤2) with criteria: adjusted R2 > 0.6 and exceeding OLS; inflection within 2003–2019; significant slope difference before/after (p < 0.05). Trends categorized into phases: rapid increase (slope > 0.1), slow increase (0 < slope ≤ 0.1), and decrease (slope ≤ 0). Social and ecological factors: Ecological factors: ecological land percentage (sum of cropland, forest, shrub, grassland, water, wetland) and ΔEVI = EVIu − EVIr to represent ecosystem quantity and quality. Social factors: population density (demographic), NLVA (economic activity proxy adjusted for vegetation), impervious area percentage (urban expansion). Population density for 2021 approximated by 2020 values. NLVA corrected to mitigate oversaturation. Driver effect quantification: Geographic detector models applied to ΔT trends. Factor detector quantifies single-factor explanatory power (q-statistic, 0–1). Interaction detector quantifies combined effects and interaction types (nonlinearly enhanced when q(X∩Y) > q(X)+q(Y); bilaterally enhanced when q(X) < q(X∩Y) < q(X)+q(Y)). Method robust to nonlinearity and collinearity. SESI construction: Social-ecological status index (SESI) oriented to UTE based on resilience-pressure concept. Indicators positively correlated with ΔT are pressure; negatively correlated indicators are resilience. Indicators normalized to 0–1. Feedback pathways computed as Ri − Pi and weighted by interaction q-statistics (Qi) from the geographic detector to emphasize stronger SES interactions: SESI = Σ Qi (Ri − Pi). SESI ranges from −1 (least desirable) to +1 (most desirable). Piecewise linear regression relates SESI to ΔT slope to identify threshold effects.

- UTE magnitude and trends: - All 136 urban areas exhibit positive UTE (urban warmer than surroundings) with mean ΔT = +2.83 °C (2000–2021). - 124 of 136 urban areas (91.18%) show significant increasing trends in ΔT (p < 0.05), with mean annual trend 0.10 ± 0.04 °C. - Nonlinearity is common: 42 urban areas (30.88%) have inflection points; some (n=3) show two inflection points (three phases). More cities transition from increase to decrease (38) than from decrease to increase (7), and many shift from faster to slower warming in recent years, indicating growing mitigation. - Phase classification across all segments: 77 rapid increase phases (slope > 0.1), 76 slow increase (0 < slope ≤ 0.1), 28 decrease (slope ≤ 0). - Spatially, decreasing trends cluster in highly urbanized/economically advanced regions (Beijing–Tianjin–Hebei, Yangtze River Delta, Pearl River Delta). - Driver effects (geographic detector q-values): - Key social factor: economic development (NLVA) q=0.560 (p < 0.01). - Key ecological factor: vegetation quality (AEVI) q=0.442 (p < 0.01). - Impervious area percentage q=0.368; ecological land percentage q=0.366 (p < 0.01), reflecting strong quantitative roles of built and ecological areas. - Social factors positively correlate with ΔT; ecological factors negatively correlate. Social and ecological factors are negatively correlated with each other (trade-offs). - Factor trends by UTE phase: - Rapid increase phase: strong social growth (mean slopes: NLVA 0.02; impervious % 1.52) and sharp ecological decline (AEVI −0.006; ecological land % −1.52). - Slow increase: moderated social growth; weaker ecological decline (AEVI −0.003; ecological land % −1.15). - Decrease: social factors still rise, but ecological degradation slows (AEVI −0.001; ecological land % −0.77), with AEVI increasing in some cities. - Interactions: - Population density shows nonlinearly enhanced interactions with AEVI (q=0.640) and ecological land % (q=0.596), indicating stronger combined effects than the sum of single effects. - NLVA and impervious % exhibit bilaterally enhanced interactions with AEVI and ecological land %; strongest interaction: NLVA ∩ AEVI q=0.777. - SESI and thresholds: - SESI differs significantly across phases (ANOVA F=76.830, p<0.01): mean SESI highest in decrease phase (0.496), lowest in rapid increase (0.035); interquartile ranges rise from rapid to slow to decrease phases. - Piecewise relation between SESI and ΔT slope reveals a threshold at SESI=0.414. Below threshold, each +0.1 SESI reduces ΔT slope by ~0.025; above threshold, reduction is ~0.095 per +0.1 SESI, indicating accelerated mitigation beyond a critical SES state. - Recent SESI spatial patterns: highest in Fuzhou (0.701), Wuhan (0.636), Urumqi (0.609); lowest in Nanchang (−0.197), Ganzhou (−0.195), Suzhou (−0.160). Northeast China averages higher SESI (~0.467); southwest lower (~0.189). 31 urban areas exceed the SESI threshold (mostly eastern China), 93 remain below. - Differences by current UTE trend: - 23 urban areas decreasing, 101 increasing. Higher SESI associates with stronger UTE mitigation in both groups; sensitivity is greater where UTE is decreasing. - Mean SESI: decreasing cities 0.496 vs increasing 0.316. Decreasing-trend cities better restrain ecological degradation and expansion: mean slopes AEVI −0.001, ecological land % −0.77, impervious % 0.75 vs increasing-trend cities AEVI −0.003, ecological land % −1.27, impervious % 1.27. - Conceptual trajectory: UTE evolution aligns with an EKC-like pattern across development stages: initial stability, rapid increase with expansion and ecological loss, slowing increase with policy interventions, and eventual decrease via synergistic socio-economic and ecological strategies.

Findings reveal widespread, nonlinear UTE evolution with growing evidence of mitigation in many Chinese urban areas, particularly in advanced urban agglomerations. This trend likely reflects the scaling of eco-environmental policies (e.g., eco-city, low-carbon, sponge city) and nature-based solutions that enhance urban vegetation, connect parks via ecological corridors, and expand urban forests, which cool via shading, higher albedo, and evapotranspiration. Vertical urban development may also reduce daytime surface temperatures by increasing shading. The multi-city analysis supports an Environmental Kuznets Curve-like relationship between urban development and UTE: rapid expansion and ecological loss drive initial intensification; subsequent policy-driven ecological protection and managed growth slow and then reverse warming as cities advance. Driver analysis underscores the dominant roles of economic activity and vegetation quality, the quantitative contributions of impervious and ecological land, and the critical importance of interactions—especially the nonlinearly enhanced effects involving population density with ecological factors and the strong NLVA–AEVI interaction—highlighting the need for integrated SES-based mitigation strategies rather than single-factor measures. Population density alone is not a consistent direct driver of UTE; its influence emerges via interactions with other social and ecological variables. Examples from Beijing and major agglomerations suggest that socio-economic growth can co-occur with ecological enhancement, enabling win-win pathways that improve heat resilience and co-benefit biodiversity and flood/drought management. The SESI threshold demonstrates that once SES development surpasses a critical state, mitigation effects accelerate, offering a practical target for policy to prioritize ecological quality improvements and manage urban expansion in tandem with economic development.

The study provides a time-series, SES-integrated assessment of UTE evolution across 136 Chinese urban areas (2000–2021). It identifies pervasive nonlinearity in UTE trends, three characteristic phases (rapid increase, slow increase, decrease), and an EKC-like evolution with urban development. Economic activity and vegetation quality emerge as primary drivers, with significant, often enhanced, interactions between social and ecological factors. The SESI quantifies SES status oriented to UTE, reveals a critical threshold (SESI=0.414) beyond which mitigation accelerates, and maps priority areas for intervention. These findings advocate for integrated, SES-based strategies—combining ecological restoration/enhancement, urban form and expansion management, and nature-based solutions—to achieve win-win outcomes for urban society and ecosystems and to mitigate heat stress. Future research should extend indicator coverage (e.g., explicit anthropogenic heat components), improve data precision, and refine SES interaction pathways across broader regions and timescales.

- Limited factor set: Only five representative social/ecological indicators were used; key anthropogenic heat sources (industry, transportation, buildings) were not explicitly quantified, potentially biasing SESI. - Data accuracy: Land cover dataset overall accuracy is 79.31%, and remote sensing product uncertainties propagate into ΔT and SESI estimates. - Affordances of time series and scale: Although 22-year series were used, some social datasets (e.g., population 2021) required approximation; spatial heterogeneity within urban areas and polycentric structures may introduce aggregation effects. - Generalizability: The framework is applied to Chinese cities; transferability requires validation in other climatic and urban development contexts with appropriate local datasets.