key: cord-0901848-truassnf
authors: Arévalo, Constanza; Amaya-Espinel, Juan David; Henríquez, Cristián; Ibarra, José Tomás; Bonacic, Cristián
title: Urban noise and surrounding city morphology influence green space occupancy by native birds in a Mediterranean-type South American metropolis
date: 2022-03-16
journal: Sci Rep
DOI: 10.1038/s41598-022-08654-7
sha: 49682c336cc0911337485eec7210253867e50d5c
doc_id: 901848
cord_uid: truassnf

Urban green spaces provide natural habitat for birds in urban landscapes, yet the effects of noise and surrounding urban morphology on bird community structure and distribution are not well understood in Latin America, the second most urbanized region in the world. Santiago of Chile is the single city belonging to the Mediterranean ecosystem in South America and is subject to extensive urbanization as seen throughout Latin America. We examined the role of 65 urban green spaces—6 large urban parks (PAR) and 59 small green spaces (SGS)—in harboring native birds during winter 2019, analyzing the quality of green areas in terms of vegetation (i.e. NDVI, native vegetation, and tree cover), exotic bird species, noise levels, and surrounding urban morphology (i.e. building height and cover). Significantly higher noise levels were detected in SGS, along with significantly greater exotic bird (n = 4) richness and abundance than PAR, which possessed significantly greater native bird (n = 25) richness and abundance. Native birds were more abundant than exotic birds in green spaces with average noise levels < 52 dB and average NDVI > 0.5. Occupancy models indicate that green space occupancy by 50% of modeled native bird species was influenced by maximum noise levels, playing a larger role than vegetation (30%) and urban morphology (0%). We stress the importance of developing networks of large green spaces in rapidly urbanizing regions, with abundant tree cover, surrounded by smaller urban morphology, and regulating noise levels to ensure the conservation of native bird communities in cities, particularly those that are threatened.

www.nature.com/scientificreports/ Native bird tolerances to urbanization conditions. Average noise levels in green spaces shared significant correlations with the richness and abundance of native urban avoider (n = 7), urban utilizer (n = 8), and urban dweller (n = 12) species, indicating an important negative trend in the richness and abundance of all native bird species as average noise levels rose in green spaces (Table 1) . Additionally, a significant positive correlation was found between NDVI and urban utilizer abundance in green spaces, a negative relation between native vegetation in green spaces and native urban utilizer richness and abundance, as well as a significant negative relation between average building height surrounding green spaces and urban avoider and urban dweller species richness and abundance. Native urban avoider species displayed the least tolerance to anthropogenic noise and building height surrounding green spaces, rarely detected in green spaces with average noise levels over 60 dB and average building height over 10 m, with most found in green spaces with average noise levels of 50-55 dB and surrounded by buildings 4-6 m tall (Fig. 1a) . Unlike most native urban utilizer species, the green-backed firecrown (Sephanoides sephaniodes) was found in most green spaces, for which reason it was considered separately from other urban utilizers (Fig. 1b) . The green-backed firecrown's tolerance to anthropogenic noise was more like that of urban dwellers, found throughout most noise levels, though more commonly in green spaces with average noise levels between 50 and 65 dB (Fig. 1c) , while other urban utilizers were most detected in green spaces with average noise levels around 55 dB, decreasing rapidly with noise levels higher than that (Fig. 1b) .

All native birds showed sensitivity to building height, with few native birds detected in green spaces surrounded by buildings more than 10 m tall on average. Meanwhile, there was not sufficient evidence that building cover surrounding green spaces, nor tree cover in green spaces, share significant relations with native urban avoider, urban utilizer, and urban dweller richness and abundance in green spaces (Table 1) . Generally, noise levels were found to be greatest in green spaces surrounded by taller urban morphology, while there was not enough evidence indicating that building cover shared a relation with noise.

Detection and occupancy probability models. Detectability remained constant across surveys and surveyed sites for the Chilean mockingbird (Mimus thenca), the only endemic bird species detected in the study area, not affected by any one variable. Detectability varied across survey sites for the remaining 13 native bird species, changing with the maximum noise level recorded during surveys (14%), vegetation cover (29%), tree cover (21%), native vegetation (14%), building cover surrounding green spaces (7%), and with a combination of average building height and maximum noise level (7%). Detectability decreased with rising noise levels, native vegetation, building height, and building cover surrounding green spaces, while increasing with vegetation cover.

Green space occupancy was most influenced by the maximum noise levels recorded in green spaces for 50% of modeled native bird species (Table 2) , with urban avoider and urban utilizer species showing the lowest noise tolerances, and exotic species displaying the highest tolerances to noise, with some species present especially in high-noise green spaces (the rock pigeon (Columba livia) and the monk parakeet (Myiopsitta monachus)). The picui ground dove (Columbina picui) and long-tailed meadowlark (Leistes loyca) were the modeled native species most sensitive to noise, their occupancy probabilities decreasing rapidly as the average maximum noise level of a green space reached 55-65 dB, approaching an occupancy probability of 0 when average maximum noise levels reached over 65 dB (Fig. 2a, b) . The common diuca finch (Diuca diuca) and rufous-collared sparrow (Zonotrichia capensis) were also quite sensitive to noise, their occupancy probabilities decreasing as average maximum noise levels reached above 55 dB (Fig. 2b, c) , but the decrease in their occupancy probabilities was more gradual than for the other two species. The austral thrush (Turdus falcklandii) displayed the least sensitivity to noise, present in all green spaces with an average maximum noise level below approximately 73 dB (Fig. 2c) . Following maximum noise level as the most influential covariate in occupancy probability, tree cover in green spaces was most influential for 20% of native birds, green space occupancy remained constant across sites for 20% of native birds, and, for the remaining 10%, occupancy was influenced by vegetation cover (Table 2) . Table 1 . Pearson/Spearman correlation coefficients and associated p-values for native urban avoider, urban utilizer, and urban dweller groups in green spaces of Santiago, Chile. *Significant correlation, given significance level α = 0.05. www.nature.com/scientificreports/

Our research determined noise to share a potentially important negative relationship with native bird richness and abundance and appears to be the most limiting factor in green space occupancy by native bird species, more so than the type and amount of vegetation present in urban green spaces, and more so than urbanization itself, represented as building height and cover surrounding green spaces. Thus, noise is potentially acting as an invisible source of habitat degradation, limiting the bird species capable of inhabiting an area, regardless of whether the appropriate vegetative conditions exist. www.nature.com/scientificreports/ As predicted, native urban avoiders reached their maximum abundances in PAR, which, given their high vegetation cover and large size, act as patches of natural habitat in cities. Native urban utilizers tended to be found in more suburban areas, and urban dwellers, both native and exotic, were detected in green spaces of all noise levels. All exotic bird species were urban dwellers, referring to their high tolerance to urbanization 5,25 , thus reaching the high abundances observed, particularly in SGS.

SGS possessed higher average noise levels and greater exotic bird abundance than PAR, which presented significantly higher numbers of native bird richness and abundance. The potential influence of noise on native bird species first becomes evident when we consider that native bird abundance tended to rise above the generally high abundance of exotic birds when average noise levels in green spaces reached below 52 dB (it should be noted that, according to the Chilean Noise Norm No. 146, the maximum allowable noise levels generated by fixed sources in residential areas of Santiago is 55 dB during the day, 7 a.m.-9 p.m.). The negative relations between noise and urban avoider, urban utilizer, and urban dweller species richness and abundance further indicate how noise may be regulating the native bird species present in green spaces, affecting urban avoider richness the most and urban dweller richness the least, while influencing the abundance of all native bird species rather similarly. Meanwhile, building height surrounding green spaces negatively influenced native urban avoider and urban dweller richness and abundance, with the greatest influence on urban dweller abundance, yet all native birds were less likely to be detected in green spaces surrounded by buildings over 10 m tall on average.

The importance of vegetation for native bird communities also cannot be denied, given that native birds reached higher abundances than exotic birds when vegetation cover reached an average NDVI value greater than 0.5. Results from this study thus suggest that exotic birds begin to replace native birds in terms of abundance as noise levels rise in urban green spaces, vegetation cover decreases, and building height surrounding green spaces increases, with native urban avoider species being the least tolerant to the influences of urbanization, and, consequently, the first to disappear when noise levels and building height become too great. The observed negative relationship between native bird species richness and maximum noise levels, and the positive relationship with vegetation cover, are comparable to results seen in other Neotropical cities 24, 26 , yet our results indicate that the relationships between these variables and bird abundance are stronger. This may indicate how bird abundance fluctuates in green spaces as some birds temporarily leave during noisy events or become quieter and more cryptic under noisy conditions 26 , while noise also negatively influences bird species richness by filtering the species that can inhabit areas of varying noise levels.

Detection probability models found native bird detectability to mostly increase with vegetation cover and tree cover in urban green spaces, except for the common diuca finch, whose detectability decreased with rising tree cover. Some of the bird species that displayed the lowest detection probabilities, such as the picui ground dove and fire-eyed diucon (Xolmis pyrope), are not frequently found in cities and possess vocalizations that are unlikely to be heard well in high-noise areas due to their low frequencies, making them more easily masked by the anthrophony, characterized by its low frequency and high intensity 31 . Consequently, birds whose vocalizations are similar in frequency and amplitude to the anthrophony were more commonly or exclusively found in green spaces that registered low noise levels, their detectability also decreasing with rising noise, as was the case with the fire-eyed diucon.

Urban green space occupancy by native bird species was mainly influenced by average maximum noise levels recorded in green spaces. Of the modeled native species, the long-tailed meadowlark and the picui ground dove, an urban avoider and an urban utilizer species respectively, were the species most sensitive to noise, their www.nature.com/scientificreports/ probability of occupying green spaces with average maximum noise levels over 55 dB decreasing rapidly and approaching zero when over 65 dB. Meanwhile, the austral thrush, an urban dweller species, was by far the most tolerant to noise of the native birds, its presence probability just beginning to decrease when average maximum noise levels reached over 73 dB in green spaces. The differing tendencies of urban avoiders, urban utilizers, and urban dwellers to occupy green spaces of varying noise levels is thus evident, with native urban dweller species more likely to occupy higher noise urban green spaces than urban avoiders and utilizers, seemingly more adapted to the high noise levels that come with inhabiting a busy city. Nonetheless, although native urban dwellers displayed greater noise tolerances than urban avoiders and utilizers, their presence in city parks can also be expected to diminish if noise levels become too high, which for the most tolerant of the native birds, means reaching an average maximum level of 73 dB or more, but 55 dB or more for less tolerant species. No relation was found between vegetation cover and noise, and some of the highest noise levels were recorded in PAR. This suggests that PAR, often considered to be quiet and peaceful areas to escape the busyness of city life, can reach noise levels as high as those recorded in SGS, reducing the quality of the greatest sources of natural habitat for birds and other wildlife in cities.

The results from this study regarding the influence of noise on bird communities support previous studies indicating that birds may be excluded from suitable habitats on account of the acoustic conditions of the local Figure 2 . Occupancy probability plots for species whose occupancy was most influenced by average maximum noise levels in green spaces in Santiago, Chile: (a) native urban avoider species, (b) native urban utilizer species, and (c) native urban dweller species. Plots are based on the models summarized in Table 2 www.nature.com/scientificreports/ environment 12, 15 . Despite abundant vegetation in PAR and some SGS, certain bird species, particularly urban avoiders and utilizers, were less likely to occupy areas that presented high noise levels. However, it is important to consider other potential influencing factors, such as predators (e.g., dogs and cats) and food availability, both of which could be linked to pedestrians and could therefore also increase noise levels in green spaces. Furthermore, in an effort to focus on the influence of anthropogenic variables on urban birds (i.e., urban morphology, noise, and vegetation type and cover), this study did not consider the size of urban green spaces as a variable in occupancy modeling, but as the results of this study suggest and others in Latin America have shown 23, 32 , green space size is likely an influencing factor that should be considered in future studies. Another variable worth considering would be road coverage, which undoubtedly plays a role in noise levels, particularly for SGS. Measures to control the COVID-19 pandemic have significantly reduced noise levels in major cities worldwide [33] [34] [35] . Noise reduction in the San Francisco Bay Area, characterized by a Mediterranean climate like Santiago, resulted in songbirds rapidly occupying newly available acoustic niches within urban soundscapes and maximizing communication through higher performance songs 35 . Consequently, native bird species not commonly found in high-noise areas, mainly urban avoider and utilizer species, may now be found in greater abundance at the community level in urban green spaces where they had been scarce or non-existent during this study, conducted pre-pandemic. Furthermore, if average noise levels dropped below 52 dB in Santiago green spaces due to region-wide shut-down measures, native birds may reach higher abundances than exotic birds. The negative effects of urban noise on bird communities are extensive, yet recent research indicating birds' rapid adaptability and improved vocal performance when noise levels are significantly lowered provides hope. Native bird species susceptible to noise may stand a chance despite growing urbanization, if noise levels in urban green spaces are regulated.

Rapid urban expansion in Latin America places natural ecosystems at great risk, reducing or altogether eliminating natural habitats for native birds and other wildlife, making urban green spaces necessary for their persistence, especially in biodiversity hotspots like central Chile. As this study illustrates, noise associated with urbanization plays a significant role in influencing green space occupancy by native bird species, and, quite possibly, other animal species dependent on acoustic signaling (e.g., amphibians and mammals). Given the recreational role of urban green spaces in cities, noise regulation within these areas should be considered, while also considering how city morphology may impact bird communities. This study exemplifies how, in addition to noise, the size of urban green spaces and the vegetation cover in them, particularly tree cover, are vital aspects to consider in city planning in order to preserve native bird communities in urban systems. Large urban parks held significantly richer bird communities than small green spaces, with greater native bird richness and abundance. Therefore, it is imperative that science and city planning collaborate to develop cities with networks of large green spaces with abundant tree cover, surrounded by smaller urban morphology, where noise is regulated and maintained at tolerable levels for native birds. There is a clear need to move towards biophilic city planning to harmonize urban growth and the protection and expansion of networks of green areas that generate habitat for birds that, in turn, provide important ecosystem services to cities.

Bird surveys and noise measurements. We conducted this study in 59 small green spaces (SGS) and 6 large urban parks (PAR) in the Metropolitan Region of Santiago (MRS) in central Chile (Fig. 3) . Located between the Andes mountains and a coastal mountain range, the MRS possesses a semi-arid Mediterranean climate, with hot, dry summers and cold, wet winters 36 . Due to its highly endemic vegetation, the Mediterranean region of central Chile is considered a biodiversity hotspot, with sclerophyllous shrubland and woodland making up the dominant vegetation 37 . The region holds about 40% of Chile's population, of which 96.3% resides in urban areas 30 . SGS, also known as "pocket parks" or "plazas", ranged in size from 0.5 to 2 ha and were independent of one another, separated by a distance of at least 250 m from other surveyed SGS and PAR 25 . We sampled 6 PAR of area > 10 ha as control points, with 1-3 survey points in each (generally located in the center of the park), depending on park size, for a total of 10 PAR sample sites, ensuring a distance of at least 250 m between survey points. Green spaces were identified using a digital layer of green area developed for the MRS at a scale of 1:5000, updated using Landsat 8 OLI Thematic Mapper satellite imagery from November 2013 25 .

We conducted point count surveys at all SGS and PAR once a month during winter 2019 (May-August), for a total of three surveys per green space. To capture the morning chorus, the point of maximum biophony activity 38 , we conducted bird surveys within four hours from sunrise. We altered the order of bird surveys each month so as to never sample a green space at the same time, to account for varying noise levels throughout the morning, which in turn could alter the bird species present during a survey 26 . We used a radius of 50 m, recording all bird species detected visually and acoustically during a 10-min period, along with the number of individuals of each species. Additionally, we determined noise levels at the beginning and end of each bird survey for a period of 30 s using the NIOSH Sound Level Meter app for smartphones combined with the Dayton Audio iMM-6 calibrated measurement microphone, measuring average continuous equivalent sound pressure levels (LA eq ) and maximum equivalent sound pressure levels (L max ) using slow A frequency weighting.

Urban matrix measurements. As a measure of the urban matrix resulting from urbanization, we considered average building height and building cover within a circular buffer of 200 m from the central point of each urban green space, obtained in 2015 25 . We calculated the proportion of building cover within the buffer using the Normalized Difference Vegetation Index (NDVI) to map the percentage of the buffer covered by vegetation and took the difference. Average building height was calculated using a raster layer with cadastral information for the study area. Building height was validated through Google Street View 3D models, estimating a height of 2.70 m per built floor. We determined average building height surrounding urban green spaces for a random www.nature.com/scientificreports/ sample (n = 15) of SGS in 2020 using the same method, but no significant differences were found between 2015 and 2020 urban morphology data (t test p value = 0.922).

Vegetation cover and composition. We determined the average NDVI for each urban green space as an estimate of vegetation cover, using QGIS 2.18.23 to process Landsat 8 OLI Thematic Mapper satellite imagery of the study area taken in September 2019, with values close to zero indicating bare soil or urban areas, and values closer to 1 implying dense vegetation. Additionally, we estimated the proportion of native vegetation and tree cover (vegetation > 2 m tall) in each green space, following 40 points along the border of 50 × 50 m plots at the center of each survey site (every 2.5 m along 4 transects of 50 m), using frequency as a measure of cover 25 .

We considered richness to be the total number of bird species detected per urban green space during the study period and abundance to be the average number of individuals of each species detected per urban green space throughout the point counts. LA eq measurements were averaged to obtain the average noise level per green space (NW), and the same was done with L max measurements, giving an average maximum noise level per green space (MNA). We tested all site variables (NW, MNA, average building height, building cover, NDVI, proportion native vegetation, and tree cover) for collinearity. In the event of high correlation (r > 0.7), only one variable was considered, as was the case with NW and MNA. Correlation analyses were also run between site variables and species richness and abundance.

We modeled native and exotic bird abundance in relation to site variables with generalized linear models (GLMs) using R version 3.6.0 and the R package ggplot2. To visualize the distribution of native birds under the influence of the most significant urban variables, we created scatterplots with marginal density plots for each urban tolerance category (urban avoiders, urban utilizers, urban dwellers) considering the urbanization variables that had the highest correlations with native bird richness and abundance. Species urban tolerances were determined following the classification method proposed by Fischer et al. (2015) , considering the spatial and temporal frequency of species observations 25 . Detection and occupancy modeling. We developed occupancy models using the R package unmarked 39 to determine the true presence and absence probabilities of bird species at sampling sites. The probability of detecting a species and its probability of occupying a green space both follow a Bernoulli model to which we incorporated data from repeated surveys, as well as site We determined the best detection probability model for each bird species detected, keeping occupancy (ψ) constant across all sites and considering all observation and site covariates. We used a predetermined set of models ("Appendix S5"), which included individual covariates as well as combinations of urban morphology and noise covariates. Models were selected based on Akaike Information Criteria (AIC), where only models with AIC ≤ 2.0 were considered in the model selection process. The model with the lowest AIC value and greatest Akaike weight (w) was selected as the best model for each bird species. The best detectability model for each bird species was then incorporated into the occupancy model selection process, where we tested all site covariates to obtain the best occupancy model. Model selection was carried out in the same manner as with detectability, using AIC to determine the best model for each bird species. Species that displayed estimated detection probabilities (p) < 0.1 or that displayed a broad confidence interval associated with their occupancy probability (ψ) were not included due to the uncertainty associated with the expected detection of those species and/or green space occupancy by those species. 15 native bird species, mainly urban avoiders and urban utilizers, were discarded from occupancy modeling due to low detection rates. However, species eliminated from occupancy modeling resulting from wide ψ confidence intervals, but that presented p > 0.1 and an appropriate associated confidence interval, were considered when determining the most influential covariates on species detectability. We created occupancy models for 10 native bird species (Table 2 ; Fig. 2 ) and 4 exotic bird species (Table 2 ). Covariates influencing detectability were considered for 14 native bird species (the 10 native species in Table 2 and Anairetes parulus, Curaeus curaeus, Troglodytes aedon, and Zenaida auriculata, which displayed appropriate p and associated confidence intervals, but inflated Ψ).

The datasets generated and analyzed during the current study, along with the R code generated to analyze the data, are available from the corresponding author upon reasonable request. www.nature.com/scientificreports/

DESA (Department of Economic and Social Affairs), Population Division

Impact of Urbanization on Birds. 235-257 in Tietze D

Behavioural adjustments for a life in the city

Land use and avian species diversity along an urban gradient

Categorizing wildlife responses to urbanization and conservation implications of terminology

De La Fuente Díaz Ordaz, A.A. & Hernández Satín, L. Caracterización ecológica de la avifauna de los parques urbanos de la Ciudad de Puebla (México)

Urbanization, biodiversity, and conservation

Ecology and Evolution of Acoustic Communication in Birds

How anthropogenic noise affects foraging

Chapter 2: Effects of noise pollution on birds: A brief review of our knowledge

Urban bioacoustics: It's not just noise

Anthropogenic noise decreases urban songbird diversity and may contribute to homogenization

Noise pollution changes avian communities and species interactions

Songs of the city: Noise-dependent spectral plasticity in the acoustic phenotype of urban birds

A phantom road experiment reveals traffic noise is an invisible source of habitat degradation

On the relation between loudness and the increased song frequency of urban birds

Bird song variations along an urban gradient: The case of the European blackbird (Turdus merula)

Effects of urban noise on song and response behaviour in great tits

Bird song and anthropogenic noise: Vocal constraints may explain why birds sing higher-frequency songs in cities

Birds sing at a higher pitch in urban noise

Birds and anthropogenic noise: Are urban songs adaptive?

Can human disturbance promote nestedness? Songbirds and noise in urban parks as a case study

Pocket parks in a compact city: How do birds respond to increasing residential abundance

Anthropogenic noise reduces bird species richness and diversity in urban parks

The influence of building abundance on Neotropical bird communities found in small urban parks

Noisy environments: Untangling the role of anthropogenic noise on bird species richness in a Neotropical city

What's new? An updated review of avian ecology in urban Latin America

Trends and knowledge gaps in field research investigating effects of anthropogenic noise

Crecimiento urbano y acceso a oportunidades: un desafío para América Latina

Estimaciones y proyecciones de la población de Chile 2002-2035 Regiones y áreas urbanorural

Toward a broader characterization of anthropogenic noise and its effects on wildlife

Relation between habitat attributes and bird richness in a western Mexico suburb

Changes in noise levels in the city of Madrid during COVID-19 lockdown in 2020

Investigating changes in noise pollution due to the COVID-19 lockdown: The case of Dublin

Singing in a silent spring: Birds respond to a halfcentury soundscape reversion during the COVID-19 shutdown

Climatic regionalisation of continental Chile

Impacts of climate change on the distribution of species and communities in the Chilean Mediterranean ecosystem

Soundscape ecology: The science of sound in the landscape

Unmarked: An R package for fitting hierarchical models of wildlife occurrence and abundance

We detected 29 bird species (Appendix S2) (86% native [n = 25] to Chile; 14% exotic [n = 4]) within the green spaces of Santiago, all categorized as being of 'least concern' in the International Union for Conservation of Nature (IUCN) Red List. Generally, native bird richness and abundance were greatest in large urban parks (PAR; area > 10 ha), while exotic bird abundance was greater in small green spaces (SGS; area < 10 ha) ("Appendix S1"). Native bird richness and abundance decreased with increasing noise levels and building height surrounding green spaces, while increasing with vegetation cover ("Appendix S3"). Exotic bird abundance increased with rising noise levels and building height, decreased with vegetation cover, and exotic bird richness decreased with tree cover ("Appendix S3"). Native birds reached lesser abundances than exotic birds when average noise levels reached over 52 dB, and, once vegetation cover in green spaces reached an average NDVI value of 0.5 or more, native birds tended to reach higher abundances than exotic birds ("Appendix S4").

We thank Fernanda Palavecino for her extensive help with obtaining urban matrix data and José Infante for his guidance with occupancy modeling. J.T.I. acknowledges the support the Center for Intercultural and Indigenous Research CIIR-ANID/FONDAP 15110006, the Center of Applied Ecology and Sustainability CAPES-ANID PIA/ BASAL FB0002 and the Cape Horn International Center for Global Change Studies and Biocultural Conservation CHIC-ANID PIA/BASAL PFB210018. This research was funded by the graduate office of the School of Agriculture and Forestry of the Pontifical Catholic University of Chile.

All authors contributed to the study conception and design. Material preparation, data collection and analysis were performed by C.A. and J.D.A.E. The first draft of the manuscript was written by C.A. and J.T.I., J.D.A.E., C.H., and C.B. participated in the review and editing of the manuscript. Project supervision was performed by C.B. All authors read and approved the final manuscript.

This study was funded by the graduate office of the School of Agriculture and Forestry of the Pontifical Catholic University of Chile.

The authors declare no competing interests.

The online version contains supplementary material available at https:// doi. org/ 10. 1038/ s41598-022-08654-7.Correspondence and requests for materials should be addressed to C.B.Reprints and permissions information is available at www.nature.com/reprints.Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.