Variation in green roof vegetation health driven by age and design

Variation in green roof vegetation health driven by age and design | 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 Variation in green roof vegetation health driven by age and design Wenxi Liao, Madison Appleby, Howard Rosenblat, Mohammad Halim, and 5 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6049604/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 08 Oct, 2025 Read the published version in Nature Cities → Version 1 posted You are reading this latest preprint version Abstract Green roofs have been increasingly implemented in cities globally to enhance urban ecosystem services degraded by climate change and rapid urbanization. However, temporal trends in green roof vegetation health and the effects of design considerations at a large scale remain unclear. Here, we used 8-cm very-high-resolution multispectral remote sensing imagery to quantify the temporal changes of vegetation health and associated design drivers across 1,380 individual green roof modules in Toronto from 2011 to 2018. Results show an average increase in vegetation health and a decline in vegetation patchiness as green roofs age. We identify module area, building height, and vegetation type as primary design factors influencing green roof vegetation health, with module area positively and building height inversely affecting vegetation health. In terms of vegetation type, sedum mats are generally healthier than woody plants and grasses on green roofs. Additionally, we identify specific thresholds, module sizes with linear dimensions of 3.2–4.8 m and building heights of 14.4 m, for which smaller and higher green roof performance abruptly declines. These findings present a robust, cost-effective analytical framework for long-term assessment and modeling of urban green infrastructure at large scales, providing valuable insights into urban greening practices. Earth and environmental sciences/Ecology Earth and environmental sciences/Environmental sciences Biological sciences/Plant sciences/Plant ecology Figures Figure 1 Figure 2 Figure 3 Figure 4 Main Urban areas are the central hubs for human settlements, characterized by dynamic integration of capital, labor, and information 1 . While urbanization provides socioeconomic benefits 2 , it has resulted in environmental degradation, including flooding 3 , water quality deterioration 4 , deforestation 5 , biodiversity losses 6 , greenhouse gas emissions 7 , and air pollution 8 . Currently, over half of the global population lives in urban areas, and it is projected that 68% of the world’s population will reside in urban areas by 2050 9 . Climate change is expected to exacerbate natural disasters and environmental hazards globally 10 – 13 . Rapid urbanization, coupled with climate change, increases urban environmental risks and poses severe threats to urban populations 14 – 16 . To address these emerging environmental challenges, the United Nations’ Sustainable Development Goals (SDGs) 11 and 13 have focused on “building inclusive, sustainable, resilient, and safe cities” and “combating climate change and its impacts” 17 . Green roofs, which are vegetated rooftop systems with engineered growing media, are increasingly developed in cities as a nature-based solution to enhance urban ecosystem services and sustainability in the long term 18 . Appropriately designed green roofs can mitigate urban environmental risks and pollution, such as urban flooding 19 , water quality degradation 20 , urban heat island effects 21 , and air and noise pollution 22 , as well as provide additional ecological (e.g., biodiversity conservation 23 ) and socioeconomic benefits (e.g., recreational and psychological values and energy efficiency improvements 24 ). Due to the multifunctionality and benefits of green roofs, vegetated roof systems have been promoted in 113 cities worldwide as of 2020, supported by policies and incentive programs, such as construction permits, financing, and tax reductions 25 . Recent studies have shown proliferating numbers of green roofs at the global scale 26 – 28 , with North America emerging as the continent with the most green roofs 26 . For example, Toronto, as the first city in North America to enact a green roof bylaw 29 , has over 1,000 green roofs 30 . As of 2016, over 700 green roofs have been developed in New York City 31 . Likewise, Chicago ( www.chicago.gov ) and Portland 32 reported approximately 500 and 400 green roofs, providing over 5.5 and 1.2 million square feet of green roof coverage as of 2013 and 2019, respectively. A 2021 survey in Seattle also shows around 2.7 million square feet of green roof coverage 33 . Beyond North America, Germany, one of the pioneers in roof greening 34 , achieved nearly 1.1–1.3 billion square feet of green roof coverage by 2020 35 . Despite the rapid increase of green roof numbers, empirical studies have found that vegetation cover on some green roofs declined and/or disappeared as roof systems age due to difficult growing environments and lack of irrigation 36 – 38 . The vegetation degradation on roofs can reduce ecosystem functions and may result in environmental pollution (e.g., substrate erosion and air pollution) 39 , 40 and economic losses (i.e., green infrastructure failure and potential misuse of funds and subsidies). However, long-term monitoring of green roof vegetation health at the city scale is lacking, which limits our understanding of vegetation performance and its impacts on roof systems. Although empirical research on individual green roofs has identified potential drivers of vegetation degradation 36 – 38 , a comprehensive understanding of design factors influencing vegetation performance in operational roof systems requires broad-scale surveys that are completely lacking. Remotely sensed imagery has been commonly used to evaluate vegetation attributes and dynamics across temporal and spatial scales 41 . In recent years, satellite imagery, typically those from Landsat and Sentinel, has been increasingly used in urban areas to quantify urban vegetation cover and its temporal changes 42 , 43 . By integrating satellite imagery with urban data (e.g., land use data), studies have also assessed urban greening potential 44 and evaluated the impacts of existing and projected greenspaces on urban environments, including their roles in stormwater management 45 , 46 and cooling effects 47 , 48 . However, the spatial resolutions of these publicly available multispectral images are insufficient for accurately identifying and assessing small and fragmented greenspace patches, such as green roofs, due to the small roof dimensions and the heterogeneity of urban landscapes 43 . Additionally, high-resolution multispectral satellite imagery (spatial resolutions of 1–5 m), such as that from WorldView, RapidEye, and SkySat, is typically commercial with limited historical archives, making long-term assessment of green roof vegetation both costly and challenging. Due to the limitations in spatial resolution and data availability, remote sensing assessments of green roof vegetation have remained challenging. The lack of remote sensing analyses has also limited investigations into the effects of green roof design features on vegetation performance in operational green roof systems. To address the challenges outlined above, we conducted a city-scale analysis to evaluate temporal changes in vegetation health on green roofs and uncover the design drivers affecting the vegetation health status. We assessed vegetation health on 188 green roof projects, comprising a total of 1,380 roof modules, in the City of Toronto from 2011 to 2018 using very-high-resolution (8 cm spatial resolution) multispectral orthoimagery. We further analyzed a wide range of system-level design specifications, including area, vegetation type, and aspect ratio of green roofs and building height, and determine the key design features influencing green roof vegetation health. Finally, we tracked changes of vegetation health with system ages on individual green roofs and investigated the potential design drivers for observed changes. We addressed the following research questions: (1) Does vegetation health show a temporal increase or decrease on green roofs? (2) What design features are the key drivers for green roof vegetation health status? (3) How does each design feature influence vegetation health on roof systems? (4) What lessons can we learn to guide future green roof design? Results Temporal patterns of green roof vegetation health and patchiness Vegetation health, assessed during the growing season when most urban deciduous trees had green leaves, generally improved as green roofs aged regardless of the green roof type (mean Normalized Difference Vegetation Index, NDVI = 0.008 × Age + 0.189; P < 0.001) (Fig. 1 a and Supplementary Fig. 1). Consistent with increasing vegetation health, the overall patchiness of vegetation decreased with green roof age or time in growing (coefficient of variation (CV) of NDVI = − 0.000006 × Time + 0.145924; P = 0.04) and pre-growing (CV of NDVI = − 0.0009 × Age + 0.0396; P < 0.001) seasons (Fig. 1 b). Similar trends were observed in other vegetation indices (VIs) for growing season data; however, a decline of vegetation health with green roof age was found for the pre-growing season when most urban deciduous trees had few leaves (Supplementary Fig. 2), particularly on extensive green roofs (Supplementary Fig. 3). The vegetation patchiness strongly declined with green roof age during the pre-growing seasons for all the VIs (Supplementary Figs. 1 and 3). In addition, the health of sedum mats and woody plants improved with green roof age, while the patchiness of woody plants declined with age or time (Supplementary Fig. 4). Similar patterns were observed for vegetation types for other VIs (Supplementary Fig. 5). Design drivers affecting green roof vegetation health and patchiness Multiple linear regression models indicate that green roof area, building height, and vegetation type are the key design drivers influencing vegetation health on green roofs (Table 1 ). Additionally, the aspect ratio of green roof modules played a critical role in vegetation health during the growing season (Table 1 ). The patchiness of vegetation was primarily influenced by the vegetation type and aspect ratio of green roofs (Supplementary Table 1). Green roof building height also affected vegetation patchiness during the pre-growing seasons (Supplementary Table 1). Likewise, the models including green roof type as a predictor revealed that module area and building height of green roofs, along with aspect ratio and green roof type during the growing seasons, affected vegetation health (Supplementary Table 2). Vegetation patchiness was mainly influenced by aspect ratio, module area during the growing season, and building height during the pre-growing season (Supplementary Table 3). Table 1 Multiple linear regressions assessing effects of design factors on the mean NDVI, mean SAVI, mean GNDVI, and mean CIg during the growing (2017) and pre-growing (2018) seasons. Asterisks indicate significance of one-way ANOVA: (*), P < 0.1 *, P < 0.05; **, P < 0.01; ***, P < 0.001. Statistically significant relationships ( P < 0.05) are in boldface type. Dashes indicate the exclusion of the predictors from the models. Predictors Mean NDVI Mean SAVI Mean GNDVI Mean CIg Growing season Pre-growing season Growing season Pre-growing season Growing season Pre-growing season Growing season Pre-growing season Intercept 0.1521*** 0.0889*** 0.1174*** 0.0777*** 0.1864*** 0.1426*** 0.5089*** 0.3694*** Roof module area 0.0002*** 0.0001** 0.0001*** 0.0001*** 0.0002*** 0.0001*** 0.0007*** 0.0003*** Building height –0.0004*** –0.0003*** –0.0003** –0.0003*** –0.0004*** –0.0005*** –0.0014*** –0.0017*** Vegetation type: mix of grass and tree (vs. grass) 0.0483 0.0108 0.0433 0.0252 0.0600 (*) 0.0391 0.2119 0.1014 Vegetation type: sedum mat (vs. grass) 0.0635*** 0.0629*** 0.0573*** 0.0514*** 0.0632*** 0.0720*** 0.2311*** 0.2268*** Vegetation type: woody plants (vs. grass) 0.0304** 0.0225** 0.0231** 0.0303*** 0.0208 (*) 0.0368*** 0.0788* 0.1178*** Roof module aspect ratio –0.0005** - –0.0004** - –0.0006*** - –0.0022*** - AIC –4653 –3911 –5085 –6691 –4675 –5910 –2011 –3042 Adjusted R 2 0.139 0.157 0.157 0.145 0.163 0.170 0.161 0.157 P -value < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 Results revealed that vegetation health strongly improved with green roof area during both growing (NDVI = 0.006 × \(\:\sqrt{area}\) + 0.146; P < 0.001) and pre-growing (NDVI = 0.004 × \(\:\sqrt{area}\) + 0.103; P < 0.001) seasons regardless of the green roof type (Fig. 2 a and Supplementary Fig. 6a and Table 4). In addition, as module area increased, the health of grasses, sedum mats, and woody plants improved (Supplementary Fig. 7). However, the patchiness of vegetation also increased with module area in both seasons (Supplementary Fig. 8 and Table 4), particularly for the green roofs with grasses and woody plants (Supplementary Fig. 9). These findings were consistent across other VIs (Supplementary Figs. 10–17 and Table 4). Vegetation health declined with green roof building height during the growing (NDVI = − 0.0003 × building height + 0.2057; P < 0.01) and pre-growing (NDVI = − 0.0005 × building height + 0.1485; P < 0.001) seasons regardless of the green roof type (Fig. 2 b and Supplementary Fig. 6b and Table 4). The health of sedum mats, as well as the health of woody plants during the pre-growing seasons, decreased with green roof building height (Supplementary Fig. 7). Similar vegetation health trends were observed for the intensive green roofs (woody plants), as well as the extensive green roofs (sedum mats and grasses) during the growing seasons, across other VIs (Supplementary Figs. 10–17). Additionally, the vegetation patchiness on extensive green roofs increased with building height during the pre-growing seasons (Supplementary Fig. 8d). Plants on extensive green roofs (growing seasons: mean NDVI = 0.204; pre-growing seasons: mean NDVI = 0.138) were healthier than those on the intensive green roofs (growing seasons: mean NDVI = 0.158; pre-growing seasons: mean NDVI = 0.121) in both seasons (Fig. 2 c and Supplementary Fig. 6c and Table 5). The sedum mats exhibited a significantly higher mean NDVI (0.182) than the woody plants (0.139) and grasses (0.11) on green roofs (Supplementary Fig. 7), with similar patterns observed for other VIs (Supplementary Figs. 14 and 15). Furthermore, the intensive green roofs showed a higher CV of NDVI and Soil Adjusted Vegetation Index (SAVI) than the extensive green roofs during the pre-growing seasons (Supplementary Figs. 8 and 13 and Table 5), while the extensive systems had higher CV of Green Normalized Difference Vegetation Index (GNDVI) and Green Chlorophyll Index (CIg) than extensive systems during the growing seasons (Supplementary Fig. 12 and Table 4). During the growing seasons, grasses showed the lowest patchiness among the vegetation types (Supplementary Fig. 9e), while no green roof type effect on vegetation patchiness was observed (Supplementary Fig. 8e). There was a positive relationship between aspect ratio and vegetation health on intensive green roofs improved in both seasons (NDVI = 0.015 × \(\:\sqrt{aspect\:ratio}\) + 0.101; P < 0.001) (Fig. 2 d and Supplementary Fig. 6d), primarily due to the enhanced health of woody plants associated with higher aspect ratios (NDVI = 0.018 × \(\:\sqrt{aspect\:ratio}\:\) + 0.097; P < 0.001) (Supplementary Fig. 7). In contrast, the health of sedum mats decreased with module aspect ratio during the growing seasons (Supplementary Fig. 7g). In general, the vegetation patchiness increased with the module aspect ratio in both seasons (Supplementary Fig. 8 and Table 4) for grasses, woody plants, and sedum mats (Supplementary Fig. 9). These patterns were generally consistent across other VIs (Supplementary Figs. 10–17). Coupling effects of green roof area and building height on vegetation health Results showed positive correlations between green roof areas and vegetation health, yet distinct patterns were observed for different building height categories. The piecewise linear regression models indicated that the critical module areas required for enhanced vegetation health on green roofs increased with building height, particularly during the growing seasons (Fig. 3 a). On high-rise buildings, the vegetation health improved rapidly until the green roof area reached approximately 10.24 m 2 (square root transformed area of 3.2 m), assuming square-shaped roof modules (Fig. 3 a). Beyond this threshold, vegetation health continued to improve with green roof area, but with a lower slope value (Fig. 3 a). Similar positive trends between green roof area and vegetation health were found for mid-rise and high-rise buildings; however, the critical green roof area thresholds for the mid-rise and high-rise buildings were around 23.04 m 2 (square root transformed area of 4.8 m) and 36 m 2 (square root transformed area of 6 m), respectively (Fig. 3 a). Overall, vegetation on low-rise buildings (mean NDVI = 0.23) was healthier than that on mid-rise (mean NDVI = 0.18) or high-rise (mean NDVI = 0.19) buildings (Fig. 3 b). Individual green roof vegetation monitoring and design drivers for temporal trends A total of 946 green roofs, each with at least three years of data, were individually analyzed for temporal trends. The descriptive statistics revealed that approximately 3% of green roofs (n = 27) showed a significant increase in vegetation health, yet around 4% of roof modules (n = 39) exhibited decreasing trends as green roofs aged (Table 2 and Fig. 4 ). Among the roof modules showing temporal decline trends (n = 39), approximately 28% were grass-based (n = 11), 62% were sedum mat-based (n = 24), and 10% were woody plant-based (n = 4) (Supplementary Table 6). The multiple linear regression models indicated that increasing building height showed a negative impact on temporal trends in vegetation health during the growing seasons (Supplementary Tables 7 and 8). Table 2 Descriptive statistics for the temporal changes of individual green roofs with at least three years of data. Significance of trends are based on P < 0.05, using values corrected by the false discovery rate procedure. Significant trend Attribute Value Increase Number of green roofs 27 Percent of green roofs (%) 2.9 Decrease Number of green roofs 39 Percent of green roofs (%) 4.1 No detectable trend Number of green roofs 880 Percent of green roofs (%) 93.0 Total number of green roofs 946 Discussion This study investigates temporal vegetation health changes and key design drivers of the vegetation health across 1,380 individual green roof modules in the City of Toronto, a globally leading city with a comprehensive green roof bylaw, over the period 2011–2018. Growing season data indicate that vegetation health improved over time on both intensive (woody plant) and extensive (grass and sedum) green roofs, with vegetation patchiness also decreasing as roof systems age. This suggests gradual vegetation development with time. Long-term vegetation development generally promotes organic matter accumulation on aged green roofs, thereby increasing organic carbon content 34 and water retention capacity 49 of substrates. Additionally, substrate nitrogen (N) content, critical for plant growth in N-limited roof systems, can increase with green roof age, due to N-fixation and N deposition 50 . Analyses reveal improved health of sedum mats over time, which aligns with previous studies that show the high adaptation of sedum species to the harsh plant growth conditions on roof systems 51 , 38 . Extensive green roofs with shallow (< 15 cm) substrates impose water-limited conditions 52 . Sedum (genera Sedum and Phedimus ) species possess remarkable resilience to drought, due to their ability to store excess water and high water use efficiency through crassulacean acid metabolism (CAM) 53 , 51 . Enhanced substrate moisture content has also been shown to mitigate substrate erosion on green roofs 40 , thereby maintaining substrate depth over the long term to support sustained vegetation development. In contrast, a temporal decline in vegetation health was observed on some green roofs, particularly those with grasses susceptible to drought stress 54 , 38 , and is likely explained by drought and/or wind exposure, especially in systems without irrigation. Analyses indicate that module area, building height, and vegetation type are important determinants of vegetation health on green roofs. Vegetation health strongly improves with increasing roof module area, which can be mainly explained by the enhanced interior microclimate resulting from smaller negative edge effects on larger green roofs. This aligns with the observed decrease in vegetation patchiness associated with lower module aspect ratios, which have shorter exposed edges. Edges, compared to interiors, typically present harsh and more variable microclimatic conditions, such as strong wind and fluctuating temperature and humidity (e.g., periodic droughts), potentially contributing to plant stress and mortality 55 , 56 . In addition, the edges of building roofs are more exposed to a higher intensity of wind-driven-rain, particularly along the windward sides of the buildings 57 in contrast to interiors of roof modules that provide microclimate buffering effects 58 . However, we also noted an increase of woody vegetation performance with increasing aspect ratio (Fig. 2 d), perhaps related to reduced competitive effects in linear arrays of woody plants. Our results present a remarkable decline in vegetation health on green roofs with increasing building height. Both wind speed and turbulence generally increase with building height 59 , resulting in abrasion-induced plant damage and increased transpiration, which can ultimately lead to plant mortality 60 . Wind damage is particularly pronounced in winter for large-leaved plants grown in sandy soils, similar to green roof substrates, primarily due to increased transpiration and leaf loss caused by cuticular abrasion 60 . This is consistent with the observed increase in vegetation patchiness with building height during spring pre-growing seasons when higher wind speeds and cold temperatures likely exacerbate wind damage 61 and chilling stress or physical injury 62 to plants. In addition, taller buildings are highly exposed to intense wind-driven-rain due to higher wind speeds and the absence of protective shielding from surrounding structures 57 , making green roof systems more susceptible to water erosion and damage. An increase of runoff coefficient has also been observed on high-rise buildings in urban areas 63 , likely enhancing water erosion on green roofs atop taller buildings. With increased building height, surface temperature fluctuations also intensify due to prolonged sunlight exposure, reduced shading, and higher wind speeds associated with low building intensity 64 , potentially suppressing plant health. While vegetation health on green roofs declines with increasing building height due to challenging climatic conditions, high-rise buildings present opportunities of generating wind energy, such as through the integration of wind generation systems (e.g., wind turbines) into the building design 65 . This highlights the importance of prioritizing green roof development on low- and mid-rise buildings to support vegetation health and enhance biodiversity 66 on green roofs. Our results also show that larger module areas are required to enhance vegetation health on high-rise buildings. Tall buildings generally present challenging environments for plant growth, characterized by strong wind 59 , wind-driven-rain 57 , temperature fluctuations 64 , and reduced pollination 66 , particularly along the edges. Increasing green roof area can enhance the microclimate within the interior of roof modules 58 to offset the negative edge effects. The increase in critical module area thresholds associated with building height provides important guidance for governments in the design and planning of future green roofs across building heights in urban areas. Our findings suggest healthier vegetation on extensive green roofs relative to intensive roof systems, with sedum mats outperforming woody plants and grasses. Sedum species, particularly Sedum acre, S. album , and S. sexangulare , are often recommended for green roofs as they well adapt to frequent and prolonged droughts on green roofs due to facultative CAM photosynthesis that reduces daytime transpiration and improve water-use-efficiency to minimize water losses 67 , 68 . In contrast, woody plants and many grasses typically have C3 photosynthesis, resulting in high transpiration rates and water use during the daytime 69 . Thus, woody plants and grasses may enhance canopy cooling on green roofs 69 , yet the cooling effect can vary with supplementary irrigation and substrate type 70 . In addition, intensive green roofs observed in this study are mostly accessible roof gardens, which are designed to have landscaping patterns with space between vegetation in individual modules (Fig. S19). The exposed substrates or walkways between vegetation could lower the remotely sensed mean vegetation index values for intensive green roofs compared to the continuous, densely formed sedum mats on extensive roof systems, which are often designed to be inaccessible and optimized for high vegetation coverage. The vegetation patchiness becomes particularly evident on intensive green roofs during pre-growing prior to deciduous tree leaf-out. Despite excellent performance of sedum mats in terms of vegetation cover, recent studies have suggested incorporating native plants into green roofs to provide additional ecosystem services, such as enhanced stormwater management, cooling effects, and biodiversity enhancement 71 . Future research should assess native plant mats and optimize substrate properties to improve vegetation cover and health on extensive green roofs. In addition, we note that the permit database used lacked detailed design information at individual system scales, such as substrate properties (e.g., substrate composition and depth) and maintenance practices (e.g., irrigation, fertilization, and weeding). There was also limited image availability, with infrequent image collection at irregular dates. Full development of near-range remote-sensing for green roof monitoring would optimally address these limitations. In conclusion, the present study demonstrates the use of very-high-resolution remotely sensed imagery to assess temporal variations of vegetation health on green roofs. The results reveal a signal of enhanced vegetation health with green roof age, although a small proportion of green roof modules showed a declining trend over time. Additionally, module area, building height, and vegetation type were identified as key design factors influencing green roof vegetation health, with vegetation health generally improving with increasing module area and decreasing building height and module aspect ratio. Our findings enable accurate assessments and modelling of environmental impacts of green roofs, including urban carbon sequestration and cycling 72 , cooling effects 73 , and stormwater management 74 , as well as their socioeconomic impacts, such as social equity and well-being 75 , 76 and energy consumption 77 . This monitoring framework presents a low-cost, efficient method that is broadly applicable to cities worldwide to assess the effectiveness of green roof policies and ensure proper allocation of green infrastructure funds. Furthermore, the findings may inform decision-makers in future design and planning of green roofs in cities, emphasizing the importance of implementing larger roof modules on mid-rise and low-rise buildings. Future research should develop automatic green roof identification and extraction methods, such as deep convolutional neutral networks 78 , to facilitate the assessment of vegetation performance on green roofs across temporal and spatial scales. Methods This study assessed long-term vegetation health on green roofs and examined design drivers influencing vegetation health status at a city scale using very-high-resolution orthoimagery and green roof design data. The framework for the data collection and analysis is shown in Extended Data Fig. 1 . First, green roofs in Toronto were identified and the very-high-resolution orthoimages for green roofs were collected. Second, green roofs were digitized based on the orthoimages and the design features for each green roof were collected. Third, the green roof orthoimages were radiometrically corrected and processed to remove shadows on green roofs through shadow masks. The temporal patterns of the vegetation health on green roofs and the key design drivers for the vegetation health status were then analyzed using linear mixed effect models and multiple linear regression models, respectively. Finally, data for individual green roofs were analyzed to explore the design factors resulting in the temporal changes in vegetation. Study area The study was conducted in the City of Toronto, Ontario, Canada. Toronto enacted the Toronto Green Roof Bylaw in 2009, which requires the new construction with a gross floor area that are greater than 2,000 m 2 to build a green roof to receive a building permit 29 . The Toronto Green Roof Bylaw 29 , in conjunction with the Toronto Green Standard 79 , provide specific requirements and recommendations for green roof design at system component levels, including the composition and maintenance of vegetation and growing media. In support of the Green Roof Bylaw, Toronto established the Eco-Roof Incentive Program to provide up to $ 100,000 CAD financial incentives for non-mandated green roof installation projects on Toronto renovation and new construction 80 . As of 2024, over 1,000 green roof permits had been issued since the inception of the green roof bylaw 30 , with an additional 600 green roof projects supported by the Eco-Roof Incentive Program 80 . Each permit or project is typically comprised of multiple non-contiguous green roofs modules. Green roof identification and orthoimage collection We used green roof building permit data updated in May 2020 from the City of Toronto ( https://open.toronto.ca/dataset/building-permits-green-roofs/ ), along with high-resolution satellite imagery retrieved in 2018 in the ArcGIS basemap to verify the documented green roofs and identify undocumented ones. The addresses of green roofs that were reported as “closed” or “inspection” progress status were geolocated as a point layer and overlayed on the satellite basemap. The installation status of each green roof was verified and recorded through visual inspection of the satellite base map. We also visually examined the satellite base map to identify the green roofs that were not documented in the Toronto green roof building permit data. The data showed that most green roofs were located in the downtown, with a proliferating number of green roofs in the area since 2018. To ensure an even spatial distribution of the green roof samples, we included all the green roofs in non-downtown regions and a subset of green roofs in the downtown. In total, 242 green roof projects in Toronto, which involved 1,759 individual green roof modules, were identified and digitized. To ensure high accuracy of vegetation indices (VIs), shadow-affected areas were removed from orthoimages, resulting in the exclusion of some green roofs that were entirely covered by shadows. Consequently, a total of 188 green roof projects, including 1,380 individual green roof modules, remained and were analyzed for vegetation health and design feature effects in this study (Extended Data Fig. 2 ). Very-high-resolution multispectral orthoimages (spatial resolutions of 5–10 cm) were collected for each green roof from 2011 to 2018 (City of Toronto Survey and Mapping Services Toronto Ontario, Canada). The remotely sensed images were retrieved from the aircraft equipped with the UltraCam camera and were geometrically calibrated based on a set of images of a defined geometry target with ground control points (Vexcel Imaging GmbH, A-8010 Graz, Austria). The images were acquired on clear, snow-free days between March and June from 2011 to 2018, with the specific timing each year contingent on prevailing snow and weather conditions. The average flight height of the aircraft was 1,201 m, with the standard deviation of 391 m. All the orthoimages were projected to NAD1983 UTM Zone 17N (EPSG: 26917) for processing and analysis. Green roof digitization and design feature collection Green roofs were digitized based on the projected very-high-resolution orthoimagery collected for each year. To mitigate potential orthoimage georeferencing issues resulting from spatial displacement due to variations in building height and the orthoimage principal point, green roofs were digitized independently from each image, thereby enhancing spatial accuracy. In the case of multiple green roofs on one building rooftop or under one building permit, green roof modules with distinct boundaries and drainage systems were digitized as individual green roofs and were assigned with new unique identification codes. Non-vegetation objects on the surface of green roof systems, such as vents, pipes, and construction waste materials, were excluded from the digitization. Green roof age and design features, including roof module type (vegetation type and green roof type) and dimension (module area and aspect ratio) of green roofs, and green roof building height, were collected and analyzed for each green roof module to understand the key design drivers for vegetation health status. Green roof ages were calculated as the number of years since system installation, determined through visual inspection of the orthoimages for the presence of each green roof across years. Vegetation type on green roofs were assessed through visual assessments of the very-high-resolution orthoimages and satellite images from Google Maps. The vegetation on green roofs was classified into four categories: sedum mat, grass, woody plants, and a mix of grass and woody plants. Green roof modules with sedum mats and grasses were grouped as extensive green roofs, corresponding to systems with shallow substrate depth and short plants 81 . Roof modules with woody plants were categorized as intensive green roofs, indicating systems with thick (> 15 cm) substrate depth and tall plants 81 . The mix of grass and woody plants was assigned to a green roof type based on the majority of vegetation type in the module. The area and perimeter of each green roof module was determined through the geometric calculation from the digitization. The aspect ratio was estimated using the geometry of each roof module, assuming the modules are quadrilateral in shape, according to Eq. ( 1 ): $$\:\text{A}\text{s}\text{p}\text{e}\text{c}\text{t}\:\text{r}\text{a}\text{t}\text{i}\text{o}=\:\frac{\text{L}}{\text{W}}=\frac{\left(\frac{\text{P}}{4}+\sqrt{\frac{{\text{P}}^{2}}{16}-\text{A}}\right)}{\left(\frac{\text{P}}{4}-\sqrt{\frac{{\text{P}}^{2}}{16}-\text{A}}\right)}$$ 1 Here, L, W, P, and A indicate length, width, perimeter, and area of the green roof module. Higher aspect ratio values indicate more elongated green roof shapes. The building height of each green roof was determined using the Toronto 3D Massing data ( https://open.toronto.ca/dataset/3d-massing/ ), which integrate LiDAR, site plan, and 3D models 82 . For green roofs with missing building height data, the building height was estimated by multiplying the building storey count obtained from Google Street View by an assumed average storey height of 3 m. Orthoimage shadow removal and radiometric correction To minimize shadow effects on VIs, binary shadow masks (shadow = 0 and non-shadow = 1) were created for the projected orthoimages to exclude shadow-affected areas on green roofs. The shadow masks were generated following the methods by Halim et al. 83 , which were improved from the techniques developed by Silva et al. 84 . The enhanced approach employed adaptive Otsu’s thresholding method with additional thresholds and morphological operations to detect and remove shadows in both dark and light colors while preserving vegetation in the images, ensuring high accuracy (~ 95%) in shadow detection and removal 83 . Green roof raster images and green-roof-shaped shadow masks were extracted using digitized green roof polygons from the projected orthoimages and binary shadow masks, respectively. Shadow-free green roof raster images were then generated by multiplying the clipped green roof raster images with the green roof-shaped shadow masks using the raster calculator in ArcGIS Pro. Radiometric correction was performed on the projected orthoimagery for temporal comparisons of VIs. Since the radiometric correction coefficients were not available for the images, the empirical line method, which is a widely used approach in the calibration of aerial and satellite imagery to convert multispectral data from digital numbers (DNs) to surface reflectance 85 , 86 , was used for radiometric correction in this study. DNs of pseudo-invariant white roofs were extracted from the orthoimagery and averaged using ArcGIS Pro. Field-measured hyperspectral reflectance of representative white roofs were collected using an ASD FieldSpec Pro spectroradiometer with a spectral range of 350–2500 nm (Analytical Spectral Devices Inc., Boulder, Colorado, USA). The reflectance data were averaged for each spectral band (green, red, and near-infrared) to enable comparison with the DNs. To account for the varied spectral responses of the camera at different wavelengths, an integrated process was employed to average the field-measured hyperspectral reflectance and synthesize the reflectance with the DNs 86 . This method considered the varied contributions at different wavelengths to accurately calculate the average reflectance for each band. The synthesized formula used in this study is shown in Eq. ( 2 ): $$\:{\text{R}}_{\text{i}}=\:\frac{{\sum\:}_{j=m}^{n}{C}_{j}{R}_{j}}{{\sum\:}_{j=m}^{n}{C}_{j}}$$ 2 Here, R i is the synthesized reflectance at spectral band i, R j is the field-measured reflectance at wavelength of j, C j is the weighting coefficient (i.e., relative spectral response) at wavelength of j, and m to n is the convolution range (i.e., spectral range) of band i. A linear regression was established between surface reflectance and DNs for each band in each year to convert DNs in the orthoimagery to reflectance (Table S9). The calibration equation was developed based on two data points: the first point was assumed to be zero at the y-intercept for the reflectance of a dark target, indicates the minimal possible surface reflectance recorded by the sensor, and the second point represents the DN and reflectance of a white pseudo-invariant target (i.e., white roofs) 87 . In cases where the DNs of the white roof exceeded 255 after the radiometric correction (e.g., in 2014 and 2017 in this study), indicating image signal saturation, a pseudo-invariant grey target (e.g., concrete surfaces of rooftops) with known surface reflectance and DNs retrieved from adjacent years was used to establish empirical linear regressions and cross-calibrate the orthoimages for surface reflectance. Vegetation health and design effect assessments We used the Normalized Difference Vegetation Index (NDVI) 88 , the most commonly used vegetation index, to assess vegetation health on green roofs. Additionally, other VIs, including Soil Adjusted Vegetation Index (SAVI) 89 , Green Normalized Difference Vegetation Index (GNDVI) 90 , and Green Chlorophyll Index (CIg) 91 , were calculated to compare to NDVI (Table S10). The strong linear correlations between the mean NDVI and the means of SAVI, GNDVI, and CIg are shown in Supplementary Fig. 19. The mean VIs for individual green roofs, indicating vegetation health status, were determined by averaging the pixel-wise VIs within the digitized green roof boundaries. The vegetation patchiness on individual green roofs was assessed using the coefficient of variation (CV). The CV for each green roof was calculated as the standard deviation divided by the mean vegetation index plus one, ensuring that the CV values remained positive without altering the data distribution. The effect of green roof age on vegetation health and patchiness, indicating the vegetation temporal changes, were analyzed with linear mixed effect models using “lme4” package in R 92 , with individual green roof modules and image collection dates as random factors. Since image collection timing varied across years, and seasonality can influence vegetation leaf status, seasonality, categorized as growing and pre-growing seasons through visual inspection of orthoimages for each year, was also included as a predictor in the models. In this study, the years with many green leaves on most urban deciduous trees, which were 2012, 2014, 2016, and 2017, were classified as growing seasons, while the years with few leaves on most urban deciduous trees, which were 2011, 2013, 2015, and 2018, were categorized as pre-growing seasons. The effects of design features on green roof vegetation health and patchiness were analyzed using multiple linear regression models for 2017 and 2018, representing growing and pre-growing seasons, respectively. All the design factors (type and dimensions of green roofs and green roof building height) were included as predictors in the conceptual models. The effects of green roof type and vegetation type were included as predictors in separate sets of models. Stepwise model selection was performed on the conceptual models using both forward selection and backward elimination to identify the key predictors influencing vegetation health and patchiness on green roofs. The impacts of green roof dimensions and building height on vegetation health and patchiness were assessed using linear regression models. The area and aspect ratio of green roof modules were normalized using a square root transformation to improve data visualization. T-tests were conducted to examine the effect of green roof type on vegetation health and patchiness. Two-way analysis of variance (ANOVA) was performed to analyze the effect of vegetation type on vegetation health and patchiness, with post-hoc Tukey’s HSD tests conducted when ANOVA results were significant. We used piecewise linear regression models in “segmented” package in R 93 to investigate the critical module area thresholds for different building height in support of healthy vegetation on green roofs. Building height was classified into low-rise (≤ 14.4 m), mid-rise (14.4–20 m), and high-rise (≥ 20 m) according to the Tall Building Design Guidelines developed by the City of Toronto 94 . Individual green roof trend analysis Linear regressions were used to examine the effect of age on vegetation health, indicated by mean NDVI, for individual green roofs with at least three years of data. P-values were adjusted using the false discovery rate method for multiple comparisons 95 . Green roofs with positive and negative slopes in the linear regressions were categorized as exhibiting temporal increases and decreases in vegetation health, respectively. Descriptive statistics, including the total number of green roofs with over three years of data and the number and percentage of green roofs showing significant temporal increases and decreases at P < 0.05, were analyzed. Multiple linear regression models were established for both increase and decrease categories to identify the key design factors influencing temporal changes in vegetation on green roofs. All the data analyses were performed in R version 4.4.1 96 . Declarations Competing interests The authors declare no competing interests. Author contributions Conceptualization: W.L., L.M., J.D., S.C.T. Methodology: W.L., M.A., H.R., J.C., J.D., S.C.T. Investigation: W.L., M.A., H.R. Formal analysis: W.L., M.A., H.R., M.A.H., C.R. Visualization: W.L. Writing – Original Draft: W.L. Funding acquisition: L.M., J.D., S.C.T. Writing – Review & Editing: all authors. Acknowledgements This study was supported by grants from the Natural Sciences and Engineering Research Council of Canada (NSERC) CREATE, NSERC Discovery, and School of Cities. The work was also funded by the Schwartz Reisman Institute for Technology and Society at the University of Toronto through a Schwartz Reisman Fellowship to W.L.. We thank Stefan Herda, Allison Smith, Alessia Collia, Giuliana Frizzi, University of Toronto Scarborough (UTSC) Facility team, and UTSC EHS team for assistance with data collection and Map and Data Library and City of Toronto for providing the data and relevant information. Data availability The data that support the findings of this study are available at Zenodo under the accession code: https://zenodo.org/records/14866857 . The Toronto Orthoimagery data was acquired from https://geo2.scholarsportal.info/ . The building height data is available at https://open.toronto.ca/dataset/3d-massing/ . Toronto building permit data used for green roof identification was acquired from https://open.toronto.ca/dataset/building-permits-green-roofs/ . Code availability The scripts used to produce the datasets and results will be available at https://github.com/wenxiliao/GRvegetation/ upon the publication of the manuscript. References Fang, C. & Yu, D. Urban agglomeration: an evolving concept of an emerging phenomenon. Landsc. Urban Plan. 162 , 126–136 (2017). Bloom, D. E., Canning, D. & Fink, G. Urbanization and the wealth of nations. Science 319 , 772–775 (2008). Yang, L. et al. Urban development pattern’s influence on extreme rainfall occurrences. Nat. Commun. 15 , 3997 (2024). Miller, J. D. & Hutchins, M. The impacts of urbanisation and climate change on urban flooding and urban water quality: a review of the evidence concerning the United Kingdom. J. Hydrol. Reg. Stud. 12 , 345–362 (2017). Curtis, P. G., Slay, C. M., Harris, N. L., Tyukavina, A. & Hansen, M. C. Classifying drivers of global forest loss. Science 361 , 1108–1111 (2018). Li, G. et al. Global impacts of future urban expansion on terrestrial vertebrate diversity. Nat. Commun. 13 , 1628 (2022). Gurney, K. R. et al. Under-reporting of greenhouse gas emissions in U.S. cities. Nat. Commun. 12 , 553 (2021). Caplin, A., Ghandehari, M., Lim, C., Glimcher, P. & Thurston, G. Advancing environmental exposure assessment science to benefit society. Nat. Commun. 10 , 1236 (2019). Ritchie, H. & Roser, M. Urbanization. OurWorldInData.org. https://ourworldindata.org/urbanization (2022). Mora, C. et al. Broad threat to humanity from cumulative climate hazards intensified by greenhouse gas emissions. Nat. Clim. Chang. 8 , 1062–1071 (2018). Fawzy, S., Osman, A. I., Doran, J. & Rooney, D. W. Strategies for mitigation of climate change: a review. Environ. Chem. Lett. 18 , 2069–2094 (2020). Wilhelm, B. et al. Impact of warmer climate periods on flood hazard in the European Alps. Nat. Geosci. 15 , 118–123 (2022). Van Vliet, M. T. H. et al. Global river water quality under climate change and hydroclimatic extremes. Nat. Rev. Earth Environ. 4 , 687–702 (2023). D’Amato, G., Cecchi, L., D’Amato, M. & Liccardi, G. Urban air pollution and climate change as environmental risk factors of respiratory allergy: an update. J. Investig. Allergol. Clin. Immunol. 20 , (2010). Wang, J. et al. Anthropogenic emissions and urbanization increase risk of compound hot extremes in cities. Nat. Clim. Chang. 11 , 1084–1089 (2021). Ma, J. & Mostafavi, A. Urban form and structure explain variability in spatial inequality of property flood risk among US counties. Commun. Earth Environ. 5 , 172 (2024). Transforming Our World: The 2030 Agenda for Sustainable Development (United Nations, accessed 25 July 2024); https://sdgs.un.org/2030agenda Mihalakakou, G. et al. Green roofs as a nature-based solution for improving urban sustainability: progress and perspectives. Renew. Sustain. Energy Rev. 180 , 113306 (2023). Ercolani, G., Chiaradia, E. A., Gandolfi, C., Castelli, F. & Masseroni, D. Evaluating performances of green roofs for stormwater runoff mitigation in a high flood risk urban catchment. J. Hydrol. 566 , 830–845 (2018). Berndtsson, J. C. Green roof performance towards management of runoff water quantity and quality: a review. Ecol. Eng. 36 , 351–360 (2010). Wong, N. H., Tan, C. L., Kolokotsa, D. D. & Takebayashi, H. Greenery as a mitigation and adaptation strategy to urban heat. Nat. Rev. Earth Environ. 2 , 166–181 (2021). Rowe, D. B. Green roofs as a means of pollution abatement. Environ. Pollut. 159 , 2100–2110 (2011). Williams, N. S. G., Lundholm, J. & Scott MacIvor, J. Do green roofs help urban biodiversity conservation? J. Appl. Ecol. 51 , 1643–1649 (2014). Zhang, G. & He, B.-J. Towards green roof implementation: drivers, motivations, barriers and recommendations. Urban For. Urban Green. 58 , 126992 (2021). Liberalesso, T., Oliveira Cruz, C., Matos Silva, C. & Manso, M. Green infrastructure and public policies: an international review of green roofs and green walls incentives. Land Use Policy 96 , 104693 (2020). Appolloni, E. et al. The global rise of urban rooftop agriculture: a review of worldwide cases. J. Clean. Prod. 296 , 126556 (2021). Williams, N. S. G. et al. Ten years of greening a wide brown land: a synthesis of Australian green roof research and roadmap forward. Urban For. Urban Green. 62 , 127179 (2021). Vasconcelos, C., & Calheiros, C. S. Enhancing Environmental Education Through Nature-Based Solutions . vol. 4 (Springer International Publishing, Cham, 2022). City of Toronto. Toronto Municipal Code Chapter 492: Green Roofs. (2017). Toronto Building. Building Permits - Green Roofs. (2024). Treglia, M. L., McPhearson, T., Sanderson, E. W., Yetman, G. & Maxwell, E. N. Examining the distribution of green roofs in New York City through a lens of social, ecological, and technological filters. Ecol. Soc. 27 , art20 (2022). Netusil, N. R., Lavelle, L., Dissanayake, S. & Ando, A. W. Valuing the public benefits of green roofs. Landsc. Urban Plan. 224 , 104426 (2022). Steinberg, J. Estimating Carbon Sequestration of Green Roofs and Elevated Green Courtyards in Seattle. (University of Washington, 2021). Thuring, C. E. & Dunnett, N. Vegetation composition of old extensive green roofs (from 1980s Germany). Ecol. Process 3 , 4 (2014). Mann, G. et al. BuGG-Market Report on Building Greening 2020 . https://www.gebaeudegruen.info/fileadmin/website/Englisch/Marketreport/BuGG_Marktreport_20210623_UK_.pdf (2020). Rowe, D. B. Effect of green roof media depth on Crassulacean plant succession over seven years. Landsc. Urban Plan. 104 , 310–319 (2012). Zhang, H. et al. Crassulacean plant succession over eight years on an unirrigated green roof in Beijing. Urban For. Urban Green. 63 , 127189 (2021). MacIvor, J. S., Appleby, M., Miotto, S., Rosenblat, H. & Margolis, L. Plant cover and biomass change on extensive green roofs over a decade and ten lessons learned. J. Environ. Manage. 360 , 121047 (2024). Janhäll, S. Review on urban vegetation and particle air pollution – deposition and dispersion. Atmos. Environ. 105 , 130–137 (2015). Liao, W., Sifton, M. A. & Thomas, S. C. Biochar granulation reduces substrate erosion on green roofs. Biochar 4 , 61 (2022). Zeng, Y. et al. Optical vegetation indices for monitoring terrestrial ecosystems globally. Nat. Rev. Earth Environ. 3 , 477–493 (2022). Chen, B. et al. Contrasting inequality in human exposure to greenspace between cities of Global North and Global South. Nat. Commun. 13 , 4636 (2022). Wu, S., Chen, B., Webster, C., Xu, B. & Gong, P. Improved human greenspace exposure equality during 21st century urbanization. Nat. Commun. 14 , 6460 (2023). Shao, H. et al. Assessing city-scale green roof development potential using Unmanned Aerial Vehicle (UAV) imagery. Urban For. Urban Green. 57 , 126954 (2021). Li, C. et al. Evaluating the runoff storage supply-demand structure of green infrastructure for urban flood management. J. Clean. Prod. 280 , 124420 (2021). Conley, G. et al. Assessing the influence of urban greenness and green stormwater infrastructure on hydrology from satellite remote sensing. Sci. Total Environ. 817 , 152723 (2022). Su, Y. et al. Estimating the cooling effect magnitude of urban vegetation in different climate zones using multi-source remote sensing. Urban Clim. 43 , 101155 (2022). Liu, K., Li, X., Wang, S. & Gao, X. Assessing the effects of urban green landscape on urban thermal environment dynamic in a semiarid city by integrated use of airborne data, satellite imagery and land surface model. Int. J. Appl. Earth Obs. Geoinf. 107 , 102674 (2022). Yang, Y. & Davidson, C. I. Green roof aging effect on physical properties and hydrologic performance. J. Sustain. Water Built Environ. 7 , 04021007 (2021). Mitchell, M. E., Emilsson, T. & Buffam, I. Carbon, nitrogen, and phosphorus variation along a green roof chronosequence: implications for green roof ecosystem development. Ecol. Eng. 164 , 106211 (2021). Bates, A. J., Sadler, J. P. & Mackay, R. Vegetation development over four years on two green roofs in the UK. Urban For. Urban Green. 12 , 98–108 (2013). Vijayaraghavan, K. Green roofs: a critical review on the role of components, benefits, limitations and trends. Renew. Sustain. Energy Rev. 57 , 740–752 (2016). Monterusso, M. A., Rowe, D. B. & Rugh, C. L. Establishment and persistence of Sedum spp. and native taxa for green roof applications. HortSci 40 , 391–396 (2005). Bates, A. J., Sadler, J. P., Greswell, R. B. & Mackay, R. Effects of varying organic matter content on the development of green roof vegetation: A six year experiment. Ecol. Eng. 82 , 301–310 (2015). Oosterhoorn, M. & Kappelle, M. Vegetation structure and composition along an interior-edge-exterior gradient in a Costa Rican montane cloud forest. For. Ecol. Manage. 126 , 291–307 (2000). Magnago, L. F. S., Rocha, M. F., Meyer, L., Martins, S. V. & Meira-Neto, J. A. A. Microclimatic conditions at forest edges have significant impacts on vegetation structure in large Atlantic forest fragments. Biodivers. Conserv. 24 , 2305–2318 (2015). Abu-Zidan, Y., Nguyen, K. & Mendis, P. Influence of building shape on wind-driven rain exposure in tall buildings. J. Archit. Eng. 27 , 04021027 (2021). Ewers, R. M. & Banks-Leite, C. Fragmentation impairs the microclimate buffering effect of tropical forests. PLoS ONE 8 , e58093 (2013). Jani, D. M., Mohd, W. M. N. W. & Salleh, S. A. Effects of high-rise residential building shape and height on the urban microclimate in a tropical region. IOP Conf. Ser.: Earth Environ. Sci. 767 , 012031 (2021). Gardiner, B., Berry, P. & Moulia, B. Review: Wind impacts on plant growth, mechanics and damage. Plant Sci. 245 , 94–118 (2016). Hoxha, B., Kuriqi, A. & Filkoski, R. V. Influence of seasonal air density fluctuations on wind speed distribution in complex terrains in the context of energy yield. Energ. Ecol. Environ. 9 , 175–187 (2024). Liu, X., Zhou, Y., Xiao, J. & Bao, F. Effects of chilling on the structure, function and development of chloroplasts. Front. Plant Sci. 9 , 1715 (2018). Gao, X., Yang, Z., Han, D., Gao, K. & Zhu, Q. The impact of wind on the rainfall–runoff relationship in urban high-rise building areas. Hydrol. Earth Syst. Sci. 25 , 6023–6039 (2021). Allegrini, J., Dorer, V. & Carmeliet, J. Influence of morphologies on the microclimate in urban neighbourhoods. J. Wind. Eng. Ind. Aerodyn. 144 , 108–117 (2015). Lu, L. & Sun, K. Wind power evaluation and utilization over a reference high-rise building in urban area. Energy Build. 68 , 339–350 (2014). Maclvor, J. S. Building height matters: nesting activity of bees and wasps on vegetated roofs. Israel J. Ecol. Evol. 62 , 88–96 (2016). Starry, O., Lea-Cox, J. D., Kim, J. & van Iersel, M. W. Photosynthesis and water use by two Sedum species in green roof substrate. Environ. Exp. Bot. 107 , 105–112 (2014). Holtum, J. A. M. The diverse diaspora of CAM: a pole-to-pole sketch. Ann. Bot. 132 , 597–625 (2023). Cao, J., Hu, S., Dong, Q., Liu, L. & Wang, Z. Green roof cooling contributed by plant species with different photosynthetic strategies. Energy Build. 195 , 45–50 (2019). MacIvor, J. S., Margolis, L., Perotto, M. & Drake, J. A. P. Air temperature cooling by extensive green roofs in Toronto Canada. Ecol. Eng. 95 , 36–42 (2016). Dvorak, B. & Volder, A. Green roof vegetation for North American ecoregions: A literature review. Landsc. Urban Plan. 96 , 197–213 (2010). Rodriguez Mendez, Q., Fuss, S., Lück, S. & Creutzig, F. Assessing global urban CO 2 removal. Nat. Cities 1 , 413–423 (2024). Massaro, E. et al. Spatially-optimized urban greening for reduction of population exposure to land surface temperature extremes. Nat. Commun. 14 , 2903 (2023). Karteris, M., Theodoridou, I., Mallinis, G., Tsiros, E. & Karteris, A. Towards a green sustainable strategy for Mediterranean cities: assessing the benefits of large-scale green roofs implementation in Thessaloniki, Northern Greece, using environmental modelling, GIS and very high spatial resolution remote sensing data. Renew. Sustain. Energy Rev. 58 , 510–525 (2016). Yin, Y., He, L., Wennberg, P. O. & Frankenberg, C. Unequal exposure to heatwaves in Los Angeles: impact of uneven green spaces. Sci. Adv. 9 , eade8501 (2023). Kim, K. et al. Inequalities in urban greenness and epigenetic aging: Different associations by race and neighborhood socioeconomic status. Sci. Adv. 9 , eadf8140 (2023). Pan, H. et al. Contribution of prioritized urban nature-based solutions allocation to carbon neutrality. Nat. Clim. Chang. 13 , 862–870 (2023). Aloysius, N. & Geetha, M. A review on deep convolutional neural networks. in 2017 International Conference on Communication and Signal Processing (ICCSP) 0588–0592 (IEEE, Chennai, 2017). doi:10.1109/ICCSP.2017.8286426. City of Toronto. Toronto green roof construction standard: supplementary guidelines. https://www.toronto.ca/wp-content/uploads/2017/08/7eb7-Toronto-Green-Roof-Construction-Standard-Supplementary-Guidelines.pdf (2017). Eco-Roof Incentive Program (City of Toronto, accessed 16 August 2024); https://www.toronto.ca/services-payments/water-environment/environmental-grants-incentives/green-your-roof/ Guidelines for the Planning, Construction and Maintenance of Green Roofs . (Landscape, Research, Development & Construction Society, Bonn, Germany, 2018). D Massing (City Planning, accessed 20 June 2024); https://open.toronto.ca/dataset/3d-massing/ Halim, M. A., Liao, W., Karim, M. R., Kayes, I. & Thomas, S. Unsupervised shadow detection for vegetated urban landscapes from RGB aerial images. (in preparation). Silva, G. F., Carneiro, G. B., Doth, R., Amaral, L. A. & Azevedo, D. F. G. de. Near real-time shadow detection and removal in aerial motion imagery application. ISPRS J. Photogramm. Remote Sens. 140 , 104–121 (2018). Smith, G. M. & Milton, E. J. The use of the empirical line method to calibrate remotely sensed data to reflectance. Int. J. Remote Sens. 20 , 2653–2662 (1999). Lu, B. & He, Y. Species classification using Unmanned Aerial Vehicle (UAV)-acquired high spatial resolution imagery in a heterogeneous grassland. ISPRS J. Photogramm. Remote Sens. 128 , 73–85 (2017). Iqbal, F., Lucieer, A. & Barry, K. Simplified radiometric calibration for UAS-mounted multispectral sensor. Eur. J. Remote Sens. 51 , 301–313 (2018). Rouse, W. & Haas, R. H. Monitoring vegetation systems in the Great Plains with ERTS. in Third ERTS-1 Symposium 309–317 (Third ERTS-1 Symposium, Washington, DC: NASA., 1974). Huete, A. R. A soil-adjusted vegetation index (SAVI). Remote Sens. Environ. 25 , 295–309 (1988). Gitelson, A. A. & Merzlyak, M. N. Remote sensing of chlorophyll concentration in higher plant leaves. Adv. Space Res. 22 , 689–692 (1998). Gitelson, A. A., Gritz †, Y. & Merzlyak, M. N. Relationships between leaf chlorophyll content and spectral reflectance and algorithms for non-destructive chlorophyll assessment in higher plant leaves. J. Plant Physiol. 160 , 271–282 (2003). Bates, D., Mächler, M., Bolker, B. & Walker, S. Fitting linear mixed-effects models using lme4. J. Stat. Softw. 67 , 1–48 (2015). Muggeo, V. M. R. Segmented: an R package to fit regression models with broken-line relationships. R news 8 , 20–25 (2008). City of Toronto. Tall Building Design Guidelines. (2013). Benjamini, Y. & Hochberg, Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J. R. Stat. Soc. Ser. B Methodol. 57 , 289–300 (1995). R Core Team. R: A Language and Environment for statistical computing (R Foundation for Statistical Computing, 2024); https://www.R-project.org/ Additional Declarations There is NO Competing Interest. Supplementary Files ExtendedData.docx Extended Data SupplementaryInformation.docx Supplementary Information Cite Share Download PDF Status: Published Journal Publication published 08 Oct, 2025 Read the published version in Nature Cities → 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-6049604","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":421093779,"identity":"cb0b741f-1420-4021-b7b0-3084c64ddf35","order_by":0,"name":"Wenxi Liao","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA4klEQVRIiWNgGAWjYLCCBAYLHgYG5gNApoQMH5FaJIBa2BJAWnjYiLRHAoh5DEAswloMjvcek3gAdA9/e8/XDT9+WQC1MD/8gFfLmXPJBiCHSZw5u+1mbx/IYWzGEvi0mN3IMXwA9suN3G23GXtAWngY8Gu5/8bgAEiL/I2cZzAtzD/w28IDscXgRg7bbYYfYC1seG2xP5NjbJBgIMFjeOaY2c3eBqAWZjYzC3xaJNvPmEn+qLCxlzve/OzGjz91cvzszY9v4NMCAQZQmrENSDATVo8M/pCmfBSMglEwCkYGAACd5j8D7GVr8gAAAABJRU5ErkJggg==","orcid":"https://orcid.org/0000-0001-6600-4395","institution":"University of Toronto and McGill University","correspondingAuthor":true,"prefix":"","firstName":"Wenxi","middleName":"","lastName":"Liao","suffix":""},{"id":421093780,"identity":"0801956f-4cd4-4127-a642-55c85ad0a24d","order_by":1,"name":"Madison Appleby","email":"","orcid":"","institution":"University of Toronto","correspondingAuthor":false,"prefix":"","firstName":"Madison","middleName":"","lastName":"Appleby","suffix":""},{"id":421093781,"identity":"91a508e1-6163-4432-8d59-072b2bfb88fb","order_by":2,"name":"Howard Rosenblat","email":"","orcid":"","institution":"University of Toronto","correspondingAuthor":false,"prefix":"","firstName":"Howard","middleName":"","lastName":"Rosenblat","suffix":""},{"id":421093782,"identity":"99f66c4f-1f39-4904-97e0-eae119b02390","order_by":3,"name":"Mohammad Halim","email":"","orcid":"https://orcid.org/0000-0003-0865-7966","institution":"University of Toronto","correspondingAuthor":false,"prefix":"","firstName":"Mohammad","middleName":"","lastName":"Halim","suffix":""},{"id":421093783,"identity":"9b7d5947-6e99-4fe3-a4cd-95514d9c2777","order_by":4,"name":"Cheryl Rogers","email":"","orcid":"https://orcid.org/0000-0003-2792-1128","institution":"Toronto Metropolitan University","correspondingAuthor":false,"prefix":"","firstName":"Cheryl","middleName":"","lastName":"Rogers","suffix":""},{"id":421093784,"identity":"10d8dbcf-986e-4ca5-be0b-b19b543494e9","order_by":5,"name":"Jing Chen","email":"","orcid":"https://orcid.org/0000-0002-8682-1293","institution":"University of Toronto","correspondingAuthor":false,"prefix":"","firstName":"Jing","middleName":"","lastName":"Chen","suffix":""},{"id":421093785,"identity":"ce5d9f9e-34c9-4f1f-b023-e589183bcff2","order_by":6,"name":"Liat Margolis","email":"","orcid":"","institution":"University of Toronto","correspondingAuthor":false,"prefix":"","firstName":"Liat","middleName":"","lastName":"Margolis","suffix":""},{"id":421093786,"identity":"8ad96de1-738d-42ae-a110-0a49c85f5ab0","order_by":7,"name":"Jennifer Drake","email":"","orcid":"https://orcid.org/0000-0001-6235-3918","institution":"Carleton University","correspondingAuthor":false,"prefix":"","firstName":"Jennifer","middleName":"","lastName":"Drake","suffix":""},{"id":421093787,"identity":"24122d65-0cde-413c-90ce-d027344bd020","order_by":8,"name":"Sean Thomas","email":"","orcid":"https://orcid.org/0000-0002-0686-2483","institution":"University of Toronto","correspondingAuthor":false,"prefix":"","firstName":"Sean","middleName":"","lastName":"Thomas","suffix":""}],"badges":[],"createdAt":"2025-02-17 16:13:58","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6049604/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6049604/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s44284-025-00331-w","type":"published","date":"2025-10-08T04:00:00+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":77415769,"identity":"ef8c30cc-117d-4dba-8373-8b4e3041bd10","added_by":"auto","created_at":"2025-02-28 11:23:06","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":164831,"visible":true,"origin":"","legend":"\u003cp\u003eTemporal changes of vegetation status with green roof age. \u003cstrong\u003ea\u003c/strong\u003e, effect of green roof age on vegetation health, shown as the mean NDVI, during the growing (green color) and pre-growing (brown color) seasons. \u003cstrong\u003eb\u003c/strong\u003e, effect of green roof age on vegetation patchiness, shown as the coefficient of variation of NDVI, during the growing (green color) and pre-growing (brown color) seasons. Lines show fitted linear regression models with 95% confidence intervals in shade. Solid and dashed lines indicate models are significant (\u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05) and non-significant (\u003cem\u003eP\u003c/em\u003e \u0026gt; 0.05), respectively.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-6049604/v1/b934f2b1ea34a4daf5f4e32c.png"},{"id":77415770,"identity":"d364573e-e076-4aee-9654-831b00e4cb26","added_by":"auto","created_at":"2025-02-28 11:23:06","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":348807,"visible":true,"origin":"","legend":"\u003cp\u003eEffects of design features on vegetation health during the growing season (2017). Extensive and intensive green roofs are shown in orange and blue colors, respectively. \u003cstrong\u003ea\u003c/strong\u003e, effect of roof module area on mean NDVI. \u003cstrong\u003eb\u003c/strong\u003e, effect of building height on mean NDVI. \u003cstrong\u003ec\u003c/strong\u003e, effect of green roof type on mean NDVI. \u003cstrong\u003ed,\u003c/strong\u003e effect of roof module aspect ratio on mean NDVI. Lines show fitted linear regression models with 95% confidence intervals in shade. Solid and dashed lines indicate models are significant (\u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05) and non-significant (\u003cem\u003eP\u003c/em\u003e \u0026gt; 0.05), respectively. Bars with different lowercase letters differ significantly at \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05 according to t-tests.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-6049604/v1/9bf1f55c8eff393b17697d4a.png"},{"id":77416783,"identity":"436629d7-d9f6-4562-b0ba-37937fed414a","added_by":"auto","created_at":"2025-02-28 11:31:06","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":117467,"visible":true,"origin":"","legend":"\u003cp\u003eEffect of roof module area on vegetation health across building height categories during the growing season (2017). \u003cstrong\u003ea\u003c/strong\u003e, piecewise linear regression models depicting roof module area effect on vegetation health on the low-rise, mid-rise, and high-rise. \u003cstrong\u003eb\u003c/strong\u003e, vegetation health on the low-rise, mid-rise, and high-rise. Vegetation health is represented as mean NDVI. Green, purple, and blue colors indicate low-rise, mid-rise, and high-rise buildings, respectively. Lines depict fitted piecewise linear regression models with 95% confidence intervals in shade. The segment points, indicated by vertical dotted lines representing square root transformed green roof area, are 3.2 m for low-rise buildings, 4.8 m for mid-rise buildings, and 6 m high-rise buildings. Solid and dashed lines indicate models are significant (\u003cem\u003eP\u003c/em\u003e\u0026lt; 0.05) and non-significant (\u003cem\u003eP\u003c/em\u003e \u0026gt; 0.05), respectively. Bars with different uppercase letters differ significantly at \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05 according to Tukey's HSD test.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-6049604/v1/de08ae1f37bf9750ab56b60a.png"},{"id":77415774,"identity":"152fb691-c9ac-4cec-88dc-d93bc4c0049c","added_by":"auto","created_at":"2025-02-28 11:23:06","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":501984,"visible":true,"origin":"","legend":"\u003cp\u003eIllustrative examples of the temporal increase and decrease of vegetation health on individual green roofs during the growing seasons. Raw RGB images indicate the green roof shaped processed orthoimages. Stretched NDVI images show the NDVI values stretched to be between –0.5 and 0.5 for consistent comparisons for the green roofs with temporal increase and decrease trends. Classified NDVI images indicate the three categories of NDVI values: NDVI ≤ 0 (red color), 0 \u0026lt; NDVI ≤ 0.2 (orange color), and NDVI \u0026gt; 0.2 (green color). Slightly different shapes of the same green roofs across years are due to the shadow removal for accurate roof module mean NDVI calculations and comparisons.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-6049604/v1/ab207c5c0e436136744dab72.png"},{"id":93107878,"identity":"a8e3a604-e9aa-4652-953c-228090c47996","added_by":"auto","created_at":"2025-10-09 07:07:21","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2349505,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6049604/v1/c642df71-960b-42c3-9880-8fe03482c0b1.pdf"},{"id":77415776,"identity":"98386c38-eb45-46dc-bc72-48b7817aef45","added_by":"auto","created_at":"2025-02-28 11:23:06","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":2968975,"visible":true,"origin":"","legend":"Extended Data","description":"","filename":"ExtendedData.docx","url":"https://assets-eu.researchsquare.com/files/rs-6049604/v1/c10d211f76e2ab99b0693280.docx"},{"id":77415785,"identity":"8590447a-1acf-493b-9dad-99c76bc438b2","added_by":"auto","created_at":"2025-02-28 11:23:06","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":28172819,"visible":true,"origin":"","legend":"Supplementary Information","description":"","filename":"SupplementaryInformation.docx","url":"https://assets-eu.researchsquare.com/files/rs-6049604/v1/2e496e1ded750be9d69f3e28.docx"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"Variation in green roof vegetation health driven by age and design","fulltext":[{"header":"Main","content":"\u003cp\u003eUrban areas are the central hubs for human settlements, characterized by dynamic integration of capital, labor, and information\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e1\u003c/span\u003e\u003c/sup\u003e. While urbanization provides socioeconomic benefits\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e, it has resulted in environmental degradation, including flooding\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e, water quality deterioration\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e, deforestation\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e, biodiversity losses\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e, greenhouse gas emissions\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e, and air pollution\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e. Currently, over half of the global population lives in urban areas, and it is projected that 68% of the world\u0026rsquo;s population will reside in urban areas by 2050\u003csup\u003e9\u003c/sup\u003e. Climate change is expected to exacerbate natural disasters and environmental hazards globally\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e10\u003c/span\u003e\u0026ndash;\u003cspan class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e. Rapid urbanization, coupled with climate change, increases urban environmental risks and poses severe threats to urban populations\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e14\u003c/span\u003e\u0026ndash;\u003cspan class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/sup\u003e. To address these emerging environmental challenges, the United Nations\u0026rsquo; Sustainable Development Goals (SDGs) 11 and 13 have focused on \u0026ldquo;building inclusive, sustainable, resilient, and safe cities\u0026rdquo; and \u0026ldquo;combating climate change and its impacts\u0026rdquo;\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e17\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eGreen roofs, which are vegetated rooftop systems with engineered growing media, are increasingly developed in cities as a nature-based solution to enhance urban ecosystem services and sustainability in the long term\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e. Appropriately designed green roofs can mitigate urban environmental risks and pollution, such as urban flooding\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e, water quality degradation\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e20\u003c/span\u003e\u003c/sup\u003e, urban heat island effects\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e, and air and noise pollution\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e, as well as provide additional ecological (e.g., biodiversity conservation\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e23\u003c/span\u003e\u003c/sup\u003e) and socioeconomic benefits (e.g., recreational and psychological values and energy efficiency improvements\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e). Due to the multifunctionality and benefits of green roofs, vegetated roof systems have been promoted in 113 cities worldwide as of 2020, supported by policies and incentive programs, such as construction permits, financing, and tax reductions\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e25\u003c/span\u003e\u003c/sup\u003e. Recent studies have shown proliferating numbers of green roofs at the global scale\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e26\u003c/span\u003e\u0026ndash;\u003cspan class=\"CitationRef\"\u003e28\u003c/span\u003e\u003c/sup\u003e, with North America emerging as the continent with the most green roofs\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e. For example, Toronto, as the first city in North America to enact a green roof bylaw\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e29\u003c/span\u003e\u003c/sup\u003e, has over 1,000 green roofs\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e30\u003c/span\u003e\u003c/sup\u003e. As of 2016, over 700 green roofs have been developed in New York City\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e31\u003c/span\u003e\u003c/sup\u003e. Likewise, Chicago (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ewww.chicago.gov\u003c/span\u003e\u003c/span\u003e) and Portland\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e32\u003c/span\u003e\u003c/sup\u003e reported approximately 500 and 400 green roofs, providing over 5.5 and 1.2\u0026nbsp;million square feet of green roof coverage as of 2013 and 2019, respectively. A 2021 survey in Seattle also shows around 2.7\u0026nbsp;million square feet of green roof coverage\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e33\u003c/span\u003e\u003c/sup\u003e. Beyond North America, Germany, one of the pioneers in roof greening\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e34\u003c/span\u003e\u003c/sup\u003e, achieved nearly 1.1\u0026ndash;1.3\u0026nbsp;billion square feet of green roof coverage by 2020\u003csup\u003e35\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eDespite the rapid increase of green roof numbers, empirical studies have found that vegetation cover on some green roofs declined and/or disappeared as roof systems age due to difficult growing environments and lack of irrigation\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e36\u003c/span\u003e\u0026ndash;\u003cspan class=\"CitationRef\"\u003e38\u003c/span\u003e\u003c/sup\u003e. The vegetation degradation on roofs can reduce ecosystem functions and may result in environmental pollution (e.g., substrate erosion and air pollution)\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e39\u003c/span\u003e,\u003cspan class=\"CitationRef\"\u003e40\u003c/span\u003e\u003c/sup\u003e and economic losses (i.e., green infrastructure failure and potential misuse of funds and subsidies). However, long-term monitoring of green roof vegetation health at the city scale is lacking, which limits our understanding of vegetation performance and its impacts on roof systems. Although empirical research on individual green roofs has identified potential drivers of vegetation degradation\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e36\u003c/span\u003e\u0026ndash;\u003cspan class=\"CitationRef\"\u003e38\u003c/span\u003e\u003c/sup\u003e, a comprehensive understanding of design factors influencing vegetation performance in operational roof systems requires broad-scale surveys that are completely lacking.\u003c/p\u003e\n\u003cp\u003eRemotely sensed imagery has been commonly used to evaluate vegetation attributes and dynamics across temporal and spatial scales\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e41\u003c/span\u003e\u003c/sup\u003e. In recent years, satellite imagery, typically those from Landsat and Sentinel, has been increasingly used in urban areas to quantify urban vegetation cover and its temporal changes\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e42\u003c/span\u003e,\u003cspan class=\"CitationRef\"\u003e43\u003c/span\u003e\u003c/sup\u003e. By integrating satellite imagery with urban data (e.g., land use data), studies have also assessed urban greening potential\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e44\u003c/span\u003e\u003c/sup\u003e and evaluated the impacts of existing and projected greenspaces on urban environments, including their roles in stormwater management\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e45\u003c/span\u003e,\u003cspan class=\"CitationRef\"\u003e46\u003c/span\u003e\u003c/sup\u003e and cooling effects\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e47\u003c/span\u003e,\u003cspan class=\"CitationRef\"\u003e48\u003c/span\u003e\u003c/sup\u003e. However, the spatial resolutions of these publicly available multispectral images are insufficient for accurately identifying and assessing small and fragmented greenspace patches, such as green roofs, due to the small roof dimensions and the heterogeneity of urban landscapes\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e43\u003c/span\u003e\u003c/sup\u003e. Additionally, high-resolution multispectral satellite imagery (spatial resolutions of 1\u0026ndash;5 m), such as that from WorldView, RapidEye, and SkySat, is typically commercial with limited historical archives, making long-term assessment of green roof vegetation both costly and challenging. Due to the limitations in spatial resolution and data availability, remote sensing assessments of green roof vegetation have remained challenging. The lack of remote sensing analyses has also limited investigations into the effects of green roof design features on vegetation performance in operational green roof systems.\u003c/p\u003e\n\u003cp\u003eTo address the challenges outlined above, we conducted a city-scale analysis to evaluate temporal changes in vegetation health on green roofs and uncover the design drivers affecting the vegetation health status. We assessed vegetation health on 188 green roof projects, comprising a total of 1,380 roof modules, in the City of Toronto from 2011 to 2018 using very-high-resolution (8 cm spatial resolution) multispectral orthoimagery. We further analyzed a wide range of system-level design specifications, including area, vegetation type, and aspect ratio of green roofs and building height, and determine the key design features influencing green roof vegetation health. Finally, we tracked changes of vegetation health with system ages on individual green roofs and investigated the potential design drivers for observed changes. We addressed the following research questions: (1) Does vegetation health show a temporal increase or decrease on green roofs? (2) What design features are the key drivers for green roof vegetation health status? (3) How does each design feature influence vegetation health on roof systems? (4) What lessons can we learn to guide future green roof design?\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eTemporal patterns of green roof vegetation health and patchiness\u003c/h2\u003e \u003cp\u003eVegetation health, assessed during the growing season when most urban deciduous trees had green leaves, generally improved as green roofs aged regardless of the green roof type (mean Normalized Difference Vegetation Index, NDVI\u0026thinsp;=\u0026thinsp;0.008 \u0026times; Age\u0026thinsp;+\u0026thinsp;0.189; \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001) (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ea and Supplementary Fig.\u0026nbsp;1). Consistent with increasing vegetation health, the overall patchiness of vegetation decreased with green roof age or time in growing (coefficient of variation (CV) of NDVI = \u0026minus;\u0026thinsp;0.000006 \u0026times; Time\u0026thinsp;+\u0026thinsp;0.145924; \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.04) and pre-growing (CV of NDVI = \u0026minus;\u0026thinsp;0.0009 \u0026times; Age\u0026thinsp;+\u0026thinsp;0.0396; \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001) seasons (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eb). Similar trends were observed in other vegetation indices (VIs) for growing season data; however, a decline of vegetation health with green roof age was found for the pre-growing season when most urban deciduous trees had few leaves (Supplementary Fig.\u0026nbsp;2), particularly on extensive green roofs (Supplementary Fig.\u0026nbsp;3). The vegetation patchiness strongly declined with green roof age during the pre-growing seasons for all the VIs (Supplementary Figs.\u0026nbsp;1 and 3). In addition, the health of sedum mats and woody plants improved with green roof age, while the patchiness of woody plants declined with age or time (Supplementary Fig.\u0026nbsp;4). Similar patterns were observed for vegetation types for other VIs (Supplementary Fig.\u0026nbsp;5).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eDesign drivers affecting green roof vegetation health and patchiness\u003c/h3\u003e\n\u003cp\u003eMultiple linear regression models indicate that green roof area, building height, and vegetation type are the key design drivers influencing vegetation health on green roofs (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Additionally, the aspect ratio of green roof modules played a critical role in vegetation health during the growing season (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). The patchiness of vegetation was primarily influenced by the vegetation type and aspect ratio of green roofs (Supplementary Table\u0026nbsp;1). Green roof building height also affected vegetation patchiness during the pre-growing seasons (Supplementary Table\u0026nbsp;1). Likewise, the models including green roof type as a predictor revealed that module area and building height of green roofs, along with aspect ratio and green roof type during the growing seasons, affected vegetation health (Supplementary Table\u0026nbsp;2). Vegetation patchiness was mainly influenced by aspect ratio, module area during the growing season, and building height during the pre-growing season (Supplementary Table\u0026nbsp;3).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eMultiple linear regressions assessing effects of design factors on the mean NDVI, mean SAVI, mean GNDVI, and mean CIg during the growing (2017) and pre-growing (2018) seasons. Asterisks indicate significance of one-way ANOVA: (*), \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.1 *, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05; **, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.01; ***, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001. Statistically significant relationships (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05) are in boldface type. Dashes indicate the exclusion of the predictors from the models.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"9\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePredictors\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003eMean NDVI\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e \u003cp\u003eMean SAVI\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e \u003cp\u003eMean GNDVI\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c9\" namest=\"c8\"\u003e \u003cp\u003eMean CIg\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eGrowing season\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePre-growing season\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eGrowing season\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003ePre-growing season\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eGrowing season\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003ePre-growing season\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003eGrowing season\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c9\"\u003e \u003cp\u003ePre-growing season\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eIntercept\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.1521***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.0889***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.1174***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.0777***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.1864***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.1426***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.5089***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.3694***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eRoof module area\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.0002***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.0001**\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.0001***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.0001***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.0002***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.0001***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.0007***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.0003***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eBuilding height\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u0026ndash;0.0004***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026ndash;0.0003***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026ndash;0.0003**\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026ndash;0.0003***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u0026ndash;0.0004***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e\u0026ndash;0.0005***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e\u0026ndash;0.0014***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e\u0026ndash;0.0017***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eVegetation type: mix of grass and tree (vs. grass)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.0483\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.0108\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.0433\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.0252\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.0600 (*)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.0391\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.2119\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.1014\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eVegetation type: sedum mat (vs. grass)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.0635***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.0629***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.0573***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.0514***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.0632***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.0720***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.2311***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.2268***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eVegetation type: woody plants (vs. grass)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.0304**\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.0225**\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.0231**\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.0303***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.0208 (*)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.0368***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.0788*\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.1178***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eRoof module aspect ratio\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u0026ndash;0.0005**\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026ndash;0.0004**\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u0026ndash;0.0006***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e\u0026ndash;0.0022***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eAIC\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u0026ndash;4653\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026ndash;3911\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026ndash;5085\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026ndash;6691\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u0026ndash;4675\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e\u0026ndash;5910\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e\u0026ndash;2011\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e\u0026ndash;3042\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eAdjusted R\u003c/b\u003e\u003csup\u003e\u003cb\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/b\u003e\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.139\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.157\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.157\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.145\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.163\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.170\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.161\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.157\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eP\u003c/b\u003e\u003cb\u003e-value\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.001\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.001\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.001\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.001\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.001\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.001\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.001\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.001\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eResults revealed that vegetation health strongly improved with green roof area during both growing (NDVI\u0026thinsp;=\u0026thinsp;0.006 \u0026times; \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\sqrt{area}\\)\u003c/span\u003e\u003c/span\u003e + 0.146; \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.001) and pre-growing (NDVI = 0.004 \u0026times; \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\sqrt{area}\\)\u003c/span\u003e\u003c/span\u003e + 0.103; \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.001) seasons regardless of the green roof type (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ea and Supplementary Fig.\u0026nbsp;6a and Table\u0026nbsp;4). In addition, as module area increased, the health of grasses, sedum mats, and woody plants improved (Supplementary Fig.\u0026nbsp;7). However, the patchiness of vegetation also increased with module area in both seasons (Supplementary Fig.\u0026nbsp;8 and Table\u0026nbsp;4), particularly for the green roofs with grasses and woody plants (Supplementary Fig.\u0026nbsp;9). These findings were consistent across other VIs (Supplementary Figs.\u0026nbsp;10\u0026ndash;17 and Table\u0026nbsp;4).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eVegetation health declined with green roof building height during the growing (NDVI = \u0026minus;\u0026thinsp;0.0003 \u0026times; building height\u0026thinsp;+\u0026thinsp;0.2057; \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.01) and pre-growing (NDVI = \u0026minus;\u0026thinsp;0.0005 \u0026times; building height\u0026thinsp;+\u0026thinsp;0.1485; \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001) seasons regardless of the green roof type (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eb and Supplementary Fig.\u0026nbsp;6b and Table\u0026nbsp;4). The health of sedum mats, as well as the health of woody plants during the pre-growing seasons, decreased with green roof building height (Supplementary Fig.\u0026nbsp;7). Similar vegetation health trends were observed for the intensive green roofs (woody plants), as well as the extensive green roofs (sedum mats and grasses) during the growing seasons, across other VIs (Supplementary Figs.\u0026nbsp;10\u0026ndash;17). Additionally, the vegetation patchiness on extensive green roofs increased with building height during the pre-growing seasons (Supplementary Fig.\u0026nbsp;8d).\u003c/p\u003e \u003cp\u003ePlants on extensive green roofs (growing seasons: mean NDVI\u0026thinsp;=\u0026thinsp;0.204; pre-growing seasons: mean NDVI\u0026thinsp;=\u0026thinsp;0.138) were healthier than those on the intensive green roofs (growing seasons: mean NDVI\u0026thinsp;=\u0026thinsp;0.158; pre-growing seasons: mean NDVI\u0026thinsp;=\u0026thinsp;0.121) in both seasons (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ec and Supplementary Fig.\u0026nbsp;6c and Table\u0026nbsp;5). The sedum mats exhibited a significantly higher mean NDVI (0.182) than the woody plants (0.139) and grasses (0.11) on green roofs (Supplementary Fig.\u0026nbsp;7), with similar patterns observed for other VIs (Supplementary Figs.\u0026nbsp;14 and 15). Furthermore, the intensive green roofs showed a higher CV of NDVI and Soil Adjusted Vegetation Index (SAVI) than the extensive green roofs during the pre-growing seasons (Supplementary Figs.\u0026nbsp;8 and 13 and Table\u0026nbsp;5), while the extensive systems had higher CV of Green Normalized Difference Vegetation Index (GNDVI) and Green Chlorophyll Index (CIg) than extensive systems during the growing seasons (Supplementary Fig.\u0026nbsp;12 and Table\u0026nbsp;4). During the growing seasons, grasses showed the lowest patchiness among the vegetation types (Supplementary Fig.\u0026nbsp;9e), while no green roof type effect on vegetation patchiness was observed (Supplementary Fig.\u0026nbsp;8e).\u003c/p\u003e \u003cp\u003eThere was a positive relationship between aspect ratio and vegetation health on intensive green roofs improved in both seasons (NDVI\u0026thinsp;=\u0026thinsp;0.015 \u0026times; \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\sqrt{aspect\\:ratio}\\)\u003c/span\u003e\u003c/span\u003e + 0.101; \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.001) (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ed and Supplementary Fig.\u0026nbsp;6d), primarily due to the enhanced health of woody plants associated with higher aspect ratios (NDVI = 0.018 \u0026times; \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\sqrt{aspect\\:ratio}\\:\\)\u003c/span\u003e\u003c/span\u003e+ 0.097; \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.001) (Supplementary Fig.\u0026nbsp;7). In contrast, the health of sedum mats decreased with module aspect ratio during the growing seasons (Supplementary Fig.\u0026nbsp;7g). In general, the vegetation patchiness increased with the module aspect ratio in both seasons (Supplementary Fig.\u0026nbsp;8 and Table\u0026nbsp;4) for grasses, woody plants, and sedum mats (Supplementary Fig.\u0026nbsp;9). These patterns were generally consistent across other VIs (Supplementary Figs.\u0026nbsp;10\u0026ndash;17).\u003c/p\u003e\n\u003ch3\u003eCoupling effects of green roof area and building height on vegetation health\u003c/h3\u003e\n\u003cp\u003eResults showed positive correlations between green roof areas and vegetation health, yet distinct patterns were observed for different building height categories. The piecewise linear regression models indicated that the critical module areas required for enhanced vegetation health on green roofs increased with building height, particularly during the growing seasons (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ea). On high-rise buildings, the vegetation health improved rapidly until the green roof area reached approximately 10.24 m\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e (square root transformed area of 3.2 m), assuming square-shaped roof modules (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ea). Beyond this threshold, vegetation health continued to improve with green roof area, but with a lower slope value (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ea). Similar positive trends between green roof area and vegetation health were found for mid-rise and high-rise buildings; however, the critical green roof area thresholds for the mid-rise and high-rise buildings were around 23.04 m\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e (square root transformed area of 4.8 m) and 36 m\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e (square root transformed area of 6 m), respectively (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ea). Overall, vegetation on low-rise buildings (mean NDVI\u0026thinsp;=\u0026thinsp;0.23) was healthier than that on mid-rise (mean NDVI\u0026thinsp;=\u0026thinsp;0.18) or high-rise (mean NDVI\u0026thinsp;=\u0026thinsp;0.19) buildings (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eb).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e\n\u003ch3\u003eIndividual green roof vegetation monitoring and design drivers for temporal trends\u003c/h3\u003e\n\u003cp\u003eA total of 946 green roofs, each with at least three years of data, were individually analyzed for temporal trends. The descriptive statistics revealed that approximately 3% of green roofs (n\u0026thinsp;=\u0026thinsp;27) showed a significant increase in vegetation health, yet around 4% of roof modules (n\u0026thinsp;=\u0026thinsp;39) exhibited decreasing trends as green roofs aged (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e and Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). Among the roof modules showing temporal decline trends (n\u0026thinsp;=\u0026thinsp;39), approximately 28% were grass-based (n\u0026thinsp;=\u0026thinsp;11), 62% were sedum mat-based (n\u0026thinsp;=\u0026thinsp;24), and 10% were woody plant-based (n\u0026thinsp;=\u0026thinsp;4) (Supplementary Table\u0026nbsp;6). The multiple linear regression models indicated that increasing building height showed a negative impact on temporal trends in vegetation health during the growing seasons (Supplementary Tables\u0026nbsp;7 and 8).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eDescriptive statistics for the temporal changes of individual green roofs with at least three years of data. Significance of trends are based on \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05, using values corrected by the false discovery rate procedure.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSignificant trend\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAttribute\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eValue\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eIncrease\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNumber of green roofs\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e27\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePercent of green roofs (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2.9\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eDecrease\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNumber of green roofs\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e39\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePercent of green roofs (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e4.1\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eNo detectable trend\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNumber of green roofs\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e880\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePercent of green roofs (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e93.0\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTotal number of green roofs\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e946\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis study investigates temporal vegetation health changes and key design drivers of the vegetation health across 1,380 individual green roof modules in the City of Toronto, a globally leading city with a comprehensive green roof bylaw, over the period 2011\u0026ndash;2018. Growing season data indicate that vegetation health improved over time on both intensive (woody plant) and extensive (grass and sedum) green roofs, with vegetation patchiness also decreasing as roof systems age. This suggests gradual vegetation development with time. Long-term vegetation development generally promotes organic matter accumulation on aged green roofs, thereby increasing organic carbon content\u003csup\u003e\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u003c/sup\u003e and water retention capacity\u003csup\u003e\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e\u003c/sup\u003e of substrates. Additionally, substrate nitrogen (N) content, critical for plant growth in N-limited roof systems, can increase with green roof age, due to N-fixation and N deposition\u003csup\u003e\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e\u003c/sup\u003e. Analyses reveal improved health of sedum mats over time, which aligns with previous studies that show the high adaptation of sedum species to the harsh plant growth conditions on roof systems\u003csup\u003e\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e,\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e\u003c/sup\u003e. Extensive green roofs with shallow (\u0026lt;\u0026thinsp;15 cm) substrates impose water-limited conditions\u003csup\u003e\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e\u003c/sup\u003e. Sedum (genera \u003cem\u003eSedum\u003c/em\u003e and \u003cem\u003ePhedimus\u003c/em\u003e) species possess remarkable resilience to drought, due to their ability to store excess water and high water use efficiency through crassulacean acid metabolism (CAM)\u003csup\u003e\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e,\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e\u003c/sup\u003e. Enhanced substrate moisture content has also been shown to mitigate substrate erosion on green roofs\u003csup\u003e\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e\u003c/sup\u003e, thereby maintaining substrate depth over the long term to support sustained vegetation development. In contrast, a temporal decline in vegetation health was observed on some green roofs, particularly those with grasses susceptible to drought stress\u003csup\u003e\u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e,\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e\u003c/sup\u003e, and is likely explained by drought and/or wind exposure, especially in systems without irrigation.\u003c/p\u003e \u003cp\u003eAnalyses indicate that module area, building height, and vegetation type are important determinants of vegetation health on green roofs. Vegetation health strongly improves with increasing roof module area, which can be mainly explained by the enhanced interior microclimate resulting from smaller negative edge effects on larger green roofs. This aligns with the observed decrease in vegetation patchiness associated with lower module aspect ratios, which have shorter exposed edges. Edges, compared to interiors, typically present harsh and more variable microclimatic conditions, such as strong wind and fluctuating temperature and humidity (e.g., periodic droughts), potentially contributing to plant stress and mortality\u003csup\u003e\u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e,\u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e56\u003c/span\u003e\u003c/sup\u003e. In addition, the edges of building roofs are more exposed to a higher intensity of wind-driven-rain, particularly along the windward sides of the buildings\u003csup\u003e\u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e57\u003c/span\u003e\u003c/sup\u003e in contrast to interiors of roof modules that provide microclimate buffering effects\u003csup\u003e\u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e58\u003c/span\u003e\u003c/sup\u003e. However, we also noted an increase of woody vegetation performance with increasing aspect ratio (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ed), perhaps related to reduced competitive effects in linear arrays of woody plants.\u003c/p\u003e \u003cp\u003eOur results present a remarkable decline in vegetation health on green roofs with increasing building height. Both wind speed and turbulence generally increase with building height\u003csup\u003e\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e\u003c/sup\u003e, resulting in abrasion-induced plant damage and increased transpiration, which can ultimately lead to plant mortality\u003csup\u003e\u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e60\u003c/span\u003e\u003c/sup\u003e. Wind damage is particularly pronounced in winter for large-leaved plants grown in sandy soils, similar to green roof substrates, primarily due to increased transpiration and leaf loss caused by cuticular abrasion\u003csup\u003e\u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e60\u003c/span\u003e\u003c/sup\u003e. This is consistent with the observed increase in vegetation patchiness with building height during spring pre-growing seasons when higher wind speeds and cold temperatures likely exacerbate wind damage\u003csup\u003e\u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e61\u003c/span\u003e\u003c/sup\u003e and chilling stress or physical injury\u003csup\u003e\u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e62\u003c/span\u003e\u003c/sup\u003e to plants. In addition, taller buildings are highly exposed to intense wind-driven-rain due to higher wind speeds and the absence of protective shielding from surrounding structures\u003csup\u003e\u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e57\u003c/span\u003e\u003c/sup\u003e, making green roof systems more susceptible to water erosion and damage. An increase of runoff coefficient has also been observed on high-rise buildings in urban areas\u003csup\u003e\u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e63\u003c/span\u003e\u003c/sup\u003e, likely enhancing water erosion on green roofs atop taller buildings. With increased building height, surface temperature fluctuations also intensify due to prolonged sunlight exposure, reduced shading, and higher wind speeds associated with low building intensity\u003csup\u003e\u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e64\u003c/span\u003e\u003c/sup\u003e, potentially suppressing plant health. While vegetation health on green roofs declines with increasing building height due to challenging climatic conditions, high-rise buildings present opportunities of generating wind energy, such as through the integration of wind generation systems (e.g., wind turbines) into the building design\u003csup\u003e\u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e65\u003c/span\u003e\u003c/sup\u003e. This highlights the importance of prioritizing green roof development on low- and mid-rise buildings to support vegetation health and enhance biodiversity\u003csup\u003e\u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e66\u003c/span\u003e\u003c/sup\u003e on green roofs.\u003c/p\u003e \u003cp\u003eOur results also show that larger module areas are required to enhance vegetation health on high-rise buildings. Tall buildings generally present challenging environments for plant growth, characterized by strong wind\u003csup\u003e\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e\u003c/sup\u003e, wind-driven-rain\u003csup\u003e\u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e57\u003c/span\u003e\u003c/sup\u003e, temperature fluctuations\u003csup\u003e\u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e64\u003c/span\u003e\u003c/sup\u003e, and reduced pollination\u003csup\u003e\u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e66\u003c/span\u003e\u003c/sup\u003e, particularly along the edges. Increasing green roof area can enhance the microclimate within the interior of roof modules\u003csup\u003e\u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e58\u003c/span\u003e\u003c/sup\u003e to offset the negative edge effects. The increase in critical module area thresholds associated with building height provides important guidance for governments in the design and planning of future green roofs across building heights in urban areas.\u003c/p\u003e \u003cp\u003eOur findings suggest healthier vegetation on extensive green roofs relative to intensive roof systems, with sedum mats outperforming woody plants and grasses. Sedum species, particularly \u003cem\u003eSedum acre, S. album\u003c/em\u003e, and \u003cem\u003eS. sexangulare\u003c/em\u003e, are often recommended for green roofs as they well adapt to frequent and prolonged droughts on green roofs due to facultative CAM photosynthesis that reduces daytime transpiration and improve water-use-efficiency to minimize water losses\u003csup\u003e\u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e67\u003c/span\u003e,\u003cspan citationid=\"CR68\" class=\"CitationRef\"\u003e68\u003c/span\u003e\u003c/sup\u003e. In contrast, woody plants and many grasses typically have C3 photosynthesis, resulting in high transpiration rates and water use during the daytime\u003csup\u003e\u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e69\u003c/span\u003e\u003c/sup\u003e. Thus, woody plants and grasses may enhance canopy cooling on green roofs\u003csup\u003e\u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e69\u003c/span\u003e\u003c/sup\u003e, yet the cooling effect can vary with supplementary irrigation and substrate type\u003csup\u003e\u003cspan citationid=\"CR70\" class=\"CitationRef\"\u003e70\u003c/span\u003e\u003c/sup\u003e. In addition, intensive green roofs observed in this study are mostly accessible roof gardens, which are designed to have landscaping patterns with space between vegetation in individual modules (Fig. S19). The exposed substrates or walkways between vegetation could lower the remotely sensed mean vegetation index values for intensive green roofs compared to the continuous, densely formed sedum mats on extensive roof systems, which are often designed to be inaccessible and optimized for high vegetation coverage. The vegetation patchiness becomes particularly evident on intensive green roofs during pre-growing prior to deciduous tree leaf-out. Despite excellent performance of sedum mats in terms of vegetation cover, recent studies have suggested incorporating native plants into green roofs to provide additional ecosystem services, such as enhanced stormwater management, cooling effects, and biodiversity enhancement\u003csup\u003e\u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e71\u003c/span\u003e\u003c/sup\u003e. Future research should assess native plant mats and optimize substrate properties to improve vegetation cover and health on extensive green roofs. In addition, we note that the permit database used lacked detailed design information at individual system scales, such as substrate properties (e.g., substrate composition and depth) and maintenance practices (e.g., irrigation, fertilization, and weeding). There was also limited image availability, with infrequent image collection at irregular dates. Full development of near-range remote-sensing for green roof monitoring would optimally address these limitations.\u003c/p\u003e \u003cp\u003eIn conclusion, the present study demonstrates the use of very-high-resolution remotely sensed imagery to assess temporal variations of vegetation health on green roofs. The results reveal a signal of enhanced vegetation health with green roof age, although a small proportion of green roof modules showed a declining trend over time. Additionally, module area, building height, and vegetation type were identified as key design factors influencing green roof vegetation health, with vegetation health generally improving with increasing module area and decreasing building height and module aspect ratio. Our findings enable accurate assessments and modelling of environmental impacts of green roofs, including urban carbon sequestration and cycling\u003csup\u003e\u003cspan citationid=\"CR72\" class=\"CitationRef\"\u003e72\u003c/span\u003e\u003c/sup\u003e, cooling effects\u003csup\u003e\u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e73\u003c/span\u003e\u003c/sup\u003e, and stormwater management\u003csup\u003e\u003cspan citationid=\"CR74\" class=\"CitationRef\"\u003e74\u003c/span\u003e\u003c/sup\u003e, as well as their socioeconomic impacts, such as social equity and well-being\u003csup\u003e\u003cspan citationid=\"CR75\" class=\"CitationRef\"\u003e75\u003c/span\u003e,\u003cspan citationid=\"CR76\" class=\"CitationRef\"\u003e76\u003c/span\u003e\u003c/sup\u003e and energy consumption\u003csup\u003e\u003cspan citationid=\"CR77\" class=\"CitationRef\"\u003e77\u003c/span\u003e\u003c/sup\u003e. This monitoring framework presents a low-cost, efficient method that is broadly applicable to cities worldwide to assess the effectiveness of green roof policies and ensure proper allocation of green infrastructure funds. Furthermore, the findings may inform decision-makers in future design and planning of green roofs in cities, emphasizing the importance of implementing larger roof modules on mid-rise and low-rise buildings. Future research should develop automatic green roof identification and extraction methods, such as deep convolutional neutral networks\u003csup\u003e\u003cspan citationid=\"CR78\" class=\"CitationRef\"\u003e78\u003c/span\u003e\u003c/sup\u003e, to facilitate the assessment of vegetation performance on green roofs across temporal and spatial scales.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003cp\u003eThis study assessed long-term vegetation health on green roofs and examined design drivers influencing vegetation health status at a city scale using very-high-resolution orthoimagery and green roof design data. The framework for the data collection and analysis is shown in Extended Data Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e. First, green roofs in Toronto were identified and the very-high-resolution orthoimages for green roofs were collected. Second, green roofs were digitized based on the orthoimages and the design features for each green roof were collected. Third, the green roof orthoimages were radiometrically corrected and processed to remove shadows on green roofs through shadow masks. The temporal patterns of the vegetation health on green roofs and the key design drivers for the vegetation health status were then analyzed using linear mixed effect models and multiple linear regression models, respectively. Finally, data for individual green roofs were analyzed to explore the design factors resulting in the temporal changes in vegetation.\u003c/p\u003e\n\u003c/div\u003e\n\u003ch3\u003eStudy area\u003c/h3\u003e\n\u003cp\u003eThe study was conducted in the City of Toronto, Ontario, Canada. Toronto enacted the Toronto Green Roof Bylaw in 2009, which requires the new construction with a gross floor area that are greater than 2,000 m\u003csup\u003e\u0026nbsp;\u003cspan class=\"CitationRef\"\u003e2\u003c/span\u003e\u0026nbsp;\u003c/sup\u003e to build a green roof to receive a building permit\u003csup\u003e\u0026nbsp;\u003cspan class=\"CitationRef\"\u003e29\u003c/span\u003e\u0026nbsp;\u003c/sup\u003e. The Toronto Green Roof Bylaw\u003csup\u003e\u0026nbsp;\u003cspan class=\"CitationRef\"\u003e29\u003c/span\u003e\u0026nbsp;\u003c/sup\u003e, in conjunction with the Toronto Green Standard\u003csup\u003e\u0026nbsp;\u003cspan class=\"CitationRef\"\u003e79\u003c/span\u003e\u0026nbsp;\u003c/sup\u003e, provide specific requirements and recommendations for green roof design at system component levels, including the composition and maintenance of vegetation and growing media. In support of the Green Roof Bylaw, Toronto established the Eco-Roof Incentive Program to provide up to \u003cspan\u003e$\u003c/span\u003e100,000 CAD financial incentives for non-mandated green roof installation projects on Toronto renovation and new construction\u003csup\u003e\u0026nbsp;\u003cspan class=\"CitationRef\"\u003e80\u003c/span\u003e\u0026nbsp;\u003c/sup\u003e. As of 2024, over 1,000 green roof permits had been issued since the inception of the green roof bylaw\u003csup\u003e\u0026nbsp;\u003cspan class=\"CitationRef\"\u003e30\u003c/span\u003e\u0026nbsp;\u003c/sup\u003e, with an additional 600 green roof projects supported by the Eco-Roof Incentive Program\u003csup\u003e\u0026nbsp;\u003cspan class=\"CitationRef\"\u003e80\u003c/span\u003e\u0026nbsp;\u003c/sup\u003e. Each permit or project is typically comprised of multiple non-contiguous green roofs modules.\u003c/p\u003e\n\u003ch3\u003eGreen roof identification and orthoimage collection\u003c/h3\u003e\n\u003cp\u003eWe used green roof building permit data updated in May 2020 from the City of Toronto (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://open.toronto.ca/dataset/building-permits-green-roofs/\u003c/span\u003e\u003c/span\u003e), along with high-resolution satellite imagery retrieved in 2018 in the ArcGIS basemap to verify the documented green roofs and identify undocumented ones. The addresses of green roofs that were reported as \u0026ldquo;closed\u0026rdquo; or \u0026ldquo;inspection\u0026rdquo; progress status were geolocated as a point layer and overlayed on the satellite basemap. The installation status of each green roof was verified and recorded through visual inspection of the satellite base map. We also visually examined the satellite base map to identify the green roofs that were not documented in the Toronto green roof building permit data. The data showed that most green roofs were located in the downtown, with a proliferating number of green roofs in the area since 2018. To ensure an even spatial distribution of the green roof samples, we included all the green roofs in non-downtown regions and a subset of green roofs in the downtown. In total, 242 green roof projects in Toronto, which involved 1,759 individual green roof modules, were identified and digitized. To ensure high accuracy of vegetation indices (VIs), shadow-affected areas were removed from orthoimages, resulting in the exclusion of some green roofs that were entirely covered by shadows. Consequently, a total of 188 green roof projects, including 1,380 individual green roof modules, remained and were analyzed for vegetation health and design feature effects in this study (Extended Data Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e\n\u003cp\u003eVery-high-resolution multispectral orthoimages (spatial resolutions of 5\u0026ndash;10 cm) were collected for each green roof from 2011 to 2018 (City of Toronto Survey and Mapping Services Toronto Ontario, Canada). The remotely sensed images were retrieved from the aircraft equipped with the UltraCam camera and were geometrically calibrated based on a set of images of a defined geometry target with ground control points (Vexcel Imaging GmbH, A-8010 Graz, Austria). The images were acquired on clear, snow-free days between March and June from 2011 to 2018, with the specific timing each year contingent on prevailing snow and weather conditions. The average flight height of the aircraft was 1,201 m, with the standard deviation of 391 m. All the orthoimages were projected to NAD1983 UTM Zone 17N (EPSG: 26917) for processing and analysis.\u003c/p\u003e\n\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\n \u003ch2\u003eGreen roof digitization and design feature collection\u003c/h2\u003e\n \u003cp\u003eGreen roofs were digitized based on the projected very-high-resolution orthoimagery collected for each year. To mitigate potential orthoimage georeferencing issues resulting from spatial displacement due to variations in building height and the orthoimage principal point, green roofs were digitized independently from each image, thereby enhancing spatial accuracy. In the case of multiple green roofs on one building rooftop or under one building permit, green roof modules with distinct boundaries and drainage systems were digitized as individual green roofs and were assigned with new unique identification codes. Non-vegetation objects on the surface of green roof systems, such as vents, pipes, and construction waste materials, were excluded from the digitization.\u003c/p\u003e\n \u003cp\u003eGreen roof age and design features, including roof module type (vegetation type and green roof type) and dimension (module area and aspect ratio) of green roofs, and green roof building height, were collected and analyzed for each green roof module to understand the key design drivers for vegetation health status. Green roof ages were calculated as the number of years since system installation, determined through visual inspection of the orthoimages for the presence of each green roof across years. Vegetation type on green roofs were assessed through visual assessments of the very-high-resolution orthoimages and satellite images from Google Maps. The vegetation on green roofs was classified into four categories: sedum mat, grass, woody plants, and a mix of grass and woody plants. Green roof modules with sedum mats and grasses were grouped as extensive green roofs, corresponding to systems with shallow substrate depth and short plants\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e81\u003c/span\u003e\u003c/sup\u003e. Roof modules with woody plants were categorized as intensive green roofs, indicating systems with thick (\u0026gt;\u0026thinsp;15 cm) substrate depth and tall plants\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e81\u003c/span\u003e\u003c/sup\u003e. The mix of grass and woody plants was assigned to a green roof type based on the majority of vegetation type in the module. The area and perimeter of each green roof module was determined through the geometric calculation from the digitization. The aspect ratio was estimated using the geometry of each roof module, assuming the modules are quadrilateral in shape, according to Eq.\u0026nbsp;(\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e):\u003c/p\u003e\n \u003cdiv id=\"Equ1\" class=\"Equation\"\u003e\n \u003cdiv class=\"mathdisplay\" id=\"FileID_Equ1\" name=\"EquationSource\"\u003e$$\\:\\text{A}\\text{s}\\text{p}\\text{e}\\text{c}\\text{t}\\:\\text{r}\\text{a}\\text{t}\\text{i}\\text{o}=\\:\\frac{\\text{L}}{\\text{W}}=\\frac{\\left(\\frac{\\text{P}}{4}+\\sqrt{\\frac{{\\text{P}}^{2}}{16}-\\text{A}}\\right)}{\\left(\\frac{\\text{P}}{4}-\\sqrt{\\frac{{\\text{P}}^{2}}{16}-\\text{A}}\\right)}$$\u003c/div\u003e\n \u003cdiv class=\"EquationNumber\"\u003e1\u003c/div\u003e\n \u003c/div\u003e\n \u003cp\u003eHere, L, W, P, and A indicate length, width, perimeter, and area of the green roof module. Higher aspect ratio values indicate more elongated green roof shapes. The building height of each green roof was determined using the Toronto 3D Massing data (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://open.toronto.ca/dataset/3d-massing/\u003c/span\u003e\u003c/span\u003e), which integrate LiDAR, site plan, and 3D models\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e82\u003c/span\u003e\u003c/sup\u003e. For green roofs with missing building height data, the building height was estimated by multiplying the building storey count obtained from Google Street View by an assumed average storey height of 3 m.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\n \u003ch2\u003eOrthoimage shadow removal and radiometric correction\u003c/h2\u003e\n \u003cp\u003eTo minimize shadow effects on VIs, binary shadow masks (shadow\u0026thinsp;=\u0026thinsp;0 and non-shadow\u0026thinsp;=\u0026thinsp;1) were created for the projected orthoimages to exclude shadow-affected areas on green roofs. The shadow masks were generated following the methods by Halim et al.\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e83\u003c/span\u003e\u003c/sup\u003e, which were improved from the techniques developed by Silva et al.\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e84\u003c/span\u003e\u003c/sup\u003e. The enhanced approach employed adaptive Otsu\u0026rsquo;s thresholding method with additional thresholds and morphological operations to detect and remove shadows in both dark and light colors while preserving vegetation in the images, ensuring high accuracy (~\u0026thinsp;95%) in shadow detection and removal\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e83\u003c/span\u003e\u003c/sup\u003e. Green roof raster images and green-roof-shaped shadow masks were extracted using digitized green roof polygons from the projected orthoimages and binary shadow masks, respectively. Shadow-free green roof raster images were then generated by multiplying the clipped green roof raster images with the green roof-shaped shadow masks using the raster calculator in ArcGIS Pro.\u003c/p\u003e\n \u003cp\u003eRadiometric correction was performed on the projected orthoimagery for temporal comparisons of VIs. Since the radiometric correction coefficients were not available for the images, the empirical line method, which is a widely used approach in the calibration of aerial and satellite imagery to convert multispectral data from digital numbers (DNs) to surface reflectance\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e85\u003c/span\u003e,\u003cspan class=\"CitationRef\"\u003e86\u003c/span\u003e\u003c/sup\u003e, was used for radiometric correction in this study. DNs of pseudo-invariant white roofs were extracted from the orthoimagery and averaged using ArcGIS Pro. Field-measured hyperspectral reflectance of representative white roofs were collected using an ASD FieldSpec Pro spectroradiometer with a spectral range of 350\u0026ndash;2500 nm (Analytical Spectral Devices Inc., Boulder, Colorado, USA). The reflectance data were averaged for each spectral band (green, red, and near-infrared) to enable comparison with the DNs. To account for the varied spectral responses of the camera at different wavelengths, an integrated process was employed to average the field-measured hyperspectral reflectance and synthesize the reflectance with the DNs\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e86\u003c/span\u003e\u003c/sup\u003e. This method considered the varied contributions at different wavelengths to accurately calculate the average reflectance for each band. The synthesized formula used in this study is shown in Eq.\u0026nbsp;(\u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e):\u003c/p\u003e\n \u003cdiv id=\"Equ2\" class=\"Equation\"\u003e\n \u003cdiv class=\"mathdisplay\" id=\"FileID_Equ2\" name=\"EquationSource\"\u003e$$\\:{\\text{R}}_{\\text{i}}=\\:\\frac{{\\sum\\:}_{j=m}^{n}{C}_{j}{R}_{j}}{{\\sum\\:}_{j=m}^{n}{C}_{j}}$$\u003c/div\u003e\n \u003cdiv class=\"EquationNumber\"\u003e2\u003c/div\u003e\n \u003c/div\u003e\n \u003cp\u003eHere, R\u003csub\u003ei\u003c/sub\u003e is the synthesized reflectance at spectral band i, R\u003csub\u003ej\u003c/sub\u003e is the field-measured reflectance at wavelength of j, C\u003csub\u003ej\u003c/sub\u003e is the weighting coefficient (i.e., relative spectral response) at wavelength of j, and m to n is the convolution range (i.e., spectral range) of band i.\u003c/p\u003e\n \u003cp\u003eA linear regression was established between surface reflectance and DNs for each band in each year to convert DNs in the orthoimagery to reflectance (Table S9). The calibration equation was developed based on two data points: the first point was assumed to be zero at the y-intercept for the reflectance of a dark target, indicates the minimal possible surface reflectance recorded by the sensor, and the second point represents the DN and reflectance of a white pseudo-invariant target (i.e., white roofs)\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e87\u003c/span\u003e\u003c/sup\u003e. In cases where the DNs of the white roof exceeded 255 after the radiometric correction (e.g., in 2014 and 2017 in this study), indicating image signal saturation, a pseudo-invariant grey target (e.g., concrete surfaces of rooftops) with known surface reflectance and DNs retrieved from adjacent years was used to establish empirical linear regressions and cross-calibrate the orthoimages for surface reflectance.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\n \u003ch2\u003eVegetation health and design effect assessments\u003c/h2\u003e\n \u003cp\u003eWe used the Normalized Difference Vegetation Index (NDVI)\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e88\u003c/span\u003e\u003c/sup\u003e, the most commonly used vegetation index, to assess vegetation health on green roofs. Additionally, other VIs, including Soil Adjusted Vegetation Index (SAVI)\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e89\u003c/span\u003e\u003c/sup\u003e, Green Normalized Difference Vegetation Index (GNDVI)\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e90\u003c/span\u003e\u003c/sup\u003e, and Green Chlorophyll Index (CIg)\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e91\u003c/span\u003e\u003c/sup\u003e, were calculated to compare to NDVI (Table S10). The strong linear correlations between the mean NDVI and the means of SAVI, GNDVI, and CIg are shown in Supplementary Fig.\u0026nbsp;19. The mean VIs for individual green roofs, indicating vegetation health status, were determined by averaging the pixel-wise VIs within the digitized green roof boundaries. The vegetation patchiness on individual green roofs was assessed using the coefficient of variation (CV). The CV for each green roof was calculated as the standard deviation divided by the mean vegetation index plus one, ensuring that the CV values remained positive without altering the data distribution.\u003c/p\u003e\n \u003cp\u003eThe effect of green roof age on vegetation health and patchiness, indicating the vegetation temporal changes, were analyzed with linear mixed effect models using \u0026ldquo;lme4\u0026rdquo; package in R\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e92\u003c/span\u003e\u003c/sup\u003e, with individual green roof modules and image collection dates as random factors. Since image collection timing varied across years, and seasonality can influence vegetation leaf status, seasonality, categorized as growing and pre-growing seasons through visual inspection of orthoimages for each year, was also included as a predictor in the models. In this study, the years with many green leaves on most urban deciduous trees, which were 2012, 2014, 2016, and 2017, were classified as growing seasons, while the years with few leaves on most urban deciduous trees, which were 2011, 2013, 2015, and 2018, were categorized as pre-growing seasons.\u003c/p\u003e\n \u003cp\u003eThe effects of design features on green roof vegetation health and patchiness were analyzed using multiple linear regression models for 2017 and 2018, representing growing and pre-growing seasons, respectively. All the design factors (type and dimensions of green roofs and green roof building height) were included as predictors in the conceptual models. The effects of green roof type and vegetation type were included as predictors in separate sets of models. Stepwise model selection was performed on the conceptual models using both forward selection and backward elimination to identify the key predictors influencing vegetation health and patchiness on green roofs. The impacts of green roof dimensions and building height on vegetation health and patchiness were assessed using linear regression models. The area and aspect ratio of green roof modules were normalized using a square root transformation to improve data visualization. T-tests were conducted to examine the effect of green roof type on vegetation health and patchiness. Two-way analysis of variance (ANOVA) was performed to analyze the effect of vegetation type on vegetation health and patchiness, with post-hoc Tukey\u0026rsquo;s HSD tests conducted when ANOVA results were significant.\u003c/p\u003e\n \u003cp\u003eWe used piecewise linear regression models in \u0026ldquo;segmented\u0026rdquo; package in R\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e93\u003c/span\u003e\u003c/sup\u003e to investigate the critical module area thresholds for different building height in support of healthy vegetation on green roofs. Building height was classified into low-rise (\u0026le;\u0026thinsp;14.4 m), mid-rise (14.4\u0026ndash;20 m), and high-rise (\u0026ge;\u0026thinsp;20 m) according to the Tall Building Design Guidelines developed by the City of Toronto\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e94\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\n \u003ch2\u003eIndividual green roof trend analysis\u003c/h2\u003e\n \u003cp\u003eLinear regressions were used to examine the effect of age on vegetation health, indicated by mean NDVI, for individual green roofs with at least three years of data. P-values were adjusted using the false discovery rate method for multiple comparisons\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e95\u003c/span\u003e\u003c/sup\u003e. Green roofs with positive and negative slopes in the linear regressions were categorized as exhibiting temporal increases and decreases in vegetation health, respectively. Descriptive statistics, including the total number of green roofs with over three years of data and the number and percentage of green roofs showing significant temporal increases and decreases at \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05, were analyzed. Multiple linear regression models were established for both increase and decrease categories to identify the key design factors influencing temporal changes in vegetation on green roofs. All the data analyses were performed in R version 4.4.1\u003csup\u003e96\u003c/sup\u003e.\u003c/p\u003e\n\u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003e \u003ch2\u003eCompeting interests\u003c/h2\u003e \u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e \u003c/p\u003e\u003ch2\u003eAuthor contributions\u003c/h2\u003e \u003cp\u003eConceptualization: W.L., L.M., J.D., S.C.T. Methodology: W.L., M.A., H.R., J.C., J.D., S.C.T. Investigation: W.L., M.A., H.R. Formal analysis: W.L., M.A., H.R., M.A.H., C.R. Visualization: W.L. Writing \u0026ndash; Original Draft: W.L. Funding acquisition: L.M., J.D., S.C.T. Writing \u0026ndash; Review \u0026amp; Editing: all authors.\u003c/p\u003e\u003ch2\u003eAcknowledgements\u003c/h2\u003e \u003cp\u003eThis study was supported by grants from the Natural Sciences and Engineering Research Council of Canada (NSERC) CREATE, NSERC Discovery, and School of Cities. The work was also funded by the Schwartz Reisman Institute for Technology and Society at the University of Toronto through a Schwartz Reisman Fellowship to W.L.. We thank Stefan Herda, Allison Smith, Alessia Collia, Giuliana Frizzi, University of Toronto Scarborough (UTSC) Facility team, and UTSC EHS team for assistance with data collection and Map and Data Library and City of Toronto for providing the data and relevant information.\u003c/p\u003e\u003ch2\u003eData availability\u003c/h2\u003e \u003cp\u003eThe data that support the findings of this study are available at Zenodo under the accession code: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://zenodo.org/records/14866857\u003c/span\u003e\u003cspan address=\"https://zenodo.org/records/14866857\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. The Toronto Orthoimagery data was acquired from \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://geo2.scholarsportal.info/\u003c/span\u003e\u003cspan address=\"https://geo2.scholarsportal.info/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. The building height data is available at \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://open.toronto.ca/dataset/3d-massing/\u003c/span\u003e\u003cspan address=\"https://open.toronto.ca/dataset/3d-massing/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Toronto building permit data used for green roof identification was acquired from \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://open.toronto.ca/dataset/building-permits-green-roofs/\u003c/span\u003e\u003cspan address=\"https://open.toronto.ca/dataset/building-permits-green-roofs/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/p\u003e\u003ch2\u003eCode availability\u003c/h2\u003e \u003cp\u003eThe scripts used to produce the datasets and results will be available at \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://github.com/wenxiliao/GRvegetation/\u003c/span\u003e\u003cspan address=\"https://github.com/wenxiliao/GRvegetation/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e upon the publication of the manuscript.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eFang, C. \u0026amp; Yu, D. Urban agglomeration: an evolving concept of an emerging phenomenon. \u003cem\u003eLandsc. Urban Plan.\u003c/em\u003e \u003cstrong\u003e162\u003c/strong\u003e, 126\u0026ndash;136 (2017).\u003c/li\u003e\n\u003cli\u003eBloom, D. E., Canning, D. \u0026amp; Fink, G. Urbanization and the wealth of nations. \u003cem\u003eScience\u003c/em\u003e \u003cstrong\u003e319\u003c/strong\u003e, 772\u0026ndash;775 (2008).\u003c/li\u003e\n\u003cli\u003eYang, L. \u003cem\u003eet al.\u003c/em\u003e Urban development pattern\u0026rsquo;s influence on extreme rainfall occurrences. \u003cem\u003eNat. Commun.\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 3997 (2024).\u003c/li\u003e\n\u003cli\u003eMiller, J. D. \u0026amp; Hutchins, M. The impacts of urbanisation and climate change on urban flooding and urban water quality: a review of the evidence concerning the United Kingdom. \u003cem\u003eJ. Hydrol. Reg. Stud.\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 345\u0026ndash;362 (2017).\u003c/li\u003e\n\u003cli\u003eCurtis, P. G., Slay, C. M., Harris, N. L., Tyukavina, A. \u0026amp; Hansen, M. C. Classifying drivers of global forest loss. \u003cem\u003eScience\u003c/em\u003e \u003cstrong\u003e361\u003c/strong\u003e, 1108\u0026ndash;1111 (2018).\u003c/li\u003e\n\u003cli\u003eLi, G. \u003cem\u003eet al.\u003c/em\u003e Global impacts of future urban expansion on terrestrial vertebrate diversity. \u003cem\u003eNat. Commun.\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 1628 (2022).\u003c/li\u003e\n\u003cli\u003eGurney, K. R. \u003cem\u003eet al.\u003c/em\u003e Under-reporting of greenhouse gas emissions in U.S. cities. \u003cem\u003eNat. Commun.\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 553 (2021).\u003c/li\u003e\n\u003cli\u003eCaplin, A., Ghandehari, M., Lim, C., Glimcher, P. \u0026amp; Thurston, G. Advancing environmental exposure assessment science to benefit society. \u003cem\u003eNat. Commun.\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 1236 (2019).\u003c/li\u003e\n\u003cli\u003eRitchie, H. \u0026amp; Roser, M. Urbanization. \u003cem\u003eOurWorldInData.org.\u003c/em\u003e https://ourworldindata.org/urbanization (2022).\u003c/li\u003e\n\u003cli\u003eMora, C. \u003cem\u003eet al.\u003c/em\u003e Broad threat to humanity from cumulative climate hazards intensified by greenhouse gas emissions. \u003cem\u003eNat. Clim. Chang.\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 1062\u0026ndash;1071 (2018).\u003c/li\u003e\n\u003cli\u003eFawzy, S., Osman, A. I., Doran, J. \u0026amp; Rooney, D. W. Strategies for mitigation of climate change: a review. \u003cem\u003eEnviron. Chem. Lett.\u003c/em\u003e \u003cstrong\u003e18\u003c/strong\u003e, 2069\u0026ndash;2094 (2020).\u003c/li\u003e\n\u003cli\u003eWilhelm, B. \u003cem\u003eet al.\u003c/em\u003e Impact of warmer climate periods on flood hazard in the European Alps. \u003cem\u003eNat. Geosci.\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 118\u0026ndash;123 (2022).\u003c/li\u003e\n\u003cli\u003eVan Vliet, M. T. H. \u003cem\u003eet al.\u003c/em\u003e Global river water quality under climate change and hydroclimatic extremes. \u003cem\u003eNat. Rev. Earth Environ.\u003c/em\u003e \u003cstrong\u003e4\u003c/strong\u003e, 687\u0026ndash;702 (2023).\u003c/li\u003e\n\u003cli\u003eD\u0026rsquo;Amato, G., Cecchi, L., D\u0026rsquo;Amato, M. \u0026amp; Liccardi, G. Urban air pollution and climate change as environmental risk factors of respiratory allergy: an update. \u003cem\u003eJ. Investig. Allergol. Clin. Immunol.\u003c/em\u003e \u003cstrong\u003e20\u003c/strong\u003e, (2010).\u003c/li\u003e\n\u003cli\u003eWang, J. \u003cem\u003eet al.\u003c/em\u003e Anthropogenic emissions and urbanization increase risk of compound hot extremes in cities. \u003cem\u003eNat. Clim. Chang.\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e, 1084\u0026ndash;1089 (2021).\u003c/li\u003e\n\u003cli\u003eMa, J. \u0026amp; Mostafavi, A. Urban form and structure explain variability in spatial inequality of property flood risk among US counties. \u003cem\u003eCommun. Earth Environ.\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, 172 (2024).\u003c/li\u003e\n\u003cli\u003eTransforming Our World: The 2030 Agenda for Sustainable Development (United Nations, accessed 25 July 2024); https://sdgs.un.org/2030agenda\u003c/li\u003e\n\u003cli\u003eMihalakakou, G. \u003cem\u003eet al.\u003c/em\u003e Green roofs as a nature-based solution for improving urban sustainability: progress and perspectives. \u003cem\u003eRenew. Sustain. Energy Rev.\u003c/em\u003e \u003cstrong\u003e180\u003c/strong\u003e, 113306 (2023).\u003c/li\u003e\n\u003cli\u003eErcolani, G., Chiaradia, E. A., Gandolfi, C., Castelli, F. \u0026amp; Masseroni, D. Evaluating performances of green roofs for stormwater runoff mitigation in a high flood risk urban catchment. \u003cem\u003eJ. Hydrol.\u003c/em\u003e \u003cstrong\u003e566\u003c/strong\u003e, 830\u0026ndash;845 (2018).\u003c/li\u003e\n\u003cli\u003eBerndtsson, J. C. Green roof performance towards management of runoff water quantity and quality: a review. \u003cem\u003eEcol. Eng.\u003c/em\u003e \u003cstrong\u003e36\u003c/strong\u003e, 351\u0026ndash;360 (2010).\u003c/li\u003e\n\u003cli\u003eWong, N. H., Tan, C. L., Kolokotsa, D. D. \u0026amp; Takebayashi, H. Greenery as a mitigation and adaptation strategy to urban heat. \u003cem\u003eNat. Rev. Earth Environ.\u003c/em\u003e \u003cstrong\u003e2\u003c/strong\u003e, 166\u0026ndash;181 (2021).\u003c/li\u003e\n\u003cli\u003eRowe, D. B. Green roofs as a means of pollution abatement. \u003cem\u003eEnviron. Pollut.\u003c/em\u003e \u003cstrong\u003e159\u003c/strong\u003e, 2100\u0026ndash;2110 (2011).\u003c/li\u003e\n\u003cli\u003eWilliams, N. S. G., Lundholm, J. \u0026amp; Scott MacIvor, J. Do green roofs help urban biodiversity conservation? \u003cem\u003eJ. Appl. Ecol.\u003c/em\u003e \u003cstrong\u003e51\u003c/strong\u003e, 1643\u0026ndash;1649 (2014).\u003c/li\u003e\n\u003cli\u003eZhang, G. \u0026amp; He, B.-J. Towards green roof implementation: drivers, motivations, barriers and recommendations. \u003cem\u003eUrban For. Urban Green. \u003c/em\u003e\u003cstrong\u003e58\u003c/strong\u003e, 126992 (2021).\u003c/li\u003e\n\u003cli\u003eLiberalesso, T., Oliveira Cruz, C., Matos Silva, C. \u0026amp; Manso, M. Green infrastructure and public policies: an international review of green roofs and green walls incentives. \u003cem\u003eLand Use Policy\u003c/em\u003e \u003cstrong\u003e96\u003c/strong\u003e, 104693 (2020).\u003c/li\u003e\n\u003cli\u003eAppolloni, E. \u003cem\u003eet al.\u003c/em\u003e The global rise of urban rooftop agriculture: a review of worldwide cases. \u003cem\u003eJ. Clean. Prod.\u003c/em\u003e \u003cstrong\u003e296\u003c/strong\u003e, 126556 (2021).\u003c/li\u003e\n\u003cli\u003eWilliams, N. S. G. \u003cem\u003eet al.\u003c/em\u003e Ten years of greening a wide brown land: a synthesis of Australian green roof research and roadmap forward. \u003cem\u003eUrban For. Urban Green.\u003c/em\u003e \u003cstrong\u003e62\u003c/strong\u003e, 127179 (2021).\u003c/li\u003e\n\u003cli\u003eVasconcelos, C., \u0026amp; Calheiros, C. S. \u003cem\u003eEnhancing Environmental Education Through Nature-Based Solutions\u003c/em\u003e. vol. 4 (Springer International Publishing, Cham, 2022).\u003c/li\u003e\n\u003cli\u003eCity of Toronto. Toronto Municipal Code Chapter 492: Green Roofs. (2017).\u003c/li\u003e\n\u003cli\u003eToronto Building. Building Permits - Green Roofs. (2024).\u003c/li\u003e\n\u003cli\u003eTreglia, M. L., McPhearson, T., Sanderson, E. W., Yetman, G. \u0026amp; Maxwell, E. N. Examining the distribution of green roofs in New York City through a lens of social, ecological, and technological filters. \u003cem\u003eEcol. Soc.\u003c/em\u003e \u003cstrong\u003e27\u003c/strong\u003e, art20 (2022).\u003c/li\u003e\n\u003cli\u003eNetusil, N. R., Lavelle, L., Dissanayake, S. \u0026amp; Ando, A. W. Valuing the public benefits of green roofs. \u003cem\u003eLandsc. Urban Plan.\u003c/em\u003e \u003cstrong\u003e224\u003c/strong\u003e, 104426 (2022).\u003c/li\u003e\n\u003cli\u003eSteinberg, J. Estimating Carbon Sequestration of Green Roofs and Elevated Green Courtyards in Seattle. (University of Washington, 2021).\u003c/li\u003e\n\u003cli\u003eThuring, C. E. \u0026amp; Dunnett, N. Vegetation composition of old extensive green roofs (from 1980s Germany). \u003cem\u003eEcol. Process\u003c/em\u003e \u003cstrong\u003e3\u003c/strong\u003e, 4 (2014).\u003c/li\u003e\n\u003cli\u003eMann, G. \u003cem\u003eet al.\u003c/em\u003e \u003cem\u003eBuGG-Market Report on Building Greening 2020\u003c/em\u003e. https://www.gebaeudegruen.info/fileadmin/website/Englisch/Marketreport/BuGG_Marktreport_20210623_UK_.pdf (2020).\u003c/li\u003e\n\u003cli\u003eRowe, D. B. Effect of green roof media depth on Crassulacean plant succession over seven years. \u003cem\u003eLandsc. Urban Plan.\u003c/em\u003e \u003cstrong\u003e104\u003c/strong\u003e, 310\u0026ndash;319 (2012).\u003c/li\u003e\n\u003cli\u003eZhang, H. \u003cem\u003eet al.\u003c/em\u003e Crassulacean plant succession over eight years on an unirrigated green roof in Beijing. \u003cem\u003eUrban For. Urban Green.\u003c/em\u003e \u003cstrong\u003e63\u003c/strong\u003e, 127189 (2021).\u003c/li\u003e\n\u003cli\u003eMacIvor, J. S., Appleby, M., Miotto, S., Rosenblat, H. \u0026amp; Margolis, L. Plant cover and biomass change on extensive green roofs over a decade and ten lessons learned. \u003cem\u003eJ. Environ. Manage.\u003c/em\u003e \u003cstrong\u003e360\u003c/strong\u003e, 121047 (2024).\u003c/li\u003e\n\u003cli\u003eJanh\u0026auml;ll, S. Review on urban vegetation and particle air pollution \u0026ndash; deposition and dispersion. \u003cem\u003eAtmos. Environ. \u003c/em\u003e\u003cstrong\u003e105\u003c/strong\u003e, 130\u0026ndash;137 (2015).\u003c/li\u003e\n\u003cli\u003eLiao, W., Sifton, M. A. \u0026amp; Thomas, S. C. Biochar granulation reduces substrate erosion on green roofs. \u003cem\u003eBiochar\u003c/em\u003e \u003cstrong\u003e4\u003c/strong\u003e, 61 (2022).\u003c/li\u003e\n\u003cli\u003eZeng, Y. \u003cem\u003eet al.\u003c/em\u003e Optical vegetation indices for monitoring terrestrial ecosystems globally. \u003cem\u003eNat. Rev. Earth Environ.\u003c/em\u003e \u003cstrong\u003e3\u003c/strong\u003e, 477\u0026ndash;493 (2022).\u003c/li\u003e\n\u003cli\u003eChen, B. \u003cem\u003eet al.\u003c/em\u003e Contrasting inequality in human exposure to greenspace between cities of Global North and Global South. \u003cem\u003eNat. Commun.\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 4636 (2022).\u003c/li\u003e\n\u003cli\u003eWu, S., Chen, B., Webster, C., Xu, B. \u0026amp; Gong, P. Improved human greenspace exposure equality during 21st century urbanization. \u003cem\u003eNat. Commun.\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 6460 (2023).\u003c/li\u003e\n\u003cli\u003eShao, H. \u003cem\u003eet al.\u003c/em\u003e Assessing city-scale green roof development potential using Unmanned Aerial Vehicle (UAV) imagery. \u003cem\u003eUrban For. Urban Green.\u003c/em\u003e \u003cstrong\u003e57\u003c/strong\u003e, 126954 (2021).\u003c/li\u003e\n\u003cli\u003eLi, C. \u003cem\u003eet al.\u003c/em\u003e Evaluating the runoff storage supply-demand structure of green infrastructure for urban flood management. \u003cem\u003eJ. Clean. Prod.\u003c/em\u003e \u003cstrong\u003e280\u003c/strong\u003e, 124420 (2021).\u003c/li\u003e\n\u003cli\u003eConley, G. \u003cem\u003eet al.\u003c/em\u003e Assessing the influence of urban greenness and green stormwater infrastructure on hydrology from satellite remote sensing. \u003cem\u003eSci. Total Environ. \u003c/em\u003e\u003cstrong\u003e817\u003c/strong\u003e, 152723 (2022).\u003c/li\u003e\n\u003cli\u003eSu, Y. \u003cem\u003eet al.\u003c/em\u003e Estimating the cooling effect magnitude of urban vegetation in different climate zones using multi-source remote sensing. \u003cem\u003eUrban Clim.\u003c/em\u003e \u003cstrong\u003e43\u003c/strong\u003e, 101155 (2022).\u003c/li\u003e\n\u003cli\u003eLiu, K., Li, X., Wang, S. \u0026amp; Gao, X. Assessing the effects of urban green landscape on urban thermal environment dynamic in a semiarid city by integrated use of airborne data, satellite imagery and land surface model. \u003cem\u003eInt. J. Appl. Earth Obs. Geoinf.\u003c/em\u003e \u003cstrong\u003e107\u003c/strong\u003e, 102674 (2022).\u003c/li\u003e\n\u003cli\u003eYang, Y. \u0026amp; Davidson, C. I. Green roof aging effect on physical properties and hydrologic performance. \u003cem\u003eJ. Sustain. Water Built Environ.\u003c/em\u003e \u003cstrong\u003e7\u003c/strong\u003e, 04021007 (2021).\u003c/li\u003e\n\u003cli\u003eMitchell, M. E., Emilsson, T. \u0026amp; Buffam, I. Carbon, nitrogen, and phosphorus variation along a green roof chronosequence: implications for green roof ecosystem development. \u003cem\u003eEcol. Eng.\u003c/em\u003e \u003cstrong\u003e164\u003c/strong\u003e, 106211 (2021).\u003c/li\u003e\n\u003cli\u003eBates, A. J., Sadler, J. P. \u0026amp; Mackay, R. Vegetation development over four years on two green roofs in the UK. \u003cem\u003eUrban For. Urban Green.\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 98\u0026ndash;108 (2013).\u003c/li\u003e\n\u003cli\u003eVijayaraghavan, K. Green roofs: a critical review on the role of components, benefits, limitations and trends. \u003cem\u003eRenew. Sustain. Energy Rev.\u003c/em\u003e \u003cstrong\u003e57\u003c/strong\u003e, 740\u0026ndash;752 (2016).\u003c/li\u003e\n\u003cli\u003eMonterusso, M. A., Rowe, D. B. \u0026amp; Rugh, C. L. Establishment and persistence of \u003cem\u003eSedum\u003c/em\u003e spp. and native taxa for green roof applications. \u003cem\u003eHortSci\u003c/em\u003e \u003cstrong\u003e40\u003c/strong\u003e, 391\u0026ndash;396 (2005).\u003c/li\u003e\n\u003cli\u003eBates, A. J., Sadler, J. P., Greswell, R. B. \u0026amp; Mackay, R. Effects of varying organic matter content on the development of green roof vegetation: A six year experiment. \u003cem\u003eEcol. Eng.\u003c/em\u003e \u003cstrong\u003e82\u003c/strong\u003e, 301\u0026ndash;310 (2015).\u003c/li\u003e\n\u003cli\u003eOosterhoorn, M. \u0026amp; Kappelle, M. Vegetation structure and composition along an interior-edge-exterior gradient in a Costa Rican montane cloud forest. \u003cem\u003eFor. Ecol. Manage. \u003c/em\u003e\u003cstrong\u003e126\u003c/strong\u003e, 291\u0026ndash;307 (2000).\u003c/li\u003e\n\u003cli\u003eMagnago, L. F. S., Rocha, M. F., Meyer, L., Martins, S. V. \u0026amp; Meira-Neto, J. A. A. Microclimatic conditions at forest edges have significant impacts on vegetation structure in large Atlantic forest fragments. \u003cem\u003eBiodivers. Conserv.\u003c/em\u003e \u003cstrong\u003e24\u003c/strong\u003e, 2305\u0026ndash;2318 (2015).\u003c/li\u003e\n\u003cli\u003eAbu-Zidan, Y., Nguyen, K. \u0026amp; Mendis, P. Influence of building shape on wind-driven rain exposure in tall buildings. \u003cem\u003eJ. Archit. Eng.\u003c/em\u003e \u003cstrong\u003e27\u003c/strong\u003e, 04021027 (2021).\u003c/li\u003e\n\u003cli\u003eEwers, R. M. \u0026amp; Banks-Leite, C. Fragmentation impairs the microclimate buffering effect of tropical forests. \u003cem\u003ePLoS ONE\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, e58093 (2013).\u003c/li\u003e\n\u003cli\u003eJani, D. M., Mohd, W. M. N. W. \u0026amp; Salleh, S. A. Effects of high-rise residential building shape and height on the urban microclimate in a tropical region. \u003cem\u003eIOP Conf. Ser.: Earth Environ. Sci.\u003c/em\u003e \u003cstrong\u003e767\u003c/strong\u003e, 012031 (2021).\u003c/li\u003e\n\u003cli\u003eGardiner, B., Berry, P. \u0026amp; Moulia, B. Review: Wind impacts on plant growth, mechanics and damage. \u003cem\u003ePlant Sci.\u003c/em\u003e \u003cstrong\u003e245\u003c/strong\u003e, 94\u0026ndash;118 (2016).\u003c/li\u003e\n\u003cli\u003eHoxha, B., Kuriqi, A. \u0026amp; Filkoski, R. V. Influence of seasonal air density fluctuations on wind speed distribution in complex terrains in the context of energy yield. \u003cem\u003eEnerg. Ecol. Environ.\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 175\u0026ndash;187 (2024).\u003c/li\u003e\n\u003cli\u003eLiu, X., Zhou, Y., Xiao, J. \u0026amp; Bao, F. Effects of chilling on the structure, function and development of chloroplasts. \u003cem\u003eFront. Plant Sci.\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 1715 (2018).\u003c/li\u003e\n\u003cli\u003eGao, X., Yang, Z., Han, D., Gao, K. \u0026amp; Zhu, Q. The impact of wind on the rainfall\u0026ndash;runoff relationship in urban high-rise building areas. \u003cem\u003eHydrol. Earth Syst. Sci.\u003c/em\u003e \u003cstrong\u003e25\u003c/strong\u003e, 6023\u0026ndash;6039 (2021).\u003c/li\u003e\n\u003cli\u003eAllegrini, J., Dorer, V. \u0026amp; Carmeliet, J. Influence of morphologies on the microclimate in urban neighbourhoods. \u003cem\u003eJ. Wind. Eng. Ind. Aerodyn.\u003c/em\u003e \u003cstrong\u003e144\u003c/strong\u003e, 108\u0026ndash;117 (2015).\u003c/li\u003e\n\u003cli\u003eLu, L. \u0026amp; Sun, K. Wind power evaluation and utilization over a reference high-rise building in urban area. \u003cem\u003eEnergy Build. \u003c/em\u003e\u003cstrong\u003e68\u003c/strong\u003e, 339\u0026ndash;350 (2014).\u003c/li\u003e\n\u003cli\u003eMaclvor, J. S. Building height matters: nesting activity of bees and wasps on vegetated roofs. \u003cem\u003eIsrael J. Ecol. Evol.\u003c/em\u003e \u003cstrong\u003e62\u003c/strong\u003e, 88\u0026ndash;96 (2016).\u003c/li\u003e\n\u003cli\u003eStarry, O., Lea-Cox, J. D., Kim, J. \u0026amp; van Iersel, M. W. Photosynthesis and water use by two Sedum species in green roof substrate. \u003cem\u003eEnviron. Exp. Bot.\u003c/em\u003e \u003cstrong\u003e107\u003c/strong\u003e, 105\u0026ndash;112 (2014).\u003c/li\u003e\n\u003cli\u003eHoltum, J. A. M. The diverse diaspora of CAM: a pole-to-pole sketch. \u003cem\u003eAnn. Bot.\u003c/em\u003e \u003cstrong\u003e132\u003c/strong\u003e, 597\u0026ndash;625 (2023).\u003c/li\u003e\n\u003cli\u003eCao, J., Hu, S., Dong, Q., Liu, L. \u0026amp; Wang, Z. Green roof cooling contributed by plant species with different photosynthetic strategies. \u003cem\u003eEnergy Build. \u003c/em\u003e\u003cstrong\u003e195\u003c/strong\u003e, 45\u0026ndash;50 (2019).\u003c/li\u003e\n\u003cli\u003eMacIvor, J. S., Margolis, L., Perotto, M. \u0026amp; Drake, J. A. P. Air temperature cooling by extensive green roofs in Toronto Canada. \u003cem\u003eEcol. Eng.\u003c/em\u003e \u003cstrong\u003e95\u003c/strong\u003e, 36\u0026ndash;42 (2016).\u003c/li\u003e\n\u003cli\u003eDvorak, B. \u0026amp; Volder, A. Green roof vegetation for North American ecoregions: A literature review. \u003cem\u003eLandsc. Urban Plan.\u003c/em\u003e \u003cstrong\u003e96\u003c/strong\u003e, 197\u0026ndash;213 (2010).\u003c/li\u003e\n\u003cli\u003eRodriguez Mendez, Q., Fuss, S., L\u0026uuml;ck, S. \u0026amp; Creutzig, F. Assessing global urban CO\u003csub\u003e2\u003c/sub\u003e removal. \u003cem\u003eNat. Cities\u003c/em\u003e \u003cstrong\u003e1\u003c/strong\u003e, 413\u0026ndash;423 (2024).\u003c/li\u003e\n\u003cli\u003eMassaro, E. \u003cem\u003eet al.\u003c/em\u003e Spatially-optimized urban greening for reduction of population exposure to land surface temperature extremes. \u003cem\u003eNat. Commun.\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 2903 (2023).\u003c/li\u003e\n\u003cli\u003eKarteris, M., Theodoridou, I., Mallinis, G., Tsiros, E. \u0026amp; Karteris, A. Towards a green sustainable strategy for Mediterranean cities: assessing the benefits of large-scale green roofs implementation in Thessaloniki, Northern Greece, using environmental modelling, GIS and very high spatial resolution remote sensing data. \u003cem\u003eRenew. Sustain. Energy Rev.\u003c/em\u003e \u003cstrong\u003e58\u003c/strong\u003e, 510\u0026ndash;525 (2016).\u003c/li\u003e\n\u003cli\u003eYin, Y., He, L., Wennberg, P. O. \u0026amp; Frankenberg, C. Unequal exposure to heatwaves in Los Angeles: impact of uneven green spaces. \u003cem\u003eSci. Adv.\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, eade8501 (2023).\u003c/li\u003e\n\u003cli\u003eKim, K. \u003cem\u003eet al.\u003c/em\u003e Inequalities in urban greenness and epigenetic aging: Different associations by race and neighborhood socioeconomic status. \u003cem\u003eSci. Adv.\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, eadf8140 (2023).\u003c/li\u003e\n\u003cli\u003ePan, H. \u003cem\u003eet al.\u003c/em\u003e Contribution of prioritized urban nature-based solutions allocation to carbon neutrality. \u003cem\u003eNat. Clim. Chang.\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 862\u0026ndash;870 (2023).\u003c/li\u003e\n\u003cli\u003eAloysius, N. \u0026amp; Geetha, M. A review on deep convolutional neural networks. in \u003cem\u003e2017 International Conference on Communication and Signal Processing (ICCSP)\u003c/em\u003e 0588\u0026ndash;0592 (IEEE, Chennai, 2017). doi:10.1109/ICCSP.2017.8286426.\u003c/li\u003e\n\u003cli\u003eCity of Toronto. Toronto green roof construction standard: supplementary guidelines. https://www.toronto.ca/wp-content/uploads/2017/08/7eb7-Toronto-Green-Roof-Construction-Standard-Supplementary-Guidelines.pdf (2017).\u003c/li\u003e\n\u003cli\u003eEco-Roof Incentive Program (City of Toronto, accessed 16 August 2024); https://www.toronto.ca/services-payments/water-environment/environmental-grants-incentives/green-your-roof/ \u003c/li\u003e\n\u003cli\u003e\u003cem\u003eGuidelines for the Planning, Construction and Maintenance of Green Roofs\u003c/em\u003e. (Landscape, Research, Development \u0026amp; Construction Society, Bonn, Germany, 2018).\u003c/li\u003e\n\u003cli\u003eD Massing (City Planning, accessed 20 June 2024); https://open.toronto.ca/dataset/3d-massing/\u003c/li\u003e\n\u003cli\u003eHalim, M. A., Liao, W., Karim, M. R., Kayes, I. \u0026amp; Thomas, S. Unsupervised shadow detection for vegetated urban landscapes from RGB aerial images. (in preparation).\u003c/li\u003e\n\u003cli\u003eSilva, G. F., Carneiro, G. B., Doth, R., Amaral, L. A. \u0026amp; Azevedo, D. F. G. de. Near real-time shadow detection and removal in aerial motion imagery application. \u003cem\u003eISPRS J. Photogramm. Remote Sens.\u003c/em\u003e \u003cstrong\u003e140\u003c/strong\u003e, 104\u0026ndash;121 (2018).\u003c/li\u003e\n\u003cli\u003eSmith, G. M. \u0026amp; Milton, E. J. The use of the empirical line method to calibrate remotely sensed data to reflectance. \u003cem\u003eInt. J. Remote Sens.\u003c/em\u003e \u003cstrong\u003e20\u003c/strong\u003e, 2653\u0026ndash;2662 (1999).\u003c/li\u003e\n\u003cli\u003eLu, B. \u0026amp; He, Y. Species classification using Unmanned Aerial Vehicle (UAV)-acquired high spatial resolution imagery in a heterogeneous grassland. \u003cem\u003eISPRS J. Photogramm. Remote Sens.\u003c/em\u003e \u003cstrong\u003e128\u003c/strong\u003e, 73\u0026ndash;85 (2017).\u003c/li\u003e\n\u003cli\u003eIqbal, F., Lucieer, A. \u0026amp; Barry, K. Simplified radiometric calibration for UAS-mounted multispectral sensor. \u003cem\u003eEur. J. Remote Sens.\u003c/em\u003e \u003cstrong\u003e51\u003c/strong\u003e, 301\u0026ndash;313 (2018).\u003c/li\u003e\n\u003cli\u003eRouse, W. \u0026amp; Haas, R. H. Monitoring vegetation systems in the Great Plains with ERTS. in \u003cem\u003eThird ERTS-1 Symposium\u003c/em\u003e 309\u0026ndash;317 (Third ERTS-1 Symposium, Washington, DC: NASA., 1974).\u003c/li\u003e\n\u003cli\u003eHuete, A. R. A soil-adjusted vegetation index (SAVI). \u003cem\u003eRemote Sens. Environ. \u003c/em\u003e\u003cstrong\u003e25\u003c/strong\u003e, 295\u0026ndash;309 (1988).\u003c/li\u003e\n\u003cli\u003eGitelson, A. A. \u0026amp; Merzlyak, M. N. Remote sensing of chlorophyll concentration in higher plant leaves. \u003cem\u003eAdv. Space Res.\u003c/em\u003e \u003cstrong\u003e22\u003c/strong\u003e, 689\u0026ndash;692 (1998).\u003c/li\u003e\n\u003cli\u003eGitelson, A. A., Gritz \u0026dagger;, Y. \u0026amp; Merzlyak, M. N. Relationships between leaf chlorophyll content and spectral reflectance and algorithms for non-destructive chlorophyll assessment in higher plant leaves. \u003cem\u003eJ. Plant Physiol.\u003c/em\u003e \u003cstrong\u003e160\u003c/strong\u003e, 271\u0026ndash;282 (2003).\u003c/li\u003e\n\u003cli\u003eBates, D., M\u0026auml;chler, M., Bolker, B. \u0026amp; Walker, S. Fitting linear mixed-effects models using lme4. \u003cem\u003eJ. Stat. Softw.\u003c/em\u003e \u003cstrong\u003e67\u003c/strong\u003e, 1\u0026ndash;48 (2015).\u003c/li\u003e\n\u003cli\u003eMuggeo, V. M. R. Segmented: an R package to fit regression models with broken-line relationships. \u003cem\u003eR news\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 20\u0026ndash;25 (2008).\u003c/li\u003e\n\u003cli\u003eCity of Toronto. Tall Building Design Guidelines. (2013).\u003c/li\u003e\n\u003cli\u003eBenjamini, Y. \u0026amp; Hochberg, Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. \u003cem\u003eJ. R. Stat. Soc. Ser. B Methodol.\u003c/em\u003e \u003cstrong\u003e57\u003c/strong\u003e, 289\u0026ndash;300 (1995).\u003c/li\u003e\n\u003cli\u003eR Core Team. \u003cem\u003eR: A Language and Environment for statistical computing\u003c/em\u003e (R Foundation for Statistical Computing, 2024); https://www.R-project.org/\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"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-6049604/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6049604/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eGreen roofs have been increasingly implemented in cities globally to enhance urban ecosystem services degraded by climate change and rapid urbanization. However, temporal trends in green roof vegetation health and the effects of design considerations at a large scale remain unclear. Here, we used 8-cm very-high-resolution multispectral remote sensing imagery to quantify the temporal changes of vegetation health and associated design drivers across 1,380 individual green roof modules in Toronto from 2011 to 2018. Results show an average increase in vegetation health and a decline in vegetation patchiness as green roofs age. We identify module area, building height, and vegetation type as primary design factors influencing green roof vegetation health, with module area positively and building height inversely affecting vegetation health. In terms of vegetation type, sedum mats are generally healthier than woody plants and grasses on green roofs. Additionally, we identify specific thresholds, module sizes with linear dimensions of 3.2\u0026ndash;4.8 m and building heights of 14.4 m, for which smaller and higher green roof performance abruptly declines. These findings present a robust, cost-effective analytical framework for long-term assessment and modeling of urban green infrastructure at large scales, providing valuable insights into urban greening practices.\u003c/p\u003e","manuscriptTitle":"Variation in green roof vegetation health driven by age and design","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-02-28 11:23:01","doi":"10.21203/rs.3.rs-6049604/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"nature-cities","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"natcities","sideBox":"Learn more about [Nature Cities](https://www.springer.com/journal/44284)","snPcode":"44284","submissionUrl":"","title":"Nature Cities","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"dd2d006a-601c-473f-b119-9ead7a24e7f7","owner":[],"postedDate":"February 28th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":44885425,"name":"Earth and environmental sciences/Ecology"},{"id":44885426,"name":"Earth and environmental sciences/Environmental sciences"},{"id":44885427,"name":"Biological sciences/Plant sciences/Plant ecology"}],"tags":[],"updatedAt":"2025-10-09T07:07:14+00:00","versionOfRecord":{"articleIdentity":"rs-6049604","link":"https://doi.org/10.1038/s44284-025-00331-w","journal":{"identity":"nature-cities","isVorOnly":false,"title":"Nature Cities"},"publishedOn":"2025-10-08 04:00:00","publishedOnDateReadable":"October 8th, 2025"},"versionCreatedAt":"2025-02-28 11:23:01","video":"","vorDoi":"10.1038/s44284-025-00331-w","vorDoiUrl":"https://doi.org/10.1038/s44284-025-00331-w","workflowStages":[]},"version":"v1","identity":"rs-6049604","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6049604","identity":"rs-6049604","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}