Philippa Kaur

Clark, M. A., Springmann, M., Hill, J. & Tilman, D. Multiple health and environmental impacts of foods. Proc. Natl Acad. Sci. USA 116, 23357 (2019).Article 
CAS 
PubMed 
PubMed Central 

Google Scholar 
Davis, K. F. et al. Assessing the sustainability of post-Green Revolution cereals in India. Proc. Natl Acad. Sci. USA 116, 25034 (2019).Article 
CAS 
PubMed 
PubMed Central 

Google Scholar 
Hoekstra, A. Y. & Wiedmann, T. O. Humanity’s unsustainable environmental footprint. Science 344, 1114 (2014).Article 
CAS 
PubMed 

Google Scholar 
O Neill, D. W., Fanning, A. L., Lamb, W. F. & Steinberger, J. K. A good life for all within planetary boundaries. Nat. Sustain. 1, 88 (2018).Article 

Google Scholar 
Steffen, W. et al. Planetary boundaries: guiding human development on a changing planet. Science 347, 1259855 (2015).Article 
PubMed 

Google Scholar 
van Dijk, M., Morley, T., Rau, M. L. & Saghai, Y. A meta-analysis of projected global food demand and population at risk of hunger for the period 2010–2050. Nat. Food 2, 494 (2021).Article 

Google Scholar 
Grassini, P., Eskridge, K. M. & Cassman, K. G. Distinguishing between yield advances and yield plateaus in historical crop production trends. Nat. Commun. 4, 2918 (2013).Article 
PubMed 

Google Scholar 
Ray, D. K., Ramankutty, N., Mueller, N. D., West, P. C. & Foley, J. A. Recent patterns of crop yield growth and stagnation. Nat. Commun. 3, 1293 (2012).Article 
PubMed 

Google Scholar 
Chen, X. et al. Integrated soil–crop system management for food security. Proc. Natl Acad. Sci. USA 108, 6399 (2011).Article 
CAS 
PubMed 
PubMed Central 

Google Scholar 
FAOSTAT. FAO http://www.fao.org/faostat/en/#home (2021).Liu, Z. et al. Optimization of China’s maize and soy production can ensure feed sufficiency at lower nitrogen and carbon footprints. Nat. Food 2, 426 (2021).Article 

Google Scholar 
Zhang, Q. et al. Outlook of China’s agriculture transforming from smallholder operation to sustainable production. Glob. Food Secur. 26, 100444 (2020).Article 

Google Scholar 
Duan, J. et al. Consolidation of agricultural land can contribute to agricultural sustainability in China. Nat. Food 2, 1014 (2021).Article 
CAS 

Google Scholar 
Cui, Z. et al. Pursuing sustainable productivity with millions of smallholder farmers. Nature 555, 363 (2018).Article 
CAS 
PubMed 

Google Scholar 
Zhou, F. et al. Deceleration of China’s human water use and its key drivers. Proc. Natl Acad. Sci. USA 117, 7702 (2020).Article 
CAS 
PubMed 
PubMed Central 

Google Scholar 
Wu, H. et al. Estimating ammonia emissions from cropland in China based on the establishment of agro-region-specific models. Agr. For. Meteorol. 303, 108373 (2021).Article 

Google Scholar 
Yue, Q. et al. Deriving emission factors and estimating direct nitrous oxide emissions for crop cultivation in China. Environ. Sci. Technol. 53, 10246 (2019).Article 
CAS 
PubMed 

Google Scholar 
Ju, X., Gu, B., Wu, Y. & Galloway, J. N. Reducing China’s fertilizer use by increasing farm size. Global Environ. Chang. 41, 26 (2016).Article 

Google Scholar 
Costanza, R. et al. Changes in the global value of ecosystem services. Global Environ. Chang. 26, 152 (2014).Article 

Google Scholar 
Mueller, N. D. et al. Closing yield gaps through nutrient and water management. Nature 490, 254 (2012).Article 
CAS 
PubMed 

Google Scholar 
Davis, K. F., Rulli, M. C., Seveso, A. & D. Odorico, P. Increased food production and reduced water use through optimized crop distribution. Nat. Geosci. 10, 919 (2017).Article 
CAS 

Google Scholar 
Chen, X. et al. Producing more grain with lower environmental costs. Nature 514, 486 (2014).Article 
CAS 
PubMed 

Google Scholar 
UN Department of Economic and Social Affairs, Population Division (2019). World Population Prospects 2019, Online Edition. Rev. 1 (2019). https://population.un.org/wpp/2019 Refinement to the 2006 IPCC Guidelines for National Greenhouse Gas Inventories (IPCC, 2019).Bowles, T. M. et al. Long-term evidence shows that crop-rotation diversification increases agricultural resilience to adverse growing conditions in North America. One Earth 2, 284 (2020).Article 

Google Scholar 
Cardinale, B. J. et al. Impacts of plant diversity on biomass production increase through time because of species complementarity. Proc. Natl Acad. Sci. USA 104, 18123 (2007).Article 
CAS 
PubMed 
PubMed Central 

Google Scholar 
Sirami, C. et al. Increasing crop heterogeneity enhances multitrophic diversity across agricultural regions. Proc. Natl Acad. Sci. USA 116, 16442 (2019).Article 
CAS 
PubMed 
PubMed Central 

Google Scholar 
Renard, D. & Tilman, D. National food production stabilized by crop diversity. Nature 571, 257 (2019).Article 
CAS 
PubMed 

Google Scholar 
Price Bureau of the National Development and Reform Commission of China. China Agricultural Products Cost–Benefit Compilation of Information 2017 (in Chinese) (China Statistics Press, 2017).Fan, S., Brzeska, J., Keyzer, M. & Halsema, A. From Subsistence to Profit: Transforming Smallholder Farms. (Inter. Food Policy Res. Inst., 2013).Wang, S. et al. Urbanization can benefit agricultural production with large-scale farming in China. Nat. Food 2, 183 (2021).Article 

Google Scholar 
Yin, Y. et al. A steady-state N balance approach for sustainable smallholder farming. Proc. Natl Acad. Sci. USA 118, e2106576118 (2021).Article 
CAS 
PubMed 
PubMed Central 

Google Scholar 
Guiding opinions of the ministry of agriculture on the adjustment of maize structure in the “sickle” area. Ministry of Agriculture and Rural Affairs of the People’s Republic of China http://www.moa.gov.cn/nybgb/2015/shiyiqi/201712/t20171219_6103893.htm (2017).Zhang, F., Chen, X. & Vitousek, P. An experiment for the world. Nature 497, 33 (2013).Article 
CAS 
PubMed 

Google Scholar 
Zhang, W. et al. Closing yield gaps in China by empowering smallholder farmers. Nature 537, 671 (2016).Article 
CAS 
PubMed 

Google Scholar 
Cyberspace Administration of China. State Council of the People’s Republic of China http://www.gov.cn/xinwen/2021-12/28/content_5664873.htm (2021).Kou, T. et al. Effects of long-term cropping regimes on soil carbon sequestration and aggregate composition in rainfed farmland of Northeast China. Soil Till. Res. 118, 132 (2012).Article 

Google Scholar 
Li, X. et al. Long-term increased grain yield and soil fertility from intercropping. Nat. Sustain. 4, 943 (2021).Article 

Google Scholar 
Damerau, K. et al. India has natural resource capacity to achieve nutrition security, reduce health risks and improve environmental sustainability. Nat. Food 1, 631 (2020).Article 

Google Scholar 
Kuang, W. et al. Cropland redistribution to marginal lands undermines environmental sustainability. Natl Sci. Rev. 9, 1 (2021).
Google Scholar 
Zhao, C. et al. Temperature increase reduces global yields of major crops in four independent estimates. Proc. Natl Acad. Sci. USA 114, 9326 (2017).Article 
CAS 
PubMed 
PubMed Central 

Google Scholar 
Ma, L. et al. Exploring future food provision scenarios for China. Environ. Sci. Technol. 53, 1385 (2018).Article 

Google Scholar 
National population development plan: 2016–2030. National Development and Reform Commission http://www.gov.cn/zhengce/content/2017-01/25/content_5163309.htm (2016).Ma, L. et al. Environmental assessment of management options for nutrient flows in the food chain in China. Environ. Sci. Technol. 47, 7260 (2013).Article 
CAS 
PubMed 

Google Scholar 
Lobell, D. B., Cassman, K. G. & Field, C. B. Crop yield gaps: their importance, magnitudes, and causes. Annu. Rev. Environ. Resour. 34, 179 (2009).Article 

Google Scholar 
Yan, X., Akiyama, H., Yagi, K. & Akimoto, H., Global estimations of the inventory and mitigation potential of methane emissions from rice cultivation conducted using the 2006 Intergovernmental Panel on Climate Change Guidelines. Global Biogeochem. Cy. https://doi.org/10.1029/2008GB003299 (2009).Smith, P., Martino, Z. & Cai, D. ‘Agriculture’, in Climate Change 2007: Mitigation (Cambridge Univ. Press, 2007).Liang, D. et al. China’s greenhouse gas emissions for cropping systems from 1978–2016. Sci. Data 8, 171 (2021).Article 
CAS 
PubMed 
PubMed Central 

Google Scholar  More

Dinosauria Owen, 1842Theropoda Marsh, 1881Dromaeosauridae Matthew and Brown, 1922Halszkaraptorinae Cau et al., 2017Revised diagnosisSmall dromaeosaurids that possess dorsoventrally flattened premaxillae, premaxillary bodies perforated by many neurovascular foramina, enlarged and closely packed premaxillary teeth that utilized delayed replacement patterns, reduced anterior maxillary teeth, dorsolateral placement of retracted external nares, greatly elongated cervical vertebrae, anterior cervical vertebrae with round lobes formed by the postzygapophyses, horizontal zygapophyses, and pronounced zygapophyseal laminae in the anterior caudal vertebrae, mediolaterally compressed ulnae with sharp posterior margins, second and third metacarpals with similar thicknesses, shelf-like supratrochanteric processes on the ilia, elongated fossae that border posterolateral ridges on the posterodistal surfaces of the femoral shafts, and third metatarsals in which the proximal halves are unconstricted and anteriorly convex.Natovenator polydontus gen. et sp. nov.HolotypeMPC-D 102/114 (Institute of Paleontology, Mongolian Academy of Sciences, Ulaanbaatar, Mongolia) is a mostly articulated skeleton with a nearly complete skull (See Supplementary Table 1 for measurements).Locality and horizonBaruungoyot Formation (Upper Cretaceous), Hermiin Tsav, Omnogovi Province, Mongolia13 (Supplementary Fig. 5).EtymologyNatovenator, from the Latin nato (swim) and venator (hunter), in reference to the hypothesized swimming behaviour and piscivorous diet of the new taxon; polydontus, from the Greek polys (many) and odous (tooth) in reference to the unusually many teeth.DiagnosisA small halszkaraptorine dromaeosaurid with the following autapomorphies: wide groove delimited by a pair of ridges on the anterodorsal surface of the premaxilla, premaxilla with an elongated internarial process that overlies nasal and extends posterior to the external naris, 13 premaxillary teeth with large and incisiviform crowns, first three anteriormost maxillary teeth are greatly reduced and are clustered together with the following tooth without any separations by interdental septa, anteroposteriorly long external naris (about 30% of the preorbital skull length), paroccipital process with a anteroposteriorly broad dorsal surface, elongate maxillary process of the palatine that extends anteriorly beyond the middle of the antorbital fenestra, pterygoid with a deep fossa on the medial surface of the quadrate ramus, distinct posterolaterally oriented projection on the lateral surface of atlas, absence of pleurocoels in cervical vertebrae (not confirmed in the missing fifth cervical centrum), posterolaterally oriented and nearly horizontal proximal shafts in the dorsal ribs, hourglass-shaped metacarpal II with distinctly concave medial and lateral surfaces.DescriptionThe skull of Natovenator is nearly complete, although the preorbital region has been affected by compression and is slightly offset from the rest of the skull (Figs. 1c, d, 2a–d and Supplementary Figs. 1, 2). Near the tip of the snout, the premaxilla is marked by a broad groove. The body of the premaxilla is also dorsoventrally low and is perforated by numerous foramina that lead into a complex network of neurovascular chambers (Supplementary Fig. 1b) as in Halszkaraptor4. Similarly, the external naris is positioned posteriorly and is level with the premaxilla-maxilla contact (Fig. 2a, b), although it is marginally behind this position in Halszkaraptor4. It is also dorsally placed compared to those of other non-avian theropods and faces dorsolaterally. The exceptionally long external naris and accordingly elongated internarial process of Natovenator (Fig. 2c) are unique among dromaeosaurids but comparable to those in aquatic toothed birds14 as well as in therizinosaurs15,16. The frontal is similar to those of other halszkaraptorines4,17 in that it is vaulted to accommodate a large orbit and has little contribution to the supratemporal fossa. A sharp nuchal crest is formed by the parietal and the squamosal (Supplementary Fig. 2a–e). The latter also produces a shelf that extends over the quadrate head as in other dromaeosaurids18. The paroccipital process curves gently on the occiput and has a broad dorsal surface that tapers laterally (Fig. 2f and Supplementary Fig. 2b, e). Its ventrolateral orientation is reminiscent of Mahakala17 but is different from the more horizontal paroccipital process of Halszkaraptor4. The occipital condyle is long and constricted at its base. A shallow dorsal tympanic recess on the lateral wall of the braincase is different from the deep one of Mahakala17. The palatine is tetraradiate with a greatly elongated maxillary process, which extends anteriorly beyond the level of the mid-antorbital fenestra. The pterygoid is missing its anterior portion (Fig. 2g and Supplementary Fig. 2a–e). A deep fossa on the medial surface of the thin quadrate ramus is not seen in any other dromaeosaurids. The mandibles of Natovenator preserve most of the elements, especially those on the left side (Fig. 1a, b, d and Supplementary Figs. 1a, 2). Each jaw is characterized by a slender dentary with nearly parallel dorsal and ventral margins, a surangular partially fused with the articular, a distinctive surangular shelf, and a fan-shaped retroarticular process that protrudes dorsomedially. The upper dentition of Natovenator is heterodont as the premaxillary teeth are morphologically distinct from the maxillary teeth (Fig. 2a, b, e and Supplementary Fig. 1a, c). There are unusually numerous premaxillary teeth tightly packed without any separation of the alveoli by bony septa. The roots of the teeth are long, and the crowns are tall and incisiviform as in Halszkaraptor4. Moreover, the large replacement teeth in the premaxilla suggest that the replacement of the premaxillary teeth was delayed as in Halszkaraptor4. However, the number of teeth in each premaxilla is 13 in Natovenator, whereas it is only 11 in Halszkaraptor4. In the maxilla, the three most anterior maxillary teeth are markedly shorter than the premaxillary teeth and the more posterior maxillary teeth. This pattern is also observed in Halszkaraptor, although the number of shorter maxillary teeth differs as it has two reduced ones7. Both the maxillary and dentary teeth have sharp fang-like crowns that lack serrations. Although posteriormost parts are poorly preserved, there are at least 23 alveoli in each of the maxilla and dentary, which suggests high numbers of teeth in both elements.The neck of Natovenator, as preserved, is twisted and includes ten elongated cervical vertebrae, although most of the 5th cervical is missing (Figs. 1, 3a–d). This elongation of the cervicals results in a noticeably longer neck than those of most dromaeosaurids and is estimated to be longer than the dorsal series. It is, however, proportionately shorter than that of Halszkaraptor, which has a neck as long as its dorsal and sacral vertebra combined4. Another peculiarity in the neck of the Natovenator is a pronounced posterolaterally extending projection on the neurapophysis of the atlas (Fig. 3a and Supplementary Fig. 2b, c, e). The postzygapophyses of each anterior cervical are fused into a single lobe-like process as in Halszkaraptor4. Pleurocoels are absent in the cervical vertebrae. In contrast, Halszkaraptor has pleurocoels on its 7th–9th cervicals4. A total of 12 dorsal vertebrae are preserved (Figs. 1a, b, 3e, 4a and Supplementary Figs. 3a–d). They all lack pleurocoels, and their parapophyses on the anterior and mid-dorsals are placed high on the anterodorsal end of each centrum. Interestingly, the positions of the parapophyses are similar to those of hesperornithiforms19,20,21 rather than other dromaeosaurids such as Deinonychus22 or Velociraptor23. The preserved dorsal ribs, articulated with the second to seventh dorsals, are flattened and posteriorly oriented (Figs. 1, 3e, 4a–d). The proximal shafts are also nearly horizontal, which is indicative of a dorsoventrally compressed ribcage. Each proximal caudal vertebra has a long centrum and horizontal zygapophyses with expanded laminae (Fig. 3f and Supplementary Fig. 3e–i), all of which are characters shared with other halszkaraptorines4,17. The forelimb elements are partially exposed (Figs. 1a, b, 2a–d, 3e, g). The nearly complete right humerus is proportionately short and distally flattened like that of Halszkaraptor4. The shaft of the ulna is mediolaterally compressed to produce a sharp posterior margin as in Halszkaraptor4 and Mahakala17. Metacarpal III is robust and is only slightly longer than metacarpal II. Similarly, metacarpal III is almost as thick and long as other second metacarpals of other halszkaraptorines4,17. The femur has a long ridge on its posterior surface, which is another characteristic shared among halszkaraptorines4. Typically for a dromaeosaurid, metatarsals II and III have ginglymoid distal articular surfaces (Fig. 3h and Supplementary Fig. 4f, h). The ventral surface of metatarsal III is invaded by a ridge near the distal end, unlike other halszkaraptorines (Fig. 3h)4,5,17,24.Phylogenetic analysisThe phylogenetic analysis found more than 99,999 most parsimonious trees (CI = 0.23, RI = 0.55) with 6574 steps. Deinonychosaurian monophyly is not supported by the strict consensus tree (Supplementary Fig. 6). Instead, Dromaeosauridae was recovered as a sister clade to a monophyletic clade formed by Troodontidae and Avialae, which is consistent with the results of Cau et al.4 and Cau7. Halszkaraptorinae is positioned at the base of Dromaeosauridae as in Cau et al.4, although there are claims that dromaeosaurid affinities of halszkaraptorines are not well supported25. Nine (seven ambiguous and two unambiguous) synapomorphies support the inclusion of Halszkaraptorinae in Dromaeosauridae. The two unambiguous synapomorphies are the anterior tympanic recess at the same level as the basipterygoid process and the presence of a ventral flange on the paroccipital process. A total of 20 synapomorphies (including one unambiguous synapomorphy) unite the four halszkaraptorines, including Natovenator (Supplementary Fig. 7). In Halszkaraptorinae, Halszkaraptor is the earliest branching taxon, and the remaining three taxa form an unresolved clade supported by three ambiguous synapomorphies (characters 121/1, 569/0, and 1153/1). Two of these synapomorphies are related to the paroccipital process (characters 121 and 569), which is not preserved in Hulsanpes5,24. The other is the presence of an expansion on the medial margin of the distal half of metatarsal III, which is not entirely preserved in the Natovenator. When scored as 0 for this character, Natovenator branches off from the unresolved clade. It suggests that the medial expansion of the dorsal surface of metatarsal III could be a derived character among halszkaraptorines. More

Rosenzweig, M. L. Habitat selection and population interactions: the search for mechanism. Am. Nat. 137, S5–S28 (1991).Article 

Google Scholar 
Binckley, C. A. & Resetarits, W. J. Habitat selection determines abundance, richness and species composition of beetles in aquatic communities. Biol. Lett. 1, 370–374 (2005).Article 
PubMed 
PubMed Central 

Google Scholar 
Foltz, S. J. & Dodson, S. I. Aquatic Hemiptera community structure in stormwater retention ponds: A watershed land cover approach. Hydrobiologia 621, 49–62 (2009).Article 

Google Scholar 
Resetarits, W. J. Habitat selection behaviour links local and regional scales in aquatic systems: Habitat selection at multiple spatial scales. Ecol. Lett. 8, 480–486 (2005).Article 
PubMed 

Google Scholar 
Leclerc, M. et al. Quantifying consistent individual differences in habitat selection. Oecologia 180, 697–705 (2016).Article 
PubMed 

Google Scholar 
Morris, D. W. Adaptation and habitat selection in the eco-evolutionary process. Proc. R. Soc. B Biol. Sci. 278, 2401–2411 (2011).Article 

Google Scholar 
Resetarits, W. J. Colonization under threat of predation: avoidance of fish by an aquatic beetle, Tropisternus lateralis (Coleoptera: Hydrophilidae). Oecologia 129, 155–160 (2001).Article 
PubMed 

Google Scholar 
Wellborn, G. A., Skelly, D. K. & Werner, E. E. Mechanisms creating community structure across a freshwater habitat gradient. Annu. Rev. Ecol. Evol. Syst. 27, 337–363 (1996).Article 

Google Scholar 
Klečka, J. & Boukal, D. S. Who eats whom in a pool? A comparative study of prey selectivity by predatory aquatic insects. PLoS ONE 7, e37741 (2012).Article 
PubMed 
PubMed Central 

Google Scholar 
Nilsson, P. A. & Brönmark, C. Prey vulnerability to a gape-size limited predator: behavioural and morphological impacts on northern pike piscivory. Oikos 88, 539–546 (2000).Article 

Google Scholar 
Šigutová, H. et al. Specialization directs habitat selection responses to a top predator in semiaquatic but not aquatic taxa. Sci. Rep. 11, 18928 (2021).Article 
PubMed 
PubMed Central 

Google Scholar 
Pintar, M. R. & Resetarits, W. J. Match and mismatch: integrating consumptive effects of predators, prey traits, and habitat selection in colonizing aquatic insects. Ecol. Evol. 11, 1902–1917 (2021).Article 
PubMed 
PubMed Central 

Google Scholar 
Pintar, M. R. & Resetarits, W. J. Jr. Out with the old, in with the new: oviposition preference matches larval success in cope’s gray treefrog Hyla chrysoscelis. J. Herpetol. 51, 186–189 (2017).Article 

Google Scholar 
Wildermuth, H. Habitat selection and oviposition site recognition by the dragonfly Aeshna juncea (L.): an experimental approach in natural habitats (Anisoptera: Aeshnidae). Odonatologica 22, 27–44 (1993).Fortin, D., Morris, D. W. & McLoughlin, P. D. Habitat selection and the evolution of specialists in heterogeneous environments. Isr. J. Ecol. Evol. 54, 311–328 (2008).Article 

Google Scholar 
McLoughlin, P. D., Boyce, M. S., Coulson, T. & Clutton-Brock, T. Lifetime reproductive success and density-dependent, multi-variable resource selection. Proc. R. Soc. B Biol. Sci. 273, 1449–1454 (2006).Article 

Google Scholar 
Morris, D. W. Scales and costs of habitat selection in heterogeneous landscapes. Evol. Ecol. 6, 412–432 (1992).Article 

Google Scholar 
McLoughlin, P. D., Morris, D. W., Fortin, D., Wal, E. V. & Contasti, A. L. Considering ecological dynamics in resource selection functions. J. Anim. Ecol. 79, 4–12 (2010).Article 
PubMed 

Google Scholar 
Leclerc, M., Dussault, C. & St-Laurent, M.-H. Behavioural strategies towards human disturbances explain individual performance in woodland caribou. Oecologia 176, 297–306 (2014).Article 
PubMed 

Google Scholar 
Bolnick, D. I. et al. The ecology of individuals: incidence and implications of individual specialization. Am. Nat. 161, 1–28 (2003).Article 
MathSciNet 
PubMed 

Google Scholar 
Sheppard, C. E. et al. Intragroup competition predicts individual foraging specialisation in a group-living mammal. Ecol. Lett. 21, 665–673 (2018).Article 
PubMed 
PubMed Central 

Google Scholar 
Forstmeier, W. & Birkhead, T. R. Repeatability of mate choice in the zebra finch: consistency within and between females. Anim. Behav. 68, 1017–1028 (2004).Article 

Google Scholar 
Gómez-Laplaza, L. M. The influence of social status on shoaling preferences in the freshwater angelfish (Pterophyllum scalare). Behaviour 142, 827–844 (2005).Article 

Google Scholar 
Gillingham, M. P. & Parker, K. L. The importance of individual variation in defining habitat selection by moose in northern British Columbia. Alces 44, 7–20 (2008).
Google Scholar 
Lesmerises, R. & St-Laurent, M.-H. Not accounting for interindividual variability can mask habitat selection patterns: a case study on black bears. Oecologia 185, 415–425 (2017).Article 
PubMed 

Google Scholar 
van Beest, F. M. et al. Increasing density leads to generalization in both coarse-grained habitat selection and fine-grained resource selection in a large mammal. J. Anim. Ecol. 83, 147–156 (2014).Article 
PubMed 

Google Scholar 
Fretwell, S. D. & Lucas, H. L. On territorial behavior and other factors influencing habitat distribution in birds I. Theoretical development. Biotheoretica 19, 16–36 (1970).Article 

Google Scholar 
Binckley, C. A. & Resetarits, W. J. Functional equivalence of non-lethal effects: generalized fish avoidance determines distribution of gray treefrog, Hyla chrysoscelis, larvae. Oikos 102, 623–629 (2003).Article 

Google Scholar 
Kraus, J. M. & Vonesh, J. R. Feedbacks between community assembly and habitat selection shape variation in local colonization. J. Anim. Ecol. 79, 795–802 (2010).PubMed 

Google Scholar 
Pollard, C. J. et al. Removal of an exotic fish influences amphibian breeding site selection: Exotic fish removal. J. Wildl. Manag. 81, 720–727 (2017).Article 

Google Scholar 
Calenge, C., Dufour, A. B. & Maillard, D. K-select analysis: a new method to analyse habitat selection in radio-tracking studies. Ecol. Model. 186, 143–153 (2005).Article 

Google Scholar 
Freitas, C., Kovacs, K. M., Lydersen, C. & Ims, R. A. A novel method for quantifying habitat selection and predicting habitat use. J. Appl. Ecol. 45, 1213–1220 (2008).
Google Scholar 
Mitchell, L. J., Kohler, T., White, P. C. L. & Arnold, K. E. High interindividual variability in habitat selection and functional habitat relationships in European nightjars over a period of habitat change. Ecol. Evol. 10, 5932–5945 (2020).Article 
PubMed 
PubMed Central 

Google Scholar 
Richter, L. et al. So close and yet so different: the importance of considering temporal dynamics to understand habitat selection. Basic Appl. Ecol. 43, 99–109 (2020).Article 

Google Scholar 
Tyler, J. A. & Rose, K. A. Individual variability and spatial heterogeneity in fish population models. Rev. Fish Biol. Fish. 4, 91–123 (1994).Article 

Google Scholar 
Córdoba-Aguilar, A. Dragonflies and Damselflies: Model Organisms for Ecological and Evolutionary Research. (Oxford University Press, 2008).Sandall, E. L. & Fischer, B. Be a professional: attend to the insects. Am. Entomol. 65, 176–179 (2019).Article 

Google Scholar 
Blaustein, L. Oviposition site selection in response to risk of predation: evidence from aquatic habitats and consequences for population dynamics and community. in Evolutionary theory and processes: modern perspectives (ed. Wasser, S. P.) 441–456 (Kluwer, 1999).Helebrandová, J. B., Pyszko, P. & Dolný, A. Behavioural phenotypic plasticity of submerged oviposition in damselflies (Insecta: Odonata). Insects 10, 124 (2019).Article 
PubMed 
PubMed Central 

Google Scholar 
Hollis, K. & Guillette, L. What associative learning in insects tells us about the evolution of learning and fixed behavior. Int. J. Comp. Psychol. 28, 25706 (2015).Article 

Google Scholar 
Papaj, D. R. & Lewis, A. C. Insect Learning: Ecological and Evolutinary Perspectives. (Chapman & Hall, 1993).Simons, M. & Tibbetts, E. Insects as models for studying the evolution of animal cognition. Curr. Opin. Insect Sci. 34, 117–122 (2019).Article 
PubMed 

Google Scholar 
Benard, M. F. Predator-induced phenotypic plasticity in organisms with complex life histories. Annu. Rev. Ecol. Evol. Syst. 35, 651–673 (2004).Article 

Google Scholar 
Cook, W. L. & Streams, F. A. Fish predation on Notonecta (Hemiptera): relationship between prey risk and habitat utilization. Oecologia 64, 177–183 (1984).Article 
CAS 
PubMed 

Google Scholar 
Larson, D. J. The predaceous water beetles (Coleoptera: Dytiscidae) of Alberta: Systematics, natural history and distribution. Quaest. Entomol. 11, 245–498 (1985).
Google Scholar 
Svensson, B. G., Tallmark, B. & Petersson, E. Habitat heterogeneity, coexistence and habitat utilization in five backswimmer species (Notonecta spp.; Hemiptera, Notonectidae). Aquat. Insects 22, 81–98 (2000).Lock, K., Adriaens, T., Meutter, F. V. D. & Goethals, P. Effect of water quality on waterbugs (Hemiptera: Gerromorpha & Nepomorpha) in Flanders (Belgium): results from a large-scale field survey. Ann. Limnol. Int. J. Limnol. 49, 121–128 (2013).Article 

Google Scholar 
Macan, T. T. A twenty-one-year study of the water-bugs in a Moorland Fishpond. J. Anim. Ecol. 45, 913–922 (1976).Article 

Google Scholar 
Boukal, D. S. et al. Catalogue of water beetles of the Czech Republic. Klapalekiana 43 (Suppl.), 1–289 (2007).Åbjörnsson, K., Wagner, B. M. A., Axelsson, A., Bjerselius, R. & Olsén, K. H. Responses of Acilius sulcatus (Coleoptera: Dytiscidae) to chemical cues from perch (Perca fluviatilis). Oecologia 111, 166–171 (1997).Article 
PubMed 

Google Scholar 
Gioria, M., Schaffers, A., Bacaro, G. & Feehan, J. The conservation value of farmland ponds: Predicting water beetle assemblages using vascular plants as a surrogate group. Biol. Conserv. 143, 1125–1133 (2010).Article 

Google Scholar 
Bergsten, J. & Miller, K. B. Taxonomic revision of the Holarctic diving beetle genus Acilius Leach (Coleoptera: Dytiscidae): Acilius taxonomic revision. Syst. Entomol. 31, 145–197 (2005).Article 

Google Scholar 
Everard, M. Britain’s Freshwater Fishes. (Princeton University Press, 2013).Miller, K. B. & Bergsten, J. Predaceous diving beetle sexual systems. in Ecology, systematics, and the natural history of predaceous diving beetles (Coleoptera: Dytiscidae) (ed. Yee, D. A.) 199–234 (Springer Netherlands, 2014).Culler, L. E., Ohba, S. & Crumrine, P. Predator-prey interactions of dytiscids. in Ecology, Systematics, and the Natural History of Predaceous Diving Beetles (Coleoptera: Dytiscidae) (ed. Yee, D. A.) 363–379 (Springer, 2014).Baines, C. B., McCauley, S. J. & Rowe, L. Dispersal depends on body condition and predation risk in the semi-aquatic insect Notonecta undulata. Ecol. Evol. 5, 2307–2316 (2015).Article 
PubMed 
PubMed Central 

Google Scholar 
Baines, C. B., Ferzoco, I. M. & McCauley, S. J. Sex-biased dispersal is independent of sex ratio in a semiaquatic insect. Behav. Ecol. Sociobiol. 71, 119 (2017).Article 

Google Scholar 
Hungerford, H. B. The biology and ecology of aquatic and semiaquatic Hemiptera. Univ. Kans. Sci. Bull. 11, 3–334 (1919).
Google Scholar 
Streams, F. A. Intrageneric predation by Notonecta (Hemiptera: Notonectidae) in the laboratory and in nature. Ann. Entomol. Soc. Am. 85, 265–273 (1992).Article 

Google Scholar 
Halekoh, U., Højsgaard, S. & Yan, J. The R Package geepack for generalized estimating equations. J. Stat. Softw. 15, 1–11 (2006).Article 

Google Scholar 
Lenth, R. V. Least-squares means: the R package lsmeans. J. Stat. Softw. 69, 1–33 (2016).Article 

Google Scholar 
Bates, A., Maechler, M., Bolker, B. & Walker, S. Fitting linear mixed-effects models using lme4. J. Stat. Softw. 67, 1–48 (2015).Article 

Google Scholar 
Stoffel, M. A., Nakagawa, S. & Schielzeth, H. rptR: repeatability estimation and variance decomposition by generalized linear mixed-effects models. Methods Ecol. Evol. 8, 1639–1644 (2017).Article 

Google Scholar 
R Core Team. R: A language and environment for statistical computing. (2021).Harvill, M. L. The antipredatory behavior of the aquatic diving beetle, Coptotomus venustus (Say)(Coleoptera: Dytiscidae) in response to fish predation. (Texas A&M University, 1994).McCauley, S. J. & Rowe, L. Notonecta exhibit threat-sensitive, predator-induced dispersal. Biol. Lett. 6, 449–452 (2010).Article 
PubMed 
PubMed Central 

Google Scholar 
Schoeppner, N. M. & Relyea, R. A. Damage, digestion, and defence: the roles of alarm cues and kairomones for inducing prey defences. Ecol. Lett. 8, 505–512 (2005).Article 
PubMed 

Google Scholar 
Roberts, G. Why individual vigilance declines as group size increases. Anim. Behav. 51, 1077–1086 (1996).Article 

Google Scholar 
Giller, P. S. Locomotory efficiency in the predation strategies of the British Notonecta (Hempitera, Heteroptera). Oecologia 52, 273–277 (1982).Article 
CAS 
PubMed 

Google Scholar 
Gittelman, S. H. Locomotion and predatory strategy in backswimmers (Hemiptera: Notonectidae). Am. Midl. Nat. 92, 496–500 (1974).Article 

Google Scholar 
Morris, D. W. Density-dependent habitat selection: testing the theory with fitness data. Evol. Ecol. 3, 80–94 (1989).Article 

Google Scholar 
Holt, R. D. Population dynamics in two-patch environments: Some anomalous consequences of an optimal habitat distribution. Theor. Popul. Biol. 28, 181–208 (1985).Article 
MathSciNet 
MATH 

Google Scholar 
Briers, R. A. Metapopulation ecology of Notonecta in small ponds. Doctoral dissertation. (1999).Popham, E. J. The migration of aquatic bugs with special reference to the Corixidae (Hemiptera Heteroptera). Arch. Für Hydrobiol. 60, 450–496 (1964).
Google Scholar 
Doligez, B., Cadet, C., Danchin, E. & Boulinier, T. When to use public information for breeding habitat selection? The role of environmental predictability and density dependence. Anim. Behav. 66, 973–988 (2003).Article 

Google Scholar 
Pintar, M. R. & Resetarits, W. J. Aquatic beetles influence colonization of disparate taxa in small lentic systems. Ecol. Evol. 10, 12170–12182 (2020).Article 
PubMed 
PubMed Central 

Google Scholar 
Sebastián-González, E., Sánchez-Zapata, J. A., Botella, F. & Ovaskainen, O. Testing the heterospecific attraction hypothesis with time-series data on species co-occurrence. Proc. R. Soc. B Biol. Sci. 277, 2983–2990 (2010).Article 

Google Scholar 
Giller, P. S. & McNeill, S. Predation strategies, resource partitioning and habitat selection in Notonecta (Hemiptera/Heteroptera). J. Anim. Ecol. 50, 789–808 (1981).Article 

Google Scholar 
Buxton, V. L., Enos, J. K., Sperry, J. H. & Ward, M. P. A review of conspecific attraction for habitat selection across taxa. Ecol. Evol. 10, 12690–12699 (2020).Article 
PubMed 
PubMed Central 

Google Scholar 
Ferzoco, I. M. C., Baines, C. B. & McCauley, S. J. Co-occurring Notonecta (Hemiptera: Heteroptera: Notonectidae) species differ in their behavioral response to cues of Belostoma (Hemiptera: Heteroptera: Belostomatidae) predation risk. Ann. Entomol. Soc. Am. 112, 402–408 (2019).Article 

Google Scholar 
Roughgarden, J. Evolution of niche width. Am. Nat. 106, 683–718 (1972).Article 

Google Scholar 
Ruckstuhl, K. E. Sexual segregation in vertebrates: proximate and ultimate causes. Integr. Comp. Biol. 47, 245–257 (2007).Article 
CAS 
PubMed 

Google Scholar 
Hochkirch, A., Gröning, J. & Krause, S. Intersexual niche segregation in Cepero’s ground-hopper Tetrix ceperoi. Evol. Ecol. 21, 727–738 (2007).Article 

Google Scholar 
Romey, W. L. & Wallace, A. C. Sex and the selfish herd: sexual segregation within nonmating whirligig groups. Behav. Ecol. 18, 910–915 (2007).Article 

Google Scholar 
Main, M. B., Weckerly, F. W. & Bleich, V. C. Sexual segregation in ungulates: new directions for research. J. Mammal. 77, 449–461 (1996).Article 

Google Scholar 
Trivers, R. L. Parental investment and sexual selection. in Sexual Selection and the Descent of Man 1871–1971 (ed. Campbell, B.) (Aldine Publishing Company, 1972).Bonduriansky, R. The evolution of male mate choice in insects: a synthesis of ideas and evidence. Biol. Rev. Camb. Philos. Soc. 76, 305–339 (2001).Article 
CAS 
PubMed 

Google Scholar 
Foster, S. E. & Soluk, D. A. Protecting more than the wetland: The importance of biased sex ratios and habitat segregation for conservation of the Hine’s emerald dragonfly Somatochlora hineana Williamson. Biol. Conserv. 127, 158–166 (2006).Article 

Google Scholar 
Miller, K. B. The phylogeny of diving beetles (Coleoptera: Dytiscidae) and the evolution of sexual conflict. Biol. J. Linn. Soc. 79, 359–388 (2003).Article 

Google Scholar 
Watson, P. J., Stallmann, R. R. & Arnqvist, G. Sexual conflict and the energetic costs of mating and mate choice in water striders. Am. Nat. 151, 46–58 (1998).Article 
CAS 
PubMed 

Google Scholar 
Rowe, L., Krupa, J. J. & Sih, A. An experimental test of condition-dependent mating behavior and habitat choice by water striders in the wild. Behav. Ecol. 7, 474–479 (1996).Article 

Google Scholar 
McLain, D. K. & Pratt, A. E. The cost of sexual coercion and heterospecific sexual harassment on the fecundity of a host-specific, seed-eating insect (Neacoryphus bicrucis). Behav. Ecol. Sociobiol. 46, 164–170 (1999).Article 

Google Scholar 
Stone, G. N. Female foraging responses to sexual harassment in the solitary bee Anthophora plumipes. Anim. Behav. 50, 405–412 (1995).Article 

Google Scholar 
Martens, A. & Rehfeldt, G. Female aggregation in Platycypha caligata (Odonata: Chlorocyphidae): A tactic to evade male interference during oviposition. Anim. Behav. 38, 369–374 (1989).Article 

Google Scholar 
Kolar, V. & Boukal, D. S. Habitat preferences of the endangered diving beetle Graphoderus bilineatus: implications for conservation management. Insect Conserv. Divers. 13, 480–494 (2020).Article 

Google Scholar 
Wilcox, C. Habitat size and isolation affect colonization of seasonal wetlands by predatory aquatic insects. Isr. J. Zool. 47, 459–475 (2001).Article 

Google Scholar 
Baines, C. B., Ferzoco, I. M. C. & McCauley, S. J. Phenotype-by-environment interactions influence dispersal. J. Anim. Ecol. 88, 1263–1274 (2019).Article 
PubMed 

Google Scholar 
Liao, W., Venn, S. & Niemelä, J. Diving beetle (Coleoptera: Dytiscidae) community dissimilarity reveals how low landscape connectivity restricts the ecological value of urban ponds. Landsc. Ecol. 37, 1049–1058 (2022).Article 

Google Scholar  More

Honey bee colony loss and parasites across space and timeHoney bee colony loss strongly depends on spatio-temporal factors33,42, which in turn have to be jointly modeled with other stressors. Focusing on CONUS climatic regions, defined by the National Centers for Environmental Information40 (see Fig. 1), this is supported by the box plots in Fig. 2 which depict appropriately normalized honey bee colony loss (upper panel) and presence of V. destructor (lower panel) quarterly between 2015 and 2021. Specifically, Fig. 2a highlights that the first quarter generally accounts for a higher and more variable proportion of losses. Average losses are typically lower and less dispersed during the second quarter, and then tend to increase again during the third and fourth quarters. The Central region, which reports the highest median losses during the first quarter (larger than 20%) exemplifies this pattern, which is in line with existing studies that link overwintering with honey bee colony loss6,29,30,31,32,33,43. On the other hand, the West North Central region follows a different pattern, where losses are typically lower during the first quarter and peak during the third. This holds, albeit less markedly, also for Northwest and Southwest regions. These differing patterns are also depicted in Fig. 3, which shows the time series of normalized colony loss for each state belonging to Central and West North Central regions – with the smoothed conditional means highlighted in black and red, respectively. Figure 2b shows that also the presence of V. destructor tends to follow a specific pattern; in most regions it increases from the first to the third quarter, and then it decreases in the fourth – with the exception of the Southwest region, where it keeps increasing. This is most likely because most beekeepers try to get V. destructor levels low by fall, so that colonies are as healthy as possible going into winter, and also because of the population dynamics of V. destructor alongside honey bee colonies – i.e., their presence typically increases as the colony grows and has more brood cycles, since this parasite develops inside honey bee brood cells44,45. The West region (which encompasses only California since Nevada was missing in the honey bee dataset; see Data) reports high levels of V. destructor throughout the year, with very small variability. A comparison of Fig. 2a and b shows that honey bee colony loss and the presence of V. destructor tend to be higher than the corresponding medians during the third quarter, suggesting a positive association. This is further confirmed in Fig. 4, which shows a scatter plot of normalized colony loss against V. destructor presence, documenting a positive association in all quarters. Although with the data at hand we are not able to capture honey bee movement across states, as well as intra-quarter losses and honey production, these preliminary findings can be useful to support commercial beekeeper strategies and require further investigation.Figure 2Empirical distribution of honey bee (Apis mellifera) colony loss (a) and Varroa destructor presence (b) across quarters (the first one being January-March) and climatic regions; red dashed lines indicate the overall medians. (a) Box plots of normalized colony loss (number of lost colonies over the maximum number of colonies) for each quarter of 2015–2021 and each climatic region. At the contiguous United States level, this follows a stable pattern across the years, with higher and more variable losses during the first quarter (see Supplementary Figs. S2-S6), but some regions do depart from this pattern (e.g., West North Central). (b) Box plots of normalized V. destructor presence (number of colonies affected by V. destructor over the maximum number of colonies) for each quarter of 2015–2021 and each climatic region. The maximum number of colonies is defined as the number of colonies at the beginning of a quarter, plus all colonies moved into that region during the same quarter.Full size imageFigure 3Comparison of normalized honey bee (Apis mellifera) colony loss (number of lost colonies over the maximum number of colonies) between Central and West North Central climatic regions for each quarter of 2015–2021 (the first quarter being January-March). (a) Trajectory of each state belonging to Central (yellow) and West North Central (blue) climatic regions. (b) Smoothed conditional means for each of the two sets of curves based on a locally weighted running line smoother where the width of the sliding window is equal to 0.2 and corresponding standard error bands are based on a 0.95 confidence level46.Full size imageFigure 4Scatter plot of normalized honey bee (Apis mellifera) colony loss (number of lost colonies over the maximum number of colonies) against normalized Varroa destructor presence (number of colonies affected by V. destructor over the maximum number of colonies) for each state and each quarter of 2015–2021 (the first quarter being January-March). Points are color-coded by quarter, and ordinary least squares fits (with corresponding standard error bands based on a 0.95 confidence level) computed by quarter are superimposed to visualize the positive association.Full size imageUp-scaling weather dataThe data sets available to us for weather related variables had a much finer spatio-temporal resolution (daily and on a (4 times 4) kilometer grid) than the colony loss data (quarterly and at the state level). Therefore, we aggregated the former to match the latter. For similar data up-scaling tasks, sums or means are commonly employed to summarize the variables available at finer resolution47. The problem with aggregating data in such a manner is that one only preserves information on the “center” of the distributions – thus losing a potentially considerable amount of information. To retain richer weather related information in our study, we considered additional summaries capturing more complex characteristics, e.g., the tails of the distributions or their entropy, to ascertain whether they may help in predicting honey bee colony loss. Within each state and quarter we therefore computed, in addition to means, indexes such as standard deviation, skewness, kurtosis, (L_2)-norm (or energy), entropy and tail indexes48. This was done for minimum and maximum temperatures, as well as precipitation data (see Data processing for details).Next, as a first way to validate the proposed weather data up-scaling approach, we performed a likelihood ratio test between nested models. Specifically, we considered a linear regression for colony loss (see Statistical model) and compared an ordinary least squares fit comprising all the computed indexes as predictors (the full model) against one comprising only means and standard deviations (the reduced model). The test showed that the use of additional indexes provides a statistically significant improvement in the fit (p-(text {value}=0.03)). This test, which can be replicated for other choices of models and estimation methods (see Supplementary Table S5), supports the use of our up-scaling approach.Figure 5 provides a spatial representation of (normalized) honey bee colony losses and of three indexes relative to the minimum temperature distribution; namely, mean, kurtosis and skewness (these all turn out to be relevant predictors based on subsequent analyses; see Table 1). For each of the four quantities, the maps are color-coded by state based on the median of first quarter values over the period 2015-2021 (first quarters typically have the highest losses, but similar patterns can be observed for other quarters; see Supplementary Figs. S12-S14). Notably, the indexes capture characteristics of the within-state distributions of minimum temperatures that do vary geographically. For example, considering minimum temperature, skewness is an index that (broadly speaking) provides information on whether the data tends to accumulate at one end or the other of the observed range of minimum temperatures (i.e., a positive/negative skewness indicates that the data accumulates towards the lower/upper range, respectively). On the other hand, kurtosis is an index that captures the presence of “extreme” values in the tails of the data (i.e., a low/high value of kurtosis indicates that the tail minimum temperatures are relatively close/very far from the typical minimum temperatures). With this in mind, going back to Fig. 5, we can see that minimum temperatures in states in the north-west present large kurtosis (a prevalence of extreme values in the tails) and negative skewness (a tendency to accumulate towards the upper values of the minimum temperature range), while the opposite is true for states in the south-east. More generally, the mean minimum temperature separates northern vs southern states, kurtosis is higher for states located in the central band of the CONUS, and skewness separates western vs eastern states.We further note that the states with lower losses during the first quarter (e.g., Montana and Wyoming) do not report extreme values in any of the considered indexes. Although these states are generally characterized by low minimum temperatures, these are somewhat “stable” (they do not show marked kurtosis or skewness in their distributions) – perhaps allowing honey bees and beekeepers to adapt to more predictable conditions. On the other hand, states with higher losses during the first quarter such as New Mexico have higher minimum temperatures as well as marked kurtosis, and thus higher chances of extreme minimum temperatures – which may indeed affect honey bee behavior and colony loss. Overall, across all quarters of the years 2015-2021, we found that normalized colony losses and mean minimum temperatures are negatively associated (the Pearson correlation is -0.17 with a p-(text {value} More

Environmental and competitive filtering is most important for future tree survivalWe find that individual functional traits of each tree were most important for individual tree survival (40–87%) for all study sites, followed by forest dynamics (16–28%) and functional diversity (10–26%) (Fig. 2). Nevertheless, importance proportions substantially varied in each study site. While individual functional traits were least important in the mixed mountainous forest under reference climate (no climate warming), individual functional traits showed highest influence in the mixed temperate forest under future climate change (Fig. 2).Figure 2Relative importance of functional diversity, forest dynamics and individual traits for individual tree survival under reference climate and future climate conditions (RCP 4.5). Panels correspond to alpine needle-leaved (A), mixed mountainous (B), mixed temperate (C) and temperate broad-leaved (D) forests. Left bars in each panel illustrate reference climate (no warming) and right bars future climate (RCP4.5). Colours indicate forest dynamics (grey), functional diversity (yellow) and individual functional traits (blue). Forest dynamics include the number of locally competing trees ( > 5 m in height) and local biomass as a proxy for the successional status.Full size imageTree survival depends on a mixture of environmental (e.g. climate) and natural competitive filtering, which excludes trees with trait combinations that underperform under local conditions16. Therefore, the high importance of individual functional traits across all study sites suggests a strong environmental and competitive filtering. Under future climate, the importance of individual functional traits generally increases or remains at high levels (Fig. 2). This shows that environmental and competitive filtering through functional traits are important processes to select best performing trees for the future, although being different for each forest type.Changing forest composition and trait shifts require large functional portfolio to secure forest resistanceUnder future climate, we observe trait shifts within plant functional types and strong changes of the forest composition (Figs. 3, 4). Especially in the alpine and mixed forests, the proportion of broad-leaved trees increases to at least ~ 70% towards the end of the twenty-first century (Fig. 3). The changing climate alters environmental and competitive filtering simultaneously, whereby broad-leaved trees become more productive, survive better and increasingly outcompete needle-leaved trees. For instance, in the two mixed forests survival probabilities of broad-leaved trees (high SLA) increase by about ~ 10%, whereas the survival of needle-leaved (low SLA) trees is reduced by 10% to 30% in a warmer climate (see Supplementary Figs. S7, S8, Panel A). Locally better adapted and competitive broad-leaved trees can replace needle-leaved trees if they die and secure the forest’s overall biomass in the future. Nevertheless, our simulated forests still contain significant amounts of needle-leaved trees in the year 2099 in the two coldest study areas (Fig. 3, red and blue lines). Therefore, mixed tree communities with high functional diversity, where broad- and needle-leaved trees coexist, contain a broad range of functional niches out of which the best suitable plant strategies emerge and result in better resistance to climate change.Figure 3Forest compositions and changes in the proportion of broad-leaved trees (summergreen and evergreen plant functional type combined) under climate change (RCP 4.5) from 2000 to 2099 for each study site. The fraction of broad-leaved trees, as simulated by LPJmL-FIT for each site, increases gradually in almost every forest type reaching at least about 70% by the end of the century. Pictures depict snapshots from visualization of model output in the years 2000 and 2099, respectively. For a full animation of all sites from 2000 to 2099 please see Supplementary Video 1.Full size imageFigure 4Trait distributions of specific leaf area in year 2000 and 2099, respectively, under future climate change (RCP 4.5). Arrows indicate trait shifts within plant functional types: BL-S Broad-leaved summergreen, BL-E Broad-leaved evergreen, T-NL Temperate needle-leaved, B-NL Boreal needle-leaved. For more detailed distributions see Supplementary Figs. S3 and S4.Full size imageSimultaneously, we observe strong trait shifts in SLA within plant functional types across all study sites under climate change (Fig. 4). In general, the community of broad-leaved trees shift to lower SLA, while boreal needle-leaved trees are strongly reduced or slowly replaced by their temperate equivalent with higher SLA (Supplementary Fig. S3A, dark blue colours). In contrast, wood density distributions remain relatively broad and do not shift strongly under climate change (Supplementary Fig. S4). Throughout the century, the increasingly warmer climate filters new trait combinations leading to changes in the community composition within and across PFTs (see Supplementary Discussion B). Those trait shifts emerge from changes in the composition within PFTs and newly establishing PFTs, and could be less drastic if trait adaptation of tree individuals was considered (see “Limitations and outlook” section). The points raised above show, that trait ranges within and between PFTs should be wide to cover potential future trait shifts that secure future forest resistance.All this suggests that functionally diverse forests are more resistant to future climate changes, due to their rich portfolio of traits. Broad trait distributions both within and between PFTs form the fundament for environmental and competitive filtering to select the most productive trees, securing the forest’s overall biomass under changing conditions. But can functional diversity further strengthen forest resistance beyond portfolio effects?Functional complementarity helps young trees to surviveWe find that, in addition to port-folio effects, functional diversity increases forest resistance by supporting the survival of young trees to changing climate conditions via trait complementarity. Our results indicate, that trees benefit from functional diversity if they grow in tree communities with high FR, high FDv and low FE (Supplementary Figs. S6–S9, Panel D–F in each figure). Here, functional traits lay highly separated (FDv and FE) and span a broad range in the functional trait space (FR), enabling functional complementarity. Under these conditions the survival of trees increases up to + 16.8% (± 1.6%) depending on the study site and climate (Table 1). This effect is highest in the alpine and mountainous forests (14–17%), whereas it is less prominent or has an opposite effect in the two temperate forests (− 7% to 6%). That suggests, that complementarity effects are stronger in cold-limited and mixed forests where a marked cold winter season fosters a co-existence between broadleaved and needle-leaved trees. Both PFTs are specialized in fixing carbon during different times of the annual cycle: Due to their leaf phenology, needle-leaved trees can already be productive when broad-leaved trees are still in progress of unfolding or shedding their leaves. On the other hand, broad-leaved trees are more productive than needle-leaved trees during warmer months. If coexistence is given, these phenological differences enable complementarity and reduce competition among PFTs. That overall increases tree survival, because trees can invest more carbon in their stems and defensive structures if competition is lower. Therefore, we argue that phenological complementary can enhance tree survival and thus forest resistance. An in-depth discussion of those mechanisms is further found in Supplementary Discussion A.Table 1 Additional survival probabilities for trees in each forest site under reference climate (central column) and future climate (RCP4.5, right column) in case FR and FDv are high, while FE is low.Full size tableSurprisingly, our results show that those complementary effects are much more important for small trees ( 10 m) in right panel, respectively. Functional diversity and forest dynamics are more important for small trees compared to large trees, whereas individual functional traits matter most for large trees. This pattern was found to be consistent across all sites (see Supplementary Table S6).Full size imageFunctionally diverse understoreys unlock the synergy of filtering and complementary effectsOur findings underline the role of functionally diverse trees in the understoreys for forest resistance. On the one hand, functional diversity supports the survival of understorey trees via functional trait complementarity. On the other hand, they form the fundament for competitive and environmental filtering. Only diverse tree communities have trait pools large enough to ensure that their tree portfolio holds trait combinations best suited for changing climate conditions. Therefore, we argue that functional diversity does not only support tree survival through complementarity, but is a prerequisite for filtering resistant trees in the first place.To profit constantly from functional diversity of the understory and ensure constant adaptation, a diverse age structure is a prerequisite. Depending on the forest type, trees are distributed in a broad range of different height and age classes in our study (Supplementary Fig. S5, Supplementary Video 1). This multi-aged structure is preserved under climate change (Supplementary Fig. S5) and allows gradual changes through constant environmental and competitive filtering in the future.In this study, we simulate forests without any human interference or management. Our results are therefore to be interpreted in the context of environmental and competitive filtering as observed in natural forests. Most managed forests lack this natural filtering effect as they are less dense and diverse in their age-structure. Functionally diverse trees in the understorey could provide the fundament for climate adapted multi-aged forests, as they constantly form new better adapted tree generations with natural competition and succession allowed. Therefore, we fully underline the importance of functionally diverse understorey trees and natural competition as the fundament for future forest resistance.Management implicationsThe results of this study highlight the importance of functionally diverse understorey trees. However, browsing by game might damage new tree saplings and limit tree diversity in the understorey. In addition, invasive species like Prunus serotina or herbaceous competition might hinder forest succession and the establishment of woody native species in European forests20,21. Therefore, regulating game, limiting the spread of invasive species and controlling herbaceous vegetation should be considered in future management practices where tree diversity in the understorey seeks to be increased or maintained. Moreover, insufficient dispersal of functionally different tree species might limit the establishment of functionally diverse trees in the understorey. Future forest management may consider to artificially plant functionally different tree species if dispersal from surrounding forests cannot be guaranteed. On the other hand, forests that already contain functionally diverse trees in the understorey should be preserved.In this study, a clear trend from needle-leaved to broad-leaved trees is captured at all sites, whereas within broad-leaved PFTs a shift to lower specific leaf area and higher leaf longevities indicates that future forests might especially benefit from longer vegetation periods (earlier leaf onset, later senescence). Therefore, forests containing broad-leaved tree individuals with high phenological plasticity could be more resistant. The broad simulated wood density ranges, which persist under climate change, imply beneficial effects for forest communities entailing a range of different growing strategies, i.e. early to late successional species. Therefore, we argue that forest fragmentation should be reduced or reversed to foster some natural dispersal of early and late successional species.This study intended to explore the potentials of functionally diverse forests as a possibility to stabilize forests under climate change over a large climatic gradient. The model used in this study operates on the more general level of functional traits and their diversity rather than on species level (see “Methods” section and Supplementary Methods A). Consequently, management implications regarding suitable specific tree species are beyond the scope of this study. However, we think that our results will stimulate the discussion on the importance of functional tree traits and their diversity for species selection.Limitations and outlookThis study focussed on identifying the importance of functional diversity for future tree survival to advance our understanding on the role of biodiversity for future forest resistance using the flexible-trait Dynamic Global Vegetation Model LPJmL-FIT. The general approach of LPJmL-FIT is to simulate biogeographic dynamics purely based on environmental and competitive filtering (see “Methods” section, Supplementary Methods A). Due to missing processes in the model and the ambiguity of former human influence, drawing site-specific implications on future forest dynamics must be taken with caution (see Supplementary Discussion D).Moreover, processes not yet captured in the LPJmL-FIT model, might play a role and could lower forest resistance in the future, which is why we recommend relying on a trait space as broad as possible.Including more climate scenarios would widen the envelope of possible future pathways by considering climate model uncertainties. Insect outbreaks and pathogens might put pressure to the already drought- and temperature-stressed trees and heavily accelerate mortality especially of needle-leaved trees, although functionally diverse forests are less vulnerable to bark beetle outbreaks22,23. Multi-layered forests showed higher growth resilience to structural disturbances such as wind-throw24, likely enhancing the importance of individual tree height and reduce the survival probability of large trees25. Belowground competition and trait plasticity could favour complementarity effects further. Variable rooting strategies could further reduce competition for soil water and thereby increase individual drought resistance of trees26. Trait plasticity can contribute to tree survival by widening niche, further increasing complementarity effects. However, trait plasticity remains one of the most challenging objectives in vegetation modelling as observational data and modelling approaches are scarce27, leaving it open how far trait relationships would hold under climate change. Considering more functional traits in our analysis might increase the overall predictive power of the random forest models. Even though the explained variance increased with the number of analysed traits explaining ecosystem properties in long-term grassland experiments, such improvement is limited as abiotic factors and their interactions with plant traits might be more important for prediction28. We conclude that simulating future forest dynamics dominated by environmental and natural competitive filtering requires to integrate both, abiotic and biotic drivers on forest dynamics. Machine learning techniques are increasingly used in forest ecological research—but mainly applied in the processing of field and forest inventory data29,30. Machine learning can help to understand the complexity of interactions and provide deeper insights into the underlying ecological process in a modelling study as we have shown here using LPJmL-FIT simulation results. Random forest analyses are suitable for a variety of data and applications because they are relatively robust to different data structures. Importance analysis can help to identify the role of underlying processes in complex models and to visualize their changes in a simple way. In doing so, model development is advanced by making use of large data sets, opening the door to further theory building and deeper understanding of plant trait ecology. More

Overview of modeling frameworkWe have used a range of econometric, economic, and agricultural land surface models to analyze the factors driving land-use change in order to assess their ecological, agricultural, climatic and economic impacts. These multi-scale models differ in their methodologies, scale of interest, and resolution, but they are very complementary and could provide a unique opportunity to analyze public policy scenario effects on land-use and resulting changes in ecosystem carbon and biodiversity.Among these models, the economic land use model Nexus Land Use (NLU)29,30 and the agricultural supply-side model Agriculture, Recomposition de l’Offre et Politique Agricole (AROPAj)31 coupled with a spatial econometric model32 have allowed us to estimate the impact on EU land-use of a scenario involving a 50% reduction in N synthetic fertilizers compared to a baseline scenario. In the present study, we use these land-use scenarios to force ORCHIDEE-crop (Organising Carbon and Hydrology in Dynamic Ecosystems), an agricultural land surface model16,33 and Projecting Responses of Ecological Diversity in Changing Terrestrial Systems (PREDICTS)34, a biodiversity model to simulate, respectively, ecosystem C and biodiversity changes across the EU covering the domain 35.25°N and 69.25°N in latitude and 9.25°W and 34.25°W in longitude. The schematic (Fig. 1) provides a brief overview of the modelling framework applied in this study.Figure 1Schematic diagram illustrating the coupling of multi-scale land-use models. The multi-scale models coupled in this study are econometric, and economic models (NLU and AROPAj), an agricultural land surface model (ORCHIDEE-crop), and a biodiversity model (PREDICTS). Coupling means, we use the output of one model as an input to other models. In addition, we have performed one-way coupling and there is no two-way interaction between models. Each economic model generates two land-use maps corresponding to Baseline and Halving-N scenario which are inputs (2 from NLU and 2 from AROPAj) to ORCHIDEE-crop and PREDICTS. The ecosystem carbon (C) sequestration is simulated by ORCHIDEE-crop and biodiversity indicators are simulated by PREDICTS model. The abbreviations ‘BaseNLU’ and ‘HaNNLU’ means Baseline and Halving-N land-use map generated by NLU model. The abbreviations ‘BaseAR’ and ‘HaNAR’ means Baseline and Halving-N land-use map generated by AROPAj model.Full size imageIn order to link the land use output data from the AROPAj and NLU models with the ORCHIDEE-crop and PREDICTS models, the first step is to match land uses and crops between the models (see Table 1). AROPAj and NLU crops are classified into ORCHIDEE-crop plant functional types (PFTs): C3 winter and summer crops, C4 summer crop and C3/C4 natural grass (see “Model descriptions” section for a detailed description of ORCHIDEE-crop PFTs). The AROPAj and NLU crops are also classified into the PREDICTS crop types: annual, perennial, N-fixing. The AROPAj and NLU “rangeland” and “pasture” categories are found in PREDICTS but in ORCHIDEE-crop they are considered to fall within the C3 natural grassland PFT. Finally, NLU and AROPAJ forest and other natural areas are classified as “primary” natural areas (with low anthropogenic use) or “secondary” (intermediate to high anthropogenic environmental use) according to the land use map of these areas35. For ORCHIDEE-crop, they are classified as natural forest PFTs. Note that the fallow areas described in AROPAj that are part of crops are classified as “grass” PFT in ORCHIDEE-crop and as “minimum” intensity annual crops in PREDICTS.Table 1 Table of correspondences between the land uses and crops represented in the AROPAJ/NLU and ORCHIDEE models and PREDICTS.Full size tableThe land-use and land cover changes described in the following sub-section are used as inputs to ORCHIDEE-crop and PREDICTS from both the NLU and AROPAj models’ output.Land-use change scenariosLand-use changes in the EU are simulated for the present day using two scenarios: (1) a business as usual scenario (Baseline) and (2) a scenario involving a policy to reduce mineral nitrogen use by 50% from the Baseline (Halving-N). The land-use changes in Halving-N and Baseline are computed by both NLU and AROPAj models. In the latter model, the computed land-use changes result from coupling between AROPAj and a spatial econometric model. Since there are differences in the nature of the models (supply-side model versus partial equilibrium model) and their underlying data, the Baseline scenarios in the NLU and AROPAj frameworks are different. A detailed description of the differences and a discussion of their implications on the production and area of different land-uses is provided in Lungarska et al.36. EU plant production is 370 and 383 MtDM (Million tons of Dry Matter) respectively based on the application of 12 TgN (Tera grams) of N fertilizer in AROPAj and NLU. Crops, grasslands, and forests cover respectively, 116, 57 and 234 Mha in NLU and respectively 94 (including fallow land), 38 and 142 Mha in AROPAj. In AROPAj and NLU, the 50% N reduction is achieved indirectly by increasing the N input price from present-day figures36.The land-use changes output from AROPAj and NLU are supplied as inputs to the ORCHIDEE-crop and PREDICTS models. The land-use changes are matched with corresponding plant functional types (PFTs) in ORCHIDEE-crop and land-uses in PREDICTS (see Table 1). “Model descriptions” section provides a detailed description of the ORCHIDEE-crop and PREDICTS models.Model descriptionsHere, we describe the ORCHIDEE-crop and PREDICTS models that quantify the impacts of halving N fertilizer consumption in the EU. Table 2 presents a brief overview of the two models.Table 2 Overview of the ORCHIDEE-crop and PREDICTS models input and output.Full size tableA detailed description of ORCHIDEE-crop: This model is a process-based agricultural land surface model that integrates crop-specific phenology based on Simulateur mulTidisciplinaire pour les Cultures Standard (STICS)37,38. Carbon allocation is based on the plant-based hybrid model from the original ORCHIDEE allocation scheme39 and a crop specific formulation of STICS providing leaf, root, and shoot biomass, grain maturity time, litter production, and litter and soil carbon decomposition. The harvest date is calculated after grains reach maturity40. The ORCHIDEE-crop model has no explicit nitrogen cycle but accounts empirically for the effect of N fertilization by increasing the maximum Rubisco- and light-limited leaf photosynthetic rates as a function of the amount of N applied, using a Michaelis–Menten function40. Also, ORCHIDEE-crop is calibrated against observations, which showed a good match between modeled observed aboveground biomass, crop yield, and daily carbon40. This version of the model currently uses three crop PFTs: C3 winter, C3 summer and C4 summer. Forests are classified as Broadleaf, Needle leaf, Deciduous, Temperate and Boreal. Up to 11 non-cropland vegetation types can co-exist with crops on a grid point of the model, according to prescribed land cover information. A gridded simulation of ORCHIDEE-crop requires 30-min time step meteorological forcing (air temperature, specific humidity, incoming shortwave and longwave radiation, rainfall), which can be interpolated in time from gridded climate analysis data or atmospheric models. In this study, this model is used to quantify the ecosystem C variables.A detailed description of PREDICTS: The PREDICTS database was collated by searching the published literature for studies where terrestrial biodiversity (including plants, fungi, vertebrates, and invertebrates) was sampled using consistent methods across multiple sites, which vary in the pressures faced. The land use and intensity of each site have been assessed and categorized in a consistent way41,42,43. Authors of studies were contacted to ask for the raw biodiversity data where this was not already available41,42. Most records in the PREDICTS database refer to the number of individuals of a species at a site; this makes it possible to compute a range of biodiversity indices. To estimate biodiversity responses to human impacts across such a global and heterogeneous dataset, linear mixed-effects models are used; random intercepts account for differences in biogeographic factors, sampling methodology and taxonomic focus, and the spatial layout of sites within studies. Using the PREDICTS database to assess the impact of human pressures on biodiversity assumes that space-for-time substitution is valid44; it assumes that the sites have reached equilibrium and so the impact of pressures on biodiversity over time can be observed across space and that the relationship between biodiversity and drivers do not vary over time.SR is calculated as the number of species at each site; it is a widely used measure of biodiversity and is both simple and intuitive. Responses of SR to land use and intensity were modelled using generalized linear mixed effects models and with a Poisson error structure; an observation-level random effect was included to account for overdispersion45. This model is then used to project SR in each grid of a 0.5° map and expressed as a percentage of the SR level in primary vegetation from land use harmonization map35.To estimate BII change with land use and intensity, two models are required. Total abundance was first calculated as the sum of all individuals at each site; it was then rescaled within the study (so that the maximum within a study is 1) and was square-root transformed before modelling as a function of land use and intensity, to account for non-normality of the model residuals (a Poisson error structure could not be used as abundance data can include non-integer data e.g. densities). Inclusion of a random slope for land use within the study was supported (based on Akaike’s Information Criterion). Compositional similarity was then calculated as the asymmetric Jaccard index, comparing each baseline site (primary vegetation) with all other sites, and logit transformed with an adjustment of 0.01 (to account for non-normality of the model residuals). Compositional similarity was then modelled as a function of land use and intensity (coarsened so that only perennial crops were allowed to differ across intensities), including the environmental and geographic distance between sites as control variables, whose effects were permitted to differ among land use and intensity levels (these variables were cube-root and log-transformed respectively to improve residual distribution). To calculate BII, total abundance (expressed as a percentage of their level in primary vegetation) and compositional similarity (expressed as a percentage of their level in primary vegetation)46 are projected for each grid of a 0.5° map; these two maps are then multiplied to give abundance-based BII19. The PREDICTS models include different levels of management (intensive, light or minimal) and different types of land cover (forest, pasture, rangeland, annual cropland, perennial cropland, and urban zones). The coefficients of these mixed-effect models and a detailed description of the link between the PREDICTS models and NLU are available in Prudhomme et al.46. The spatial predictions of biodiversity were computed using a python pipeline, which was developed specifically for the PREDICTS project (https://github.com/ricardog/raster-project).In our modeling framework, the impact of halving N fertilizer goes through two steps: (i) we calculate the effect of this reduction of N fertilizer on agricultural yield, and (ii) calculate the effect of the yield reduction on biodiversity. By keeping yield as a proxy of agricultural land use intensification as proposed in Prudhomme et al.46, we include not only the direct effect of the reduction of N fertilizer on biodiversity but also the effects correlated to this reduction of N fertilizer such as the reduction of other chemical inputs (P and K fertilizers and pesticides). While the effect of the change in N fertilization on yield is calculated by the classical concave production function in agronomy29, the effect of the change in yield is calculated by coupling the NLU land use model and the PREDICTS biodiversity model46. For each category of crops (annual, perennial, leguminous), the coupling consists of estimating (using a Generalized Additive Model [GAM]) the share of each intensity class (minimum, light, intense) as a function of the average calorie yield based on the average crop yield maps from a plant growth model. The maps describing the share of land use intensities are from Newbold et al.19 Similarly for pasture, the share of each intensity class (light, intense) is estimated with the help of a GAM as a function of ruminant density.SimulationsOur experimental design focuses on assessing the effects of a 50% reduction in present-day N fertilizer use levels across the EU. The choice of halving N fertilizer in EU agriculture is related to the “Farm to Fork” strategy, which puts forward the ambition for 2030 to reduce nutrient losses to the environment from both organic and mineral fertilizers by at least 50%. The results from NLU (and its nitrogen balance module) show that this level of reduction corresponds to a 50% reduction in nutrient losses (nitrogen and phosphorus) aimed by the Farm to Fork strategy as a part of the European Green Deal. AROPAj models exclusively the EU countries (in 2012, there were 28 member states) while NLU simulations cover the EU and the rest of the world (EU being a part of the European region as represented by the model). However, the N reduction policy implemented in the EU alone and the comparison of the results conducted only for the EU. All EU member states are considered but for some of them we present results. A total of four simulations corresponding to four land-use maps (two from AROPAj and two from NLU, see Fig. 1) are performed in the ORCHIDEE-crop model and also in the PREDICTS model. In addition to changes in the area of different land-uses, changes in mineral N input are accounted for in both models. However, changes in organic N input and crop rotations are not accounted for. In ORCHIDEE-crop 55% of the carbon harvested from croplands is exported but the remaining residues are returned to the soils.ORCHIDEE-crop simulation details: the model simulations are performed over a domain covering the EU. Four idealized simulations are carried out using the ORCHIDEE-crop model by forcing present-day meteorological data (2006–2010), levels of N fertilizer (150 KgN/ha) and atmospheric CO2 concentration (385 ppm). The four simulations include Halving-N and Baseline corresponding to AROPAj and NLU land-use scenarios (two ORCHIDEE-crop simulations per economic model). All four simulations start from the year 2010 climate and carbon cycle conditions with a recycled climate (2006–2010) for 150 years. For the year 2010, climate and carbon cycle conditions are obtained from the output of historical simulations. Historical simulations from the year 1901 to the year 2010 are performed for both AROPAj and NLU Baseline scenario land-use land cover maps. In addition, these historical simulations started from an equilibrium state of soil carbon, energy and water cycle variables corresponding to the year 1901. The 1901 equilibrium state is determined by running a 350-year spin-up simulation corresponding to a recycled climate (1901–1910). The observation-based climate forcing data from the Global Soil Wetness Project was only available starting from the year 1901. The drift in soil carbon over the last 100 years of the 350-year simulations is less than 1%. The equilibrium state simulations corresponding to the year 1901 were necessary to have stabilized biophysical and ecosystem C variables across the EU. Other forcing variables, e.g. atmospheric CO2 concentration (296.57 ppm), N-fertilization rate (32 KgN/ha), harvest index (0.25), and also the phenology parameters for short-cycle variety winter and summer crops16 corresponding to the year 1901 were prescribed.PREDICTS simulation details: the PREDICTS model represents changes in broad-sense biodiversity in different land-uses and intensities of land-use relative to a reference land-use (as the biodiversity metrics assessed include all terrestrial biodiversity for which data are present in the PREDICTS database including plants, fungi, vertebrates and invertebrates). Here the reference ecosystem is a primary natural ecosystem. Biodiversity changes are then reported as a percentage by dividing the obtained biodiversity levels by the level of biodiversity present in the primary natural ecosystem. This simulation is performed for each grid point on a map of the EU for land-use scenarios corresponding to Baseline and Halving-N for both economic models, AROPAj and NLU (Fig. 1).Breakdown method for biodiversity and carbon changesThe Halving-N and Baseline scenarios provide contrasted land-use maps according to the assumptions of economic and land-use models36. This results in different plant and animal production, and different land-uses at the European scale in each model. A price shock on inputs, as represented in the Halving-N scenario compared to the Baseline scenario, can induce (1) a spatial reallocation of production or (2) production changes47. Here, we separate out the effects of these two mechanisms on biodiversity (species richness) and carbon indicators (NPP and soil carbon) by decomposing the overall environmental differences between the Halving-N and the Baseline scenarios. The breakdown is not possible for the BII indicator because this indicator is the product of two indicators: abundance and a similarity indicator of ecological communities.First, we breakdown the carbon and biodiversity differences by land-use type. The breakdown for carbon is straightforward because the carbon changes are computed for each land-use. The biodiversity changes associated with each land-use are computed by setting no changes in the other PREDICTS model land-uses. The sum of the biodiversity changes for each land-use is thus equal to the overall change in biodiversity.For each land-use i (forest, grassland and cropland), we separate out the carbon and biodiversity differences between the Halving-N and the Baseline scenarios into two effects in accordance with Eq. (1): (i) the carbon and biodiversity difference associated with the area difference—called “Area effect”, and (ii) the carbon and biodiversity difference associated with the difference in biodiversity and carbon sequestration per unit area—called “Intensity effect”. The “Area effect” corresponds to the change in carbon sequestration and biodiversity associated with a change in the land-use area. For example, a reduction in grassland area leads to reduction in the C sequestration and biodiversity associated with this area. The “Intensity effect” corresponds to a change in the C sequestration and biodiversity per unit area. For example, a reallocation of production toward places with high soil C content leads to an increase in the carbon stock per hectare or an increase in crop yield leads to a reduction in the biodiversity per unit of cropland. Thus, the “Intensity effect” corresponds to the effect of a production reallocation on C sequestration, and the effect of land-use intensity on biodiversity.We use the Logarithmic Mean Division Index (LMDI) method, which breaks down the target values into several main influencing factors based on mathematical identity transformation48 as follows.$$Delta {E}_{i}=Delta {E}_{i}^{A}+Delta {E}_{i}^{I}$$
(1)
(Delta {E}_{i}) is the difference in the environmental indicator between the Halving-N and the Baseline scenarios. Superscript ‘A’ denotes area effect and ‘I’ denotes intensity effect. Subscript ‘i’ denotes different land-use (e.g. forests, grassland, cropland etc.). (Delta {E}_{i}^{A}) is the difference in the environmental indicator between the Halving-N and the Baseline scenarios associated with the difference in area. (Delta {E}_{i}^{I}) is the difference between the Halving-N and the Baseline scenarios associated with the different intensity per unit of area of the environmental indicator.$$Delta {E}_{i}^{A}=frac{{E}_{i}^{hN}-{E}_{i}^{b}}{ln({E}_{i}^{hN})-{ln(E}_{i}^{b})}times lnleft(frac{{A}_{i}^{hN}}{{A}_{i}^{b}}right)$$
(2)
({E}_{i}^{hN}) is the level of the environmental indicator in the Halving-N (superscript hN) scenario. ({E}_{i}^{b}) is the level of the environmental indicator in the Baseline (superscript b). ({A}_{i}^{hN}) is the area of land-use i in the Halving-N scenario. ({A}_{i}^{b}) is the area of land-use i in the Baseline$$Delta {E}_{i}^{I}=frac{{E}_{i}^{hN}-{E}_{i}^{b}}{ln({E}_{i}^{hN})-ln({E}_{i}^{b})}times lnleft(frac{{e}_{i}^{hN}}{{e}_{i}^{b}}right)$$
(3)
Equation (3) is same as Eq. (2) but for the intensity of the environmental indicator ({e}_{i}).The breakdown of the differences in the environmental indicators is performed between the Halving-N scenario and the Baseline. A positive variation ((Delta {E}_{i} >0)) indicates a higher environmental indicator in the Halving-N scenario compared to the Baseline without implying any temporal variation since the scenarios compare the environmental indicator status in 2012 in the AROPAj and in the NLU land-uses. Conversely, a negative variation ((Delta {E}_{i} More

UPRLIMET is our response to a need for a consistent method for predicting the upper extent of trout in all streams across land ownerships within our region. By developing and implementing the model using LiDAR-derived flowline hydrography, we offer a standardized, spatially explicit, spatially contiguous (where LiDAR hydrography is available), and high-quality fish-distribution layer based on the probability of fish presence. UPRLIMET maps both the probability of trout and the upper limit of trout across landscapes, ownerships, and jurisdictions, and better captures the upper extent of fish in headwater reaches relative to previous approaches allowing for a cross-boundary distribution map on which decision-makers and managers can base policies and regulations.This work provides a transferable prediction modeling framework for systematically and comprehensively estimating the upper distribution limit of fish, which could be calibrated and implemented in watersheds and for fish species around the globe. Although the dependency on LiDAR-derived data here may be seen as a limitation to broader implementation of this method, the method is scalable to any resolution, and LiDAR is becoming increasingly ubiquitous in the United States through the U.S. Geological Survey 3D Elevation Program, which is funding LiDAR acquisitions across the United States. Furthermore, LiDAR data is available globally via data from GEDI and ICESAT-2 satellites that offer coarser resolution (~ 25-m) data that are still superior to either ASTER or SRTM derived-DEMs26, 27.Minimizing prediction errors for the upper limit of trout is important to decision support and management planning because it ensures that forest-harvest regulations and management prescriptions are aligned. It is important to note that the prediction error estimates from this study are derived from the NSpCV process, except for models using 20% slope thresholds or unaltered parameterization of Fransen’s model13, because it is likely that the NSpCV estimates are conservative. They tended to overestimate error, as evidenced by the fact that the Refit model (i.e. Fransen’s optimal model13 refit to our data) exhibited a larger MAE than the unchanged optimal Fransen model13. This unexpected result was likely due to applying the NSpCV routine on the Refit model, resulting in the use of many intermediate models to characterize predictive performance using randomized subsets of independent training and test data. In contrast, the optimal Fransen model13 was developed independently using the data in this study and thus error could be evaluated directly without subsampling imposed by NSpCV.The relatively low error for the two-stage model that becomes UPRLIMET suggests that it more accurately characterizes the upper limit of fish than all other models considered in this study, including the Fransen model13, which has been used for estimating upper limit of fish regionally. Although some of the models exhibited relatively small differences in error relative to the model that became UPRLIMET, small differences in predicted upper limit locations when considered in aggregate across multiple watersheds can potentially alter management decisions and expected outcomes. Differences in predictive performance and error between UPRLIMET and the optimal Fransen model13 are likely attributed to high-accuracy hydrography and hydro-topographic data (as LiDAR-derived DEMs were not available in western Washington in 2006), which allowed a finer-scale of analysis (i.e., 5-m vs 10-m reaches). Additionally, the fact that UPRLIMET was fit to data solely from western Oregon likely offers predictive performance gains when applied to western Oregon when compared to the Fransen model13 that was fit to data from western Washington.Quantifying the predicted accuracy associated with applying UPRLIMET to western Washington will require new data and is outside the intended scope of this study. However, we think it is reasonable to infer findings from UPRLIMET across regions with similar climatic and hydro-topographic conditions including northwestern California, western Oregon, western Washington, and southwestern British Columbia, especially given the broad availability of LiDAR-derived DEMs. This conclusion is supported the fact that both the Fransen13 and Refit models produced similar logistic regression coefficients (Data S5) and similar Matthews Correlation Coefficients (Data S6), suggesting that feature space of the two models is similar. This evidence is further corroborated by the high degree of overlap observed among the distributions of each of the four predictor variables for both western Oregon and Washington. We acknowledge that UPRLIMET does not contain identical predictor variables to Fransen’s model13 but maintain that they are similar enough in purpose that it is reasonable to assume that the feature space similarities are retained.When we undertook this study, we hypothesized that a prediction model based on RF would offer superior predictive performance over those based on LR, given the availability of 67 predictor variables and RF’s demonstrated superior predictive performance in ecological applications23,24,25. However, our results suggests no improvement is offered by including more than four of the 67 environmental predictors examined, and that no clear advantage is offered by employing the more complex RF model, as evidenced by the top three of the top five prediction models being four-variable LR model algorithms (Fig. 3; Data S3.) The general importance of these variables to so many models is likely due to the strong linear relationships in the response of fish or no fish in logit space given the slopes of the curves in the partial dependence profiles (Fig. 4). This finding is congruent with the fundamental premise of LR, which is to explain and predict a response with a functional relationship, whereas RF deliberately focuses only on maximizing prediction accuracy with many decision trees28. Additional advantages to prediction models based on LR include the following: relatively better extrapolation performance over RF29, the simplicity of transferring a LR model to another processing platform using the model coefficients (versus the black box of RF decisions), and the immensely reduced computational processing times associated with LR model fitting and prediction. These advantages are especially key to this work, where there may be a desire to implement the model on other landscapes without the requisite expertise in doing so using the R software30. However, there are tradeoffs, as LR is more sensitive to the influence of outliers and multi-collinearity among variables, and overfitting is an increasing concern as the number of predictor variables increase, whereas RF tends to be robust to these concerns, but is more likely to produce a high-variance, low-bias prediction model31.Although there is no single, general explanation for distribution limits of species32, the intersection of stream size, slope, and elevation together locate the upper limit of fish. Stream size corresponds to major ecosystem changes along a stream continuum including for energy sources, ecosystem metabolism, habitat characteristics, and biodiversity33, as well as the upper distribution limit of fish, as shown here. As expected, stream size accounts for the top two variables in the model suggesting that it is the major driver of the upper distribution limit of fish with the probability of trout increasing with increasing upstream stream length and upstream drainage area. Our finding proposes that downstream stream reaches are more likely to have fish. Although the underlying mechanisms have multiple influences, factors related to increasing stream size, such as increasing habitat size, habitat complexity, stability, or temperature variability34 have been shown to be important. Similarly, stream size is the most sensitive factor in intrinsic potential models for Chinook Salmon (O. tshawytscha35). Slope, the next variable of importance influencing the upper extent of fish, exerts control on physical habitats in streams, including channel morphology, hydraulics, sediment transport, substrate, and habitat36. Steep slopes drastically prevent trout from reaching areas above waterfalls or impassable chutes of over 25% slope, but trout can be found in streams channels without barriers at slopes as high as 28%7, 14, 37. Other fishes, such as Coho Salmon (O. kisutch) and steelhead (O. mykiss) are generally not found above 12% slope38. Interestingly, survival of fishes that make it upstream or are introduced above barriers may be facilitated by a geomorphic setting that is less prone to debris flows and other episodic sediment fluxes and has a greater resilience to flooding resulting from wider valley and greater floodplain connectivity39. Elevation or vertical topographic position may indirectly integrate broad influences of other landscape-scale or climate factors or also indirectly capture stream size, influencing the likelihood of fish presence. Frequently, species richness increases at lower elevations40, and we suggest that elevation also contributes to species distribution limits, as is the case for the Endangered Species Act listed Bull Trout (Salvelinus confluentus)41. The multiple factors associated with elevation correspond to the relationship found for stream size that smaller streams are less likely to have fish. Ultimately, the intersection of stream size, slope, and elevation guide us to finding the upper extent of fish in streams.Physical influences have been proposed to be more limiting to fish distributions upstream, such as near the upper extent of fish, whereas biological factors are probably more important downstream33. Although 67 environmental predictor variables representing geologic, soil, climatic, and hydro-topographic conditions at local and patch scales are evaluated (Data S1), only the hydro-topographic variables of stream size, slope, and elevation are important to predicting the upper limit of fish in UPRLIMET. In fact, the top 9 models (Fig. 3; Data S3) relied on just four to five hydro-topographic variables, most of which were patch-scale variables or elevation at 1000 m, all of which incorporate a broader extent of influence. This suggests that local scale variables that contribute to fish limits, including slope or riparian influences may need to be further explored. In addition, some of the remaining 63 variables present in UPRLIMET, such as precipitation and air temperature, are important drivers of within-network trout distributions and contribute to their connectivity. Some of these predictor variables appear in the 10th ranked 26-variable RF-O-SR1 model (Data S2; Data S4; Data S8), but the influence appears to be dubious for isolating the upper limit and explaining variation in fish occurrence because MAE of upper limit was substantially higher than the 9 models with lower MAEs (Fig. 3; Data S3), and the lower MCC of the associated RF-O sub-model (Data S6). It is likely that other combinations of the 67 predictor variables, including precipitation, may be more important when this model development and evaluation framework is applied elsewhere, especially if those areas contain fishes or are places that are vulnerable to changing water temperatures and streamflow regimes. In addition, biological factors may be a concern in other watersheds, including invasive species and fish stocking which can limit the longitudinal distribution and the upstream extent of fishes.Given the large geographic extent of this study, we expected other variables such as precipitation to be more important drivers, however due to a combination of a wet water year, a lack of precipitation gradient in the study area, coarse grain data, and location of fish in streams this was not the case. For example, 2017 was a wetter than normal water year53, and it may be that the gradient of precipitation variation in western Oregon was not strong enough to explain the variation in the spatial distribution of trout occurrence. All climate data, including the precipitation data were sourced from relatively coarse-scale (800 m) PRISM data. The inability to adequately downscale precipitation to characterize how precipitation truly varies within and between patches, especially along elevational gradients, likely confounded how the model interprets the influence of precipitation. Trout occurrence was on perennial streams, which is likely far enough downstream of locations where variation in precipitation was the dominant influence on streamflow permanence and consequently would not have been a factor.Stream network structure plays a key role in the upper limits of fish. Upper limits for fish can occur at either lateral or terminal points13 and when mapping these points, differences were seen for UPRLIMET relative to other datasets. Lateral limits end in the tributary stream just above where it connects with a mainstem stream. Terminal limits include both mid-stream terminal limits where fish drop out in the middle of a stream channel owing to a soft (i.e., transient barrier or puttering out) or hard (i.e., waterfall) edge, and confluence terminal limits where the upper limit of fish ends at the confluence. For example, when closely examining the 14 watersheds where we have overlapping information across various datasets and models, UPRLIMET and the Fransen optimal model13 exhibit substantial agreement in their lateral limits. However, the largest differences are in their terminal ends, especially terminal mid-stream limits, probably owing to hydro-topographic changes that contribute to fish occurrence at confluences, which are more pronounced than mid-stream. Accordingly, the logic in the stopping rule is likely important in identifying specific upper extent of fish distributions in reaches that end mid-stream.Differences among databases for the upper distribution limits of fish come from both the upper limit points and depiction of fish-bearing reaches, underscoring the importance of having a shared map with common coverage of the fish extent across landscapes and ownerships. Differences among mapped distributions can result from source information, relating to whether it is modeled or occurrence data. Models, such as UPRLIMET, can be applied across a broad extent based on model parameters and training data, thereby offering broad coverage for distributions (and quantifiable error) across the landscape, ownerships, and jurisdictions. However, models are limited by accuracy and fit. As such, they can incorrectly predict distributions in some areas, especially if there are prediction features not yet trained with the model data where prediction would require extrapolation of the model. This makes both the training dataset and modeled extent important considerations, as models are only as good as the data used to develop them. Updating UPRLIMET with new data as it becomes available will help to expand the prediction domain, improve accuracy, and allow the model to do more interpolation than extrapolation.Distributions based on occurrence information depend heavily on data availability, data quality, and access. Differences in data availability can lead to inconsistent coverage across landscapes and ownerships, with high coverage in some watersheds and low to no coverage in others. Inconsistent coverage can lead to errors that are difficult to quantify across landscapes, ownerships, and survey crews. Occurrence information also depends on the ability to survey watersheds and gain access across ownership types, including on private lands that do not have the same assurances of access as public lands, resulting in information asymmetry42, 43. Data quality also depends on the spatial accuracy of the points of uppermost fish, which are a function of GPS quality and error, and can drastically change the modeled results, as these points are used in the training dataset. Differences among mapped distribution limits also result from differences in field protocols on designating last fish. For example, some crews note fish distribution limits where they visually see the last fish, whereas others note it upstream of where they saw last fish, based on habitat features that would limit fish. With the advent of LiDAR-derived DEMs and associated LiDAR-derived stream hydrography, like those available in much of western Oregon, have revealed additional flowlines in watersheds compared to previous topographic maps, which adds more potential tributaries to survey for fish-distribution assessments. When these new previously unmapped tributaries are paired with a model, such as UPRLIMET, a common information set is available across landowners, managers, and agencies for the upper extent of fish. This helps policymakers determine where to apply regulations that support fisheries and forest management, based on the upper fish limit.Next steps for applying and expanding the model include addressing current data gaps. More information and observations about the upper distribution limits of fish beyond western Oregon would be needed to properly expand the spatial scope of the model. The upper extent of fish is at the detection limit of many current technologies, including global nativation satellite system (GNSS), geographic information systems (GIS), and LiDAR, especially in forested landscapes. Better precision of GNSS coordinates from observations would help greatly. From an ecological perspective, we could focus on fish distribution limits that vary seasonally or interannually to better understand which stream features and hydrologic parameters influence those endpoints. We also need information related to locations of barriers, including culverts, waterfalls, and knickpoints to understand their influence on contemporary distributions. Incorporating variables representing riparian conditions as well as leveraging higher-resolution DEMs ( More

Research Group for Genomic Epidemiology, Technical University of Denmark, Kgs, Lyngby, DenmarkPatrick Munk, Christian Brinch, Frederik Duus Møller, Thomas N. Petersen, Rene S. Hendriksen, Anne Mette Seyfarth, Jette S. Kjeldgaard, Christina Aaby Svendsen & Frank M. AarestrupCentre for Immunity, Infection and Evolution, University of Edinburgh, Edinburgh, UKBram van Bunnik & Mark WoolhouseCentre for Antibiotic Resistance Research (CARe), University of Gothenburg, Gothenburg, SwedenFanny Berglund & D. G. Joakim LarssonDepartment of Viroscience, Erasmus MC, Rotterdam, The NetherlandsMarion KoopmansInstitute of Public Health, Tirana, AlbaniaArtan BegoUniversidad de Buenos Aires, Buenos Aires, ArgentinaPablo PowerMelbourne Water Corporation, Melbourne, AustraliaCatherine Rees & Kris CoventryCharles Darwin University, Darwin, AustraliaDionisia LambrinidisUniversity of Copenhagen, Frederiksberg C, DenmarkElizabeth Heather Jakobsen Neilson & Yaovi Mahuton Gildas HounmanouCharles Darwin University, Darwin Northern Territory, AustraliaKaren GibbCanberra Hospital, Canberra, AustraliaPeter CollignonALS Water, Scoresby, AustraliaSusan CassarAustrian Agency for Health and Food Safety (AGES), Vienna, AustriaFranz AllerbergerUniversity of Dhaka, Dhaka, BangladeshAnowara Begum & Zenat Zebin HossainEnvironmental Protection Department, Bridgetown, St. Michael, BarbadosCarlon WorrellLaboratoire Hospitalier Universitaire de Bruxelles (LHUB-ULB), Brussels, BelgiumOlivier VandenbergAQUAFIN NV, Aartselaar, BelgiumIlse PietersPolytechnic School of Abomey-Calavi, Abomey-Calavi, BeninDougnon Tamègnon VictorienUniversidad Catσlica Boliviana San Pablo, La Paz, BoliviaAngela Daniela Salazar Gutierrez & Freddy SoriaPublic Health Institute of the Republic of Srpska, Faculty of Medicine University of Banja Luka, Banja Luka, Bosnia and HerzegovinaVesna Rudić GrujićPublic Health Institute of the Republic of Srpska, Banja Luka, Bosnia and HerzegovinaNataša MazalicaBotswana International University of Science and Technology, Palapye, BotswanaTeddie O. RahubeUniversidade Federal de Minas Gerais, Belo Horizonte, BrazilCarlos Alberto Tagliati & Larissa Camila Ribeiro de SouzaOswaldo Cruz Institute, Rio de Janeiro, BrazilDalia RodriguesVale Institute of Technology, Belιm, PA, BrazilGuilherme OliveiraNational Center of Infectious and Parasitic Diseases, Sofia, BulgariaIvan IvanovUniversity of Ouagadougou, Ouagadougou, Burkina FasoBonkoungou Isidore Juste & Traoré OumarInstitut Pasteur du Cambodge, Phnom Penh, CambodiaThet Sopheak & Yith VuthyCentre Pasteur du Cameroun, Yaoundι, CameroonAntoinette Ngandijo, Ariane Nzouankeu & Ziem A. Abah Jacques OlivierUniversity of Regina, Regina, CanadaChristopher K. YostEau Terre Environnement Research Centre (INRS-ETE), Quebec City G1K 9A9, Canada and Indian Institute of Technology, Jammu, IndiaPratik KumarEau Terre Environnement Research Centre (INRS-ETE), Quebec City G1K 9A9, Canada and Lassonde School of Enginerring, York University, Toronto, CanadaSatinder Kaur BrarUniversity of N’Djamena, N’Djamena, ChadDjim-Adjim TaboEscuela de Medicina Veterinaria, Facultad de Ciencias de la Vida, Universidad Andres Bello, Santiago, ChileAiko D. AdellInstitute of Public Health, Santiago, ChileEsteban Paredes-Osses & Maria Cristina MartinezCentro de Biotecnologνa de los Recursos Naturales, Facultad de Ciencias Agrarias y Forestales, Talca, ChileSara Cuadros-OrellanaGuangdong Provincial Center for Disease Control and Prevention, Guangzhou, ChinaChangwen Ke, Huanying Zheng & Li BaishengThe Hong Kong Polytechnic University, Hong Kong, ChinaLok Ting Lau & Teresa ChungShantou University Medical College, Shantou, ChinaXiaoyang JiaoNanjing University of Information Science and Technology, Nanjing, ChinaYongjie YuCenter for Disease Control and Prevention of Henan province, Zhengzhou, ChinaZhao JiaYongColombian Integrated Program for Antimicrobial Resistance Surveillance – Coipars, CI Tibaitatα, Corporaciσn Colombiana de Investigaciσn Agropecuaria (AGROSAVIA), Tibaitatα – Mosquera, Cundinamarca, ColombiaJohan F. Bernal Morales, Maria Fernanda Valencia & Pilar Donado-GodoyInstitut Pasteur de Côte d’Ivoire, Abidjan, Côte d’IvoireKalpy Julien CoulibalyUniversity of Zagreb, Zagreb, CroatiaJasna HrenovicAndrija Stampar Teaching Institute of Public Health, Zagreb, CroatiaMatijana JergovićVeterinary Research Institute, Brno, Czech RepublicRenáta KarpíškováCentre de Recherche en Sciences Naturelles de Lwiro (CRSN-LWIRO), Bukavu, Democratic Republic of CongoZozo Nyarukweba DeogratiasBIOFOS A/S, Copenhagen K, DenmarkBodil ElsborgTechnical University of Denmark, Kgs., Lyngby, DenmarkLisbeth Truelstrup Hansen & Pernille Erland JensenSuez Canal University, Ismailia, EgyptMohamed AbouelnagaUniversity of Sadat City, Sadat City, EgyptMohamed Fathy SalemMinistry of Health, Environmental Microbiology, Tallinn, EstoniaMarliin KoolmeisterAddis Ababa University, Addis Ababa, EthiopiaMengistu Legesse & Tadesse EgualeUniversity of Helsinki, Helsinki, FinlandAnnamari HeikinheimoFrench Institute Search Pour L’exploitation De La Mer (Ifremer), Nantes, FranceSoizick Le Guyader & Julien SchaefferInstituto Nacional de Investigaciσn en Salud Pϊblica-INSPI (CRNRAM), Galαpagos, Quito, EcuadorJose Eduardo VillacisNational Public Health Laboratories, Ministry of Health and Social Welfare, Kotu, GambiaBakary SannehNational Center for Disease Control and Public Health, Tbilisi, GeorgiaLile MalaniaRobert Koch Institute, Berlin, GermanyAndreas Nitsche & Annika BrinkmannTechnische Universitδt Dresden, Institute of Hydrobiology, Dresden, GermanySara Schubert, Sina Hesse & Thomas U. BerendonkUniversity for Development Studies, Tamale, GhanaCourage Kosi Setsoafia SabaUniversity of Ghana, Accra, GhanaJibril MohammedKwame Nkrumah University of Science and Technology, Kumasi, PMB, GhanaPatrick Kwame FegloCouncil for Scientific and Industrial Research Water Research Institute, Accra, GhanaRegina Ama BanuVeterinary Research Institute of Thessaloniki, Hellenic Agricultural Organisation-DEMETER, Thermi, GreeceCharalampos KotzamanidisAthens Water Supply and Sewerage Company (EYDAP S.A.), Athens, GreeceEfthymios LytrasUniversidad de San Carlos de Guatemala, Guatemala City, GuatemalaSergio A. LickesSemmelweis University, Institute of Medical Microbiology, Budapest, HungaryBela KocsisUniversity of Veterinary Medicine, Budapest, HungaryNorbert SolymosiUniversity of Iceland, Reykjavνk, IcelandThorunn R. ThorsteinsdottirCochin University of Science and Technology, Cochin, IndiaAbdulla Mohamed HathaKasturba Medical College, Manipal, IndiaMamatha BallalApollo Diagnostics, Mangalore, IndiaSohan Rodney BangeraShiraz University of Medical Sciences, Shiraz, IranFereshteh FaniShahid Beheshti University of Medical Sciences, Tehran, IranMasoud AlebouyehNational University of Ireland Galway, Galway, IrelandDearbhaile Morris, Louise O’Connor & Martin CormicanBen Gurion University of the Negev and Ministry of Health, Beer-Sheva, IsraelJacob Moran-GiladIstituto Zooprofilattico Sperimentale del Lazio e della Toscana, Rome, ItalyAntonio Battisti, Elena Lavinia Diaconu & Patricia AlbaCNR – Water Research Institute, Verbania, ItalyGianluca Corno & Andrea Di CesareNational Institute of Infectious Diseases, Tokyo, JapanJunzo Hisatsune, Liansheng Yu, Makoto Kuroda, Motoyuki Sugai & Shizuo KayamaNational Center of Expertise, Taldykorgan, KazakhstanZeinegul ShakenovaMount Kenya University, Thika, KenyaCiira KiiyukiaKenya Medical Research Institute, Nairobi, KenyaEric Ng’enoUniversity of Prishtina “Hasan Prishtina” & National Institute of Public Health of Kosovo, Pristina, KosovoLul RakaKuwait Institute for Scientific Research, Kuwait City, KuwaitKazi Jamil, Saja Adel Fakhraldeen & Tareq AlaatiInstitute of Food Safety, Riga, LatviaAivars Bērziņš, Jeļena Avsejenko, Kristina Kokina, Madara Streikisa & Vadims BartkevicsAmerican University of Beirut, Beirut, LebanonGhassan M. MatarCentral Michigan University & Michigan Health Clinics, Saginaw, MI, USAZiad DaoudNational Food and Veterinary Risk Assessment Institute, Vilnius, LithuaniaAsta Pereckienė & Ceslova Butrimaite-AmbrozevicieneLuxembourg Institute of Science and Technology, Belvaux, LuxembourgChristian PennyInstitut Pasteur de Madagascar, Antananarivo, MadagascarAlexandra Bastaraud & Jean-Marc CollardUniversity of Antananarivo, Centre d’Infectiologie Charles Mιrieux, Antananarivo, MadagascarTiavina Rasolofoarison, Luc Hervé Samison & Mala Rakoto AndrianariveloUniversity of Malawi, Blantyre, MalawiDaniel Lawadi BandaMalaysian Genomics Resource Centre Berhad, Kuala Lumpur, MalaysiaArshana AminAIMST University, COMBio, Kedah, MalaysiaHeraa Rajandas & Sivachandran ParimannanWater Services Corporation, Luqa, MaltaDavid SpiteriEnvironmental Health Directorate, St. Venera, MaltaMalcolm Vella HaberUniversity of Mauritius, Reduit, MauritiusSunita J. SantchurnInstitute for Public Health Montenegro, Podgorica, MontenegroAleksandar Vujacic & Dijana DjurovicInstitut Pasteur du Maroc, Casablanca, MoroccoBrahim Bouchrif & Bouchra KarraouanCentro de Investigaηγo em Saϊde de Manhiηa (CISM), Maputo, MozambiqueDelfino Carlos VubilAgriculture and Forestry University, Kathmandu, NepalPushkar PalNational Institute for Public, Health and the Environment (RIVM), Bilthoven, The NetherlandsHeike Schmitt & Mark van PasselUniversity of Otago, Dunedin, New ZealandGert-Jan Jeunen & Neil GemmellUniversity of Otago, Christchurch, New ZealandStephen T. ChambersUniversity of Central America, Managua, NicaraguaFania Perez Mendoza & Jorge Huete-PιrezUniversidad Nacional Autσnoma de Nicaragua-Leσn, Leσn, NicaraguaSamuel VilchezUniversity of Ilorin, Ilorin, NigeriaAkeem Olayiwola Ahmed, Ibrahim Raufu Adisa & Ismail Ayoade OdetokunUniversity of Ibadan, Ibadan, NigeriaKayode FashaeNorwegian Institute of Public Health, Oslo, NorwayAnne-Marie Sørgaard & Astrid Louise WesterVEAS, Slemmestad, NorwayPia Ryrfors & Rune HolmstadUniversity of Agriculture, Faisalabad, PakistanMashkoor MohsinAga Khan University, Karachi, PakistanRumina Hasan & Sadia ShakoorLaboratorio Central de Salud Publica, Asuncion, ParaguayNatalie Weiler Gustafson & Claudia Huber SchillInstituto Nacional de Salud, Lima, PeruMaria Luz Zamudio RojasUniversidad de Piura, Piura, PeruJorge Echevarria Velasquez & Felipe Campos YauceWHO Environmental and Occupational Health, Manila, PhilippinesBonifacio B. MagtibayMaynilad Water Services, Inc., Quezon City, PhilippinesKris Catangcatang & Ruby SibuloNational Veterinary Research Institute, Pulawy, PolandDariusz WasylUniversidade Catσlica Portuguesa, CBQF – Centro de Biotecnologia e Quνmica Fina – Laboratσrio Associado, Escola Superior de Biotecnologia, Porto, PortugalCelia Manaia & Jaqueline RochaAguas do Tejo Atlantico, Lisboa, PortugalJose Martins & Pedro ÁlvaroGwangju Institute of Science and Technology, Gwangju, Republic of KoreaDoris Di Yoong Wen, Hanseob Shin & Hor-Gil HurKorea Advanced Institute of Science and Technology, Daejeon, Republic of KoreaSukhwan YoonInstitute of Public Health of the Republic of North Macedonia, Skopje, Republic of North MacedoniaGolubinka Bosevska & Mihail KochubovskiState Medical and Pharmaceutical University, Chișinău, Republic of MoldovaRadu CojocaruNational Agency for Public Health, Chișinău, Republic of MoldovaOlga BurduniucKing Abdullah University of Science and Technology, Thuwal, Saudi ArabiaPei-Ying HongUniversity of Edinburgh, Edinburgh, Scotland, UKMeghan Rose PerryInstitut Pasteur de Dakar, Dakar, SenegalAmy GassamaInstitute of Veterinary Medicine of Serbia, Belgrade, SerbiaVladimir RadosavljevicNanyang Technological University, Singapore, SingaporeMoon Y. F. Tay, Rogelio Zuniga-Montanez & Stefan WuertzPublic Health Authority of the Slovak Republic, Bratislava, SlovakiaDagmar Gavačová, Katarína Pastuchová & Peter TruskaNational Laboratory of Health, Environment and Food, Ljubljana, SloveniaMarija TrkovIndependent consultant, Johannesburg, South AfricaKaren KeddyDaspoort Waste Water Treatment Works, Pretoria, South AfricaKerneels EsterhuyseKorea Advanced Institute of Science and Technology, Daejeon, South KoreaMin Joon SongSchool of Veterinary Sciences, Lugo, SpainMarcos Quintela-BalujaLabaqua, Santiago de Compostela, SpainMariano Gomez LopezIRTA, Centre de Recerca en Sanitat Animal (CReSA, IRTA-UAB), Campus de la Universitat Autonoma de Barcelona, Bellaterra, SpainMarta Cerdà-CuéllarUniversity of Kelaniya, Ragama, Sri LankaR. R. D. P. Perera, N. K. B. K. R. G. W. Bandara & H. I. PremasiriMedical Research Institute, Colombo, Sri LankaSujatha PathirageCaribbean Public Health Agency, Catries, Saint LuciaKareem CharlemagneThe Sahlgrenska Academy at the University of Gothenburg, Gothenburg, SwedenCarolin RutgerssonSwedish University of Agricultural Sciences, Uppsala, SwedenLeif Norrgren & Stefan ÖrnFederal Food Safety and Veterinary Office (FSVO), Bern, SwitzerlandRenate BossAra Region Bern AG, Herrenschwanden, SwitzerlandTanja Van der HeijdenCenters for Disease Control, Taipei, TaiwanYu-Ping HongKilimanjaro Clinical Research Institute, Moshi, TanzaniaHappiness Houka KumburuSokoine University of Agriculture, Morogoro, TanzaniaRobinson Hammerthon MdegelaFaculty of Science and Technology, Suratthani Rajabhat University, Surat Thani, ThailandKaknokrat ChonsinFaculty of Public Health, Mahidol University, Bangkok, ThailandOrasa SuthienkulFaculty of Medicine Siriraj Hospital, Bangkok, ThailandVisanu ThamlikitkulNational Institute for Public Health and the Environment (RIVM), Bilthoven, NetherlandsAna Maria de Roda HusmanNational Institute of Hygiene, Lomι, TogoBawimodom BidjadaAgence de Mιdecine Prιventive, Dapaong, TogoBerthe-Marie Njanpop-LafourcadeDivision of Integrated Surveillance of Health Emergencies and Response, Lomι, TogoSomtinda Christelle Nikiema-PessinabaPublic Health Institution of Turkey, Ankara, TurkeyBelkis LeventHatay Mustafa Kemal University, Hatay, TurkeyCemil KurekciMakerere University, Kampala, UgandaFrancis Ejobi & John Bosco KaluleAbu Dhabi Public Health Center, Abu Dhai, United Arab EmiratesJens ThomsenDubai municipality, WWTP Al Aweer, Dubai, UAEOuidiane ObaidiRashid Hospital, Dubai, UAELaila Mohamed JassimNorthumbrian Water, Northumbria House, Abbey Road, Pity Me, Durham, UKAndrew MooreUniversity of Exeter Medical School, Cornwall, UKAnne Leonard, Lihong Zhang & William H. GazeNewcastle University, Newcastle upon Tyne, UKDavid W. Graham & Joshua T. BunceBrightwater Treatment Plant, Woodinville, WA, USABrett LeforDepartment of Environmental Sciences and Engineering, University of North Carolina at Chapel Hill, Chapel Hill, NC, USADrew Capone & Joe BrownUniversity of North Carolina, Chapel Hill, USAEmanuele Sozzi & Mark D. SobseyUniversity of Washington, Seattle, WA, USAJohn Scott Meschke, Nicola Koren Beck, Pardi Sukapanpatharam & Phuong TruongBaylor University, Waco, USAMichael DavisColumbia Boulevard WWTP, Portland, USARonald LilienthalEastern Illinois University, Charleston, USASanghoon KangThe Ohio State University, Columbus Ohio, USAThomas E. WittumLaboratorio Tecnolσgico del Uruguay, Montevideo, UruguayNatalia Rigamonti & Patricia BaklayanInstitute of Public Health in Ho Chi Minh City, Ho Chi Minh, VietnamChinh Dang Van, Doan Minh Nguyen Tran & Nguyen Do PhucUniversity of Zambia, Lusaka, ZambiaGeoffrey KwendaF.M.A., M.K., and M.W. conceived the study and secured funding. R.S.H., A.M.S., C.A.A.S., and J.S.K. organized sample collection, material transfer, and logistics. F.D.M., P.M., and C.B. did quality control, sample selection, and outlier detection. P.M., C.B., F.D.M., T.N.P., and F.B. performed bioinformatics analyses. P.M. and C.B. carried out data and statistical analyses and visualization. P.M. and F.M.A. drafted the initial manuscript with input from C.B., B.v.B., D.G.J.L., M.W., and M.K. The Global Sewage Consortium authors carried out sewage sampling, filled in metadata and shipped the samples to DTU. All authors helped to review and improve the manuscript. More