Ecology

Forbes, B. C. & Kumpula, T. The ecological role and geography of reindeer (Rangifer tarandus) in Northern Eurasia. Geogr. Compass 3, 1356–1380 (2009).Article 

Google Scholar 
Post, E. & Pedersen, C. Opposing plant community responses to warming with and without herbivores. Proc. Natl Acad. Sci. USA 105, 12353–12358 (2008).Article 
CAS 

Google Scholar 
Berkes, F., Colding, J. & Folke, C. Navigating Social-Ecological Systems: Building Resilience for Complexity and Change (Cambridge Univ. Press, 2003).Tremblay, R., Landry-Cuerrier, M. & Humphries, M. M. Culture and the social-ecology of local food use by Indigenous communities in northern North America. Ecol. Soc. 25, 8 (2020).Kenny, T.-A., Fillion, M., Simpkin, S., Wesche, S. & Chan, L. Caribou (Rangifer tarandus) and Inuit nutrition security in Canada. Ecohealth 15, 590–607 (2018).Article 

Google Scholar 
Benson, K. Gwich’in Knowledge of Porcupine Caribou: State of Current Knowledge and Gaps Assessment (Department of Cultural Heritage, Gwich’in Tribal Council, 2019); https://thelastgreatherd.com/wp-content/uploads/2020/06/GTC-current-knowledge-and-gaps-assessment.pdfParlee, B. & Caine, K. When the Caribou Do Not Come: Indigenous Knowledge and Adaptive Management in the Western Arctic (UBC Press, 2018).Herds: Status of Herds (CircumArctic Rangifer Monitoring and Assessment Network, accessed 3 November 2021); https://carma.caff.is/herdsFesta-Bianchet, M., Ray, J. C., Boutin, S., Côté, S. D. & Gunn, A. Conservation of caribou (Rangifer tarandus) in Canada: an uncertain future. Can. J. Zool. 89, 419–434 (2011).Article 

Google Scholar 
Gunn, A. Voles, lemmings and caribou: population cycles revisited? Rangifer 23, 105–111 (2003).Article 

Google Scholar 
Ferguson, M. A. D., Williamson, R. G. & Messier, F. Inuit knowledge of long-term changes in a population of Arctic tundra caribou. Arctic 51, 201–219 (1998).Article 

Google Scholar 
Beaulieu, D. Dene traditional knowledge about caribou cycles in the Northwest Territories. Rangifer 32, 59–67 (2012).Article 

Google Scholar 
Mallory, C. D. & Boyce, M. S. Observed and predicted effects of climate change on Arctic caribou and reindeer. Environ. Rev. 26, 13–25 (2018).Article 

Google Scholar 
Uboni, A. et al. Long-term trends and role of climate in the population dynamics of Eurasian reindeer. PLoS ONE 11, e0158359 (2016).Article 

Google Scholar 
Chapin, F. S. III et al. Directional changes in ecological communities and social-ecological systems: a framework for prediction based on Alaskan examples. Am. Nat. 168, S36–S49 (2006).Article 

Google Scholar 
Tengö, M. et al. Weaving knowledge systems in IPBES, CBD and beyond – lessons learned for sustainability. Curr. Opin. Environ. Sustain. 26, 17–25 (2017).Article 

Google Scholar 
Berkes, F. Sacred Ecology 4th edn (Routledge, 2018).Stuart Chapin, F. III et al. Earth stewardship: science for action to sustain the human-earth system. Ecosphere 2, 89 (2011).Parlee, B. L., Sandlos, J. & Natcher, D. C. Undermining subsistence: barren-ground caribou in a ‘tragedy of open access’. Sci. Adv. 4, e1701611 (2018).Article 

Google Scholar 
Johnson, J. T. et al. Weaving Indigenous and sustainability sciences to diversify our methods. Sustain. Sci. 11, 1–11 (2016).Article 
CAS 

Google Scholar 
Reid, A. J. et al. ‘Two-eyed seeing’: an Indigenous framework to transform fisheries research and management. Fish Fish. 22, 243–261 (2021).Article 

Google Scholar 
Tengö, M., Brondizio, E. S., Elmqvist, T., Malmer, P. & Spierenburg, M. Connecting diverse knowledge systems for enhanced ecosystem governance: the multiple evidence base approach. AMBIO 43, 579–591 (2014).Article 

Google Scholar 
Aminpour, P. et al. The diversity bonus in pooling local knowledge about complex problems. Proc. Natl Acad. Sci. USA 118, e2016887118 (2021).Article 
CAS 

Google Scholar 
Henri, D. A. et al. Weaving Indigenous knowledge systems and Western sciences in terrestrial research, monitoring and management in Canada: a protocol for a systematic map. Ecol. Solut. Evid. 2, e12057 (2021).Article 

Google Scholar 
Ljubicic, G. J., Mearns, R., Okpakok, S. & Robertson, S. Nunami iliharniq (learning from the land): reflecting on relational accountability in land-based learning and cross-cultural research in Uqšuqtuuq (Gjoa Haven, Nunavut). Arct. Sci. 8, 252–291 (2022).Article 

Google Scholar 
Stern, E. R. & Humphries, M. M. Interweaving local, expert, and Indigenous knowledge into quantitative wildlife analyses: a systematic review. Biol. Conserv. 266, 109444 (2022).Article 

Google Scholar 
Bourgeon, L., Burke, A. & Higham, T. Earliest human presence in North America dated to the last glacial maximum: new radiocarbon dates from Bluefish Caves, Canada. PLoS ONE 12, e0169486 (2017).Article 

Google Scholar 
Kuhnlein, H. V., McDonald, M., Spigelski, D., Vittrekwa, E. & Erasmus, B. in Indigenous Peoples’ Food Systems: the Many Dimensions of Culture, Diversity and Environment for Nutrition and Health (eds Kuhnlein, H. V. et al.) Ch. 3 (FAO, Centre for Indigenous Peoples’ Nutrition and Environment, 2009).Porcupine Caribou Technical Committee. Porcupine Caribou Annual Summary Report 2018–2019 (Porcupine Caribou Management Board, Whitehorse, Yukon, 2019); https://pcmb.ca/wp-content/uploads/2020/06/PCH_annual_summ_report_Nov29_2019_FINAL.pdfIPCC. Climate Change 2013: The Physical Science Basis (eds Stocker, T. F. et al.) (Cambridge Univ. Press, 2013).Zhang, X. et al. in Canada’s Changing Climate Report (eds Bush, E. & Lemmen, D. S.) Ch. 4 (Government of Canada, 2019).Griffith, B. et al. in Arctic Refuge Coastal Plain Terrestrial Wildlife Research Summaries Biological Science Report USGS/BRD BSR-2002-0001 (eds Douglas, D. C. et al.) 8–37 (US Geological Survey, 2002).Russell, D. & Gunn, A. Vulnerability Analysis of the Porcupine Caribou Herd to Potential Development of the 1002 lands in the Arctic National Wildlife Refuge, Alaska (Environment Yukon, Canadian Wildlife Service and GNWT Department of Environment and Natural Resources, 2019); https://pcmb.ca/wp-content/uploads/2021/10/Russell-and-Gunn-PCH-vulnerability-analysis-2019.pdfKruse, J. A. et al. Modeling sustainability of Arctic communities: an interdisciplinary collaboration of researchers and local knowledge holders. Ecosystems 7, 815–828 (2004).Berman, M., Nicolson, C., Fofinas, G., Tetlichi, J. & Martin, S. Adaptation and sustainability in a small Arctic community: results of an agent-based simulation model. Arctic 57, 401–414 (2004).Article 

Google Scholar 
Kofinas, G., Aklavik, Arctic Village, Old Crow & Fort McPherson. in The Earth is Faster Now: Indigenous Observations of Arctic Environmental Change (eds Krupnik, I. & Jolly, D.) 55–91 (Arctic Research Consortium of the United States, 2002).Eamer, J. in Bridging Scales and Knowledge Systems: Concepts and Applications in Ecosystem Assessment (eds Reid, W. V. et al.) 185–206 (Island Press, 2006).Shipley, B. Cause and Correlation in Biology: a User’s Guide to Path Analysis, Structural Equations and Causal Inference with R 2nd edn (Cambridge Univ. Press, 2016).Parlee, B. & Furgal, C. Well-being and environmental change in the Arctic: a synthesis of selected research from Canada’s International Polar Year program. Clim. Change 115, 13–34 (2012).Article 

Google Scholar 
Kofinas, G. P. The Costs of Power Sharing: Community Involvment in Canadian Porcuine Caribou Co-management. PhD thesis, Univ. of British Columbia (1998).Ford, J. D. et al. Including indigenous knowledge and experience in IPCC assessment reports. Nat. Clim. Change 6, 349–353 (2016).Article 

Google Scholar 
Brinkman, T. J. et al. Arctic communities perceive climate impacts on access as a critical challenge to availability of subsistence resources. Clim. Change 139, 413–427 (2016).Article 

Google Scholar 
McNeil, P., Russell, D. E., Griffith, B., Gunn, A. & Kofinas, G. Where the wild things are: seasonal variation in caribou distribution in relation to climate change. Rangifer 25, 51–63 (2005).Berman, M. & Kofinas, G. Hunting for models: grounded and rational choice approaches to analyzing climate effects on subsistence hunting in an Arctic community. Ecol. Econ. 49, 31–46 (2004).Article 

Google Scholar 
Hansen, B. B. et al. Climate events synchronize the dynamics of a resident vertebrate community in the High Arctic. Science 339, 313–315 (2013).Article 
CAS 

Google Scholar 
Collings, P., Marten, M. G., Pearce, T. & Young, A. G. Country food sharing networks, household structure, and implications for understanding food insecurity in Arctic Canada. Ecol. Food Nutr. 55, 30–49 (2016).Article 

Google Scholar 
BurnSilver, S., Magdanz, J., Stotts, R., Berman, M. & Kofinas, G. Are mixed economies persistent or transitional? Evidence using social networks from Arctic Alaska. Am. Anthropol. 118, 121–129 (2016).Article 

Google Scholar 
Baggio, J. A. et al. Multiplex social ecological network analysis reveals how social changes affect community robustness more than resource depletion. Proc. Natl Acad. Sci. USA 113, 13708–13713 (2016).Article 
CAS 

Google Scholar 
Gagnon, C. A. et al. Merging Indigenous and scientific knowledge links climate with the growth of a large migratory caribou population. J. Appl. Ecol. 57, 1644–1655 (2020).Article 

Google Scholar 
Houde, N. The six faces of traditional ecological knowledge: challenges and opportunities for Canadian co-management arrangements. Ecol. Soc. 12, 34 (2007).Article 

Google Scholar 
Fancy, S. G., Pank, L. F., Whitten, K. R. & Regelin, W. L. Seasonal movements of caribou in Arctic Alaska as determined by satellite. Can. J. Zool. 67, 644–650 (1989).Article 

Google Scholar 
Porcupine Caribou Technical Committee. Porcupine Caribou Annual Summary Report 2014 (Porcupine Caribou Management Board, Whitehorse, Yukon, 2014); https://pcmb.ca/wp-content/uploads/2021/07/PCH_annual_summ_report_2014_2015_NOV19_FINAL.pdfEastland, W. G. Influence of Weather on Movements and Migrations of Caribou. PhD thesis, Univ. of Alaska (1991).Tyler, N. J. C. Climate, snow, ice, crashes, and declines in populations of reindeer and caribou (Rangifer tarandus L.). Ecol. Monogr. 80, 197–219 (2010).Article 

Google Scholar 
Hansen, B. B., Aanes, R. & Saether, B. E. Feeding-crater selection by high-arctic reindeer facing ice-blocked pastures. Can. J. Zool. 88, 170–177 (2010).Article 

Google Scholar 
Solberg, E. J. et al. Effects of density-dependence and climate on the dynamics of a Svalbard reindeer population. Ecography 24, 441–451 (2001).Article 

Google Scholar 
Hansen, B. B., Aanes, R., Herfindal, I., Kohler, J. & Sæther, B.-E. Climate, icing, and wild arctic reindeer: past relationships and future prospects. Ecology 92, 1917–1923 (2011).Article 

Google Scholar 
Langlois, A. et al. Detection of rain-on-snow (ROS) events and ice layer formation using passive microwave radiometry: a context for Peary caribou habitat in the Canadian Arctic. Remote Sens. Environ. 189, 84–95 (2017).Article 

Google Scholar 
Russell, D. E., Gunn, A. & White, R. G. CircumArctic collaboration to monitor caribou and wild reindeer. Arctic 68, 6–10 (2015).Article 

Google Scholar 
Russell, D. E. et al. CARMA’s MERRA-based caribou range climate database. Rangifer 33, 145–152 (2013).Article 

Google Scholar 
ArcGIS version 10 (Environmental Systems Resource Institute, 2010).Cai, J., Russell, D. & Whitfield, P. Methodology and Algorithms for Constructing CARMA Bio-climate Tables (Simon Fraser Univ., 2011).Stenseth, N. C. & Mysterud, A. Weather packages: finding the right scale and composition of climate in ecology. J. Anim. Ecol. 74, 1195–1198 (2005).Article 

Google Scholar 
Pebesma, E. J. & Bivand, R. S. Classes and methods for spatial data in R. R News 5, 9–13 (2005).Bivand, R. S., Pebesma, E. J. & Gomez-Rubio, V. Applied Spatial Data Analysis with R 2nd edn (Springer, 2013).Bivand, R. S., Keitt, T. & Rowlingson, B. Rgdal: Bindings for the Geospatial Data Abstraction Library. R package version 0.8-16 (R Foundation for Statistical Computing, 2014); https://cran.r-project.org/web/packages/rgdal/index.htmlBivand, R. S. & Rundel, C. Rgeos: Interface to Geometry Engine – Open Source (GEOS). R package version 0.3-4 (R Foundation for Statistical Computing, 2014); https://cran.r-project.org/web/packages/rgeos/index.htmlLefcheck, J. S. piecewise SEM: piecewise structural equation modelling in R for ecology, evolution and systematics. Methods Ecol. Evol. 7, 573–579 (2016).Article 

Google Scholar 
Shipley, B. Confirmatory path analysis in a generalized multilevel context. Ecology 90, 363–368 (2009).Article 

Google Scholar 
Thomas, D. W. et al. Common paths link food abundance and ectoparasite loads to physiological performance and recruitment in nestling blue tits. Funct. Ecol. 21, 947–955 (2007).Article 

Google Scholar 
Shipley, B. The AIC model selection method applied to path analytic models compared using a d-separation test. Ecology 94, 560–564 (2013).Article 

Google Scholar 
Burnham, K. P. & Anderson, D. R. Model Selection and Multimodel Inference: a Practical Information-Theoretic Approach (Springer, 2002).Schielzeth, H. Simple means to improve the interpretability of regression coefficients. Methods Ecol. Evol. 593, 103–113 (2010).Article 

Google Scholar  More

Goal and scope of the studyThe goal of this study was to assess the planetary footprints of GHG mitigation strategies for the global production of plastics. To calculate planetary footprints, we apply LCA in combination with the planetary boundaries framework as proposed by ref. 22. As GHG mitigation strategies, we consider recycling, bio-based production and production via CCU, and compare their planetary footprints to the planetary footprints of fossil-based plastics. We use a bottom-up model covering >90% of global plastic production for 2030 (and 2050, Supplementary Information, section 3). The bottom-up model builds on the plastic production system from ref. 10 and includes plastic production, the supply chain and the disposal of plastics at the end of life.Functional unitIn LCA, the functional unit quantifies the functions of the investigated product system. In this study, the function of the product system is the production and disposal of >90% of global plastics. To cover >90% of global plastics, we define the functional unit as the yearly global production and disposal of 14 large-volume plastics (summarized in Supplementary Table 5). We estimated the yearly production volumes for 2030 and 2050 based on the production volumes in 2015 and the annual growth rates shown in Supplementary Table 5.Our assessment includes plastic disposal. However, the production and disposal of plastics do not necessarily occur in the same year. For instance, while polyolefins used for plastic packaging have an average lifetime of 6 months, the average lifetime of polyurethane used in construction is 35 years11. Including the lifetime of plastics, and hence, the temporal difference between production and disposal, would lead to an increasing plastic stock. An increasing stock, in turn, represents a carbon sink during the production year that appears to enable the production of net-negative GHG emission plastics based on biomass or CCU. However, the plastic stock is not a permanent carbon sink, which would be required for producing net-negative GHG emission plastics55. To avoid misleading conclusions about net-negative bio- and CCU-based plastics, we assign the planetary footprints from disposal to the year of plastic production. Thereby, we conservatively assess the planetary footprints of plastics.In addition, we address the challenge highlighted in ref. 56 that the increasing demand for plastics renders determining the absolute sustainability of plastics difficult. We meet this challenge by assuming a steady-state production system with a recurring functional unit in the same amount every year. We thereby analyse discrete scenarios with constant consumption levels for plastics. Therefore, our conclusions depend on the accuracy of the demand forecasts and apply only to the production volumes considered.System boundariesWe use cradle-to-grave system boundaries, including plastic production and supply chain, potential recycling and final disposal at the end of life. Assessing the use phase of plastics is not possible because of a lack of data. The versatile properties of plastics result in a wide range of applications that cannot be represented in a single study. Furthermore, it would be necessary to consider not only the emissions of the use phase (probably relatively small) but also the system-wide environmental consequences of using plastics in each application compared to other materials. Thus, a consequential assessment of the plastic use phase is desirable but beyond the scope of this study.The plastics supply chain includes several intermediate chemicals such as monomers, solvents or other reactants. The bottom-up model covers the production of all intermediate chemicals in the foreground system. As a background system, we use aggregated datasets from the LCA database ecoinvent. A list of all intermediate chemicals and all aggregated datasets can be found in Supplementary Information, section 1. In addition, the foreground system of the bottom-up model does not include environmental impacts from infrastructure and transportation because of a lack of data. However, we consider the environmental impacts of infrastructure and transportation from other industrial sectors by aggregated datasets, for example, from electricity generation and biomass cultivation.The bottom-up model includes the best available fossil-based technologies and the following technologies for plastic disposal and virgin production based on biomass and CCU.Plastic waste disposalThe bottom-up model includes three options for plastic waste disposal: landfilling, incineration with energy recovery and recycling. Plastic waste can occur in several forms: as sorted fraction of municipal solid waste, as mixed plastics and residues from sorting, and as residues from mechanical recycling. For all fractions, we include waste incineration with energy recovery and landfilling.Landfilled plastic waste is assumed to degrade by approximately 1% of the contained carbon, which is in line with the ecoinvent database45. Mechanical recycling is only modelled for sorted fractions of packaging waste owing to impurities of mixed and non-packaging wastes. In contrast, chemical recycling can be applied to all plastic fractions. In this study, we model chemical recycling as pyrolysis to refinery feedstock, that is, naphtha. The pyrolysis has yields of 29 to 69% depending on the type of plastic (details in Supplementary Information, section 1). Furthermore, we include options for chemical recycling of plastic waste to monomers, which are still early-stage technologies. To derive the minimal necessary recycling rate in Fig. 5, we apply an optimistic scenario with a 95% yield of chemical recycling processes following common modelling in life-cycle inventories of chemicals (Supplementary Information, section 3)57. All calculations are constrained to maximum recycling rates of 94% as the remaining 6% are assumed to be the minimal landfilling rate until the middle of the century11. The assumption is based on historical trends in end-of-life treatment of plastics.Bio-based productionBio-based GHG mitigation is frequently discussed in the literature and is often associated with competition with the food industry58. To avoid competition with the food industry, the bottom-up model is restricted to lignocellulosic biomass as feedstock, that is, energy crops, forest residues and by-products from other industrial biomass processes (for example, bagasse). In this study, unless mentioned otherwise, we model biomass as energy crops because of their potential for large-scale application (Supplementary Information, section 3). However, we conduct a sensitivity analysis for other lignocellulosic biomass sources to assess the sustainability of bio-based plastics in more detail.For each biomass type, we account for the carbon uptake during the biomass growth phase by giving a credit corresponding to the biomass carbon content. We do not consider land use change emissions as current literature lacks an assessment of land use change effects on other Earth-system processes besides climate change.For biomass processing, we include the following high-maturity processes: gasification to syngas and fermentation to ethanol, and the subsequent conversion to methanol and ethylene (Supplementary Table 1). Methanol and ethylene can be further converted to propylene and aromatics, which all together represent the building blocks for all plastics in this study.CCU-based productionCCU-based plastic production particularly requires CO2 and hydrogen. For CO2 supply, we consider CO2 capture from highly concentrated point sources within the plastics supply chain. Highly concentrated point sources include the conventional fossil-based processes, ammonia production, steam methane reforming, ethylene oxide production, the bio-based processes for ethanol and syngas, and plastic waste incineration. Capturing from processes within the plastics supply chain is limited by the amount of CO2 emitted by these processes and avoids the corresponding emissions. For these processes, we considered the energy demand for compressing the CO2 with 0.4 MJ of electricity59. For waste incineration, we consider a decrease in energy output when capturing CO2. All further CO2 sources are conservatively approximated by direct air capture. For 1 kg CO2 captured via direct air capture, we include an uptake of 1 kg of CO2 equivalent while considering the energy demand of 1.29 MJ electricity and 4.19 MJ heat60.Hydrogen for CCU is produced by water electrolysis, with an overall efficiency of 67%61. Previous studies have already shown that renewable electricity is required for CCU to be environmentally beneficial13. Thus, we conduct a sensitivity analysis for multiple electricity technologies to assess their influence on the sustainability of CCU-based plastics (Supplementary Information).For CCU-based production, we include high-maturity technologies, such as CO2-based methanol and methane, as well as subsequent production of olefins and aromatics (Supplementary Table 1). We do not consider CCS as an additional scenario, as fossil resources and storage capacities are ultimately limited. Therefore, CCS may serve as an interim solution for GHG mitigation but stands in contrast to long-term sustainability as the goal of this study.Pathway definitionWe assess nine pathways for the plastics industry towards sustainability. Pathway 1 is fossil-based plastic production (current recycling rate of 23%) that serves as a reference. We also include two pathways that combine all circular technologies: Pathway 2, which minimizes the climate change impact (climate-optimal), and pathway 3, which minimizes the maximal transgression of the share of SOS of the plastics industry (balanced) (Fig. 2). To assess the impact of switching from fossil to renewable feedstocks, we introduce pathway 4, which is bio-based, and pathway 5, which is CCU-based (Fig. 3). Pathways 4 and 5 include the current recycling rate of 23%. In addition, we introduce three pathways with the maximum recycling rates of 94%: pathway 6, in which the remaining virgin production is based on fossil resources; pathway 7, in which it is based on biomass; and pathway 8, in which it is based on CO2 (Fig. 3). Pathway 9 combines biomass, CCU and recycling, and additionally includes chemical recycling of polymers to monomers to calculate the minimal recycling rate to achieve sustainable plastics (Fig. 5).The planetary boundaries frameworkWe follow the recommendations for absolute environmental sustainability assessment in ref. 29 and choose the planetary boundaries framework for the assessment. The planetary boundaries framework suits the goal of the study best because of its precautionary principles for the definition of environmental thresholds, the SOS. We assess eight of the nine Earth-system processes suggested in ref. 21, namely, climate change, ocean acidification, changes in biosphere integrity, the biogeochemical flow of nitrogen and phosphorus (referred to as N cycle and P cycle), aerosol loading, freshwater use, stratospheric ozone depletion, and land-system change. We do not assess the Earth-system process of novel entities since neither control variables nor the boundary itself is yet adequately defined22. We consider the global boundaries for the Earth-system processes in line with the scope of this study. These global boundaries and the corresponding calculation of planetary footprints are subject to assumptions and thus incorporate uncertainty (Supplementary Information, section 2).For the two subprocesses for climate change (namely, atmospheric CO2 concentration and energy imbalance at the top-of-atmosphere), we only consider the energy imbalance at the top-of-atmosphere quantified by radiative forcing. We focus on radiative forcing, as the control variable is more inclusive and fundamental, and the global limits are stricter than for atmospheric CO2 concentration21. Thereby, we conservatively assess climate change.Biosphere integrity is divided into functional and genetic diversity of species. Preserving functional diversity ensures a stable ecosystem by maintaining all ecosystem services. We assess the functional diversity of species using the method proposed in ref. 18. The method covers the mean species abundance loss caused by the two main stressors, direct land use and GHG emissions, as a proxy for the biodiversity intactness index. Genetic diversity provides the long-term ability of the biosphere to persist under and adapt to gradual changes of the environment21. Genetic diversity is often approximated by the global extinction rate. However, using the global extinction rate does not fully cover variation of genetic composition, resulting in high uncertainties when quantifying genetic diversity18. Thus, we focus on functional diversity.Downscaling of the safe operating spaceAs the plastics industry accounts for only a fraction of all human activities, we assign a share of the SOS to plastics. The plastics industry should operate within its assigned share to be considered environmentally sustainable. To assign a share of SOS to the plastics industry, we apply utilitarian downscaling principles. Utilitarian downscaling principles are tailored to maximize welfare in society29. We approximate welfare by consumption expenditure on plastics as an economic indicator for consumer preferences and human needs62. An extensive discussion on the other downscaling principles and their implications can be found in Supplementary Information.Although the final consumption expenditures on plastics are negligible, the industry consumes plastics to produce other goods and services. Accordingly, plastics are produced mostly in the upstream supply chain to support the final consumption of other goods and services. Thus, consuming other goods and services induces plastic production. To account for this inducement of plastic production, we used the total global plastic production xplastics to represent the global intermediate and final consumption expenditure on plastics. For this purpose, we use the gross output vector x of the product-by-product input–output table of EXIOBASE for the year 2020 (ref. 63). To calculate the share of SOS of the plastics industry, we divide the total global plastic production xplastics by the gross world product. The gross world product equals the total global final consumption expenditure. Analogously, we also consider the end-of-life treatment of plastics to be consistent with the system boundaries of the environmental assessment.We estimate the share of SOS for the plastics industry for 2030 and 2050 based on data for the year 2020. Accordingly, we assume that the market share of the plastics industry and, therefore, its share of SOS do not change in the coming years despite the increasing production volume of plastics. Thereby, we implicitly assume that all industries grow equally economically. Alternatively, economic forecasting models could estimate future market shares of plastics. However, applying economic forecasting models is complex, and the results would still be highly uncertain, especially if industry pursues low-carbon technology pathways. Therefore, estimating future market shares is beyond the scope of this study.Technology choice modelTo calculate the planetary footprint of plastics, we use a bottom-up model of the plastics industry. The model builds on the technology choice model (TCM) that allows for linear optimization of production systems27. The TCM represents the production system based on the following elements: technologies, intermediate flows, elementary flows and final demands. Ref. 27 describes each element in detail.The TCM is based on the established computational structure of LCA64. This structure arranges the data that represent the physical production system in the technology matrix A and the elementary flow matrix B. In the technology matrix A, columns represent technologies, and rows represent intermediate flows. Therefore, the coefficient aij of the A matrix corresponds to an intermediate flow i that is either produced (aij  > 0) or consumed (aij  More

Higham, T. E. et al. Linking ecomechanical models and functional traits to understand phenotypic diversity. Trends Ecol. Evol. 36, 860–873 (2021).Article 
PubMed 

Google Scholar 
Kearney, M. R., Jusup, M., McGeoch, M. A., Kooijman, S. A. & Chown, S. L. Where do functional traits come from? The role of theory and models. Funct. Ecol. 35, 1385–1396 (2021).Article 
CAS 

Google Scholar 
Mayr, E. Geographical character gradients and climatic adaptation. Evolution 10, 105–108 (1956).Article 

Google Scholar 
Gaston, K. J., Chown, S. L. & Evans, K. L. Ecogeographical rules: elements of a synthesis. J. Biogeogr. 35, 483–500 (2008).Article 

Google Scholar 
Chown, S. L. & Gaston, K. J. Macrophysiology for a changing world. Proc. R. Soc. B 275, 1469–1478 (2008).Article 
PubMed 
PubMed Central 

Google Scholar 
Rubalcaba, J. G. & Jimeno, B. Biophysical models unravel associations between glucocorticoids and thermoregulatory costs across avian species. Funct. Ecol. 36, 64–72 (2022).Article 
CAS 

Google Scholar 
Anderson, R. O., White, C. R., Chapple, D. G. & Kearney, M. R. A hierarchical approach to understanding physiological associations with climate. Glob. Ecol. Biogeogr. 31, 332–346 (2022).Article 

Google Scholar 
Angilletta, M. J. Jr, Niewiarowski, P. H. & Navas, C. A. The evolution of thermal physiology in ectotherms. J. Therm. Biol. 27, 249–268 (2002).Article 

Google Scholar 
Olalla‐Tárraga, M. Á., Rodríguez, M. Á. & Hawkins, B. A. Broad‐scale patterns of body size in squamate reptiles of Europe and North America. J. Biogeogr. 33, 781–793 (2006).Article 

Google Scholar 
Amado, T., Moreno Pinto, M. G. & Olalla‐Tárraga, M. Á. Anuran 3D models reveal the relationship between surface area‐to‐volume ratio and climate. J. Biogeogr. 46, 1429–1437 (2019).
Google Scholar 
Castro, K. M. S. A. et al. Water constraints drive allometric patterns in the body shape of tree frogs. Sci. Rep. 11, 1218 (2021).Clusella-Trullas, S., Terblanche, J. S., Blackburn, T. M. & Chown, S. L. Testing the thermal melanism hypothesis: a macrophysiological approach. Funct. Ecol. 22, 232–238 (2008).Ghalambor, C. K., Huey, R. B., Martin, P. R., Tewksbury, J. J. & Wang, G. Are mountain passes higher in the tropics? Janzen’s hypothesis revisited. Integr. Comp. Biol. 46, 5–17 (2006).Article 
PubMed 

Google Scholar 
Bennett, J. M. et al. The evolution of critical thermal limits of life on Earth. Nat. Commun. 12, 1198 (2021).Sunday, J. M. et al. Thermal-safety margins and the necessity of thermoregulatory behavior across latitude and elevation. Proc. Natl Acad. Sci. USA 111, 5610–5615 (2014).Article 
CAS 
PubMed 
PubMed Central 

Google Scholar 
Muñoz, M. M. The Bogert effect, a factor in evolution. Evolution 76, 49–66 (2021).Article 
PubMed 

Google Scholar 
Bogert, C. M. Thermoregulation in reptiles, a factor in evolution. Evolution 3, 195–211 (1949).Article 
CAS 
PubMed 

Google Scholar 
Huey, R. B., Hertz, P. E. & Sinervo, B. Behavioral drive versus behavioral inertia in evolution: a null model approach. Am. Nat. 161, 357–366 (2003).Article 
PubMed 

Google Scholar 
Kearney, M. R. & Porter, W. P. NicheMapR—an R package for biophysical modelling: the microclimate model. Ecography 40, 664–674 (2017).Article 

Google Scholar 
Messier, J., McGill, B. J., Enquist, B. J. & Lechowicz, M. J. Trait variation and integration across scales: is the leaf economic spectrum present at local scales? Ecography 40, 685–697 (2017).Article 

Google Scholar 
Ricklefs, R. E. & Schluter, D. (eds) Species Diversity in Ecological Communities: Historical and Geographical Perspectives (Univ. Chicago Press, 1993).Angilletta, M. J. Jr, Steury, T. D. & Sears, M. W. Temperature, growth rate, and body size in ectotherms: fitting pieces of a life-history puzzle. Integr. Comp. Biol. 44, 498–509 (2004).Article 
PubMed 

Google Scholar 
Pincheira-Donoso, D. The balance between predictions and evidence and the search for universal macroecological patterns: taking Bergmann’s rule back to its endothermic origin. Theory Biosci. 129, 247–253 (2010).Article 
PubMed 

Google Scholar 
Slavenko, A. et al. Global patterns of body size evolution in squamate reptiles are not driven by climate. Glob. Ecol. Biogeogr. 28, 471–483 (2019).Article 

Google Scholar 
Stevenson, R. D. Body size and limits to the daily range of body temperature in terrestrial ectotherms. Am. Nat. 125, 102–117 (1985).Article 

Google Scholar 
Rubalcaba, J. G., Gouveia, S. F. & Olalla‐Tárraga, M. A. A mechanistic model to scale up biophysical processes into geographical size gradients in ectotherms. Glob. Ecol. Biogeogr. 28, 793–803 (2019).Article 

Google Scholar 
Rubalcaba, J. G. & Olalla‐Tárraga, M. Á. The biogeography of thermal risk for terrestrial ectotherms: scaling of thermal tolerance with body size and latitude. J. Anim. Ecol. 89, 1277–1285 (2020).Article 
PubMed 

Google Scholar 
Pincheira-Donoso, D., Hodgson, D. J. & Tregenza, T. The evolution of body size under environmental gradients in ectotherms: why should Bergmann’s rule apply to lizards? BMC Evol. Biol. 8, 68 (2008).Jablonski, D. Biotic interactions and macroevolution: extensions and mismatches across scales and levels. Evolution 62, 715–739 (2008).Article 
PubMed 

Google Scholar 
Kearney, M. R., Porter, W. P. & Huey, R. B. Modelling the joint effects of body size and microclimate on heat budgets and foraging opportunities of ectotherms. Methods Ecol. Evol. 12, 458–467 (2021).Article 

Google Scholar 
Campbell-Staton, S. C., Bare, A., Losos, J. B., Edwards, S. V. & Cheviron, Z. A. Physiological and regulatory underpinnings of geographic variation in reptilian cold tolerance across a latitudinal cline. Mol. Ecol. 27, 2243–2255 (2018).Article 
CAS 
PubMed 

Google Scholar 
Boretto, J. M., Fernández, J. B., Cabezas-Cartes, F., Medina, M. S. & Ibargüengoytía, N. R. in Lizards of Patagonia (eds Morando, M. & Avila, L. J.) 335–371 (Springer, 2020).Deutsch, C. A. et al. Impacts of climate warming on terrestrial ectotherms across latitude. Proc. Natl Acad. Sci. USA 105, 6668–6672 (2008).Article 
CAS 
PubMed 
PubMed Central 

Google Scholar 
Araújo, M. B. et al. Heat freezes niche evolution. Ecol. Lett. 16, 1206–1219 (2013).Article 
PubMed 

Google Scholar 
Sunday, J. M., Bates, A. E. & Dulvy, N. K. Global analysis of thermal tolerance and latitude in ectotherms. Proc. R. Soc. B 278, 1823–1830 (2011).Article 
PubMed 

Google Scholar 
Hoffmann, A. A., Chown, S. L. & Clusella‐Trullas, S. Upper thermal limits in terrestrial ectotherms: how constrained are they? Funct. Ecol. 27, 934–949 (2013).Article 

Google Scholar 
Sunday, J. et al. Thermal tolerance patterns across latitude and elevation. Philos. Trans. R. Soc. B 374, 20190036 (2019).Article 

Google Scholar 
Huey, R. B. & Slatkin, M. Cost and benefits of lizard thermoregulation. Q. Rev. Biol. 51, 363–384 (1976).Article 
CAS 
PubMed 

Google Scholar 
Vasseur, D. A. et al. Increased temperature variation poses a greater risk to species than climate warming. Proc. R. Soc. B 281, 20132612 (2014).Article 
PubMed 
PubMed Central 

Google Scholar 
Porter, W. P., Mitchell, J. W., Beckman, W. A. & DeWitt, C. B. Behavioral implications of mechanistic ecology. Oecologia 13, 1–54 (1973).Article 
CAS 
PubMed 

Google Scholar 
Hertz, P. E., Huey, R. B. & Stevenson, R. D. Evaluating temperature regulation by field-active ectotherms: the fallacy of the inappropriate question. Am. Nat. 142, 796–818 (1993).Article 
CAS 
PubMed 

Google Scholar 
Fey, S. B. et al. Opportunities for behavioral rescue under rapid environmental change. Glob. Change Biol. 25, 3110–3120 (2019).Article 

Google Scholar 
Martin, T. L. & Huey, R. B. Why ‘suboptimal’ is optimal: Jensen’s inequality and ectotherm thermal preferences. Am. Nat. 171, E102–E118 (2008).Article 
PubMed 

Google Scholar 
R Core Team. A Language and Environment for Statistical Computing (R Foundation for Statistical Computing, 2019).Campbell, G. S. & Norman, J. M. An Introduction to Environmental Biophysics 2nd edn (Springer-Verlag, 1998).Mao, J. & Yan, B. Global Monthly Mean Leaf Area Index Climatology, 1981–2015 (ORNL DAAC, 2019).Meiri, S. et al. Are lizards feeling the heat? A tale of ecology and evolution under two temperatures. Glob. Ecol. Biogeogr. 22, 834–845 (2013).Article 

Google Scholar 
Marino, S., Hogue, I. B., Ray, C. J. & Kirschner, D. E. A methodology for performing global uncertainty and sensitivity analysis in systems biology. J. Theor. Biol. 254, 178–196 (2008).Article 
PubMed 
PubMed Central 

Google Scholar 
Renardy, M., Hult, C., Evans, S., Linderman, J. J. & Kirschner, D. E. Global sensitivity analysis of biological multiscale models. Curr. Opin. Biomed. Eng. 11, 109–116 (2019).Article 
PubMed 
PubMed Central 

Google Scholar 
Carnell, R. lhs: Latin hypercube samples. R package version 1.1.1 (2020).Meiri, S. Traits of lizards of the world: variation around a successful evolutionary design. Glob. Ecol. Biogeogr. 27, 1168–1172 (2018).Article 

Google Scholar 
Clusella-Trullas, S., Blackburn, T. M. & Chown, S. L. Climatic predictors of temperature performance curve parameters in ectotherms imply complex responses to climate change. Am. Nat. 177, 738–751 (2011).Article 
PubMed 

Google Scholar 
Bennett, J. M. et al. GlobTherm, a global database on thermal tolerances for aquatic and terrestrial organisms. Sci. Data 5, 180022 (2018).Roll, U. et al. The global distribution of tetrapods reveals a need for targeted reptile conservation. Nat. Ecol. Evol. 1, 1677–1682 (2017).Article 
PubMed 

Google Scholar 
Tonini, J. F. R., Beard, K. H., Ferreira, R. B., Jetz, W. & Pyron, R. A. Fully-sampled phylogenies of squamates reveal evolutionary patterns in threat status. Biol. Conserv. 204, 23–31 (2016).Article 

Google Scholar 
Ives, A. R. R2s for correlated data: phylogenetic models, LMMs, and GLLMs. Syst. Biol. 68, 234–251 (2019).Article 
PubMed 

Google Scholar 
Johnson, T. F., Isaac, N. J. B., Paviolo, A. & González-Suárez, M. Handling missing values in trait data. Glob. Ecol. Biogeogr. 30, 51–62 (2020).Article 

Google Scholar 
Goolsby, E. W., Bruggeman, J. & Ané, C. Rphylopars: fast multivariate phylogenetic comparative methods for missing data and within‐species variation. Methods Ecol. Evol. 8, 22–27 (2017).Article 

Google Scholar 
Koenker, R. et al. Package ‘quantreg’ (R-CRAN, 2018); https://cran.r-project.org/web/packages/quantreg/quantreg.pdfGriffith, D. A. & Peres-Neto, P. R. Spatial modeling in ecology: the flexibility of eigenfunction spatial analyses. Ecology 87, 2603–2613 (2006).Article 
PubMed 

Google Scholar 
Bivand, R. R packages for analyzing spatial data: a comparative case study with areal data. Geogr. Anal. 54, 488–518 (2022).Article 

Google Scholar 
Rubalcaba, J. G. et al. Data: ‘Climate drives global functional trait variation in lizards’. figshare https://doi.org/10.6084/m9.figshare.19949315 (2022). More

Artificial intelligence (AI) and machine learning could revolutionize the search for life on other planets. But before these tools can tackle distant locales such as Mars, they need to be tested here on Earth.A team of researchers have successfully trained an AI to map biosignatures — any feature which provides evidence of past or present life — in a three-square-kilometre area of Chile’s Atacama Desert. The AI substantially reduced the area the team needed to search and boosted the likelihood of finding living organisms in one of the driest places on the planet. The results were reported on 6 March in Nature Astronomy1.Kimberley Warren-Rhodes, a senior research scientist at the SETI Institute in Mountain View, California, and lead author on the paper, has been chasing biosignatures since the early 2000s, when she realized how few tools existed to study the biology of other planets. She wanted to combine her background in statistical ecology with emerging technologies such as AI to help mission scientists, “who are under a lot of pressure to find biosignatures” but tightly constrained in how they do so. Rovers that are controlled remotely from Earth, for example, can travel only limited distances and collect relatively few specimens, placing a premium on sampling locations that are the most likely to yield life. Mission scientists base these predictions in part on Mars analogues on Earth, where scientists scour extreme habitats to determine how and where living organisms thrive.Searching for lifeBeginning in 2016, Warren-Rhodes’ group travelled to the high, parched plateau of the Atacama Desert — a proposed Mars analogue at an elevation of around 3,500 metres in the Chilean Andes — to search for rock-dwelling, photosynthetic organisms called endoliths. To fully characterize the environment, the researchers collected everything from drone footage to geochemical analyses to DNA sequences. Together, this data set mimics the types of information researchers are collecting on Mars with orbital satellites, drones and rovers.Warren-Rhodes’ team fed its data into an AI-based convolutional neural network (CNN) and a machine-learning algorithm that in turn predicted where life was most likely to be found in the Atacama.

Aerial view (left) and ground view from a rover of a biosignature probability map of the same area.Credit: M. Phillips, K. A. Warren-Rhodes & F. Kalaitzis

By targeting their sample collection on the basis of AI feedback, the researchers were able to reduce their search area by up to 97% and increase their likelihood of finding life by up to 88%. “At the end, you could plop us down, and instead of wandering around for a long time, it would take us a minute to find life,” Warren-Rhodes says. Specifically, the team found that endoliths in the Atacama were most often found in a mineral called alabaster — which is porous and retains water — and tended to aggregate in transitional areas between various microhabitats, such as where sand and alabaster crystals abut one another.“I’m very impressed and very happy to see this suite of work,” says Kennda Lynch, an astrobiologist at the Lunar and Planetary Institute in Houston, Texas, who studies biosignatures. “It’s really cool that they can show some success with an AI to help predict where to go and look.”Graham Lau, an astrobiologist at the Blue Marble Space Institute of Science who is based in Boulder, Colorado, worked on another Mars analogue in the Canadian Arctic as a graduate student, to study how biology influences the formation of rare minerals that can serve as biosignatures on other planets. “Ever since I first read Frank Herbert’s Dune as a young child, I was struck by this idea of applying ecology to planets,” he says. But up until the last decade or so, the tools and data weren’t available to address such questions with scientific rigour. “The place where we have almost unlimited data possibilities is through these orbital observations and drone imaging,” he says, “and I do see this paper as being one of many pieces along the pathway to doing these larger analyses.”Deceptively simpleThe new method will need to be verified across multiple ecosystems, Lau and Lynch say, including those with more complex geology and greater biodiversity. The Atacama, Lau notes, is relatively simple in terms of the habitats and the types of life that are likely to be found there. And on Mars, the high level of ultraviolet radiation striking the planet’s surface means that scientists might need to detect clues that hint at life below ground.

NASA’s Perseverance rover collected its first rock sample from an area in Mars’ Jezero Crater.Credit: NASA/JPL-Caltech/ASU/MSSS

Ultimately, Warren-Rhodes says she would like to see a comprehensive database of different Mars analogues that could feed valuable information to mission scientists planning their next sampling run. Her team’s advance, she adds, might appear “deceptively simple” to anyone who grew up watching Star Trek explorers scanning alien worlds with a tricorder. But, it represents an important advance in extraterrestrial research, in which biology has often lagged behind chemistry and geology. Imagine, for instance, virtual-reality headsets that feed mission scientists real-time data as they scan a surface, using a rover’s ‘eyes’ to direct their activities. “To have our team make one of these first steps towards reliably detecting biosignatures using AI is exciting,” she says. “It’s really a momentous time.” More

SettingThis study was conducted in approximately 40-year-old naturally regenerated P. densiflora stands in Wola National Experimental Forest in Gyeongnam province in South Korea (35°12′ N, 128°10′ E; Table 1). The productivity of this forest is low, with a dominant tree height of 10 m at 20 years of age. Over the last 10 years, the mean annual precipitation was 1490 mm, of which one third fell during summer (July–August), and the mean temperature was 13.1 °C. The vegetation growing season generally lasts for approximately 200 days, extending from early April to October. The soil texture is a silt loam originating from sandstone and shale (clay 13.0 ± 0.8%, silt 44.1 ± 1.3%, sand 42.9 ± 1.6%; n = 18). The given texture results in volumetric water contents at 13.4 ± 0.7% (m3 m−3) at permanent wilting point (1500 kPa) and 40.7 ± 1.2% at field capacity (10 kPa)55. The understory is covered with Lespedeza spp., Quercus variabilis, Q. serrata, Smilax china, and Lindera glauca.In 2010, we selected two adjacent P. densiflora stands approximately 100 m apart from each other (180 m and 195 m above sea level, on slopes of 15° and 33°, both stands face south). Following a completely randomized design, we established nine plots (10 × 10 m2 with a 5 m untreated buffer) within each stand, of which three were randomly assigned to annual NPK fertilization, three to PK fertilization, and the rest to a control treatment without fertilization. The fertilizer, composed of urea, fused superphosphate and potassium chloride (N3P4K1) or P4K1 was added manually by deposition on the forest floor for 3 years in April 2011, April 2012, and March 2013. Over these 3 years, the NPK plots received 33.9 g N, 45 g P, and 11.1 g K m−2, while the PK plots received 45 g P and 11.1 g K m−2.Tree and stand measurementsThe standing biomass of trees was estimated using a combination of site-specific allometric equations based on destructive harvesting56 and repeated measurements of the dimensions of all trees in each plot (5–18 trees plot−1). The stem diameter at 1.2 m (D) was measured for all trees in each plot for which D was ≥ 6 cm. Selecting a representative tree in size for each plot within the 4 × 4 m2 center of the plot, we measured the tree height (H) and crown base for the representative trees. Measurements were performed in April and September 2011, September 2012–2014, and October 2021. We observed no effect of fertilization on the relationship between D and H or between D and crown base, so we assumed no effect on the allometric functions for foliage or branch biomass. A dendrometer band (Series 5 Manual Band, Forestry Suppliers Inc., Jackson, MS, USA) was installed on 18 representative trees (one per plot) to monitor radial growth monthly.Three 0.25 m2 circular litter traps were installed 60 cm above the forest floor in each plot in April 2011. Litter was collected at 3-month intervals between June 2011 and March 2015. The litter from each trap was transported to the laboratory and then oven-dried at 65 °C for 48 h. All dried samples were separated into needles, bark, cones, branches, and miscellaneous components, and weighed separately.In September 2014, we estimated the biomass of understory vegetation, separately for woody plants and herbaceous plants. All woody plants  More

Description of the phenological, vegetative and yield traits of the accessions per habitatThe process of data management included the computation of mean squares for the assessed phenological, vegetative and yield traits of the accessions with the sampling habitats considered as the treatment factor. The error mean squares served in the multiple comparison of means reported in Table 1.Table 1 Means of the measured phonological, vegetative and flowering and yield traits of Cucurbita moschata genotypes sampled from seven habitats.Full size tableRegarding the phenological traits, the accessions from the habitat of Zh have the longest period from seeding to first male (102.39 d) and first female (108.14 d) flower appearances, and the longest period from seeding to physiological maturity (153.95 d). For those traits, the accessions from Tiassale and Soubre are not significantly different from those of Zh. And, accessions from Tiassale and Zh have the longest periods from seeding to 50% flowering. On the other hand, accessions from Korho, Ferke, Bondu and Burki develop their first male and female flowers and attain 50% flowering in a very short period. They also reach physiological maturity faster. Accessions from Korho, however, have the longest period from seeding to 50% emergence (6.07 d) and accessions from Bondu have the longest period from first female flower appearance to physiological maturity (53.04 d).
For the vegetative traits, accessions from Tiassale and Soubre have the largest girth size (4.43 cm and 4.63 cm, respectively). Accessions from Tiassale have the longest (24.98 cm) and widest (19.94 cm) leaves, the longest male (16.2 cm) and female (4.03 cm) peduncles and the longest petioles (34.94 cm). The measures for those organs on accessions from Soubre rank second to those of Tiassale. On the other hand, accessions from Korho, Ferke, Bondu and Burki are characterized by smaller girth size, smaller leaves, smaller petioles and smaller peduncles of male and female flowers. But the accessions from Bondu are the tallest (586.91 cm) followed by the accessions from Ferke (489.20 cm). And the accessions from Zh are the shortest (417.38 cm).For the flowering and yield traits, accessions from Tiassale and Soubre show the largest numbers of male (27.33 units and 22.58 units, respectively) and female (5.22 units and 6.05 units, respectively) flowers per plant, largest numbers of fruits per plant (2.78 units and 2.53 units, respectively) and largest measures of all fruit-related traits. Their seeds are very large, but in small numbers. In contrast, accessions from Korho, Ferke, Bondu and Burki have the smallest numbers of male and female flowers per plant, the smallest numbers of fruits per plant and the smallest measures of fruit-related traits. They have large numbers of seeds, but their seeds are smaller, except the seeds of the accessions from Burki. Refer to Table 1 for more detailed information.
Variability of the phenological, vegetative and yield traitsTable 2 shows the spread of the phenological and morphological traits of the assessed accessions of C. moschata. All the evaluated traits showed very wide ranges of distribution of the observations. Some conspicuously wide ranges of traits include number of days to 50% flowering (DTF) that goes from 52 to 152 d, plant height with a minimum of 48 cm and a maximum of 1510 cm, diameter of the fruit that is between 5.8 cm and 35 cm, weight of the fruit that varies between 150 g and 10,930 g and number of seeds per fruit that spreads in the interval from 32 units per fruit to 729 units per fruit. Excluding the number of days to 50% emergence (DTE), all the other assessed traits have remarkably wide ranges of phenotypic expressions (Table 2). All the traits but DTE, DTF, days from first female flower appearance to fruit maturity, fruit length and length of the dry seed, had outliers. The number of outliers ranged from 1 to 67. Except the outliers observed with the width of the dry seed, all the outliers were above 1.5*IQR + Q3 where IQR is the inter-quartile range and Q3 is the third quartile. The presence of outliers is indicative of the richness and large variability of the population of accessions. The outliers are exceptional performances that fall outside the normal distribution of the observations. They are a stock of unusual traits that can be used in a crop improvement program when beneficial. For example, the observed outliers for diameter of the fruit, weight of the fruit or thickness of the pulp can be used in a breeding program for the improvement of fruit yield. Similarly, outliers for beneficial traits related to the seed can be used to improve C. moschata crop for seed yield. Besides, the computed mean squares (data not reported) showed highly significant variations between accessions for the assessed traits. They all yielded p-values less than 0.01, providing additional support to the evidence of large variability among the accessions of C. moschata of Cote d’Ivoire. The computed standard deviation, and median absolute deviation for each trait are additional evidence. We should note that in most cases, the mean squares associated to year (data not reported) were not significant, indicating the relative stability of the assessed traits.Table 2 Minimum (Min), first quartile (Q1), median, third quartile (Q3), maximum (Max), standard deviation (SD), median absolute deviation (MAD) and outliers obtained from the phenological, vegetative and flowering and yield traits of 663 accessions of C. moschata.Full size tableThe components of variance, the quantitative genetic differentiation, the overall mean, and the coefficients of variation are reported in Table 3. The lme4 package37 used in the determination of the components of variance, does not provide p-values in the analysis of mixed or random models. The reported quantities in Table 3 are not accompanied with tests of significance. It is worth mentioning that the respective units of measure of the assessed traits are squared for the variances and the evaluated estimates will be reported without the units of measure. The phenotypic variance ((sigma_{p}^{2})) is partitioned into variance between morphotypes or genotypic variance ((sigma_{g}^{2})), and within morphotypes or residual variance ((sigma_{e}^{2})). For the class of phenological traits, considerable genotypic variances were observed with days to 50% flowering (266.21) and days to first male flower appearance (254.40), compared with their respective residual variances (148.13 and 199.50). Regarding the class of vegetative traits, only the peduncle length of male flowers had a genotypic variance (9.22) greater than its residual variance (8.86). In the class of flowering and yield traits, 8 of the 15 traits assessed showed large genotypic variances in comparison with their respective residual variances. They are number of female flowers per plant ((sigma_{g}^{2}) = 3.02 versus (sigma_{e}^{2}) = 2.36), length of the fruit ((sigma_{g}^{2}) = 53.96 versus (sigma_{e}^{2}) = 48.97), diameter of the fruit ((sigma_{g}^{2}) = 37.17 versus (sigma_{e}^{2}) = 16.76), volume of the fruit ((sigma_{g}^{2}) = 10,713,468 versus (sigma_{e}^{2}) = 3,904,590), weight of the fruit ((sigma_{g}^{2}) = 5,413,819 versus (sigma_{e}^{2}) = 1,420,187), diameter of the cavity enclosing the seed ((sigma_{g}^{2}) = 19.12 versus (sigma_{e}^{2}) = 7.75), thickness of the fruit pulp ((sigma_{g}^{2}) = 1.11 versus (sigma_{e}^{2}) = 0.94) and weight of the fruit pulp ((sigma_{g}^{2}) = 5,979,212 versus (sigma_{e}^{2}) = 1,088,750). For a trait to have a lager genotypic variance than the residual variance is synonymous to a relative ease of improvement of the crop for that trait through a breeding program.Table 3 Components of variances ((sigma_{p}^{2}), (sigma_{g}^{2}), (sigma_{e}^{2}), (sigma_{a}^{2})), quantitative genetic differentiation ((Q_{ST})), overall mean ((mu)), and coefficients of variation (%) ((CV_{p}),(CV_{g}),(CV_{e})), of the measured phenological, vegetative and yield traits of the accessions of C. moschata of Cote d’Ivoire.Full size tableThe coefficient of variation (CV) is another statistic that measures variation. It is actually the dispersion of a trait per unit measure of its mean, which can be used to compare variations of traits with different measurement units or different scales. As a rule-of-thumb, a coefficient of variation greater than 20% is indicative of large variation for the trait. The phenotypic coefficient of variation is considerably high for 25 of the 28 assessed traits. Only the number of days from seeding to physiological maturity, the first and second longest axes of the dry seed show coefficients of variation less than 20%. Traits with very large phenotypic coefficients of variation include the peduncle length of female flowers ((CV_{p}) = 93.98%), weight of the pulp ((CV_{p}) = 92.96%), volume of the fruit ((CV_{p}) = 89.17%), weight of the fruit ((CV_{p}) = 78.30%) and number of female flowers per plant ((CV_{p}) = 65.81%). With respect to the residual coefficients of variation, only the number of days from seeding to 50% emergence and number of days from first female flower appearance to physiological maturity have residual coefficients of variation greater than 20%, among the phenological traits. All the vegetative traits have residual coefficients of variation greater than 20%, and show a near-perfect linear relation (r = 0.98; p  More

Preparation of the pao cai soupFirst, 35 kg of white radish (Raphanus sativus), 35 kg of cabbage (Brassica oleracea), 2 kg of chili pepper (Capsicum frutescens), 1 kg of ginger (Zingiber officinale), 1 kg of peppercorns (Zanthoxylum bungeanum), 2.5 kg of rock sugar, and 210 kg of cold boiled water (containing 6% salt) were divided into six ceramic jars. After 7 days of natural fermentation at room temperature, the pao cai was filtered out with sterile gauze to obtain 200 kg of pao cai soup. To ensure an even distribution of microorganisms in the soup, the soup was mixed well and then left to rest for 12 h, the supernatant was taken, and the soup was left to rest for 12 h again.The plants used in this study were cultivated vegetables which purchased from the vegetable market at the study site. All local, national or international guidelines and legislation were adhered to in the production of this study.Establishment of the microcosm systemSeventy-eight for each size of 10 ml, 20 ml, 50 ml, 100 ml, 250 ml, 500 ml, and 1000 ml sterile glass culture flasks were filled with pao cai soup, the bottle mouth was sealed with sterile sealing film, and the bottle was capped without leaving any air (Fig. 1). Each flask became a microcosm and was cultured in a 25 °C incubator.Figure 1Schematic diagram of the establishment of the microcosmic system.Full size imageSample collectionBefore the microcosm system was established, a sample of well-mixed pao cai soup was taken as a reference to establish background biodiversity. The microbial community dynamics should change the fastest at the beginning of the microcosm system establishment and gradually become slower over time. Considering the workload and cost, this study collected samples daily for 1–10 day after the establishment of the microcosm and then collected every 2 days for 10–30 day and every 5 days for 30–60 day. Three different microcosms of the same volume were established. Monitoring was carried out for 60 days, and a total of 546 samples of 7 volumetric gradients were obtained at 26 time points. At the time of sampling, the pao cai soup in the microcosm was mixed, and 50 mL of sample (10 mL of sample was collected for microcosm systems with a volume of less than 50 mL) was collected. The sample was centrifuged at 8000 rpm for 10 min, the supernatant was collected for pH determination, and the pellet was stored in a − 80 °C freezer.Microbial analysesMicrobial DNA was extracted from pao cai samples using the E.Z.N.A.® Soil DNA Kit (Omega Biotek, Norcross, GA, U.S.) according to the manufacturer’s protocols. For bacteria, we targeted the V3-V4 region of the 16S ribosomal RNA (rRNA) gene, using the 338F (5′-ACTCCTACGGGAGGCAGCAG-3′) and 806R (5′-GGACTACHVGGGTWTCTAAT-3′) primer pairs. For fungi, we targeted the ITS1-1F region of the nuclear ribosomal internal transcribed spacer region (ITS rDNA) gene, using ITS1-1F-F (5′-CTTGGTCATTTAGAGGAAGTAA-3′) and ITS-1F-R (5′-GCTGCGTTCTTCATCGATGC-3′). PCRs were performed in triplicate in a 20 μL mixture containing 4 μL of 5 × FastPfu Buffer, 2 μL of 2.5 mM dNTPs, 0.8 μL of each primer (5 μM), 0.4 μL of FastPfu Polymerase and 10 ng of template DNA. The PCR program for the 16S rRNA gene was as follows: 3 min of denaturation at 95 °C; 27 cycles of 30 s at 95 °C, 30 s of annealing at 55 °C, and 45 s of elongation at 72 °C; and a final extension at 72 °C for 10 min. For the ITS1-1F region, the PCR program was as follows: samples were initially denatured at 98 °C for 1 min, followed by 30 cycles of denaturation at 98 °C for 10 s, primer annealing at 50 °C for 30 s, and extension at 72 °C for 30 s. A final extension step of 5 min at 72 °C was added to ensure complete amplification of the target region. The resulting PCR products were extracted from a 2% agarose gel, further purified using the AxyPrep DNA Gel Extraction Kit (Axygen Biosciences, Union City, CA, USA) and quantified using QuantiFluor™-ST (Promega, Madison, WI, USA).Purified amplicons were pooled in equimolar amounts and paired-end sequenced (2 × 300) on an Illumina NovaSeq platform (Illumina, San Diego, CA, USA) according to standard protocols. The analysis was conducted by following the “Atacama soil microbiome tutorial” of QIIME2 docs along with customized program scripts (https://docs.qiime2.org/2019.1/). Briefly, raw data FASTQ files were imported in the QIIME2 system using the qiime tools import program. Demultiplexed sequences from each sample were quality filtered, trimmed, denoised, and merged, and then the chimeric sequences were identified and removed using the QIIME2 DADA2 plugin to obtain the feature table of amplicon sequence variants (ASVs)24. Compared with traditional OTU that clusters at 97% similarity, ASV has higher accuracy, equivalent to 99% similarity clustering. The QIIME2 feature-classifier plugin was then used to align ASV sequences to the pretrained GREENGENES 13_8 99% database (trimmed to the V3-V4 region bound by the 338F/806R primer pair for bacteria) and UNITE database (for fungi) to generate the taxonomy table25. Any contaminating mitochondrial and chloroplast sequences were filtered using the QIIME2 feature-table plugin. Based on the sequence number of the lowest sample, perform the resampling to make the sequence number equal for each sample. Due to the random nature of sequencing, ASVs specific to each sample in this study were present. To reduce the uncertainty introduced by the sequencing process, we filtered out rare ASVs with less than 0.001% of the total sequence volume.Data analysisIn this study, the data of fungi and bacteria were integrated and analyzed, and all microbial diversity appearing in the text represent the sum of all fungi and bacteria. Species richness is equal to the number of taxa, which is equal to the total number of all bacterial and fungal ASVs. The vegan package in R 4.2.1 was used to calculate the species richness of each sample based on the ASV feature table26. Using flask volume instead of area, SAR fitting was performed using a semi-logarithmic model, and its significance was tested. The semi-logarithmic model is the function S = c + b*logA, where S is species richness, A is area (in this case, volume is used instead), and b and c are fit parameters27.The microcosmic system in this study is hermetically sealed, and all microorganisms originate from a single portion of well-mixed paocai soup (ie species pool). The speciation process in the 60-day experimental system should be negligible due to the short experimental period. The extinction rate of a microcosm system is equal to the number of ASVs lost in the microcosm system compared to the species pool divided by the total number of ASVs in the species pool. The extinction rate is the number of extinct ASVs in each system compared to the species pool. Pearson correlation analysis was performed with volume as the independent variable and extinction rate as the dependent variable to determine the correlation between volume and extinction rate at each time point. When microorganisms of a microcosmic system disappear entirely or cannot be detected, the microcosm is recorded as an annihilated microcosm. The annihilation rate at a time point is equal to the number of microcosms annihilated at that time, divided by the total number of microcosms. The difference between the extinction rate and annihilation rate defined in this paper is that the extinction rate is for ASVs within each sample, and the annihilation rate is for microcosmic system at each sampling time point. The two indicators jointly characterize the local extinction of microorganisms from different perspectives. Non-linear regression with a bell-shaped form was performed with time as an independent variable and pH and annihilation rate as dependent variables, and regression lines were plotted based on R 4.2.1.According to the taxonomy table, bacterial ASVs were divided into acid-producing and non-acid-producing categories, and their extinction rates were calculated separately. The agricola, ggplot2, vegan and ggpubr packages were used to draw alpha diversity box plots and perform the Wilcoxon rank sum test for differences between groups26,28,29,30. Non-metric multidimensional scaling (NMDS) analysis was performed with the vegan package based on Bray–Curtis dissimilarity. In addition, the potential Kyoto Encyclopedia of Genes and Genomes (KEGG) orthologue (KO) functional profiles of microbial communities were predicted with PICRUSt31. Resistance-related genes were screened using the gene function predictions. The relationship between the relative abundance of resistance-related genes and the volume of the microcosm was analysed by Pearson correlation, and a forest map was plotted to present the results. More

We investigated the complex relationships between the environmental characteristics of blue spaces and visit-related mental well-being in a multi-country study including 17 bluespace types and four facets of subjective well-being. Our aim was to operationalise, and consider the utility of, the Bratman et al.9 conceptual model that links ecosystem services (ESS) with mental health.Consistent with the proposed conceptual model, mental well-being outcomes relied on a complex interplay of individual, environmental, and visit characteristics.Summary of findingsOverall, bluespace visits were associated with better subjective mental well-being outcomes if the visits took place in nearby coastal areas or rural rivers, were perceived as safe and to have good water quality, and had a long duration. They could involve a range of activities such as playing with children, socialising, or walking. The degree to which the perceived presence of wildlife predicted visit satisfaction varied depending on the bluespace type, suggesting that the importance of ecosystem features such as biodiversity may vary by the setting.We can also identify the combination of environmental and visit characteristics associated with particularly high levels of well-being for a specific outcome. For example, an optimal visit in terms of happiness might be to sandy beaches where there are high levels of perceived safety and excellent water quality; with a visit lasting at least three hours; and possibly involving playing with children, socialising, sunbathing/paddling and/or walking with a dog; and has short travel times that do not involve public transport.RQ1—natural and environmental featuresResearch question 1a—Which bluespace type(s) were associated with the highest levels of recalled visit mental well-being?Four of the five bluespace types associated with the highest levels of visit satisfaction were coastal (sea cliffs, rocky shore, sandy beaches, rural river and seaside promenade), indicating that these environments may be particularly beneficial for well-being. Visits to these environments were also associated with the lowest levels of visit anxiety, with the exception of seaside promenade and sea cliffs, which were not significantly different to the grand mean. Seaside promenade was the only urban environment in the top five.In addition, only coastal sites were associated with significantly higher levels of visit happiness (compared to the grand mean), further highlighting the potential importance of these environments. Although not explored here, coastal scenes tend to be associated with particularly high aesthetic and scenic value25,26 which may also be positively related to subjective well-being.These findings are broadly consistent with other studies from the UK17,27, but are extended here to our international sample. White et al.28 also used data from the BlueHealth International Survey (BIS) and explored visit frequency to different environments and associations with general mental health and well-being outcomes, including the World Health Organisation five-item Well-being index referring to the two weeks prior to the survey. Consistent with the results here, they found that visit frequency to “coastal blue” environments was more strongly associated with psychological well-being in general than visit frequency to “inland blue” environments. Our study adds to these more general findings by showing that these associations may come as a direct result of the recalled well-being experienced on specific visits to these locations.Confidence in our results was strengthened as we included general mental well-being in our analysis to adjust for whether happier people tend to visit sandy beaches, for example. The results for visit anxiety were not always the inverse of the trends observed in the positive measures of well-being, supporting the need to look at multiple aspects of mental well-being when considering the effects of nature contact.Research question 1b—Which bluespace qualities were associated with the highest levels of recalled visit mental well-being?Of the range of qualities that we investigated as predictors, perceived safety and ‘excellent’ water quality (vs. ‘sufficient’) consistently exhibited the strongest relationships with subjective mental well-being. Perceived safety has been found to be important when visiting blue spaces in several qualitative studies29,30,31, as well as a quantitative study with older adults in Hong Kong14. Blue spaces have particular safety issues with respect to drowning32,33, but fear of crime29,30,33 or pedestrian safety34 may also be relevant.Water quality has also been found to be important in previous economic valuation studies of recreational use and enjoyment of lakes and estuaries in the USA and Australia35,36 as well as a contingent behaviour experiment carried out as part of the BlueHealth International Survey (in European countries only)37. We recognise that here we used a metric of perceived water quality, rather than measures based on biological or toxicological sampling. Nevertheless, perceptions have been reported to positively correlate with sampled water quality parameters38, although assessments can vary systematically such as by bluespace type39. Highly visible harmful algal blooms, for instance, have also been found to affect experiences of blue spaces40.Further, and again consistent with earlier work15,41,42, the presence of facilities and wildlife, and absence of litter, were generally associated with better subjective mental well-being. Both perceived presence of wildlife and facilities were also associated with higher levels of anxiety however, indicating complexities between environmental qualities and well-being. Some wildlife may be deemed unpleasant or an ecosystem disservice, for example. The presence of good facilities may indicate the presence of more people; and visitor density in natural environments can be related to preference43. These results highlight the importance of environmental quality and not just type, consistent with other frameworks12,37.Research question 2—How is exposure, as operationalised by visit duration, related to recalled visit mental well-being?Broadly consistent with research in the green and bluespace literature14,17,44, we found that mental well-being outcomes were generally higher with greater exposure as indicated by visit duration. For decreasing visit anxiety, this was only significant when visits were longer than an hour and a half. As we did not measure pre-visit anxiety levels, we are cautious about identifying this as a potential temporal threshold for reducing anxiety at this stage.Similarly, also using the BlueHealth International Survey, White et al.28 found that well-being outcomes were higher with greater visit exposure to green and blue spaces using a metric of visit frequency. However, in contrast to this and other research which looked at overall weekly aggregated time in nature (e.g.28,45), we have no evidence of diminishing marginal returns as the effect sizes associated with specific visit duration continued to increase with increasing duration.Research question 3—What experiences in blue spaces, in terms of activities (3a) and companions (3b), are associated with the most positive recalled visit mental well-being outcomes?Although walking was the most popular activity, the activity with the highest mental well-being ratings was playing with children, especially in certain locations such as beaches (Fig. 4). However, we also find that anxiety tended to be higher when children were present. We speculate that the purpose of the visit may be important. For example, many who go to the beach with children do so in order to play. However, if children are present on more adult-oriented activities such as hiking, this may increase adult anxiety during the visit. From a representative sample of English adults, White et al.17 found that recent nature visits with children were associated with the lowest levels of well-being. Therefore, visits with children may be associated with a more complex set of emotions, being both slightly more stressful, but also potentially more rewarding and ‘meaningful’46. Ecosystem features of beaches may be particularly supportive of high well-being activities. A qualitative study in the UK, for instance, highlighted the particular opportunities for adults and children to play together at the beach, including rock-pooling and making sandcastles as well as water-based activities47.Visits with other adults were associated with higher levels of both visit satisfaction and worthwhile-ness, and socialising as an activity was associated with better visit well-being for all outcomes compared to the grand mean. This is consistent with studies using the day reconstruction method, which link activities with experiential well-being, in the USA48 and Germany49 where socialising was associated with the highest, or second highest, levels of well-being for all the activities assessed. Further, social interactions have been recognised as an important benefit by many of those visiting freshwater blue spaces in a previous study18.Research question 4—Does the relationship between wildlife presence and recalled visit well-being vary by bluespace settings?The relationship between the presence of wildlife and visit satisfaction varied with bluespace type. The strongest positive association was found for fen, marsh and bog areas, which may also be related to the purpose of visit. For instance, those who visit places such as fens, marshes and bogs, may do so for the explicit purpose of observing wildlife (often birds) and the presence of wildlife would therefore be important for satisfaction with the visit.Perceptions towards wildlife have been found to vary by location in other studies. For example, in Sweden, greater prior experience with geese at beaches was associated with a negative attitude towards geese50. Further, the species present are likely to vary across different environments. In three urban areas in the UK, green spaces correlated with the abundance and species richness of birds considered to provide cultural services (songbirds and woodpeckers), while an abundance of birds considered to provide disservices (e.g. some gull species, feral pigeons) was independent of green spaces51. Preferences for some species over others may explain some of the negative or null relationships between the presence of wildlife at different blue spaces. These examples from the literature, alongside our own results, indicate the potential for benefits from the management of wildlife for psychological ecosystem services differentially across environments, although these should be considered alongside other conservation and ESS goals.MechanismsSeveral mechanisms potentially explain the beneficial effects of visiting blue spaces on mental health and well-being12, including the provision of opportunities for physical activity52,53; social interaction18; cognitive restoration and stress reduction17,54; emotion regulation55 and connecting with nature12. Consistent with these mechanisms, we found that respondents were using blue spaces for both physical activity and social interaction; and that playing with children and socialising were associated with particularly high levels of well-being.In addition to the positive association we find between some ESS and well-being, including presence of wildlife and water quality, additional bluespace ESS not considered here, may also affect mental health and well-being12. For example, the provision of a cooling effect56 and air pollution mitigation57.Strengths and limitationsA key strength is our operationalisation of the Bratman et al.9 conceptual model for mental health using data from a large, 18 country survey that included 17 different bluespace types, five quality metrics and four subjective mental well-being outcomes. The relatively high explanatory power of our models suggests all the variables we explored were important for subjective well-being.Despite the strengths, however, there were also several limitations. The survey was cross-sectional and causality cannot be inferred. For example, happier people may choose to visit a beach rather than another location, although we also controlled for general levels of subjective mental well-being in an attempt to control for this possibility (See Supplemental Materials). Further, although the majority of respondents (53%) recalled a visit within the last 7 days, some were recalling visits up to a month ago, with potential memory biases increasing in line with length of recall.Although our data were collected by an international market research company to be representative of age, gender and region within country, our online sample may not be fully representative across more characteristics and any country-level conclusions need to be treated with caution. We also acknowledge that there were no results from Africa, the Middle East or South America; and Hong Kong was the only representative from Asia. This suggests far more research is needed in other regions to better understand how bluespace ecosystems interact with subjective well-being globally.There may also be socioeconomic confounds that we did not include in our models which may account for some of the effects. Not everyone visits nature for recreation58, including about 4000 people here who did not visit a bluespace in the four weeks prior to responding to the survey. Some groups may therefore have been under-represented; and we should be careful in assuming that our findings generalise to all sub-population groups.Nevertheless, our visit sub-sample distributions were generally similar to that of the weighted percentages in the full sample, with the exception of age where those aged over 60 were under-represented (Table S2); therefore, we suspect these issues were not too influential for the overall results, although care needs to be extended to inferences with respect to older adults.A further limitation was that we only considered the qualities of places where people reported making recreational visits, with respondents presumably less likely to visit places where they feel really unsafe or lacking in facilities29. Further research may want to study responses to a broader range of bluespace settings, including those that are less visited, to determine the generalisability of the generally positive results found here. Such studies could use pre-existing tools to objectively assess the quality of blue spaces59.ImplicationsOur finding that coastal environments are particularly beneficial adds to the body of evidence linking coastal environments with health and well-being and suggests this is consistent across many countries. Previous research has found that greater proximity to blue spaces, especially coastal settings, predicts visit frequency14,60,61 as well as other health outcomes—e.g. reduced risk of mortality and better general health, well-being and physical activity53,62. Here, we found that shorter travel times also predict visit well-being, highlighting the importance of having equitable access to good quality natural environments near to people’s homes.We also identified that different types of coastal and inland blue spaces (e.g. seaside promande, rural rivers), with different qualities (e.g. wildlife present), involving particular types of activities in specific social configurations (e.g. playing with children), were especially good at promoting well-being. This moves beyond a simple location-based assessment of benefit to one that recognises the complex interplay between location, behavioural and social processes. Numerous commentators63 (including Bratman et al.9 on which we have based this paper) have argued that we need to go beyond the determinate effects of green and blue spaces and develop a far richer, more nuanced understanding. The approach we have taken here is intended as a step in this direction.In terms of policy applications, these results provide support for the potential health benefits of efforts to improve equitable access to high quality environments, such as the English Coast Path (https://englandcoastpath.co.uk/) and the creation of beaches in Barcelona with the Olympic project in 199264. Our results also hint at the importance of high-level legislation, such as the EU’s Bathing Waters Directive65 for mental well-being37. If conducted on a more fine-grained geographical level, results could have the potential to leverage public support for more localised conservation initiatives. Furthermore, such results could be used as a basis for integration into more systematic conservation planning66.Further researchAlthough we incorporate a range of variables in our analysis, and our pseudo-R2 values are relatively high for a social research context, considerable variation remains unexplained. Although other individual characteristics may be important, such as nature connectedness67 and memories68, further research could explore the specific ecosystem features and social contexts associated with the particular positive results from coastal spaces, which would be of interest to policy makers and environmental managers. We also speculated that purpose of visit may explain some of our findings. Further research could explore the interactions between motivations and location, experience, and well-being outcomes.The presence of wildlife was differentially important across bluespace types and further research could unpack this. Exploring similar possibilities for the other quality metrics, as well as considering additional ecosystem characteristics, would also be informative. For example, identifying which factors are important in perceptions of safety in blue spaces. Bratman et al.9 also considered effect modification by visitor characteristics and further research could include interactions, or sub-group analysis, by socio-demographic factors.Further research could also explore longer-term benefits of these features over repeated visits; the potential for ecosystem disservices, such as the relationships we find between an interaction of wildlife and ice rinks and well-being; the potential for negative outcomes associated with ecosystem degradation69; and the potential for positive mental health outcomes from ecological restoration70.We have demonstrated some of the complexities involved in the human-nature relationship and that many factors are related to the outcome from a visit. The conceptual model applied allows the investigation of a wide range of variables including natural features and other environmental qualities, and characteristics of the exposure and experience, as well as individual parameters. We suggest that other researchers can apply this conceptual model and design data collection accordingly to target specific research questions and hypotheses (as opposed to where we have fitted already collected data). More