The study addresses what the authors describe as an epistemological gap between the objective thermodynamic performance of urban wetlands and the subjective thermal well-being of their visitors. A park can register impressively low land surface temperatures on a Landsat image while still leaving visitors sweltering along exposed pathways, or it can feel pleasantly cool in spots that satellites classify as only moderately green. Traditional studies typically measure one side of this divide or the other, rarely both at scale. Zhao and Wang’s answer was to synthesize Landsat-derived environmental data with large-scale crowd-sourced digital footprints, treating the reviews posted on the Dianping platform as a form of social sensing that captures in-situ thermal experience.

The technical core of the approach lies in the use of large language models to extract and classify temperature perception sentiments from unstructured public reviews. The researchers processed 11,109 online reviews, employing the AI model to identify statements in which visitors described feeling hot, cool, or comfortable in relation to specific park environments. Because AI-generated classifications can be error-prone, the team subjected the output to rigorous human verification, including a five percent random sample check that achieved a Cohen’s kappa of 0.82, a level of agreement generally considered to indicate strong reliability. All user identifiers were removed during data processing, and only anonymized and aggregated perception indicators were analyzed, which meant the study involved no direct interaction with human participants and required no formal ethical approval.

On the physical side, the researchers derived a battery of landscape metrics from satellite imagery, including land surface temperature profiles, patch richness density, and the density of impervious-surface patches. Patch richness density, a measure of how many different land-cover types are packed into a given area, emerged as strongly associated with cooling efficiency. The finding suggests that landscape heterogeneity may enhance thermal well-being through two parallel channels: physically, by promoting air exchange between diverse surfaces, and psychologically, through the restorative benefits of visual complexity. A park with a rich mosaic of water, woodland, wetland, and lawn appears to do more for human comfort than a uniform expanse of any single cooling surface.

Perhaps the most striking result is what the authors call an accessibility paradox. Conventional wisdom in urban climate design holds that pavement should be minimized, since impervious surfaces absorb heat and raise local temperatures. Yet the analysis found that a moderate density of impervious-surface patches is associated with higher perceptions of coolness among park visitors, with an exploratory empirical turning point around 32.3 patches per hectare in the Shenzhen wetland-park sample. Below that threshold, well-distributed access infrastructure appears to act as a behavioral mediator, facilitating visitor movement toward the cooling cores of the parks. Paved paths, in other words, are not merely heat sources; they are the routes by which people reach the shaded, waterside zones where cooling is most intense.

The authors are careful to frame this pattern as hypothesis-generating rather than prescriptive. The identified turning point is exploratory and subject to site-specific ecological and urban-morphological conditions, meaning that a figure derived from Shenzhen’s high-density subtropical context cannot simply be transplanted to wetland parks in other climates or urban forms. Still, the finding challenges the conventional minimize-pavement assumption and suggests that thermal satisfaction depends on a strategic balance between ecological integrity and recreational accessibility. A park that is ecologically pristine but difficult to traverse may deliver less human cooling benefit than one that pairs robust green infrastructure with thoughtfully distributed access paths.

The study also identified a context-dependent comfort threshold in the physical temperature data. Comfortable perception showed an empirical turning point around a land surface temperature of 28.8 degrees Celsius, indicating that visitors’ sense of thermal comfort operates within an adaptation range rather than conforming to a universal standard. This resonates with a growing body of thermal comfort research showing that people in different climates, cultures, and built environments calibrate their expectations differently. In Shenzhen’s humid subtropical environment, a satellite-measured surface temperature below roughly 29 degrees appears to fall within the range visitors describe as comfortable, while temperatures above it shift sentiment toward discomfort.

Methodologically, the work contributes what the authors call a replicable, AI-enhanced paradigm for aligning climate resilience research with public thermal perception. The computational social science framework could in principle be applied to any city where remote sensing data and large volumes of public online reviews coexist, offering urban planners a way to audit not just how cool their green infrastructure is on paper but how cool it feels to the people who use it. The authors note that AI tools were used to improve readability and grammatical accuracy of the manuscript but were not used to generate scientific conclusions, fabricate data, or create figures, and the researchers reviewed and edited all AI-processed content while taking full responsibility for the publication’s integrity.

The implications for urban governance are significant at a moment when cities worldwide are investing heavily in blue-green infrastructure as a climate adaptation strategy. If the findings hold up across further sites, they imply that design benchmarks built purely on physical metrics, such as maximizing vegetation cover and minimizing pavement, may misallocate resources. Instead, planners might aim for a portfolio approach: heterogeneous landscapes rich in patch types, cooling cores that are genuinely reachable, and access infrastructure distributed at densities that mediate behavior without eroding ecological function. The non-linear patterns identified in the study offer indicative design references rather than prescriptive rules, positioning thermal well-being as an emergent property of the interaction between landscape, climate, and human movement.

The research, supported by the Guangdong Basic and Applied Basic Research Foundation, arrives as an open-access article with a permanent DOI, and its supplementary datasets, including land surface temperature profiles and normalized cooling metrics, are publicly available for other researchers to scrutinize and extend. As heat waves intensify and urban populations grow, the study offers a glimpse of a future in which the thermal experience of city dwellers is measured not only from orbit but through the words they leave behind, decoded at scale by machines and translated into evidence for human-centric urban design.

Subject of Research: Environmental Science

Article Title: Decoding urban thermal well-being: bridging the gap between physical cooling and social perception in wetlands via AI-enhanced sensing

Article References: Zhao, G., & Wang, W. (2026). Decoding urban thermal well-being: bridging the gap between physical cooling and social perception in wetlands via AI-enhanced sensing. Humanities and Social Sciences Communications. https://doi.org/10.1057/s41599-026-09466-0

Image Credits: AI Generated

DOI: 10.1057/s41599-026-09466-0

Keywords: urban heat, wetland parks, AI, remote sensing, thermal comfort, Shenzhen, urban planning, climate resilience, Decoding, urban, thermal, well-being