Phenological Metrics in the Mitigation of Urban Heat: Timing and Thresholds

Phenological Metrics in the Mitigation of Urban Heat: Timing and Thresholds | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Phenological Metrics in the Mitigation of Urban Heat: Timing and Thresholds Qihao Weng, Baoling Gui, Josep Peñuelas, James Voogt, Anshuman Bhardwaj, and 4 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6948672/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted You are reading this latest preprint version Abstract Urban vegetation is critical for mitigating summer heat, but previous studies have largely relied on static greenness metrics, leaving a gap in understanding of how vegetation phenology (its seasonal life cycle) regulates urban temperatures on a global scale. Here, we utilize interpretable machine learning and satellite data to quantify the influence of key phenological metrics, encompassing growth intensity (e.g., peak greenness), timing (e.g., start of season), and duration, on summer Land Surface Temperature (LST) across 24 major global cities. We found that: 1) a significant temporal mismatch exists in over 80% of cities, with vegetation green-up lagging seasonal surface warming by 50–100 days, creating a window of thermal vulnerability; 2) seasonal accumulation of Enhanced Vegetation Index (EVI) provides stable, linear cooling, whereas Maximum EVI (MEV) and EVI amplitude (EA) exhibit a distinct threshold effect, with their cooling benefits diminishing or even reversing beyond a critical point; 3) vegetation's cooling effect changes with context, delivering roughly 25% greater cooling in the top 10% of temperature extremes compared to moderate conditions; and 4) in certain contexts, vegetation's cooling effect is observationally weakened or even offset when it is masked by the dominant influence of a positively correlated warming factor, such as high elevation. These findings provide mechanistic evidence that simply increasing green cover is insufficient; future urban heat mitigation must shift to "phenology-aware" designs that synchronize vegetation's life cycle with seasonal heat peaks to achieve maximum cooling benefits. Earth and environmental sciences/Climate sciences/Climate change/Climate-change mitigation Earth and environmental sciences/Ecology/Urban ecology Earth and environmental sciences/Environmental social sciences/Sustainability Earth and environmental sciences/Environmental social sciences/Climate-change mitigation Figures Figure 1 Figure 2 Figure 3 Figure 4 1. Introduction Rapid urbanization has intensified the urban heat island (UHI) effect, threatening public health, raising energy demand, and disrupting vegetation phenology 1 – 3 . During summer, in particular, excessive urban heating can exacerbate thermal stress, deteriorate air quality, and alter ecosystem functions, thereby affecting the timing and length of vegetation growth cycles 3 – 5 . In contrast, urban vegetation helps moderate surface heat through shading and evapotranspiration, and higher vegetation cover is generally associated with lower land surface temperature (LST), especially during summer when such cooling effects are strongest 6 – 8 . However, this cooling is neither spatially uniform nor linearly proportional to vegetation density. Prior studies have identified threshold effects, where substantial cooling benefits occur only once green space coverage exceeds a critical proportion of the urban landscape. Below this threshold, additional greening may yield limited or negligible thermal mitigation—especially in compact, densely built urban cores 9 . These limitations underscore the need to better understand the conditions, magnitudes, and timings under which vegetation most effectively mitigates urban heat. Numerous studies have confirmed a strong inverse relationship between vegetation cover and LST, with the cooling effect being especially pronounced in the growing season. Areas with higher fractional vegetation cover (FVC) consistently have lower summer LSTs than do sparsely vegetated zones, and the intensity of the surface UHI effect often decreases as vegetation density increases 10 , 11 . This cooling effect, however, is not static but varies substantially across seasons, climatic zones, and vegetation functional states. The correlation between vegetation cover and LST weakens during winter or dry periods when vegetation is dormant or leafless 12 – 14 . This weakening highlights the importance of phenology in determining the capacity of seasonal thermal regulation. When plants green up earlier in spring or delay senescence in autumn, the duration of active cooling periods can be extended, thereby increasing the benefits of thermal mitigation in seasons 15 . In contrast, disruptions in phenological rhythms, such as early senescence induced by drought, can suppress transpiration and weaken the ability of vegetation to cool when needed the most during summer peaks 16 – 18 . Recent studies have increasingly recognized the interactions between phenological dynamics and LST 5 , 19 , however, key uncertainties remain. Most notably, existing work has focused on individual cities or homogenous climate zones 9 , 20 , limiting the ability to capture global variability in urban form, vegetation structure, and seasonal heat exposure. As cities worldwide confront escalating heat risks, there is an urgent need to develop a globally comparable understanding of how phenological traits modulate urban climate, especially where species-level vegetation data are scarce. Research on urban phenology has largely emphasized temporal indicators—such as the start and end of season (SOS and EOS)—while underappreciating amplitude-based traits like peak greenness, cumulative productivity, and the rate of vegetation growth or decline. These attributes are crucial for assessing the scale and persistence of vegetative cooling 19 , 21 , particularly in cities where vegetation is fragmented or human-managed, and where thermal resilience depends on both timing and intensity of green activation. The biophysical pathways linking phenology and thermal regulation crucially remain poorly understood. Vegetative cooling may only occur beyond specific thresholds of greenness, saturate at high canopy densities, or lag behind seasonal increases in temperature 1 , 22 . Responses may also vary across climates: vegetation may cool more effectively in hot and dry regions but underperform in mesic regions where evapotranspiration is less limiting. Such asymmetries challenge the assumption of the efficiency of uniform cooling and underscore the need for nonlinear analytical frameworks that detect thresholds, saturation, and time-lagged effects 4 , 23 . To address these gaps, we investigated phenological–thermal coupling across 24 global megacities using satellite-derived vegetation metrics, harmonized data for surface temperature, and interpretable machine-learning techniques. Specifically, we: (1) quantified the relative influence of both timing- and magnitude-based phenological traits on urban summer LST, (2) examined intercity variability and spatial heterogeneity in interactions between the vegetation and temperature, and (3) identify nonlinear, threshold, and lagged phenological responses under thermal stress. These insights, while based on surface-temperature proxies, provide mechanistic evidence to guide temperature-responsive, phenology-aware greening strategies for urban climate adaptation. 2. Results 2.1 Magnitude and directionality of LST responses to vegetation phenological changes We used SHapley Additive exPlanations (SHAP) values derived from Categorical Boosting (CatBoost) regression models for the 24 cities to quantify the contribution of vegetation phenology and environmental factors in determining urban LST in summer (Fig. 1 ). R² for these models varied across cities, ranging from 0.47 (Istanbul) to 0.97 (Nairobi and Tehran), indicating the diverse explanatory power of the selected features depending on the local geography and ecology. An analysis of feature importance across the 24 global cities identified both universal and region-specific mechanisms driving urban summer LST. Nighttime light intensity was the most frequent dominant predictor (Fig. 1 ), ranking as one of the three most frequent predictors for 17 cities, underscoring the strong and consistent impact of anthropogenic activity on urban heat. Elevation followed closely, ranking as one of the three most frequent predictors for 15 cities, particularly in high-relief cities such as Mexico City, Tehran, and Nairobi, where topographic variability strongly influenced local patterns of temperature. Vegetation indicators also played a critical role: Enhanced Vegetation Index (EVI)-based metrics, three phenological variables (maximum EVI (MEV), seasonal EVI accumulation (SEA), and EVI amplitude (EA)) were the most influential in modulating LST across most cities and were among the three most frequent factors in 13, 8, and 6 cities, respectively, highlighting the importance of both overall greenness and seasonal variation in modulating urban temperatures. Proximity to water bodies ranked highly in arid and hydrometeorologically sensitive cities such as Cairo, Lima, and Riyadh, indicating the crucial cooling effect of water bodies in dry climates. Interestingly, variables of phenological timing (e.g. SOS and peak timing) were less often among the most frequent predictors, suggesting that their thermal regulatory impact may be more spatially heterogeneous and context-dependent but important in some cities (e.g. Tehran). Identifying the most important phenological variables and assessing how these variables nonlinearly affect temperature, how their influence varies under different temperatures ranges, and how they interact with other urban environmental factors are essential to fully understand the role of phenology in regulating urban summer LST. The following integrates analyses of sensitivity and partial-dependence diagnostics to explore both the magnitude and directionality of the responses of LST to changes in vegetation, highlighting critical thresholds, saturation points, and joint effects with covariates such as elevation, nighttime light, and distance to water. The findings indicated that interactions between vegetation and temperature were much more complex than linear, univariate interpretations. Figure 2 shows that, across nearly every city, increasing MEV, SEA, or EA consistently yields a negative ∆LST, indicating a cooling effect regardless of whether pixels are in the moderate (60–80%) or high-temperature (top 20%) subset. In every case, the high-temperature curve lies below the moderate-temperature curve, meaning that, for the same perturbation, cooling is stronger (more negative ∆LST) under extreme heat than under moderate heat. SEA invariably follows a near linear decline in both strata, with only modest offsets (typically < 0.3°C) between moderate and high curves, demonstrating its stable cooling contribution. MEV and EA, by contrast, exhibit a clear threshold effect: for negative perturbations or small ∆ values, the response is relatively flat, and cooling is modest. As ∆ approaches zero and becomes positive, cooling accelerates rapidly, producing a steeper drop in ∆LST. Beyond ∆ ≈ 0.1–0.2, both curves begin to plateau, indicating diminishing marginal cooling (e.g., Riyadh (Fig. 2 a)). Not every city conforms perfectly to this pattern. First, a few MEV or EA curves slope upward as ∆ increases—contradicting the expected cooling trend. For instance, Nairobi’s moderate-temperature MEV rises into positive territory for ∆ > 0.1 (Fig. 2 b), and Tehran’s moderate-temperature MEV climbs slightly for ∆ around 0 (Fig. 2 c). Second, some curves cross ∆LST = 0, indicating a switch from warming to cooling (or vice versa). For instance, Guangzhou’s moderate-temperature MEV crosses at a similar range (Fig. 2 d), whereas their high-temperature curves remain negative. Third, certain moderate-temperature trajectories remain nearly flat at ∆LST ≈ 0 across all ∆—for example, Sydney’s SEA barely moves above or below zero (Fig. 2 e), suggesting minimal net temperature change despite perturbations. Finally, although every city’s moderate curve lies above its high curve, the vertical gap and overall ∆LST range differ remarkably: Cairo and Nairobi often show mid–high separations ≲ 0.1°C (Fig. 2 f,b), whereas Sydney can exceed 0.6–1.0°C (Fig. 2 e). These discrepancies reveal substantial heterogeneity in how extra greenness cools under moderate versus extreme heat. 2.2 Joint effects of vegetation and urban environments on LST in summer The preceding section highlighted divergent and occasionally counterintuitive behaviors, such as when increased vegetation indices were associated with increasing temperatures, but these phenomena require a deeper exploration of the underlying mechanisms. We used partial-dependence plots (PDPs) to distinguish between such anomalies, which enabled the visualization of marginal effects and patterns of interaction between individual predictors and LST. Unlike global analyses of sensitivity that assess average impacts, PDPs allow us to identify nonlinear transitions and conditional dependencies. They helped us to determine how vegetation variables, such as EA or maximum greenness, had either cooling or warming effects depending on co-occurring environmental factors such as urban built-up density, elevation, or distance to water. This analysis found that the vegetation did not function in isolation; its capacity for thermal regulation was instead determined, and sometimes constrained, by its interaction with co-occurring factors like elevation and urban heat intensity. In Beijing (Fig. 3 a), we observed that higher MEV values tend to coincide spatially with pixels located at relatively higher elevations. While elevation is typically associated with decreasing air temperature due to the lapse rate, it is important to note that land surface temperature (LST) responds primarily to surface energy balance rather than air mass characteristics. In the urban, elevation differences are modest, but elevated areas may receive more direct solar radiation due to reduced obstruction, greater sky exposure, and lower shading from surrounding structures or vegetation. These factors can lead to increased daytime LST, even at slightly higher altitudes. As a result, the expected cooling from greater MEV in these areas is partially or fully masked by the stronger surface warming associated with topographic exposure. Consequently, the net ∆LST associated with MEV appears neutral or even slightly positive in these zones—not because MEV fails to cool, but because elevation-linked radiative effects dominate the LST response. In Jakarta (Fig. 3 b), a similar confounding phenomenon emerges between EA and nighttime light intensity. The PDP for EA versus nighttime light shows that the higher EA values are clustered in areas with central urban zones of more bright night light readings characterized by dense built ups and strong anthropogenic heat emissions. Within those zones, incremental increases in EA no longer translate to additional cooling; instead, the persistent waste heat from traffic, air-conditioning, and concrete surfaces dominates. As a result, when EA reaches these urban peaks, the observed ∆LST actually rises or remains flat, despite greater vegetation amplitude 24 . 2.3. Phenology–temperature interplay along urban-rural gradients This section examines the spatial coordination and temporal alignment between the phenological transitions of vegetation and the dynamics of summer LST along urban–rural gradients. We analyzed both the north-south and east-west cross-sectional profiles of EVI-derived phenological markers and LST curves to determine whether vegetation and temperature responded synchronously across space and whether cities had coherent thermal-phenological coupling. Figure 4 identifies consistent urban–rural differences, quantifies timing mismatches between biophysical and thermal signals, and assesses how urbanisation modifies vegetation functioning. Our transect-based phenological and thermal profiles across 24 global megacities identified both consistent spatial patterns and city-specific deviations that together highlight the complex interplay between vegetative activity and urban thermal dynamics. Figure 4 demonstrates systematic and city-wide temporal mismatches between vegetation phenology and surface temperature trajectories, as reflected in the timing of three key seasonal markers. Across more than 80% of cities (as shown in Section 4 of the Supplementary Information), the EVI-based SOS lags behind the LST-based onset of seasonal warming by 50 to 100 days such as Mexico City (Fig. 4a), particularly within urban cores. In contrast, the timing of phenological peak and EOS relative to LST shows more complex and divergent patterns. In some cities, such as Paris and Cario (Fig. 4b,c), EVI peak appears significantly earlier than LST peak (by up to 80–100 days), while in others (e.g., Mexico City (Fig. 4a)), EVI peak lags behind, indicating delayed vegetation responses to sustained warming. For EOS, a typical pattern is that EVI-based EOS precedes thermal decline, especially in cities with strong heat retention like Paris or Cario (Fig. 4b,c), where LST remains high even after vegetation enters senescence. Building on this temporal perspective, our transect-based spatial analysis reveals considerable heterogeneity in how these timing mismatches manifest across the urban–rural gradients. In cities like Mexico City (Fig. 4a), we observed parabolic (U-shaped) spatial profiles, where vegetation SOS and peaks occur earlier in city centers and later toward the outskirts, while EOS is delayed at the core. This pattern creates an extended growing season within urban cores but does not necessarily align with temperature cycles, as LST patterns are relatively flatter across the transects. In fact, in most cities—including Cario, and Beijing (Fig. 4c,d)—LST shows limited spatial variability between urban and suburban areas, while phenological metrics vary by more than 60–100 days. This contrast underscores that while surface heat accumulation may follow a relatively coherent seasonal trajectory, vegetation activation is far more heterogeneous and sensitive to local conditions. 3. Discussion This study provides a globally comparable, mechanistic understanding of how vegetation phenology modulates summer LST across 24 global cities spanning diverse climates, topographies, and urban forms. In contrast to earlier works that rely predominantly on static greenness metrics or categorical land cover types 9 , 20 , 25 , this study introduces a phenology-informed perspective that jointly considers the effects of both timing and intensity of vegetation activity. Our findings not only establish the relative importance of vegetation phenology in shaping urban thermal environments but also lay theoretical foundation for designing phenology-informed greening strategies that are seasonally aligned and climate responsive. Recent studies have explored greening strategies to reduce urban heat 20 , 26 , but few have integrated the seasonality and synchronicity of vegetation response with temperature dynamics at moderate spatial resolution. Our work thus moves beyond correlational patterns and establishes a functional link between phenological dynamics and LST, echoing calls for mechanistic metrics in studies of urban sustainability 27 , 28 . Machine-learning modelling and analysis indicated that cities that had higher model performance (e.g. Nairobi, and Mexico City), as indicated by higher R² values, typically possess distinct elevation gradients or pronounced spatial heterogeneity in their green space structures. These physical and ecological variations enhance the detectability of the vegetation–temperature relationship by introducing stronger spatial contrasts 29 . For example, in Nairobi, the combination of topographic variation and clustered green space facilitates clear spatial transitions in phenology–LST dynamics. In contrast, cities such as Istanbul or New York—with fragmented morphologies and thermally relatively uniform environments—tend to show weaker statistical associations, due likely to microclimatic noise and anthropogenic influences that obscure vegetation signals 30 – 33 . Temperature-stratified SHAP and interaction analyses further reveals the nonlinear and threshold-dependent cooling responses of three key vegetation phenology parameters—MEV, SEA, and EA—under different thermal regimes across diverse global cities. Overall, increases in these parameters are associated with negative ∆LST (i.e., cooling), yet with distinct behavioral differences. Among them, SEA exhibits a remarkably linear and consistent relationship with LST across both moderate- and high-temperature conditions, with minimal offsets (generally < 0.3°C). Unlike MEV and EA, SEA is less susceptible to nonlinear breakdowns from water stress or thermal saturation. This finding establishes SEA as a robust metric for urban greening assessment and a quantitative reference for long-term heat mitigation 34 , 35 . In contrast, MEV and EA follow a distinct activation–saturation pattern: under low perturbations (∆ < 0), their cooling impact is negligible; after surpassing a critical threshold (∆ ≈ 0), cooling accelerates rapidly before plateauing. This nonlinearity likely reflects key ecophysiological limits: at high densities, self-shading and water competition can reduce marginal cooling, while extreme heat may trigger stomatal closure, negating the benefits of high greenness 36 – 38 . This pattern is particularly pronounced in arid and semi-arid cities—such as those in Southwest Asia, the Indian subcontinent, and the southwestern U.S.—where evapotranspiration is tightly coupled with soil moisture availability 39 , 40 . These results align with ecophysiological studies 41 that identify a leaf area–dependent trigger for transpiration and confirm, for the first time at a global pixel scale, that greenness-related cooling is highly contingent on local hydrological constraints. In humid or well-watered cities, such as those in Western Europe or coastal and southern hemisphere regions, this threshold response weakens, and marginal cooling ramps up more gradually demonstrating that excess greenness in such environments yields consistent, non-abrupt thermal benefits. We identify and quantify pixel-scale cooling thresholds that demarcate activation and saturation phases of MEV/EA—revealing spatially explicit tipping points not previously mapped. We further ground our analysis of the phenology-LST relationship in ecological niche theory and propose new evaluation metrics 42 . We propose a set of new evaluation metrics: the Cooling Activation Threshold (CAT), defined as the minimum MEV/EA level at which net LST reduction begins; the Saturation Cooling Point (SCP), beyond which additional greening yields negligible further cooling; and the Linear Cooling Index (LCI), characterizing SEA's continuous and stable contribution to LST moderation across thermal strata. The perturbation values used in our models can be scaled back to actual EVI or amplitude units via the inverse of z-score normalization (i.e., multiplying the perturbation by the standard deviation of the original data and then adding the mean), which allows direct mapping from modeled perturbations to feasible vegetation enhancements. This not only translates our findings into tangible benchmarks for greening design, assessing whether a plant's traits fit the city's unique thermal niche but also lays the groundwork for temperature-oriented phenological benchmarks in urban greening design. Our results challenge the hypothesis that increased greenness universally yields lower surface temperatures. In several contexts, especially in compact urban cores with strong anthropogenic heat, elevated terrain, or low hydrological connectivity, we observe localized warming or flat ∆LST responses to vegetation increases, particularly in densely built areas. Moreover, limited Sky View Factor (SVF) in compact urban areas may reduce the effectiveness of vegetative cooling by constraining radiative loss and air circulation, further modulating the phenological impacts on surface temperature 43 . These findings suggest that vegetation cooling is not univariate but modulated by interactions with topography, urban morphology, and human activity intensity. Such outcomes extend prior work 25 , which identified the importance of irrigation and land cover type but lacked the resolution to pinpoint thresholds or reversals. Our analysis offers new insight into where and why greening may fail or even backfire, such as when dense vegetation co-occurs with poor ventilation or high nocturnal heat emission. These mechanisms had been hinted in studies like Kim et al. 21 , yet not explicitly measured at the pixel scale across global major cities. Our findings most notably identified a pervasive temporal decoupling between vegetation activity and early-season increases in temperature across most of the cities. EVI-based SOS lagged behind the initial seasonal increase in LST by 50–100 d in > 80% of the cities, including Beijing, New Delhi, Riyadh, and Lagos. This lag is particularly pronounced in cities located in monsoonal, arid, or equatorial climate zones, where delayed vegetation onset often fails to align with periods of peak thermal accumulation 44 . One potential explanation is that, in dry environments, LST may respond more rapidly than air temperature (Tair) due to the early-season exposure of bare soil and impervious surfaces, which heat quickly and are strongly detected by thermal sensors 45 , 46 . In contrast, plant phenology often tracks accumulated Tair and soil moisture availability, leading to temporal mismatches in observed thermal and vegetative signals 47 . This temporal mismatch often stems from distinct urban drivers, including altered microclimates that disrupt natural cues, the selection of non-native species with unsynchronized phenology, and management practices like irrigation 48 , 49 . While prior research has established that urbanization influences plant phenology 50 , 51 , our study advances this understanding by quantitatively linking phenological timing mismatches to urban thermal risk. Rather than treating phenological shifts as isolated ecological patterns, we identify their functional consequences for temperature regulation, introducing a new framework for assessing thermal vulnerability windows based on plant–climate asynchrony. Crucially, this mismatch is not limited to SOS but spans the entire growing season. We observe persistent offsets between peak greenness and peak surface temperature, and frequent cases where EVI-based EOS occurs well before the dissipation of heat—particularly in hot-arid environments. Such temporal asynchronies expose urban surfaces to extended periods of high temperatures in the absence of active vegetation, thereby weakening the effectiveness of urban greening interventions 4 , 21 . Our transect-based analysis reveals that these lags are not randomly distributed but spatially structured, often intensifying toward urban fringes. These findings challenge the conventional assumption that increasing vegetative cover automatically enhances thermal resilience and instead highlight that the timing of vegetation activity is as critical as its quantity 35 , 52 . Moreover, in cities with high levels of artificial irrigation or engineered landscapes, we find phenological cues may be decoupled from climatic forcing altogether. In such cases, vegetation patterns visible in satellite-based EVI may reflect human management rather than physiological adaptation, leading to overestimated assessment of functional cooling 53 . This underscores the limitations of relying on static greenness metrics as proxies for climate-responsive behavior. Building on our findings, we propose that effective urban greening strategies must move towards a new approach: incorporating both trait-based and temporally adaptive design by selecting species with high transpiration capacity and well-timed canopy development 20 , 54 . Only by synchronizing vegetation activation with periods of thermal demand can cities enhance the true cooling potential of green infrastructure in the face of accelerating climate extremes. Finally, our results address the broader challenge of designing climatically responsive and phenologically synchronized greening interventions in an era of accelerating warming and urbanization. Consistent with recent calls for integrating phenology into adaptive urban planning 46 , 53 , our study offers a transferable diagnostic approach for aligning ecological function with climatic need—bridging the gap between satellite-detected signals (pixels) and policy-relevant urban design (people). Previous efforts relying solely on NDVI have struggled to capture vegetation’s actual capacity to regulate urban temperatures 34 , underscoring the limitations of static greenness proxies. By introducing seasonally dynamic and ecophysiologically meaningful phenological metrics, our framework helps overcome this gap and advances towards theoretically informed remote sensing applications in cities. The variability and occasional ineffectiveness of vegetative cooling—especially in compact, arid, or heavily anthropogenic zones—underscore the need to integrate engineered solutions alongside ecological strategies. Cooling materials such as reflective surfaces, high-albedo roofs, and permeable pavements can provide complementary thermal benefits where vegetation alone is insufficient 55 , 56 4. Limitations and future perspectives This study offers a globally consistent and spatially explicit perspective on how phenology modulates urban summer LST, but several limitations remain. First, the static observational design limits inference on causality and feedback. The bidirectional dynamics between LST and vegetation, i.e., how heat stress delays greening or how vegetation alters microclimates, remain underexplored. Causal-inference frameworks and physiological models of the response of plant stress 57 , 58 could help uncover these interactions. Second, our model framework, despite its interpretability and nonlinear capability, does not explicitly account for spatial dependencies inherent in urban systems. Relationships between vegetation and temperature are often modulated by microclimatic factors such as shading, irrigation, and soil variability. Incorporating spatial regression or spatially aware deep-learning architectures may improve spatial resolution and context sensitivity (such as how the temperature of a vegetated pixel is influenced by its surrounding built-up land) 52 , 59 – 61 . The phenological parameters also identified the temporal dynamics of greenness but omitted other key biophysical traits such as leaf area index (LAI), canopy height, albedo, or the efficiency of evapotranspiration, which are crucial for explaining cooling mechanisms. Future efforts may benefit from integrating LiDAR, hyperspectral, or flux-tower data to identify the functional complexity of vegetation. The use of moderate-resolution satellite products (e.g. MODIS) offers a practical and robust framework for identifying relationships between vegetation and temperature at the city scale 48 . Crucially, such data not only provide spatial smoothing of greenness and thermal signals but also help mitigate the influence of fine-scale heterogeneity—such as individual tree placement or small built structures—that can distort assessments of vegetation cooling at larger scales. Rather than attributing cooling effects to isolated landscape elements, this approach emphasizes the collective thermal regulation exerted by vegetation within defined spatial units, aligning with the study’s objective of identifying how much phenological capacity is required within a given neighborhood to achieve measurable cooling. By focusing on aggregated effects, moderate-resolution imagery supports scalable and interpretable assessments across structurally diverse urban environments 62 – 64 . High-resolution satellite imagery holds promise for advancing intra-urban analyses by resolving finer spatial variations in vegetation structure and LST 65 , 66 . Such data could help characterize localized edge effects, small green patch performance, or spatial discontinuities missed at coarser scales. These benefits, however, come with trade-offs, including increased computational demands, greater sensitivity to noise, and reduced generalizability 67 . Moving forward, a hybrid, multi-resolution framework that combines the spatial richness of high-resolution inputs with the statistical robustness and transferability of coarser products may offer a more effective path for advancing phenology-based urban climate research 61 , 68 . Complementary integration with canopy-level measurements and numerical modeling could further refine assessments by capturing three-dimensional structure, radiative properties, and physiological processes beyond what satellites alone can observe. 5. Methods 5.1 Study area and data We targeted 24 global megacities that together span the full gamut of urban conditions: arid, tropical, temperate and cold climates; coastal, inland and high-density settlement patterns; and a spectrum of vegetation configurations from heavily fragmented street trees to extensive park systems 69 . Each city’s 40 km radius buffer (centred on its geographic centroid) captures both dense urban cores and peri-urban fringes. We chose this set to ensure that our analysis reflects the diversity of thermal, hydrological and morphological contexts in which cities—and their greening strategies—operate worldwide. City locations and climatic information are shown in Fig. S1 in Supplementary Information Section 2 . Details of the criteria for selecting cities, a metadata table, and the spatial layout are provided in Table S1 in Supplementary Information Section 1 . We acquired five years (2017–2021) of remotely sensed data for vegetation and temperature for each city. Specifically, we used the MODIS enhanced vegetation index (MOD13A2, 1-km resolution, 16-d composite) and MODIS land-surface temperature (MOD11A2, 1 km, 8-d composite). To represent the synergistic effects of various vegetation parameters with urban microclimatic and topographical backgrounds on urban thermal regulation, we used nighttime light intensity (Visible Infrared Imaging Radiometer Suite (VIIRS), as a proxy for urban activity and impervious surface coverage), elevation and slope from the Shuttle Radar Topography Mission (SRTM) Digital Elevation Model (DEM), and distance to water bodies derived from the JRC Global Surface Water data set. All raster layers were resampled, temporally aligned, and projected onto a common grid for integration and modelling. All remotely sensed data sets were harmonized at a resolution of 1 km and pre-processed to remove non-vegetated pixels and outliers. See Section 1 of Supplementary Information for a full description of data filtering, temporal compositing, and protocols of spatial projection. 5.2 Data processing and feature extraction We constructed a standardized, multisource data set combining metrics of phenology, Land Surface Temperature (LST), and auxiliary environmental covariates to support the analysis of the interactions between vegetation and temperature across the 24 cities. Vegetation dynamics were derived from the MODIS enhanced vegetation index (EVI) from 2017 to 2021. We extracted key phenological parameters for each pixel by fitting smoothed EVI time series to a double logistic function. We calculated the maximum EVI (MEV), the seasonal EVI accumulation (SEA), which represents total vegetation productivity, the greening rate (i.e., the slope of EVI increase from Start of the Growing Season (SOS) to MEV), and the senescence rate (i.e., the slope of EVI decline from MEV to End of the Growing Season (EOS)), EVI amplitude (EA) (MEV, minimum EVI), and growing-season length (GSL) ) (EOS - SOS). These parameters were averaged over five years to ensure stability, and outliers were excluded. Before fitting, EVI values flagged as “fill” or outside [0,1] were masked. We required at least 80% valid observations per pixel; any pixel with > 20% missing composites was excluded. After parameter extraction, we applied a sigma-clipping step, removing metric values beyond three standard deviations of the city-level distribution, and used the interquartile range (IQR) method to discard any remaining extreme outliers (values more than 1.5 × IQR above the third or below the first quartile). All phenological metrics were then averaged over five years to ensure stability. An illustration of all phenological indices and their ecological interpretation is shown in Fig. S2 in Section 2 of Supplementary Information, with description of full calculation procedures. Daytime LST was derived from MOD11A2 and limited to each city’s climatological summer months (e.g., June to August in temperate zones). Raw brightness temperatures were converted to Celsius, filtered to remove cloud-contaminated retrievals, and bilinearly resampled to 250 m to align with the phenology grid; values outside a plausible range (− 10°C to 60°C) or more than three standard deviations from the seasonal mean were clipped. We also incorporated VIIRS nighttime light intensity (2021 monthly composites) as a proxy for urban activity, SRTM elevation and slope, and permanent water occurrence from the JRC Global Surface Water dataset. Each layer was resampled to 250 m, aligned to the same grid, and underwent a similar outlier screening (sigma-clipping and IQR filtering) to ensure consistent, high-quality inputs for our machine-learning and cross-sectional analyses. 5.3 Machine learning modelling and analysis To capture the complex, nonlinear interactions between vegetation phenology, environmental covariates, and summer LST, we selected CatBoost Regressor—a gradient-boosting decision-tree algorithm that excels in handling heterogeneous feature types, robustly manages missing values, and natively implements ordered boosting to reduce overfitting. Compared to alternative methods such as Random Forest or eXtreme Gradient Boosting, CatBoost often delivers superior accuracy and faster convergence on tabular data, especially when features exhibit varying distributions and mutual dependencies 70 , 71 . The explanatory variables included a broad range of phenological parameters such as SOS, EOS, GSL, peak EVI time, slopes of greening and senescence, MSV, EA, SEA, and auxiliary environmental factors, including nighttime light intensity (representing anthropogenic activity), elevation, and presence of water bodies. We split each city’s pixel-level dataset into training (80%) and testing (20%) subsets with a fixed random seed (42) to ensure reproducibility and evaluated model performance via R² and RMSE on the held-out test set. Hyperparameters (500 iterations, learning rate = 0.1, tree depth = 8, GPU acceleration) were selected based on grid searches optimizing test-set RMSE. To avoid confounding by purely natural vegetation far from human influence, we masked out all pixels with VIIRS nighttime radiance below a conservative threshold—thereby focusing training on urban and peri-urban contexts where anthropogenic heat and managed greenspaces co-occur. Model performance was evaluated using R², with high predictive accuracy across cities, which confirmed that the selected features identified substantial spatial variability in summer LST. This strong model foundation enabled an analysis of interpretability using SHAP-based approaches. We used multiple SHAP (SHapley Additive exPlanations) tools to identify how features associated with vegetation influenced surface temperature and to explore potential nonlinearities and interactive effects. (1) Ranking of importance of global features: We used the calculation of native SHAP in CatBoost to calculate the average contribution of each feature to the model predictions. This procedure allowed us to identify dominant drivers of urban summer temperature and to assess the relative importance of phenological variables vs environmental factors such as elevation and urban intensity. (2) Temperature Sensitivity Analysis of Vegetation Phenology Parameters: We performed a perturbation-based sensitivity analysis to assess how phenological traits modulate urban LST under differing thermal regimes. First, training pixels were divided into high-temperature (top 20% LST) and moderate-temperature (60–80% LST) subsets. Within each subset, we incrementally adjusted each vegetation parameter by Δ = ±0.3 (in its native units) and recorded the resulting change in the model’s predicted LST. By comparing ΔLST curves for high- versus moderate-temperature samples, we identified where a trait continued to impart cooling (ΔLST 0), and whether any saturation or reversal points existed. This approach highlights both positive and negative associations between phenological metrics and surface temperature (Detailed explanations of methods and interpretations of charts can be found in Section 3 of Supplementary Information). (3) SHAP Dependence Plots (with Interaction Effects): We generated SHAP dependence plots to visualize how the relationship between a vegetation metric and LST is affected by a second variable. In these plots, a primary feature's value is on the x-axis, its impact on LST is on the y-axis, and the points are colored by the value of an interacting feature (e.g., elevation). This method directly reveals context-dependent effects 72 . 5.4 Cross-sectional node analysis To explore the mechanisms of spatial responses between phenology and urban thermal environments, we conducted a cross-sectional analysis by extracting pixel-level phenological and thermal timing across both horizontal and vertical transects that intersected the urban centres. These transects spanned the entire study areas, from the city cores to the peripheries, allowing us to identify directionally comprehensive gradients of ecological and thermal processes. We first derived a central north-south and east-west line for each city that passed through the image centre and then extracted EVI and LST time series for all pixels along these lines. The EVI time series were aggregated across five years (2017–2021), smoothed using a double logistic function, and used to calculate the timing of key phenological events (SOS, peak of season (Peak), and EOS) based on the characteristics of the derivative and amplitude of the fitted curve. For LST, we converted the original values to actual temperatures, applied Gaussian smoothing, and similarly determined SOS, Peak, and EOS based on the gradients and thermal peaks for each pixel. All extracted timing variables were converted to day of year (DOY) and plotted against their corresponding distances from the city centres (in kilometres), with positive and negative distances indicating positions in opposite directions from the cores. A detailed illustration of this method is provided in Fig. S4 in Section 4 of Supplementary Information. We fitted second-order polynomial curves to the spatial profiles of SOS, Peak, and EOS for both EVI and LST along both transects to enhance interpretability. The final visualisation consisted of six smoothed curves overlaid on a single plot: three curves representing phenological timing and three curves representing thermal timing. Raw timing values are displayed as transparent scatter points, and polynomial trend lines emphasize the continuous variation in each metric across space. We also marked the centre point (distance = 0 km) for reference and converted pixel distances into physical distances (in kilometres) based on image resolution. This dual-transect approach allowed us to analyze direction-specific patterns of thermal–phenological coupling and to identify whether key events were delayed, advanced, or distorted in different location. This method ultimately offered a spatially resolved, phenologically aware representation of urban thermal behavior and provided a mechanistic understanding of how vegetation growth and surface heat interacted along urban–rural gradients on both axes. Declarations Data Availability Statement All remote sensing datasets used in this study are publicly available. Enhanced Vegetation Index (EVI) and land surface temperature (LST) data were obtained from the Moderate Resolution Imaging Spectroradiometer (MODIS) products (MOD13A2 and MOD11A2) provided by the National Aeronautics and Space Administration’s (NASA) Land Processes Distributed Active Archive Center (LP DAAC) ( https://lpdaac.usgs.gov/ ). Nighttime light data were sourced from the Visible Infrared Imaging Radiometer Suite Day/Night Band (VIIRS-DNB) product (VNP46A2), while elevation data were retrieved from the Shuttle Radar Topography Mission (SRTM) 30 m Digital Elevation Model (DEM) via the United States Geological Survey (USGS) EarthExplorer. All data was first obtained and processed using the Google Earth Engine platform. Derived variables and processed results can be made available upon request. Code availability The Python code used to generate the figures and analyses of this manuscript is available at GitHub: https://github.com/baoling123/Vegetation-Phenology-.git References Du H et al (2025) Exacerbated heat stress induced by urban browning in the Global South. Nat Cities 2:157–169 Huang WTK et al (2023) Economic valuation of temperature-related mortality attributed to urban heat islands in European cities. Nat Commun 14:1–12 Yang J et al (2023) Simulating urban expansion using cellular automata model with spatiotemporally explicit representation of urban demand. Landsc Urban Plann 231:104640 Esperon-Rodriguez M et al (2022) Climate change increases global risk to urban forests. Nat Clim Chang 12:950–955 Jia W et al (2021) Urbanization imprint on land surface phenology: The urban–rural gradient analysis for Chinese cities. Glob Change Biol 27:2895–2904 Jia S, Weng Q, Yoo C, Xiao H, Zhong Q (2024) Building energy savings by green roofs and cool roofs in current and future climates. npj Urban Sustain 4:1–13 Ren Z et al (2024) The cooling capacity of urban vegetation and its driving force under extreme hot weather: A comparative study between dry-hot and humid-hot cities. Build Environ 263:111901 Shiflett SA et al (2017) Variation in the urban vegetation, surface temperature, air temperature nexus. Sci Total Environ 579:495–505 Li Y et al (2024) Green spaces provide substantial but unequal urban cooling globally. Nat Commun 15:1–13 Wu B et al (2024) Mitigation of urban heat island in China (2000–2020) through vegetation-induced cooling. Sustainable Cities Soc 112:105599 Zhao J et al (2025) Investigating the quantitative impact of the vegetation indices on the urban thermal comfort based on machine learning: A case study of the Qinhuai River Basin, China. Sustainable Cities Soc 125:106357 Teskey R et al (2015) Responses of tree species to heat waves and extreme heat events. Plant Cell Environ 38:1699–1712 Zhao L et al (2018) Interactions between urban heat islands and heat waves. Environ Res Lett 13:034003 Zhou D et al (2016) Spatiotemporal trends of urban heat island effect along the urban development intensity gradient in China. Sci Total Environ 544:617–626 Anderegg WRL, Kane JM, Anderegg L (2013) D. L. Consequences of widespread tree mortality triggered by drought and temperature stress. Nat Clim Change 3:30–36 Berg A et al (2016) Land–atmosphere feedbacks amplify aridity increase over land under global warming. Nat Clim Change 6:869–874 Seneviratne SI et al (2010) Investigating soil moisture–climate interactions in a changing climate: A review. Earth Sci Rev 99:125–161 Zhao M, Running SW (2010) Drought-Induced Reduction in Global Terrestrial Net Primary Production from 2000 Through 2009. Science 329:940–943 Wu Z et al (2025) Tree species composition governs urban phenological responses to warming. Nat Commun 16:1–11 Massaro E et al (2023) Spatially-optimized urban greening for reduction of population exposure to land surface temperature extremes. Nat Commun 14:1–10 Kim J, Sohn S, Wang Z, Kim Y (2024) Nonuniform response of vegetation phenology to daytime and nighttime warming in urban areas. Commun Earth Environ 5:1–7 Yin Y, He L, Wennberg PO, Frankenberg C (2023) Unequal exposure to heatwaves in Los Angeles: Impact of uneven green spaces. Sci Adv 9:eade8501 Wahid A, Gelani S, Ashraf M, Foolad MR (2007) Heat tolerance in plants: An overview. Environ Exp Bot 61:199–223 Shepherd M (2022) The Curious Relationship Between COVID-19 Lockdowns and Urban Heat Islands. Geophysical Research Letters 49, e2022GL098198 Li L, Sun S, Zhong L, Han J, Qian X (2025) Novel spatiotemporal nonlinear regression approach for unveiling the impact of urban spatial morphology on carbon emissions. Sustainable Cities Soc 125:106381 Schwaab J et al (2021) The role of urban trees in reducing land surface temperatures in European cities. Nat Commun 12:6763 Grimm NB et al (2008) Global Change and the Ecology of Cities. Science 319:756–760 Zhang L et al (2022) Direct and indirect impacts of urbanization on vegetation growth across the world’s cities. Sci Adv 8:eabo0095 Debbage N, Shepherd JM (2015) The urban heat island effect and city contiguity. Comput Environ Urban Syst 54:181–194 Back Y et al (2024) Current Interventions Are Inadequate to Maintain Cities’ Resilience During Concurrent Drought and Excessive Heat. Earth’s Future 13, eEF005208 (2025) Mateo F et al (2013) Machine learning methods to forecast temperature in buildings. Expert Syst Appl 40:1061–1068 Richards D, Fung T, Belcher R, Edwards P (2020) Differential air temperature cooling performance of urban vegetation types in the tropics. Urban Forestry Urban Green 50:126651 Zhu L et al (2024) Street trees: The contribution of latent heat flux to cooling dense urban areas. Urban Clim 58:102147 Zipper SC et al (2016) Urban heat island impacts on plant phenology: intra-urban variability and response to land cover. Environ Res Lett 11:054023 Manoli G et al (2019) Magnitude of urban heat islands largely explained by climate and population. Nature 573:55–60 Thornley JHM, Cannell MGR (2000) Modelling the Components of Plant Respiration: Representation and Realism. Ann Botany 85:55–67 Sun L et al (2020) Evaluation of seasonal patterns of hydraulic redistribution in a humid subtropical area, East China. Hydrol Process 34:1052–1062 Overdieck D (2016) Water Use Efficiency and Stomatal Conductance. In: Overdieck D (ed) CO2, Temperature, and Trees: Experimental Approaches. Springer, Singapore, pp 57–64. doi: 10.1007/978-981-10-1860-2_5 . Crawford B, Kelsey K, Ibsen P, Rees A, Charobee A (2024) Intra-urban variations in land surface phenology in a semi-arid environment. Environ Res Lett 20:014036 Ibsen PC et al (2024) Urban tree cover provides consistent mitigation of extreme heat in arid but not humid cities. Sustainable Cities Soc 113:105677 Prevéy J et al (2017) Greater temperature sensitivity of plant phenology at colder sites: implications for convergence across northern latitudes. Glob Change Biol 23:2660–2671 Pocheville A (2015) The Ecological Niche: History and Recent Controversies. In: Heams T, Huneman P, Lecointre G, Silberstein M (eds) Handbook of Evolutionary Thinking in the Sciences. Springer Netherlands, Dordrecht, pp 547–586. doi: 10.1007/978-94-017-9014-7_26 . Moyer AN, Hawkins TW (2017) River effects on the heat island of a small urban area. Urban Clim 21:262–277 Seneviratne SI et al (2010) Investigating soil moisture–climate interactions in a changing climate: A review. Earth Sci Rev 99:125–161 Meng L et al (2020) Urban warming advances spring phenology but reduces the response of phenology to temperature in the conterminous United States. Proceedings of the National Academy of Sciences 117, 4228–4233 Fu YH et al (2015) Declining global warming effects on the phenology of spring leaf unfolding. Nature 526:104–107 Zipper SC et al (2016) Urban heat island impacts on plant phenology: intra-urban variability and response to land cover. Environ Res Lett 11:054023 Kabano P, Lindley S, Harris A (2021) Evidence of urban heat island impacts on the vegetation growing season length in a tropical city. Landsc Urban Plann 206:103989 Meng L et al (2020) Urban warming advances spring phenology but reduces the response of phenology to temperature in the conterminous United States. Proceedings of the National Academy of Sciences 117, 4228–4233 Vitasse Y, Signarbieux C, Fu YH (2018) Global warming leads to more uniform spring phenology across elevations. Proceedings of the National Academy of Sciences 115, 1004–1008 Wang Y et al (2024) Thermal, water, and land cover factors led to contrasting urban and rural vegetation resilience to extreme hot months. PNAS Nexus 3:pgae147 Wang J, Zhou W, Pickett STA, Qian Y (2024) A scaling law for predicting urban trees canopy cooling efficiency. Proceedings of the National Academy of Sciences 121, e2401210121 Zhao S, Liu S, Zhou D (2016) Prevalent vegetation growth enhancement in urban environment. Proceedings of the National Academy of Sciences 113, 6313–6318 Wang S et al (2019) Urban – rural gradients reveal joint control of elevated CO2 and temperature on extended photosynthetic seasons. Nat Ecol Evol 3:1076–1085 Gao K, Feng J, Santamouris M (2024) Are grand tree planting initiatives meeting expectations in mitigating urban overheating during heat waves? Sustainable Cities Soc 113:105671 Li H et al (2024) Cooling efficacy of trees across cities is determined by background climate, urban morphology, and tree trait. Commun Earth Environ 5:754 Kelloway EK (1995) Structural equation modelling in perspective. J Organizational Behav 16:215–224 Shojaie A, Fox EB (2022) Granger Causality: A Review and Recent Advances. Annual Rev Stat Its Application 9:289–319 Anselin L (2002) Under the hood Issues in the specification and interpretation of spatial regression models. Agric Econ 27:247–267 Marcińczak S, Iglesias-Pascual R, Kopeć D, Wróbel K, Mooses V (2025) Landscapes of thermal inequality: Exploring patterns of climate justice across multiple spatial scales in Spain. Landsc Urban Plann 254:105255 Tian T et al (2025) Assessing the role of spatial aggregation schemes with varying campaign durations of mobile measurements on land use regression models for estimating nitrogen dioxide. Environ Pollut 368:125689 Gong P et al (2020) Annual maps of global artificial impervious area (GAIA) between 1985 and 2018. Remote Sens Environ 236:111510 Lu Y, Wu J, Liu M (2025) Decoding the cooling potential of urban green spaces: A cross-city investigation of driving factors in 311 Chinese cities under varying climate zones. Sustainable Cities Soc 126:106410 Wu S, Yu W, An J, Lin C, Chen B (2023) Remote sensing of urban greenspace exposure and equality: Scaling effects from greenspace and population mapping. Urban Forestry Urban Green 90:128136 Weng Q (2009) Thermal infrared remote sensing for urban climate and environmental studies: Methods, applications, and trends. ISPRS J Photogrammetry Remote Sens 64:335–344 Weng Q, Fu P, Gao F (2014) Generating daily land surface temperature at Landsat resolution by fusing Landsat and MODIS data. Remote Sens Environ 145:55–67 Li H et al (2024) Cooling efficacy of trees across cities is determined by background climate, urban morphology, and tree trait. Commun Earth Environ 5:1–14 Gong Z, Ge W, Guo J, Liu J (2024) Satellite remote sensing of vegetation phenology: Progress, challenges, and opportunities. ISPRS J Photogrammetry Remote Sens 217:149–164 Gui B, Bhardwaj A, Sam L (2025) Comparative analysis of different machine learning algorithms for urban footprint extraction in diverse urban contexts using high-resolution remote sensing imagery. J Geogr Sci 35:664–696 Hsu S-C, Sharma AK, Tanone R, Ye Y-T (2024) Predicting Rainfall Using Random Forest and CatBoost Models. 10.11159/icgre24.146 Li Y et al (2024) Development and Evaluation of Machine Learning Models for Air-to-Land Temperature Conversion Using the Newly Established Kunlun Mountain Gradient Observation System. Land 13:1903 Kim Y, Kim Y (2022) Explainable heat-related mortality with random forest and SHapley Additive exPlanations (SHAP) models. Sustainable Cities Soc 79:103677 Additional Declarations There is NO Competing Interest. Supplementary Files SupplementaryInformation.docx Phenological Metrics in the Mitigation of Urban Heat: Timing and Thresholds Cite Share Download PDF Status: Under Review Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6948672","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":503604828,"identity":"a92d6146-11f3-41e4-a98f-27db574f8e60","order_by":0,"name":"Qihao Weng","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABAElEQVRIiWNgGAWjYDACZh446wADQwGYZcDA2JBAjBa2BJBiIrQwwLXwGBCnhe8478EPHxgOy5nzr/n24IPB4cQG9uZtEow70nBqkTzMlyw5g+GwseWMt9sNZ4C08Bwrk2A8k4NTi8FhHgNpHobbiRtunN0mzQPSIpFjJsHYVoFPi/FvoJb6DTfOPINokX9DUIsZyJYEg/M9bFBbeEBacDtMEqjFcobBf8MNN9jMgX5JN27jSSu2SGzD7X2+82eMb3yoSJM3OH/42YMPFday/eyHN9742JaMUwvDAbDzgFgigQ1INjOASIYE3BqgWkCA/wBIcR0+taNgFIyCUTBCAQBTrVVQz4qFVQAAAABJRU5ErkJggg==","orcid":"https://orcid.org/0000-0002-2498-0934","institution":"The Hong Kong Polytechnic University","correspondingAuthor":true,"prefix":"","firstName":"Qihao","middleName":"","lastName":"Weng","suffix":""},{"id":503604829,"identity":"276fd11a-6348-4ed9-90a4-d715242fc863","order_by":1,"name":"Baoling Gui","email":"","orcid":"","institution":"University of Aberdeen","correspondingAuthor":false,"prefix":"","firstName":"Baoling","middleName":"","lastName":"Gui","suffix":""},{"id":503604830,"identity":"33a6a44c-5250-4770-87c4-68f0c5dba5c1","order_by":2,"name":"Josep Peñuelas","email":"","orcid":"https://orcid.org/0000-0002-7215-0150","institution":"Global Ecology Unit CREAF-CSIC-UAB","correspondingAuthor":false,"prefix":"","firstName":"Josep","middleName":"","lastName":"Peñuelas","suffix":""},{"id":503604831,"identity":"f2ac6603-9e34-453c-89aa-7eefdd3a1914","order_by":3,"name":"James Voogt","email":"","orcid":"https://orcid.org/0000-0002-7975-0052","institution":"Western University Canada","correspondingAuthor":false,"prefix":"","firstName":"James","middleName":"","lastName":"Voogt","suffix":""},{"id":503604832,"identity":"78daa3a6-42fc-495e-97b4-71721505db81","order_by":4,"name":"Anshuman Bhardwaj","email":"","orcid":"","institution":"University of Aberdeen","correspondingAuthor":false,"prefix":"","firstName":"Anshuman","middleName":"","lastName":"Bhardwaj","suffix":""},{"id":503604833,"identity":"54f516a2-7bf2-47f0-b406-e7b56ccda440","order_by":5,"name":"Lydia Sam","email":"","orcid":"","institution":"University of Aberdeen","correspondingAuthor":false,"prefix":"","firstName":"Lydia","middleName":"","lastName":"Sam","suffix":""},{"id":503604834,"identity":"054eaa4d-2142-49b7-b713-65c311de6e28","order_by":6,"name":"Xiaobo Yin","email":"","orcid":"","institution":"The University of Hong Kong","correspondingAuthor":false,"prefix":"","firstName":"Xiaobo","middleName":"","lastName":"Yin","suffix":""},{"id":503604835,"identity":"ae29d69e-f2bf-4833-9fe2-aff2b7b068c2","order_by":7,"name":"J. Shepherd","email":"","orcid":"","institution":"University of Georgia","correspondingAuthor":false,"prefix":"","firstName":"J.","middleName":"","lastName":"Shepherd","suffix":""},{"id":503604836,"identity":"da58b2eb-2596-46d3-8bf9-b379ea707239","order_by":8,"name":"Dev Niyogi","email":"","orcid":"","institution":"Purdue University","correspondingAuthor":false,"prefix":"","firstName":"Dev","middleName":"","lastName":"Niyogi","suffix":""}],"badges":[],"createdAt":"2025-06-22 09:20:24","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6948672/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6948672/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":89627903,"identity":"029c33ce-d404-482e-956c-cecbfe7ca50d","added_by":"auto","created_at":"2025-08-22 06:16:52","extension":"jpeg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":664362,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eRelative importance of vegetation phenology and environmental drivers in explaining urban summer land surface temperature (LST) across 24 global cities\u003c/strong\u003e. Each cell represents the normalized SHAP value of a given feature within a specific city, indicating the relative contribution of that factor to LST variation. Values have been normalized per city to highlight within-city feature rankings. A blue gradient reflects the relative strength of importance, with darker shades indicating stronger influence. The R² value displayed in the right of panel quantifies model performance, reflecting how well the predictors explain temperature variations in that city. The 12 features evaluated in the model include three environmental drivers—elevation, nighttime light, and distance to water—and nine vegetation phenology metrics derived from the Enhanced Vegetation Index (EVI). The phenological metrics based on magnitude include Maximum EVI (MEV), Seasonal EVI Accumulation (SEA) representing total productivity, and EVI amplitude (EA). Metrics based on timing include the Start of the Season (SOS), the End of the Season (EOS), the Growing Season Length (GSL), and the timing of peak EVI (Peak time). Finally, the rates of change, greening rate and senescence rate, were also included.\u003c/p\u003e","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-6948672/v1/6057ee009a0302c202cf4cdc.jpeg"},{"id":89628852,"identity":"97ab79f2-03d3-46c3-a08a-60fd654e1fcd","added_by":"auto","created_at":"2025-08-22 06:24:52","extension":"jpeg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":991492,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eTemperature sensitivity of vegetation phenology parameters under moderate and high summer land surface temperature (LST) conditions across six cities\u003c/strong\u003e. The results are represented by panels a to c, which correspond to the cities of Riyadh, Nairobi, Lagos, Tehran, Guangzhou, Sydney and Cairo. Panels each show one city’s perturbation‐based sensitivity curves for the three key phenology metrics (MEV, SEA, and EA). On the horizontal axis, Δ ∈ [–0.3, +0.3] represents incremental increases or decreases in the target vegetation variable; on the vertical axis, ∆ Prediction LST is the change in model-predicted summer LST resulting from that perturbation. Blue curves correspond to the moderate-temperature subset and red curves to the high-temperature subset. A downward slope (∆ LST \u0026lt; 0) indicates a cooling effect; the steeper the drop when Δ \u0026gt; 0, the stronger the cooling as greenness increases. Plateaus or inflection points near Δ ≈0 signal thresholds beyond which additional vegetation gains produce diminishing or no further cooling. A vertical offset between blue and red curves shows how the same perturbation yields different cooling magnitudes under moderate versus extreme heat (larger offsets imply greater sensitivity loss at high LST). Detailed information for each of the other 24 cities is provided in Section 5 of the Supplementary Information.\u003c/p\u003e","description":"","filename":"floatimage2.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-6948672/v1/5d1b7c8f527b78e71b330d3b.jpeg"},{"id":89627904,"identity":"276d6d73-3c95-4e0a-84d4-b44bea909432","added_by":"auto","created_at":"2025-08-22 06:16:52","extension":"jpeg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":490314,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSensitivity analysis diagram of two cities (left) compared with partial dependency diagram (right)\u003c/strong\u003e. a–b, Combined visualization of SHapley Additive exPlanations (SHAP) SHAP-based marginal sensitivity (left) and partial dependence plots (PDPs; right) for the two representative cities: Beijing (a), and Jakarta (b). On the left of each panel, SHAP values indicate the direction and magnitude of the marginal impact of vegetation variables—such as maximum greenness (MEV) and enhanced amplitude (EA)—on LST under different temperature conditions. Positive values reflect warming effects, while negative values represent cooling. Stratification by thermal regime (moderate vs. high temperature) reveals the context-dependent attenuation of vegetation cooling under heat stress. On the right of each panel, PDPs illustrate the nonlinear and conditional interactions between vegetation variables and key environmental factors.\u003c/p\u003e","description":"","filename":"floatimage3.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-6948672/v1/2ccbbc15e8420d6f3860278a.jpeg"},{"id":89627905,"identity":"68d818a9-1273-40d0-aa41-e13a4b1bd096","added_by":"auto","created_at":"2025-08-22 06:16:52","extension":"jpeg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":1218956,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCross-sectional temporal profiles of EVI phenology and LST across urban–rural gradients\u003c/strong\u003e. The results are represented by a to d, which correspond to the cities of Mexico City, Paris, Cairo, and Beijing, respectively. The horizontal axis represents the distance from the city centre, with the city centre as its origin. The vertical axis represents the time at three nodes. North-South (each subgraph at the top) and East-West (each subgraph at the bottom) transects depict the spatial distribution of SOS, peak time, and EOS for both EVI and LST across representative cities, highlighting urban–rural timing disparities and vegetation–temperature misalignment. The second-order nonlinear fitting is provided for reference only; accurate analysis should be based on the nodes. Through this figure, we can reveal the lag between LST and vegetation phenology nodes in different cities, while also reflecting the urban-rural differences in the arrival times of different nodes based on distance. Detailed information for each of the remaining 24 cities is provided in Section 5 of the Supplementary Information.\u003c/p\u003e","description":"","filename":"floatimage4.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-6948672/v1/ff08fc8f2c179b2e6f4dfdc1.jpeg"},{"id":103506616,"identity":"a77211d1-3730-49fc-8584-0ad311d72eb7","added_by":"auto","created_at":"2026-02-26 13:38:06","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":4247725,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6948672/v1/b95d8fd0-c453-4f19-a800-13751ae7a2de.pdf"},{"id":89627914,"identity":"a8afea2a-dca4-480b-9eee-620f55a56c84","added_by":"auto","created_at":"2025-08-22 06:16:54","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":100964252,"visible":true,"origin":"","legend":"Phenological Metrics in the Mitigation of Urban Heat: Timing and Thresholds","description":"","filename":"SupplementaryInformation.docx","url":"https://assets-eu.researchsquare.com/files/rs-6948672/v1/e23b2ceace92fa653c9cd0fe.docx"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"Phenological Metrics in the Mitigation of Urban Heat: Timing and Thresholds","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eRapid urbanization has intensified the urban heat island (UHI) effect, threatening public health, raising energy demand, and disrupting vegetation phenology \u003csup\u003e\u003cspan additionalcitationids=\"CR2\" citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e. During summer, in particular, excessive urban heating can exacerbate thermal stress, deteriorate air quality, and alter ecosystem functions, thereby affecting the timing and length of vegetation growth cycles \u003csup\u003e\u003cspan additionalcitationids=\"CR4\" citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e. In contrast, urban vegetation helps moderate surface heat through shading and evapotranspiration, and higher vegetation cover is generally associated with lower land surface temperature (LST), especially during summer when such cooling effects are strongest \u003csup\u003e\u003cspan additionalcitationids=\"CR7\" citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e. However, this cooling is neither spatially uniform nor linearly proportional to vegetation density. Prior studies have identified threshold effects, where substantial cooling benefits occur only once green space coverage exceeds a critical proportion of the urban landscape. Below this threshold, additional greening may yield limited or negligible thermal mitigation\u0026mdash;especially in compact, densely built urban cores\u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u003c/sup\u003e. These limitations underscore the need to better understand the conditions, magnitudes, and timings under which vegetation most effectively mitigates urban heat.\u003c/p\u003e \u003cp\u003eNumerous studies have confirmed a strong inverse relationship between vegetation cover and LST, with the cooling effect being especially pronounced in the growing season. Areas with higher fractional vegetation cover (FVC) consistently have lower summer LSTs than do sparsely vegetated zones, and the intensity of the surface UHI effect often decreases as vegetation density increases \u003csup\u003e\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e,\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e. This cooling effect, however, is not static but varies substantially across seasons, climatic zones, and vegetation functional states. The correlation between vegetation cover and LST weakens during winter or dry periods when vegetation is dormant or leafless \u003csup\u003e\u003cspan additionalcitationids=\"CR13\" citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u003c/sup\u003e. This weakening highlights the importance of phenology in determining the capacity of seasonal thermal regulation. When plants green up earlier in spring or delay senescence in autumn, the duration of active cooling periods can be extended, thereby increasing the benefits of thermal mitigation in seasons \u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e. In contrast, disruptions in phenological rhythms, such as early senescence induced by drought, can suppress transpiration and weaken the ability of vegetation to cool when needed the most during summer peaks \u003csup\u003e\u003cspan additionalcitationids=\"CR17\" citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eRecent studies have increasingly recognized the interactions between phenological dynamics and LST \u003csup\u003e\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e,\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e, however, key uncertainties remain. Most notably, existing work has focused on individual cities or homogenous climate zones \u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e,\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u003c/sup\u003e, limiting the ability to capture global variability in urban form, vegetation structure, and seasonal heat exposure. As cities worldwide confront escalating heat risks, there is an urgent need to develop a globally comparable understanding of how phenological traits modulate urban climate, especially where species-level vegetation data are scarce.\u003c/p\u003e \u003cp\u003eResearch on urban phenology has largely emphasized temporal indicators\u0026mdash;such as the start and end of season (SOS and EOS)\u0026mdash;while underappreciating amplitude-based traits like peak greenness, cumulative productivity, and the rate of vegetation growth or decline. These attributes are crucial for assessing the scale and persistence of vegetative cooling \u003csup\u003e\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e,\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e, particularly in cities where vegetation is fragmented or human-managed, and where thermal resilience depends on both timing and intensity of green activation.\u003c/p\u003e \u003cp\u003eThe biophysical pathways linking phenology and thermal regulation crucially remain poorly understood. Vegetative cooling may only occur beyond specific thresholds of greenness, saturate at high canopy densities, or lag behind seasonal increases in temperature \u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e,\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e. Responses may also vary across climates: vegetation may cool more effectively in hot and dry regions but underperform in mesic regions where evapotranspiration is less limiting. Such asymmetries challenge the assumption of the efficiency of uniform cooling and underscore the need for nonlinear analytical frameworks that detect thresholds, saturation, and time-lagged effects \u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e,\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eTo address these gaps, we investigated phenological\u0026ndash;thermal coupling across 24 global megacities using satellite-derived vegetation metrics, harmonized data for surface temperature, and interpretable machine-learning techniques. Specifically, we: (1) quantified the relative influence of both timing- and magnitude-based phenological traits on urban summer LST, (2) examined intercity variability and spatial heterogeneity in interactions between the vegetation and temperature, and (3) identify nonlinear, threshold, and lagged phenological responses under thermal stress. These insights, while based on surface-temperature proxies, provide mechanistic evidence to guide temperature-responsive, phenology-aware greening strategies for urban climate adaptation.\u003c/p\u003e"},{"header":"2. Results","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1 Magnitude and directionality of LST responses to vegetation phenological changes\u003c/h2\u003e \u003cp\u003eWe used SHapley Additive exPlanations (SHAP) values derived from Categorical Boosting (CatBoost) regression models for the 24 cities to quantify the contribution of vegetation phenology and environmental factors in determining urban LST in summer (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). R\u0026sup2; for these models varied across cities, ranging from 0.47 (Istanbul) to 0.97 (Nairobi and Tehran), indicating the diverse explanatory power of the selected features depending on the local geography and ecology.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eAn analysis of feature importance across the 24 global cities identified both universal and region-specific mechanisms driving urban summer LST. Nighttime light intensity was the most frequent dominant predictor (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e), ranking as one of the three most frequent predictors for 17 cities, underscoring the strong and consistent impact of anthropogenic activity on urban heat. Elevation followed closely, ranking as one of the three most frequent predictors for 15 cities, particularly in high-relief cities such as Mexico City, Tehran, and Nairobi, where topographic variability strongly influenced local patterns of temperature. Vegetation indicators also played a critical role: Enhanced Vegetation Index (EVI)-based metrics, three phenological variables (maximum EVI (MEV), seasonal EVI accumulation (SEA), and EVI amplitude (EA)) were the most influential in modulating LST across most cities and were among the three most frequent factors in 13, 8, and 6 cities, respectively, highlighting the importance of both overall greenness and seasonal variation in modulating urban temperatures. Proximity to water bodies ranked highly in arid and hydrometeorologically sensitive cities such as Cairo, Lima, and Riyadh, indicating the crucial cooling effect of water bodies in dry climates. Interestingly, variables of phenological timing (e.g. SOS and peak timing) were less often among the most frequent predictors, suggesting that their thermal regulatory impact may be more spatially heterogeneous and context-dependent but important in some cities (e.g. Tehran).\u003c/p\u003e \u003cp\u003eIdentifying the most important phenological variables and assessing how these variables nonlinearly affect temperature, how their influence varies under different temperatures ranges, and how they interact with other urban environmental factors are essential to fully understand the role of phenology in regulating urban summer LST. The following integrates analyses of sensitivity and partial-dependence diagnostics to explore both the magnitude and directionality of the responses of LST to changes in vegetation, highlighting critical thresholds, saturation points, and joint effects with covariates such as elevation, nighttime light, and distance to water. The findings indicated that interactions between vegetation and temperature were much more complex than linear, univariate interpretations.\u003c/p\u003e \u003cp\u003eFigure\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e shows that, across nearly every city, increasing MEV, SEA, or EA consistently yields a negative ∆LST, indicating a cooling effect regardless of whether pixels are in the moderate (60\u0026ndash;80%) or high-temperature (top 20%) subset. In every case, the high-temperature curve lies below the moderate-temperature curve, meaning that, for the same perturbation, cooling is stronger (more negative ∆LST) under extreme heat than under moderate heat. SEA invariably follows a near linear decline in both strata, with only modest offsets (typically\u0026thinsp;\u0026lt;\u0026thinsp;0.3\u0026deg;C) between moderate and high curves, demonstrating its stable cooling contribution. MEV and EA, by contrast, exhibit a clear threshold effect: for negative perturbations or small ∆ values, the response is relatively flat, and cooling is modest. As ∆ approaches zero and becomes positive, cooling accelerates rapidly, producing a steeper drop in ∆LST. Beyond ∆ \u0026asymp; 0.1\u0026ndash;0.2, both curves begin to plateau, indicating diminishing marginal cooling (e.g., Riyadh (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ea)).\u003c/p\u003e \u003cp\u003eNot every city conforms perfectly to this pattern. First, a few MEV or EA curves slope upward as ∆ increases\u0026mdash;contradicting the expected cooling trend. For instance, Nairobi\u0026rsquo;s moderate-temperature MEV rises into positive territory for ∆ \u0026gt; 0.1 (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eb), and Tehran\u0026rsquo;s moderate-temperature MEV climbs slightly for ∆ around 0 (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ec). Second, some curves cross ∆LST\u0026thinsp;=\u0026thinsp;0, indicating a switch from warming to cooling (or vice versa). For instance, Guangzhou\u0026rsquo;s moderate-temperature MEV crosses at a similar range (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ed), whereas their high-temperature curves remain negative. Third, certain moderate-temperature trajectories remain nearly flat at ∆LST\u0026thinsp;\u0026asymp;\u0026thinsp;0 across all ∆\u0026mdash;for example, Sydney\u0026rsquo;s SEA barely moves above or below zero (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ee), suggesting minimal net temperature change despite perturbations. Finally, although every city\u0026rsquo;s moderate curve lies above its high curve, the vertical gap and overall ∆LST range differ remarkably: Cairo and Nairobi often show mid\u0026ndash;high separations\u0026thinsp;≲\u0026thinsp;0.1\u0026deg;C (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ef,b), whereas Sydney can exceed 0.6\u0026ndash;1.0\u0026deg;C (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ee). These discrepancies reveal substantial heterogeneity in how extra greenness cools under moderate versus extreme heat.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e2.2 Joint effects of vegetation and urban environments on LST in summer\u003c/h2\u003e \u003cp\u003eThe preceding section highlighted divergent and occasionally counterintuitive behaviors, such as when increased vegetation indices were associated with increasing temperatures, but these phenomena require a deeper exploration of the underlying mechanisms. We used partial-dependence plots (PDPs) to distinguish between such anomalies, which enabled the visualization of marginal effects and patterns of interaction between individual predictors and LST. Unlike global analyses of sensitivity that assess average impacts, PDPs allow us to identify nonlinear transitions and conditional dependencies. They helped us to determine how vegetation variables, such as EA or maximum greenness, had either cooling or warming effects depending on co-occurring environmental factors such as urban built-up density, elevation, or distance to water. This analysis found that the vegetation did not function in isolation; its capacity for thermal regulation was instead determined, and sometimes constrained, by its interaction with co-occurring factors like elevation and urban heat intensity.\u003c/p\u003e \u003cp\u003eIn Beijing (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ea), we observed that higher MEV values tend to coincide spatially with pixels located at relatively higher elevations. While elevation is typically associated with decreasing air temperature due to the lapse rate, it is important to note that land surface temperature (LST) responds primarily to surface energy balance rather than air mass characteristics. In the urban, elevation differences are modest, but elevated areas may receive more direct solar radiation due to reduced obstruction, greater sky exposure, and lower shading from surrounding structures or vegetation. These factors can lead to increased daytime LST, even at slightly higher altitudes. As a result, the expected cooling from greater MEV in these areas is partially or fully masked by the stronger surface warming associated with topographic exposure. Consequently, the net ∆LST associated with MEV appears neutral or even slightly positive in these zones\u0026mdash;not because MEV fails to cool, but because elevation-linked radiative effects dominate the LST response.\u003c/p\u003e \u003cp\u003eIn Jakarta (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eb), a similar confounding phenomenon emerges between EA and nighttime light intensity. The PDP for EA versus nighttime light shows that the higher EA values are clustered in areas with central urban zones of more bright night light readings characterized by dense built ups and strong anthropogenic heat emissions. Within those zones, incremental increases in EA no longer translate to additional cooling; instead, the persistent waste heat from traffic, air-conditioning, and concrete surfaces dominates. As a result, when EA reaches these urban peaks, the observed ∆LST actually rises or remains flat, despite greater vegetation amplitude \u003csup\u003e\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003e2.3. Phenology\u0026ndash;temperature interplay along urban-rural gradients\u003c/h2\u003e \u003cp\u003eThis section examines the spatial coordination and temporal alignment between the phenological transitions of vegetation and the dynamics of summer LST along urban\u0026ndash;rural gradients. We analyzed both the north-south and east-west cross-sectional profiles of EVI-derived phenological markers and LST curves to determine whether vegetation and temperature responded synchronously across space and whether cities had coherent thermal-phenological coupling. Figure\u0026nbsp;4 identifies consistent urban\u0026ndash;rural differences, quantifies timing mismatches between biophysical and thermal signals, and assesses how urbanisation modifies vegetation functioning.\u003c/p\u003e \u003cp\u003eOur transect-based phenological and thermal profiles across 24 global megacities identified both consistent spatial patterns and city-specific deviations that together highlight the complex interplay between vegetative activity and urban thermal dynamics.\u003c/p\u003e \u003cp\u003eFigure\u0026nbsp;4 demonstrates systematic and city-wide temporal mismatches between vegetation phenology and surface temperature trajectories, as reflected in the timing of three key seasonal markers. Across more than 80% of cities (as shown in Section \u003cspan refid=\"Sec7\" class=\"InternalRef\"\u003e4\u003c/span\u003e of the Supplementary Information), the EVI-based SOS lags behind the LST-based onset of seasonal warming by 50 to 100 days such as Mexico City (Fig.\u0026nbsp;4a), particularly within urban cores. In contrast, the timing of phenological peak and EOS relative to LST shows more complex and divergent patterns. In some cities, such as Paris and Cario (Fig.\u0026nbsp;4b,c), EVI peak appears significantly earlier than LST peak (by up to 80\u0026ndash;100 days), while in others (e.g., Mexico City (Fig.\u0026nbsp;4a)), EVI peak lags behind, indicating delayed vegetation responses to sustained warming. For EOS, a typical pattern is that EVI-based EOS precedes thermal decline, especially in cities with strong heat retention like Paris or Cario (Fig.\u0026nbsp;4b,c), where LST remains high even after vegetation enters senescence.\u003c/p\u003e \u003cp\u003eBuilding on this temporal perspective, our transect-based spatial analysis reveals considerable heterogeneity in how these timing mismatches manifest across the urban\u0026ndash;rural gradients. In cities like Mexico City (Fig.\u0026nbsp;4a), we observed parabolic (U-shaped) spatial profiles, where vegetation SOS and peaks occur earlier in city centers and later toward the outskirts, while EOS is delayed at the core. This pattern creates an extended growing season within urban cores but does not necessarily align with temperature cycles, as LST patterns are relatively flatter across the transects. In fact, in most cities\u0026mdash;including Cario, and Beijing (Fig.\u0026nbsp;4c,d)\u0026mdash;LST shows limited spatial variability between urban and suburban areas, while phenological metrics vary by more than 60\u0026ndash;100 days. This contrast underscores that while surface heat accumulation may follow a relatively coherent seasonal trajectory, vegetation activation is far more heterogeneous and sensitive to local conditions.\u003c/p\u003e \u003c/div\u003e"},{"header":"3. Discussion","content":"\u003cp\u003eThis study provides a globally comparable, mechanistic understanding of how vegetation phenology modulates summer LST across 24 global cities spanning diverse climates, topographies, and urban forms. In contrast to earlier works that rely predominantly on static greenness metrics or categorical land cover types \u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e,\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e,\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e\u003c/sup\u003e, this study introduces a phenology-informed perspective that jointly considers the effects of both timing and intensity of vegetation activity. Our findings not only establish the relative importance of vegetation phenology in shaping urban thermal environments but also lay theoretical foundation for designing phenology-informed greening strategies that are seasonally aligned and climate responsive. Recent studies have explored greening strategies to reduce urban heat \u003csup\u003e\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e,\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e, but few have integrated the seasonality and synchronicity of vegetation response with temperature dynamics at moderate spatial resolution. Our work thus moves beyond correlational patterns and establishes a functional link between phenological dynamics and LST, echoing calls for mechanistic metrics in studies of urban sustainability \u003csup\u003e\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e,\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eMachine-learning modelling and analysis indicated that cities that had higher model performance (e.g. Nairobi, and Mexico City), as indicated by higher R\u0026sup2; values, typically possess distinct elevation gradients or pronounced spatial heterogeneity in their green space structures. These physical and ecological variations enhance the detectability of the vegetation\u0026ndash;temperature relationship by introducing stronger spatial contrasts \u003csup\u003e\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u003c/sup\u003e. For example, in Nairobi, the combination of topographic variation and clustered green space facilitates clear spatial transitions in phenology\u0026ndash;LST dynamics. In contrast, cities such as Istanbul or New York\u0026mdash;with fragmented morphologies and thermally relatively uniform environments\u0026mdash;tend to show weaker statistical associations, due likely to microclimatic noise and anthropogenic influences that obscure vegetation signals \u003csup\u003e\u003cspan additionalcitationids=\"CR31 CR32\" citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eTemperature-stratified SHAP and interaction analyses further reveals the nonlinear and threshold-dependent cooling responses of three key vegetation phenology parameters\u0026mdash;MEV, SEA, and EA\u0026mdash;under different thermal regimes across diverse global cities. Overall, increases in these parameters are associated with negative ∆LST (i.e., cooling), yet with distinct behavioral differences. Among them, SEA exhibits a remarkably linear and consistent relationship with LST across both moderate- and high-temperature conditions, with minimal offsets (generally\u0026thinsp;\u0026lt;\u0026thinsp;0.3\u0026deg;C). Unlike MEV and EA, SEA is less susceptible to nonlinear breakdowns from water stress or thermal saturation. This finding establishes SEA as a robust metric for urban greening assessment and a quantitative reference for long-term heat mitigation \u003csup\u003e\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e,\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e\u003c/sup\u003e. In contrast, MEV and EA follow a distinct activation\u0026ndash;saturation pattern: under low perturbations (∆ \u0026lt; 0), their cooling impact is negligible; after surpassing a critical threshold (∆ \u0026asymp; 0), cooling accelerates rapidly before plateauing. This nonlinearity likely reflects key ecophysiological limits: at high densities, self-shading and water competition can reduce marginal cooling, while extreme heat may trigger stomatal closure, negating the benefits of high greenness \u003csup\u003e\u003cspan additionalcitationids=\"CR37\" citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e\u003c/sup\u003e. This pattern is particularly pronounced in arid and semi-arid cities\u0026mdash;such as those in Southwest Asia, the Indian subcontinent, and the southwestern U.S.\u0026mdash;where evapotranspiration is tightly coupled with soil moisture availability \u003csup\u003e\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e,\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e\u003c/sup\u003e. These results align with ecophysiological studies \u003csup\u003e\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e\u003c/sup\u003e that identify a leaf area\u0026ndash;dependent trigger for transpiration and confirm, for the first time at a global pixel scale, that greenness-related cooling is highly contingent on local hydrological constraints. In humid or well-watered cities, such as those in Western Europe or coastal and southern hemisphere regions, this threshold response weakens, and marginal cooling ramps up more gradually demonstrating that excess greenness in such environments yields consistent, non-abrupt thermal benefits. We identify and quantify pixel-scale cooling thresholds that demarcate activation and saturation phases of MEV/EA\u0026mdash;revealing spatially explicit tipping points not previously mapped.\u003c/p\u003e \u003cp\u003eWe further ground our analysis of the phenology-LST relationship in ecological niche theory and propose new evaluation metrics \u003csup\u003e\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e\u003c/sup\u003e. We propose a set of new evaluation metrics: the Cooling Activation Threshold (CAT), defined as the minimum MEV/EA level at which net LST reduction begins; the Saturation Cooling Point (SCP), beyond which additional greening yields negligible further cooling; and the Linear Cooling Index (LCI), characterizing SEA's continuous and stable contribution to LST moderation across thermal strata. The perturbation values used in our models can be scaled back to actual EVI or amplitude units via the inverse of z-score normalization (i.e., multiplying the perturbation by the standard deviation of the original data and then adding the mean), which allows direct mapping from modeled perturbations to feasible vegetation enhancements. This not only translates our findings into tangible benchmarks for greening design, assessing whether a plant's traits fit the city's unique thermal niche but also lays the groundwork for temperature-oriented phenological benchmarks in urban greening design.\u003c/p\u003e \u003cp\u003eOur results challenge the hypothesis that increased greenness universally yields lower surface temperatures. In several contexts, especially in compact urban cores with strong anthropogenic heat, elevated terrain, or low hydrological connectivity, we observe localized warming or flat ∆LST responses to vegetation increases, particularly in densely built areas. Moreover, limited Sky View Factor (SVF) in compact urban areas may reduce the effectiveness of vegetative cooling by constraining radiative loss and air circulation, further modulating the phenological impacts on surface temperature \u003csup\u003e\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e\u003c/sup\u003e. These findings suggest that vegetation cooling is not univariate but modulated by interactions with topography, urban morphology, and human activity intensity. Such outcomes extend prior work \u003csup\u003e\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e\u003c/sup\u003e, which identified the importance of irrigation and land cover type but lacked the resolution to pinpoint thresholds or reversals. Our analysis offers new insight into where and why greening may fail or even backfire, such as when dense vegetation co-occurs with poor ventilation or high nocturnal heat emission. These mechanisms had been hinted in studies like Kim et al. \u003csup\u003e\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e, yet not explicitly measured at the pixel scale across global major cities.\u003c/p\u003e \u003cp\u003eOur findings most notably identified a pervasive temporal decoupling between vegetation activity and early-season increases in temperature across most of the cities. EVI-based SOS lagged behind the initial seasonal increase in LST by 50\u0026ndash;100 d in \u0026gt;\u0026thinsp;80% of the cities, including Beijing, New Delhi, Riyadh, and Lagos. This lag is particularly pronounced in cities located in monsoonal, arid, or equatorial climate zones, where delayed vegetation onset often fails to align with periods of peak thermal accumulation \u003csup\u003e\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e\u003c/sup\u003e. One potential explanation is that, in dry environments, LST may respond more rapidly than air temperature (Tair) due to the early-season exposure of bare soil and impervious surfaces, which heat quickly and are strongly detected by thermal sensors \u003csup\u003e\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e,\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e\u003c/sup\u003e. In contrast, plant phenology often tracks accumulated Tair and soil moisture availability, leading to temporal mismatches in observed thermal and vegetative signals \u003csup\u003e\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e\u003c/sup\u003e. This temporal mismatch often stems from distinct urban drivers, including altered microclimates that disrupt natural cues, the selection of non-native species with unsynchronized phenology, and management practices like irrigation \u003csup\u003e\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e,\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e\u003c/sup\u003e. While prior research has established that urbanization influences plant phenology \u003csup\u003e\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e,\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e\u003c/sup\u003e, our study advances this understanding by quantitatively linking phenological timing mismatches to urban thermal risk. Rather than treating phenological shifts as isolated ecological patterns, we identify their functional consequences for temperature regulation, introducing a new framework for assessing thermal vulnerability windows based on plant\u0026ndash;climate asynchrony. Crucially, this mismatch is not limited to SOS but spans the entire growing season. We observe persistent offsets between peak greenness and peak surface temperature, and frequent cases where EVI-based EOS occurs well before the dissipation of heat\u0026mdash;particularly in hot-arid environments. Such temporal asynchronies expose urban surfaces to extended periods of high temperatures in the absence of active vegetation, thereby weakening the effectiveness of urban greening interventions \u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e,\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eOur transect-based analysis reveals that these lags are not randomly distributed but spatially structured, often intensifying toward urban fringes. These findings challenge the conventional assumption that increasing vegetative cover automatically enhances thermal resilience and instead highlight that the timing of vegetation activity is as critical as its quantity \u003csup\u003e\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e,\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e\u003c/sup\u003e. Moreover, in cities with high levels of artificial irrigation or engineered landscapes, we find phenological cues may be decoupled from climatic forcing altogether. In such cases, vegetation patterns visible in satellite-based EVI may reflect human management rather than physiological adaptation, leading to overestimated assessment of functional cooling \u003csup\u003e\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e\u003c/sup\u003e. This underscores the limitations of relying on static greenness metrics as proxies for climate-responsive behavior. Building on our findings, we propose that effective urban greening strategies must move towards a new approach: incorporating both trait-based and temporally adaptive design by selecting species with high transpiration capacity and well-timed canopy development \u003csup\u003e\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e,\u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e\u003c/sup\u003e. Only by synchronizing vegetation activation with periods of thermal demand can cities enhance the true cooling potential of green infrastructure in the face of accelerating climate extremes.\u003c/p\u003e \u003cp\u003eFinally, our results address the broader challenge of designing climatically responsive and phenologically synchronized greening interventions in an era of accelerating warming and urbanization. Consistent with recent calls for integrating phenology into adaptive urban planning \u003csup\u003e\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e,\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e\u003c/sup\u003e, our study offers a transferable diagnostic approach for aligning ecological function with climatic need\u0026mdash;bridging the gap between satellite-detected signals (pixels) and policy-relevant urban design (people). Previous efforts relying solely on NDVI have struggled to capture vegetation\u0026rsquo;s actual capacity to regulate urban temperatures \u003csup\u003e\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u003c/sup\u003e, underscoring the limitations of static greenness proxies. By introducing seasonally dynamic and ecophysiologically meaningful phenological metrics, our framework helps overcome this gap and advances towards theoretically informed remote sensing applications in cities. The variability and occasional ineffectiveness of vegetative cooling\u0026mdash;especially in compact, arid, or heavily anthropogenic zones\u0026mdash;underscore the need to integrate engineered solutions alongside ecological strategies. Cooling materials such as reflective surfaces, high-albedo roofs, and permeable pavements can provide complementary thermal benefits where vegetation alone is insufficient \u003csup\u003e\u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e,\u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e56\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e"},{"header":"4. Limitations and future perspectives","content":"\u003cp\u003eThis study offers a globally consistent and spatially explicit perspective on how phenology modulates urban summer LST, but several limitations remain. First, the static observational design limits inference on causality and feedback. The bidirectional dynamics between LST and vegetation, i.e., how heat stress delays greening or how vegetation alters microclimates, remain underexplored. Causal-inference frameworks and physiological models of the response of plant stress \u003csup\u003e\u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e57\u003c/span\u003e,\u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e58\u003c/span\u003e\u003c/sup\u003e could help uncover these interactions. Second, our model framework, despite its interpretability and nonlinear capability, does not explicitly account for spatial dependencies inherent in urban systems. Relationships between vegetation and temperature are often modulated by microclimatic factors such as shading, irrigation, and soil variability. Incorporating spatial regression or spatially aware deep-learning architectures may improve spatial resolution and context sensitivity (such as how the temperature of a vegetated pixel is influenced by its surrounding built-up land) \u003csup\u003e\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e,\u003cspan additionalcitationids=\"CR60\" citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e61\u003c/span\u003e\u003c/sup\u003e. The phenological parameters also identified the temporal dynamics of greenness but omitted other key biophysical traits such as leaf area index (LAI), canopy height, albedo, or the efficiency of evapotranspiration, which are crucial for explaining cooling mechanisms. Future efforts may benefit from integrating LiDAR, hyperspectral, or flux-tower data to identify the functional complexity of vegetation.\u003c/p\u003e \u003cp\u003eThe use of moderate-resolution satellite products (e.g. MODIS) offers a practical and robust framework for identifying relationships between vegetation and temperature at the city scale \u003csup\u003e\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e\u003c/sup\u003e. Crucially, such data not only provide spatial smoothing of greenness and thermal signals but also help mitigate the influence of fine-scale heterogeneity\u0026mdash;such as individual tree placement or small built structures\u0026mdash;that can distort assessments of vegetation cooling at larger scales. Rather than attributing cooling effects to isolated landscape elements, this approach emphasizes the collective thermal regulation exerted by vegetation within defined spatial units, aligning with the study\u0026rsquo;s objective of identifying how much phenological capacity is required within a given neighborhood to achieve measurable cooling. By focusing on aggregated effects, moderate-resolution imagery supports scalable and interpretable assessments across structurally diverse urban environments \u003csup\u003e\u003cspan additionalcitationids=\"CR63\" citationid=\"CR62\" class=\"CitationRef\"\u003e62\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e64\u003c/span\u003e\u003c/sup\u003e. High-resolution satellite imagery holds promise for advancing intra-urban analyses by resolving finer spatial variations in vegetation structure and LST \u003csup\u003e\u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e65\u003c/span\u003e,\u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e66\u003c/span\u003e\u003c/sup\u003e. Such data could help characterize localized edge effects, small green patch performance, or spatial discontinuities missed at coarser scales. These benefits, however, come with trade-offs, including increased computational demands, greater sensitivity to noise, and reduced generalizability \u003csup\u003e\u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e67\u003c/span\u003e\u003c/sup\u003e. Moving forward, a hybrid, multi-resolution framework that combines the spatial richness of high-resolution inputs with the statistical robustness and transferability of coarser products may offer a more effective path for advancing phenology-based urban climate research \u003csup\u003e\u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e61\u003c/span\u003e,\u003cspan citationid=\"CR68\" class=\"CitationRef\"\u003e68\u003c/span\u003e\u003c/sup\u003e. Complementary integration with canopy-level measurements and numerical modeling could further refine assessments by capturing three-dimensional structure, radiative properties, and physiological processes beyond what satellites alone can observe.\u003c/p\u003e"},{"header":"5. Methods","content":"\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003e5.1 Study area and data\u003c/h2\u003e \u003cp\u003eWe targeted 24 global megacities that together span the full gamut of urban conditions: arid, tropical, temperate and cold climates; coastal, inland and high-density settlement patterns; and a spectrum of vegetation configurations from heavily fragmented street trees to extensive park systems \u003csup\u003e\u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e69\u003c/span\u003e\u003c/sup\u003e. Each city\u0026rsquo;s 40 km radius buffer (centred on its geographic centroid) captures both dense urban cores and peri-urban fringes. We chose this set to ensure that our analysis reflects the diversity of thermal, hydrological and morphological contexts in which cities\u0026mdash;and their greening strategies\u0026mdash;operate worldwide. City locations and climatic information are shown in Fig. \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e in Supplementary Information Section \u003cspan refid=\"Sec2\" class=\"InternalRef\"\u003e2\u003c/span\u003e. Details of the criteria for selecting cities, a metadata table, and the spatial layout are provided in Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e in Supplementary Information Section \u003cspan refid=\"Sec1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. We acquired five years (2017\u0026ndash;2021) of remotely sensed data for vegetation and temperature for each city. Specifically, we used the MODIS enhanced vegetation index (MOD13A2, 1-km resolution, 16-d composite) and MODIS land-surface temperature (MOD11A2, 1 km, 8-d composite). To represent the synergistic effects of various vegetation parameters with urban microclimatic and topographical backgrounds on urban thermal regulation, we used nighttime light intensity (Visible Infrared Imaging Radiometer Suite (VIIRS), as a proxy for urban activity and impervious surface coverage), elevation and slope from the Shuttle Radar Topography Mission (SRTM) Digital Elevation Model (DEM), and distance to water bodies derived from the JRC Global Surface Water data set. All raster layers were resampled, temporally aligned, and projected onto a common grid for integration and modelling. All remotely sensed data sets were harmonized at a resolution of 1 km and pre-processed to remove non-vegetated pixels and outliers. See Section \u003cspan refid=\"Sec1\" class=\"InternalRef\"\u003e1\u003c/span\u003e of Supplementary Information for a full description of data filtering, temporal compositing, and protocols of spatial projection.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003e5.2 Data processing and feature extraction\u003c/h2\u003e \u003cp\u003eWe constructed a standardized, multisource data set combining metrics of phenology, Land Surface Temperature (LST), and auxiliary environmental covariates to support the analysis of the interactions between vegetation and temperature across the 24 cities. Vegetation dynamics were derived from the MODIS enhanced vegetation index (EVI) from 2017 to 2021. We extracted key phenological parameters for each pixel by fitting smoothed EVI time series to a double logistic function. We calculated the maximum EVI (MEV), the seasonal EVI accumulation (SEA), which represents total vegetation productivity, the greening rate (i.e., the slope of EVI increase from Start of the Growing Season (SOS) to MEV), and the senescence rate (i.e., the slope of EVI decline from MEV to End of the Growing Season (EOS)), EVI amplitude (EA) (MEV, minimum EVI), and growing-season length (GSL) ) (EOS - SOS). These parameters were averaged over five years to ensure stability, and outliers were excluded. Before fitting, EVI values flagged as \u0026ldquo;fill\u0026rdquo; or outside [0,1] were masked. We required at least 80% valid observations per pixel; any pixel with \u0026gt;\u0026thinsp;20% missing composites was excluded. After parameter extraction, we applied a sigma-clipping step, removing metric values beyond three standard deviations of the city-level distribution, and used the interquartile range (IQR) method to discard any remaining extreme outliers (values more than 1.5 \u0026times; IQR above the third or below the first quartile). All phenological metrics were then averaged over five years to ensure stability. An illustration of all phenological indices and their ecological interpretation is shown in Fig. S2 in Section \u003cspan refid=\"Sec2\" class=\"InternalRef\"\u003e2\u003c/span\u003e of Supplementary Information, with description of full calculation procedures.\u003c/p\u003e \u003cp\u003eDaytime LST was derived from MOD11A2 and limited to each city\u0026rsquo;s climatological summer months (e.g., June to August in temperate zones). Raw brightness temperatures were converted to Celsius, filtered to remove cloud-contaminated retrievals, and bilinearly resampled to 250 m to align with the phenology grid; values outside a plausible range (\u0026minus;\u0026thinsp;10\u0026deg;C to 60\u0026deg;C) or more than three standard deviations from the seasonal mean were clipped. We also incorporated VIIRS nighttime light intensity (2021 monthly composites) as a proxy for urban activity, SRTM elevation and slope, and permanent water occurrence from the JRC Global Surface Water dataset. Each layer was resampled to 250 m, aligned to the same grid, and underwent a similar outlier screening (sigma-clipping and IQR filtering) to ensure consistent, high-quality inputs for our machine-learning and cross-sectional analyses.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003e5.3 Machine learning modelling and analysis\u003c/h2\u003e \u003cp\u003eTo capture the complex, nonlinear interactions between vegetation phenology, environmental covariates, and summer LST, we selected CatBoost Regressor\u0026mdash;a gradient-boosting decision-tree algorithm that excels in handling heterogeneous feature types, robustly manages missing values, and natively implements ordered boosting to reduce overfitting. Compared to alternative methods such as Random Forest or eXtreme Gradient Boosting, CatBoost often delivers superior accuracy and faster convergence on tabular data, especially when features exhibit varying distributions and mutual dependencies \u003csup\u003e\u003cspan citationid=\"CR70\" class=\"CitationRef\"\u003e70\u003c/span\u003e,\u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e71\u003c/span\u003e\u003c/sup\u003e. The explanatory variables included a broad range of phenological parameters such as SOS, EOS, GSL, peak EVI time, slopes of greening and senescence, MSV, EA, SEA, and auxiliary environmental factors, including nighttime light intensity (representing anthropogenic activity), elevation, and presence of water bodies. We split each city\u0026rsquo;s pixel-level dataset into training (80%) and testing (20%) subsets with a fixed random seed (42) to ensure reproducibility and evaluated model performance via R\u0026sup2; and RMSE on the held-out test set. Hyperparameters (500 iterations, learning rate\u0026thinsp;=\u0026thinsp;0.1, tree depth\u0026thinsp;=\u0026thinsp;8, GPU acceleration) were selected based on grid searches optimizing test-set RMSE. To avoid confounding by purely natural vegetation far from human influence, we masked out all pixels with VIIRS nighttime radiance below a conservative threshold\u0026mdash;thereby focusing training on urban and peri-urban contexts where anthropogenic heat and managed greenspaces co-occur.\u003c/p\u003e \u003cp\u003eModel performance was evaluated using R\u0026sup2;, with high predictive accuracy across cities, which confirmed that the selected features identified substantial spatial variability in summer LST. This strong model foundation enabled an analysis of interpretability using SHAP-based approaches. We used multiple SHAP (SHapley Additive exPlanations) tools to identify how features associated with vegetation influenced surface temperature and to explore potential nonlinearities and interactive effects.\u003c/p\u003e \u003cp\u003e(1) Ranking of importance of global features: We used the calculation of native SHAP in CatBoost to calculate the average contribution of each feature to the model predictions. This procedure allowed us to identify dominant drivers of urban summer temperature and to assess the relative importance of phenological variables vs environmental factors such as elevation and urban intensity.\u003c/p\u003e \u003cp\u003e(2) Temperature Sensitivity Analysis of Vegetation Phenology Parameters: We performed a perturbation-based sensitivity analysis to assess how phenological traits modulate urban LST under differing thermal regimes. First, training pixels were divided into high-temperature (top 20% LST) and moderate-temperature (60\u0026ndash;80% LST) subsets. Within each subset, we incrementally adjusted each vegetation parameter by Δ = \u0026plusmn;0.3 (in its native units) and recorded the resulting change in the model\u0026rsquo;s predicted LST. By comparing ΔLST curves for high- versus moderate-temperature samples, we identified where a trait continued to impart cooling (ΔLST\u0026thinsp;\u0026lt;\u0026thinsp;0), where it became neutral or led to warming (ΔLST\u0026thinsp;\u0026gt;\u0026thinsp;0), and whether any saturation or reversal points existed. This approach highlights both positive and negative associations between phenological metrics and surface temperature (Detailed explanations of methods and interpretations of charts can be found in Section \u003cspan refid=\"Sec6\" class=\"InternalRef\"\u003e3\u003c/span\u003e of Supplementary Information).\u003c/p\u003e \u003cp\u003e(3) SHAP Dependence Plots (with Interaction Effects): We generated SHAP dependence plots to visualize how the relationship between a vegetation metric and LST is affected by a second variable. In these plots, a primary feature's value is on the x-axis, its impact on LST is on the y-axis, and the points are colored by the value of an interacting feature (e.g., elevation). This method directly reveals context-dependent effects \u003csup\u003e\u003cspan citationid=\"CR72\" class=\"CitationRef\"\u003e72\u003c/span\u003e\u003c/sup\u003e .\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003e5.4 Cross-sectional node analysis\u003c/h2\u003e \u003cp\u003eTo explore the mechanisms of spatial responses between phenology and urban thermal environments, we conducted a cross-sectional analysis by extracting pixel-level phenological and thermal timing across both horizontal and vertical transects that intersected the urban centres. These transects spanned the entire study areas, from the city cores to the peripheries, allowing us to identify directionally comprehensive gradients of ecological and thermal processes. We first derived a central north-south and east-west line for each city that passed through the image centre and then extracted EVI and LST time series for all pixels along these lines. The EVI time series were aggregated across five years (2017\u0026ndash;2021), smoothed using a double logistic function, and used to calculate the timing of key phenological events (SOS, peak of season (Peak), and EOS) based on the characteristics of the derivative and amplitude of the fitted curve. For LST, we converted the original values to actual temperatures, applied Gaussian smoothing, and similarly determined SOS, Peak, and EOS based on the gradients and thermal peaks for each pixel. All extracted timing variables were converted to day of year (DOY) and plotted against their corresponding distances from the city centres (in kilometres), with positive and negative distances indicating positions in opposite directions from the cores. A detailed illustration of this method is provided in Fig. S4 in Section \u003cspan refid=\"Sec7\" class=\"InternalRef\"\u003e4\u003c/span\u003e of Supplementary Information.\u003c/p\u003e \u003cp\u003eWe fitted second-order polynomial curves to the spatial profiles of SOS, Peak, and EOS for both EVI and LST along both transects to enhance interpretability. The final visualisation consisted of six smoothed curves overlaid on a single plot: three curves representing phenological timing and three curves representing thermal timing. Raw timing values are displayed as transparent scatter points, and polynomial trend lines emphasize the continuous variation in each metric across space. We also marked the centre point (distance\u0026thinsp;=\u0026thinsp;0 km) for reference and converted pixel distances into physical distances (in kilometres) based on image resolution. This dual-transect approach allowed us to analyze direction-specific patterns of thermal\u0026ndash;phenological coupling and to identify whether key events were delayed, advanced, or distorted in different location. This method ultimately offered a spatially resolved, phenologically aware representation of urban thermal behavior and provided a mechanistic understanding of how vegetation growth and surface heat interacted along urban\u0026ndash;rural gradients on both axes.\u003c/p\u003e \u003c/div\u003e"},{"header":"Declarations","content":"\u003ch2\u003eData Availability Statement\u003c/h2\u003e \u003cp\u003eAll remote sensing datasets used in this study are publicly available. Enhanced Vegetation Index (EVI) and land surface temperature (LST) data were obtained from the Moderate Resolution Imaging Spectroradiometer (MODIS) products (MOD13A2 and MOD11A2) provided by the National Aeronautics and Space Administration\u0026rsquo;s (NASA) Land Processes Distributed Active Archive Center (LP DAAC) (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://lpdaac.usgs.gov/\u003c/span\u003e\u003cspan address=\"https://lpdaac.usgs.gov/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). Nighttime light data were sourced from the Visible Infrared Imaging Radiometer Suite Day/Night Band (VIIRS-DNB) product (VNP46A2), while elevation data were retrieved from the Shuttle Radar Topography Mission (SRTM) 30 m Digital Elevation Model (DEM) via the United States Geological Survey (USGS) EarthExplorer. All data was first obtained and processed using the Google Earth Engine platform. Derived variables and processed results can be made available upon request.\u003c/p\u003e\u003ch2\u003eCode availability\u003c/h2\u003e \u003cp\u003eThe Python code used to generate the figures and analyses of this manuscript is available at GitHub: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://github.com/baoling123/Vegetation-Phenology-.git\u003c/span\u003e\u003cspan address=\"https://github.com/baoling123/Vegetation-Phenology-.git\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eDu H et al (2025) Exacerbated heat stress induced by urban browning in the Global South. Nat Cities 2:157\u0026ndash;169\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHuang WTK et al (2023) Economic valuation of temperature-related mortality attributed to urban heat islands in European cities. Nat Commun 14:1\u0026ndash;12\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYang J et al (2023) Simulating urban expansion using cellular automata model with spatiotemporally explicit representation of urban demand. Landsc Urban Plann 231:104640\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEsperon-Rodriguez M et al (2022) Climate change increases global risk to urban forests. Nat Clim Chang 12:950\u0026ndash;955\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJia W et al (2021) Urbanization imprint on land surface phenology: The urban\u0026ndash;rural gradient analysis for Chinese cities. Glob Change Biol 27:2895\u0026ndash;2904\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJia S, Weng Q, Yoo C, Xiao H, Zhong Q (2024) Building energy savings by green roofs and cool roofs in current and future climates. npj Urban Sustain 4:1\u0026ndash;13\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRen Z et al (2024) The cooling capacity of urban vegetation and its driving force under extreme hot weather: A comparative study between dry-hot and humid-hot cities. Build Environ 263:111901\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShiflett SA et al (2017) Variation in the urban vegetation, surface temperature, air temperature nexus. Sci Total Environ 579:495\u0026ndash;505\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi Y et al (2024) Green spaces provide substantial but unequal urban cooling globally. Nat Commun 15:1\u0026ndash;13\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWu B et al (2024) Mitigation of urban heat island in China (2000\u0026ndash;2020) through vegetation-induced cooling. Sustainable Cities Soc 112:105599\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhao J et al (2025) Investigating the quantitative impact of the vegetation indices on the urban thermal comfort based on machine learning: A case study of the Qinhuai River Basin, China. Sustainable Cities Soc 125:106357\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTeskey R et al (2015) Responses of tree species to heat waves and extreme heat events. Plant Cell Environ 38:1699\u0026ndash;1712\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhao L et al (2018) Interactions between urban heat islands and heat waves. Environ Res Lett 13:034003\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhou D et al (2016) Spatiotemporal trends of urban heat island effect along the urban development intensity gradient in China. Sci Total Environ 544:617\u0026ndash;626\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAnderegg WRL, Kane JM, Anderegg L (2013) D. L. Consequences of widespread tree mortality triggered by drought and temperature stress. Nat Clim Change 3:30\u0026ndash;36\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBerg A et al (2016) Land\u0026ndash;atmosphere feedbacks amplify aridity increase over land under global warming. Nat Clim Change 6:869\u0026ndash;874\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSeneviratne SI et al (2010) Investigating soil moisture\u0026ndash;climate interactions in a changing climate: A review. Earth Sci Rev 99:125\u0026ndash;161\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhao M, Running SW (2010) Drought-Induced Reduction in Global Terrestrial Net Primary Production from 2000 Through 2009. Science 329:940\u0026ndash;943\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWu Z et al (2025) Tree species composition governs urban phenological responses to warming. Nat Commun 16:1\u0026ndash;11\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMassaro E et al (2023) Spatially-optimized urban greening for reduction of population exposure to land surface temperature extremes. Nat Commun 14:1\u0026ndash;10\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKim J, Sohn S, Wang Z, Kim Y (2024) Nonuniform response of vegetation phenology to daytime and nighttime warming in urban areas. Commun Earth Environ 5:1\u0026ndash;7\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYin Y, He L, Wennberg PO, Frankenberg C (2023) Unequal exposure to heatwaves in Los Angeles: Impact of uneven green spaces. Sci Adv 9:eade8501\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWahid A, Gelani S, Ashraf M, Foolad MR (2007) Heat tolerance in plants: An overview. Environ Exp Bot 61:199\u0026ndash;223\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShepherd M (2022) The Curious Relationship Between COVID-19 Lockdowns and Urban Heat Islands. \u003cem\u003eGeophysical Research Letters\u003c/em\u003e 49, e2022GL098198\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi L, Sun S, Zhong L, Han J, Qian X (2025) Novel spatiotemporal nonlinear regression approach for unveiling the impact of urban spatial morphology on carbon emissions. Sustainable Cities Soc 125:106381\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchwaab J et al (2021) The role of urban trees in reducing land surface temperatures in European cities. Nat Commun 12:6763\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGrimm NB et al (2008) Global Change and the Ecology of Cities. Science 319:756\u0026ndash;760\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang L et al (2022) Direct and indirect impacts of urbanization on vegetation growth across the world\u0026rsquo;s cities. Sci Adv 8:eabo0095\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDebbage N, Shepherd JM (2015) The urban heat island effect and city contiguity. Comput Environ Urban Syst 54:181\u0026ndash;194\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBack Y et al (2024) Current Interventions Are Inadequate to Maintain Cities\u0026rsquo; Resilience During Concurrent Drought and Excessive Heat. \u003cem\u003eEarth\u0026rsquo;s Future\u003c/em\u003e 13, eEF005208 (2025)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMateo F et al (2013) Machine learning methods to forecast temperature in buildings. Expert Syst Appl 40:1061\u0026ndash;1068\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRichards D, Fung T, Belcher R, Edwards P (2020) Differential air temperature cooling performance of urban vegetation types in the tropics. Urban Forestry Urban Green 50:126651\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhu L et al (2024) Street trees: The contribution of latent heat flux to cooling dense urban areas. Urban Clim 58:102147\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZipper SC et al (2016) Urban heat island impacts on plant phenology: intra-urban variability and response to land cover. Environ Res Lett 11:054023\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eManoli G et al (2019) Magnitude of urban heat islands largely explained by climate and population. Nature 573:55\u0026ndash;60\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eThornley JHM, Cannell MGR (2000) Modelling the Components of Plant Respiration: Representation and Realism. Ann Botany 85:55\u0026ndash;67\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSun L et al (2020) Evaluation of seasonal patterns of hydraulic redistribution in a humid subtropical area, East China. Hydrol Process 34:1052\u0026ndash;1062\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOverdieck D (2016) Water Use Efficiency and Stomatal Conductance. In: Overdieck D (ed) CO2, Temperature, and Trees: Experimental Approaches. Springer, Singapore, pp 57\u0026ndash;64. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/978-981-10-1860-2_5\u003c/span\u003e\u003cspan address=\"10.1007/978-981-10-1860-2_5\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCrawford B, Kelsey K, Ibsen P, Rees A, Charobee A (2024) Intra-urban variations in land surface phenology in a semi-arid environment. Environ Res Lett 20:014036\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIbsen PC et al (2024) Urban tree cover provides consistent mitigation of extreme heat in arid but not humid cities. Sustainable Cities Soc 113:105677\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePrev\u0026eacute;y J et al (2017) Greater temperature sensitivity of plant phenology at colder sites: implications for convergence across northern latitudes. Glob Change Biol 23:2660\u0026ndash;2671\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePocheville A (2015) The Ecological Niche: History and Recent Controversies. In: Heams T, Huneman P, Lecointre G, Silberstein M (eds) Handbook of Evolutionary Thinking in the Sciences. Springer Netherlands, Dordrecht, pp 547\u0026ndash;586. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/978-94-017-9014-7_26\u003c/span\u003e\u003cspan address=\"10.1007/978-94-017-9014-7_26\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMoyer AN, Hawkins TW (2017) River effects on the heat island of a small urban area. Urban Clim 21:262\u0026ndash;277\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSeneviratne SI et al (2010) Investigating soil moisture\u0026ndash;climate interactions in a changing climate: A review. Earth Sci Rev 99:125\u0026ndash;161\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMeng L et al (2020) Urban warming advances spring phenology but reduces the response of phenology to temperature in the conterminous United States. \u003cem\u003eProceedings of the National Academy of Sciences\u003c/em\u003e 117, 4228\u0026ndash;4233\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFu YH et al (2015) Declining global warming effects on the phenology of spring leaf unfolding. Nature 526:104\u0026ndash;107\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZipper SC et al (2016) Urban heat island impacts on plant phenology: intra-urban variability and response to land cover. Environ Res Lett 11:054023\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKabano P, Lindley S, Harris A (2021) Evidence of urban heat island impacts on the vegetation growing season length in a tropical city. Landsc Urban Plann 206:103989\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMeng L et al (2020) Urban warming advances spring phenology but reduces the response of phenology to temperature in the conterminous United States. \u003cem\u003eProceedings of the National Academy of Sciences\u003c/em\u003e 117, 4228\u0026ndash;4233\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVitasse Y, Signarbieux C, Fu YH (2018) Global warming leads to more uniform spring phenology across elevations. \u003cem\u003eProceedings of the National Academy of Sciences\u003c/em\u003e 115, 1004\u0026ndash;1008\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang Y et al (2024) Thermal, water, and land cover factors led to contrasting urban and rural vegetation resilience to extreme hot months. PNAS Nexus 3:pgae147\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang J, Zhou W, Pickett STA, Qian Y (2024) A scaling law for predicting urban trees canopy cooling efficiency. \u003cem\u003eProceedings of the National Academy of Sciences\u003c/em\u003e 121, e2401210121\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhao S, Liu S, Zhou D (2016) Prevalent vegetation growth enhancement in urban environment. \u003cem\u003eProceedings of the National Academy of Sciences\u003c/em\u003e 113, 6313\u0026ndash;6318\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang S et al (2019) Urban\u0026thinsp;\u0026ndash;\u0026thinsp;rural gradients reveal joint control of elevated CO2 and temperature on extended photosynthetic seasons. Nat Ecol Evol 3:1076\u0026ndash;1085\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGao K, Feng J, Santamouris M (2024) Are grand tree planting initiatives meeting expectations in mitigating urban overheating during heat waves? Sustainable Cities Soc 113:105671\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi H et al (2024) Cooling efficacy of trees across cities is determined by background climate, urban morphology, and tree trait. Commun Earth Environ 5:754\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKelloway EK (1995) Structural equation modelling in perspective. J Organizational Behav 16:215\u0026ndash;224\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShojaie A, Fox EB (2022) Granger Causality: A Review and Recent Advances. Annual Rev Stat Its Application 9:289\u0026ndash;319\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAnselin L (2002) Under the hood Issues in the specification and interpretation of spatial regression models. Agric Econ 27:247\u0026ndash;267\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMarcińczak S, Iglesias-Pascual R, Kopeć D, Wr\u0026oacute;bel K, Mooses V (2025) Landscapes of thermal inequality: Exploring patterns of climate justice across multiple spatial scales in Spain. Landsc Urban Plann 254:105255\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTian T et al (2025) Assessing the role of spatial aggregation schemes with varying campaign durations of mobile measurements on land use regression models for estimating nitrogen dioxide. Environ Pollut 368:125689\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGong P et al (2020) Annual maps of global artificial impervious area (GAIA) between 1985 and 2018. Remote Sens Environ 236:111510\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLu Y, Wu J, Liu M (2025) Decoding the cooling potential of urban green spaces: A cross-city investigation of driving factors in 311 Chinese cities under varying climate zones. Sustainable Cities Soc 126:106410\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWu S, Yu W, An J, Lin C, Chen B (2023) Remote sensing of urban greenspace exposure and equality: Scaling effects from greenspace and population mapping. Urban Forestry Urban Green 90:128136\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWeng Q (2009) Thermal infrared remote sensing for urban climate and environmental studies: Methods, applications, and trends. ISPRS J Photogrammetry Remote Sens 64:335\u0026ndash;344\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWeng Q, Fu P, Gao F (2014) Generating daily land surface temperature at Landsat resolution by fusing Landsat and MODIS data. Remote Sens Environ 145:55\u0026ndash;67\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi H et al (2024) Cooling efficacy of trees across cities is determined by background climate, urban morphology, and tree trait. Commun Earth Environ 5:1\u0026ndash;14\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGong Z, Ge W, Guo J, Liu J (2024) Satellite remote sensing of vegetation phenology: Progress, challenges, and opportunities. ISPRS J Photogrammetry Remote Sens 217:149\u0026ndash;164\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGui B, Bhardwaj A, Sam L (2025) Comparative analysis of different machine learning algorithms for urban footprint extraction in diverse urban contexts using high-resolution remote sensing imagery. J Geogr Sci 35:664\u0026ndash;696\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHsu S-C, Sharma AK, Tanone R, Ye Y-T (2024) Predicting Rainfall Using Random Forest and CatBoost Models. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.11159/icgre24.146\u003c/span\u003e\u003cspan address=\"10.11159/icgre24.146\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi Y et al (2024) Development and Evaluation of Machine Learning Models for Air-to-Land Temperature Conversion Using the Newly Established Kunlun Mountain Gradient Observation System. Land 13:1903\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKim Y, Kim Y (2022) Explainable heat-related mortality with random forest and SHapley Additive exPlanations (SHAP) models. Sustainable Cities Soc 79:103677\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"nature-portfolio","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"","title":"Nature Portfolio","twitterHandle":"","acdcEnabled":false,"dfaEnabled":false,"editorialSystem":"ejp","reportingPortfolio":"","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-6948672/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6948672/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eUrban vegetation is critical for mitigating summer heat, but previous studies have largely relied on static greenness metrics, leaving a gap in understanding of how vegetation phenology (its seasonal life cycle) regulates urban temperatures on a global scale. Here, we utilize interpretable machine learning and satellite data to quantify the influence of key phenological metrics, encompassing growth intensity (e.g., peak greenness), timing (e.g., start of season), and duration, on summer Land Surface Temperature (LST) across 24 major global cities. We found that: 1) a significant temporal mismatch exists in over 80% of cities, with vegetation green-up lagging seasonal surface warming by 50\u0026ndash;100 days, creating a window of thermal vulnerability; 2) seasonal accumulation of Enhanced Vegetation Index (EVI) provides stable, linear cooling, whereas Maximum EVI (MEV) and EVI amplitude (EA) exhibit a distinct threshold effect, with their cooling benefits diminishing or even reversing beyond a critical point; 3) vegetation's cooling effect changes with context, delivering roughly 25% greater cooling in the top 10% of temperature extremes compared to moderate conditions; and 4) in certain contexts, vegetation's cooling effect is observationally weakened or even offset when it is masked by the dominant influence of a positively correlated warming factor, such as high elevation. These findings provide mechanistic evidence that simply increasing green cover is insufficient; future urban heat mitigation must shift to \"phenology-aware\" designs that synchronize vegetation's life cycle with seasonal heat peaks to achieve maximum cooling benefits.\u003c/p\u003e","manuscriptTitle":"Phenological Metrics in the Mitigation of Urban Heat: Timing and Thresholds","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-08-22 06:16:48","doi":"10.21203/rs.3.rs-6948672/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"nature-communications","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"NCOMMS","sideBox":"Learn more about [Nature Communications](http://www.nature.com/ncomms/)","snPcode":"","submissionUrl":"https://mts-ncomms.nature.com/","title":"Nature Communications","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature Communications","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"109882b8-1d9a-4f5d-a543-e28795d39861","owner":[],"postedDate":"August 22nd, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":53503052,"name":"Earth and environmental sciences/Climate sciences/Climate change/Climate-change mitigation"},{"id":53503053,"name":"Earth and environmental sciences/Ecology/Urban ecology"},{"id":53503054,"name":"Earth and environmental sciences/Environmental social sciences/Sustainability"},{"id":53503055,"name":"Earth and environmental sciences/Environmental social sciences/Climate-change mitigation"}],"tags":[],"updatedAt":"2026-04-02T13:20:54+00:00","versionOfRecord":[],"versionCreatedAt":"2025-08-22 06:16:48","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6948672","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6948672","identity":"rs-6948672","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}