Landscape and Urban Planning 228 (2022) 104570 Contents lists available at ScienceDirect Landscape and Urban Planning journal homepage: www.elsevier.com/locate/landurbplan Research Paper Wildness and habitat quality drive spatial patterns of urban biodiversity Celina Aznarez a, b, *, Jens-Christian Svenning c, German Taveira d, Francesc Baró a, e, f, Unai Pascual b, g, h a Institute of Environmental Science and Technology (ICTA), Universitat Autònoma de Barcelona, Cerdanyola del Vallès, Spain b Basque Centre for Climate Change (BC3), Scientific Campus of the University of the Basque Country, Leioa, Spain c Centre for Biodiversity Dynamics in a Changing World, Department of Biology, Aarhus University, Aarhus, Denmark d Departamento de Territorio, Ambiente y Paisaje, Centro Universitario de la Región Este (CURE), Universidad de la República (UdelaR), Maldonado, Uruguay e Vrije Universiteit Brussel (VUB), Geography Department, Brussels, Belgium f Vrije Universiteit Brussel (VUB), Sociology Department, Brussels, Belgium g Basque Foundation for Science, Ikerbasque, Bilbao, Spain h Centre for Development and Environment, University of Bern, Bern, Switzerland H I G H L I G H T S • We tested wildness and habitat quality suitability as urban biodiversity indicators. • Both indicators predicted biodiversity, particularly in urban green and blue spaces. • Wildness and habitat quality correlated with biodiversity. • Peripheral urban green and blue spaces showed higher biodiversity than centric ones. • Rewilding initiatives on large centric areas could enhance urban biodiversity. A B S T R A C T Urban green and blue spaces (UGBS) are key for biodiversity conservation. Many studies focus on UGBS benefits for well-being, but how UGBS ecological and quality influence urban biodiversity is still poorly understood. We analysed the predictive accuracy of urban wildness (UW) and habitat quality (HQ) spatial patterns to biodiversity in the city of Vitoria-Gasteiz, Basque Country. Using GIS techniques, we mapped relative UW as a landscape quality, considering remoteness, challenging terrain, and perceived naturalness. We further evaluated HQ using the InVEST habitat quality module, including data on habitat sensitivity to threats (e.g. population density, light and noise pollution, accessibility) and suitability for biodiversity support, based on a parametrization by expert consultation. We compared UW and HQ to observed species richness obtained from crowd-sourced databases as a biodiversity proxy. UW and HQ models predicted general biodiversity urban patterns, being particularly adequate in UGBS. Peripheral UGBS were associated with higher UW and HQ and positively correlated to biodiversity, as opposed to the smaller-sized centrally located UGBS, more exposed to threats. Both predictors significantly explained biodiversity, and HQ better accounted for threat susceptibility in UGBS. Our findings suggest that small-sized UGBS, such as parks and squares, fail to effectively support urban biodiversity, due to their high exposure and vulnerability to threats, particularly in centric areas. Emphasizing efforts in larger centric UGBS with rewilding strategies (e.g. lowering management frequency) and reducing exposure to threats is essential to increase the habitat quality of UGBS and thus support urban biodiversity. 1. Introduction contributions to human well-being, including biodiversity support, urban cooling, runoff control, air and water quality improvement, op­ Given the continuous urban development around the world with portunities for recreation and building social cohesion (Baró et al., 2019; anthropogenic activities increasingly shaping ecological processes and Díaz et al., 2018; Gómez-Baggethun & Barton, 2013). There is growing patterns, there is a growing need to enhance the multiple functions and evidence that exposure and access to UGBS areas encompass multiple benefits supplied by urban green and blue spaces (UGBS) (Mansur et al., health benefits: mental health, lowering disease and mortality risk, 2022; Reyes-Riveros et al., 2021). UGBS supplies multiple functions and among others (Ribeiro et al., 2021; Engemann et al., 2019; Twohig- * Corresponding author at: Institute of Environmental Science and Technology (ICTA), Universitat Autònoma de Barcelona, Cerdanyola del Vallès, Spain. E-mail addresses: celina.aznarez@bc3research.org (C. Aznarez), svenning@bio.au.dk (J.-C. Svenning), german.taveira@cure.edu.uy (G. Taveira), francesc.baro@ vub.be (F. Baró), unai.pascual@bc3research.org (U. Pascual). https://doi.org/10.1016/j.landurbplan.2022.104570 Received 14 April 2022; Received in revised form 26 August 2022; Accepted 3 September 2022 Available online 14 September 2022 0169-2046/© 2022 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by- nc-nd/4.0/). C. Aznarez et al. Landscape and Urban Planning 228 (2022) 104570 Bennett & Jones, 2018; Triguero-Mas et al., 2017). By providing people influence). One of these models is the Habitat Quality module from the with opportunities to connect with nature, UGBS are key to decreasing Integrated Valuation of Environmental Services and Trade-offs (InVEST- the ‘extinction of experience’ sensu Soga and Gaston (2016) and the HQ) which has been used to analyse the degradation status of non-urban alienation from nature, which are major drivers of biodiversity loss and landscapes (Di Febbraro et al., 2018; Terrado et al., 2016), to identify environmental injustice (Noss, 2020; Lepczyk et al., 2017; Soga and specific conservation targets (Bhagabati et al., 2014), and to study Gaston, 2016; Shanahan et al., 2015; Dunn et al., 2006). natural or protected areas (Moreira et al., 2018; Sallustio et al., 2017). UGBS have a key role in supporting biodiversity, by harbouring As biodiversity patterns are inherently spatial (Sharp et al., 2020) the native and non-native species and thus promoting conservation both at use of spatially explicit indicators, such as urban wildness or HQ can regional and global scales (Lepczyk et al., 2017). Yet, biodiversity pat­ play an important role in predicting the suitability of a landscape to host terns in cities rely on the size, quantity and quality characteristics of biodiversity and ecological functions. The use of these models is cost- UGBS, ultimately influencing the provision of related health and well- effective, can be done remotely and can be linked to characteristics of being benefits (Houlden et al., 2021; Marselle et al., 2019; Lepczyk the ecosystem’s structure and functioning, such as naturalness (Müller et al., 2017; Fuller et al., 2007). Assessing urban biodiversity is, there­ et al., 2018; Di Febbraro et al., 2018). Although the application of fore, key to better understanding the role of UGBS in providing functions InVEST HQ model has proved to be useful to assess biodiversity con­ and contributions to well-being in urban socio-ecosystems. Even if servation status at large scales, it often oversimplifies the variability of biodiversity and UGBS are highly intertwined, the spatial and ecological habitat types at smaller scales as is the case of urban regions (Sallustio characteristics of UGBS influencing urban biodiversity remain poorly et al., 2017). To our knowledge, InVEST HQ has not yet been applied to understood (Marselle et al., 2021; Schwarz et al., 2017). estimate HQ for urban land uses, which can be often considered in the Urban planners usually perceive UGBS as multifunctional structures model as threats to biodiversity instead of potential habitat sources (Wu and it is often assumed that more UGBS will support more ecosystem et al., 2019; Han et al., 2019; Sallustio et al., 2017). Although wildness services and host higher biodiversity (Schwarz et al., 2017). But and HQ has been extensively assessed in large scales and often low although urban areas have the potential to provide habitat and sustain anthropized areas, there are limitations on their applications at a finer wildlife (Russo & Holzer, 2021) as human-altered habitats, their resolution in smaller geographical areas (Cao et al., 2019; Sallustio et al., ecological functioning is in continuous interaction with anthropogenic 2017). Areas with high land cover heterogeneity such as urban areas are impacts. Moreover, the way societies conceive different ecological usually oversimplified in landscape assessments of wildness and HQ, by processes as beneficial or detrimental to human well-being depends on homogenizing land uses and habitats types at coarser scales (Cao et al., social-ecological contexts which ultimately determines how societies 2019; Sallustio et al., 2017). This may lead to the underrepresentation of might value the conservation of urban biodiversity (Marselle et al., locally important areas for biodiversity conservation and affect man­ 2021; Dunn et al., 2006). Preserving and enhancing urban biodiversity agement opportunities to enhance naturalness and wildness. Cities are and raising public awareness of its societal values requires an under­ increasingly important areas worldwide and UGBS are key on the rela­ standing of the drivers and effects of changes to urban biodiversity tionship societies – nature, but have traditionally been seen as opposite (Knapp et al., 2021; Sallustio et al., 2017; Terrado et al., 2016). Thus, to wilderness. Therefore, new approaches are necessary to better address there is a clear need to analyse UGBS not only in terms of their quantity the opportunities and challenges associated with rewilding UGBS in but also in UGBS quality attributes related to the conservation value of cities. Identifying factors influencing the biodiversity and their rela­ the naturalness present and their capacity to harbour urban biodiversity tionship with wildness and habitat quality at smaller scales at the urban (Reyes-Riveros et al., 2021; Lepczyk et al., 2017). end of the wilderness continuum is a research avenue that calls for UGBS may include areas with different management regimes, further exploration (Carver & Fritz, 2016; Dymond et al., 2003; ranging from heavily managed formal UGBS such as urban parks to Kowarik, 2011). barely managed semi-natural and ruderal areas (Aguilera et al., 2019). Research on urban ecology is usually mostly focused on large There is a growing interest in understanding how to reduce intensive metropolitan regions and capital cities, despite the fast growth and management in UGBS and restore relatively ‘wild’, self-managing eco­ dominance of small and medium-sized cities in many regions such as systems to avoid and reverse urban biodiversity loss (Aguilera et al., Europe (Borsekova et al., 2018; Boulton et al., 2018). Hence, the 2019; Müller et al., 2018). Wildness is the relative quality of being wild assessment of how UGBS and its capacity to harbour biodiversity in mid- or undomesticated, encompassing a broad spectrum of landscape con­ sized cities is still understudied (Boulton et al., 2018). Here, we analyse texts, anthropogenic influence and scales: from remnant patches of the suitability of the spatially-explicit HQ and wildness models, as pre­ urban vegetation to vast pristine areas, disturbance regimes and dictors of biodiversity when adapted to a mid-size urban area. Particu­ biogeochemical processes (Noss, 2020). Wildness, as a landscape char­ larly, we: (i) spatially assessed relative urban wildness as a landscape acter can be spatially represented. Yet, it cannot be directly separated quality following Müller et al. (2018); (ii) adapted the InVEST habitat from non-wildness, but rather expressed from less to more ‘wild’ along a quality module for the assessment of habitat quality on the urban continuum of anthropogenic impacts (Zoderer et al., 2020; Carver & landscape, considering the effects of anthropogenic impacts on the Fritz, 2016). There have been many recent wildness mapping studies urban landscape and particularly across UGBS; (iii) identified potential (Müller et al., 2015) almost exclusively referred to non-urban land­ UGBS that may have a key role in terms of biodiversity maintenance by scapes, often minimally affected by human action (Zoderer et al., 2020; using the outputs from (i) and (ii); and (iv) analysed the correlation Radford et al., 2019; Carver et al., 2012, 2013). Even though only one between wildness, HQ, and species richness as biodiversity proxies. study has applied wildness mapping in cities (Müller et al., 2018), wildness can be a cost-effective indicator for management to be used in 2. Materials and methods urban contexts: to enhance biodiversity, screening for priority conser­ vation areas, while improving human well-being in cities (Mansur et al., 2.1. Study area 2022; Jalkanen et al., 2020; Noss, 2020; Kowarik, 2018; Müller et al., 2018). We conducted our study in Vitoria Gasteiz, a middle-sized European In addition, habitat quality (hereafter HQ) indicates the capacity of city, and the administrative capital of the Basque Country (248.087 an ecosystem to deliver the resources and conditions needed for wildlife inhabitants, Eustat, 2020), located in the North of the Iberian Peninsula and is a key determinant of biodiversity (Terrado et al., 2016; Hall et al., (Fig. 1). Vitoria-Gasteiz is situated at the centre of a transitional 1997). As a way to assess HQ, ecological spatial models are gaining biogeographic region (Mediterranean and Eurosiberian) (De la Hera, research interest to evaluate habitat attributes, such as key resources 2019) with an extensive plain delimited by mountains connecting the and the constraints impairing the use of resources (i.e. anthropogenic Cantabria (West) and the Pyrenean (East) mountain ranges, two 2 C. Aznarez et al. Landscape and Urban Planning 228 (2022) 104570 Fig. 1. Location of Vitoria-Gasteiz municipality, city and main urban green and blue spaces within city boundaries defined by urban land uses and delimited by the green belt. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) biodiversity hotspots maintaining the ecological connectivity of north­ by the Environmental Studies Centre (CEA), a public agency of the ern Iberia, and part of the Pan-European Ecological Network (CEA, Vitoria-Gasteiz City Council. Brownfields and vacant land were classi­ 2014). Vitoria-Gasteiz has a remarkable share of urban and periurban fied based on the perceived naturalness as “Recreative areas” or “Rela­ green and blue spaces (Fig. 1), including a 35 km-long green belt sur­ tively extensive open landscapes” according to the reconversion plan or rounding the core city (Orive & Lema, 2012). The green belt was created use assigned for each area in the inventory of NBS provided by the City during the 1990s as a way to transform peripheral degraded areas into Council. UGBS along with enhancing biodiversity by restoring habitats and Remoteness was mapped by combining distance from mechanised increasing ecological connectivity (Newman & Cabanek, 2020). It con­ access (shortest walking distances from main roads) and noise exposure sists of a system of peri-urban semi-natural green spaces, promenades, to any pixel on the map as in Carver et al. (2013) and Müller et al. wetlands, streams, and ponds in the interface between the city and the (2018). We reclassified the previous composite map based on the land countryside (Monclús, 2018). Because of their natural value, Salburua use data into a cost surface building on Müller et al. 2018 (Table S3, wetlands, Vitoria Mounts and the River Zadorra have been protected as Supplementary material). We estimated the time it takes a pedestrian to community interest sites and included as part of the European Natura move across each pixel based on the assumed travel times for each land 2000 network (CEA, 2014, Fig. 1). With 20 m2 of UGBS per inhabitant, use type. The minimum time needed by a pedestrian to access any pixel Vitoria-Gasteiz is considered a “green” city (CEA, 2014). Because of its in the study area from a vehicle access point was then calculated by the pioneering and ambitious greening strategies, Vitoria-Gasteiz gained Path distance tool of ArcGIS. The weighted means of noise values (Lden) international recognition and was awarded in 2012 as a European Green retrieved for the study area (CEA, 2017) were used for the calculation of Capital (CEA, 2014). In addition, Vitoria-Gasteiz offers a great test-bed noise exposure. The pixels without noise data were assigned 30 dB, for research given the high availability environmental data, including comparable to a quiet garden (Müller et al., 2018). We reclassified these the spatial data required in our modelling approach. into inverse values, so high values show low decibel values making sure that the indicator correlated positively to UW. Next, we combined both datasets, remoteness from mechanised access and noise exposure to 2.2. Urban wildness spatial model create the final remoteness map. Challenging terrain was depicted by combining terrain ruggedness We mapped urban wildness (UW) at the city level along a continuum index and occurrence of wetlands, again building on previous ap­ of historically human-modified landscapes, based on three parameters proaches (Müller et al., 2018; Carver et al., 2013). The ruggedness index building on Carver et al. (2013), Müller et al. (2015), Müller et al. was considered as the standard deviation (SD) of terrain curvature (2018), and Radford et al. (2019): i) Perceived naturalness, ii) within a 250 m radius of the observer since it has been generally Remoteness and iii) Challenging terrain. assumed that people perceive that area of their surroundings (Müller Perceived naturalness was defined as the vegetation and land cover et al., 2018; Scottish Natural Heritage, 2014; Carver et al., 2012). We pattern created by land (un)management, appearing natural to the ca­ calculated the ruggedness index from a high-resolution (10 m cell size) sual observer (Carver et al., 2013). To estimate perceived naturalness for digital elevation model (DEM) data available for the study site (CEA, the whole study area, we joined the following spatial datasets: Urban 2017), with a higher score indicating steep and rough terrain. At this Land Use Map 2020, Green Belt Land Use Map 2020, and Nature-Based resolution, the DEM would also capture anthropogenic infrastructure (e. Solutions Inventory (2020) (Table S1, Supplementary material) on a g. buildings), so when calculating the SD of the terrain curvature such composite land use dataset (hereafter ‘land use map’) at a resolution of buildings would tend to return high values (Müller et al., 2018). To 10 m. Land uses were reclassified into 18 naturalness classes following avoid the misleading effect of anthropogenic infrastructure on the index, Müller et al. (2018) (Table S2, Supplementary material). Naturalness we first excluded all pixels containing built infrastructure (e.g. roads, values ranged from 1, representing ‘completely sealed areas’, to 18 as railways, buildings), from the dataset, and then the SD was calculated ‘land cover under the least human influence’. All datasets were provided 3 C. Aznarez et al. Landscape and Urban Planning 228 (2022) 104570 for each pixel using a 250 m neighbourhood buffer. As a previously ‘unnatural’) approach (Müller et al., 2018), we considered the relative identified limitation, the method captures ruggedness rather than the habitat suitability score for each land use class organized along a challenging character of the terrain more generally, so flatter areas such naturalness continuum. To account for threats, we used spatial infor­ as wetlands and bogs tend to be underestimated (Müller et al., 2018; mation of railways and traffic intensity sorted by main, secondary, and Scottish Natural Heritage, 2014). Thus, to cover all challenging terrain residential roads (Vitoria-Gasteiz City Council, 2020). We also included in terms of topographic attributes, we considered the occurrence of raster information on noise pollution decomposed by dB intensity wetlands from the land use map to the terrain ruggedness index. (Vitoria-Gasteiz City council, 2018) and population density data (Vito­ Therefore, if a pixel cell led within a wetland, we added the average SD ria-Gasteiz city council, 2016) as additional potential threats. of terrain ruggedness to the pixel cell value (Müller et al., 2018). A relative habitat suitability score Hj (scaled from 0 to 1), was We obtained the final relative UW map by Vitoria-Gasteiz by assigned to each land use class, where 1 indicates land use classes with combining these three indicator maps (perceived naturalness, remote­ the highest suitability for biodiversity (Sharp et al., 2020). Hj and all ness, and challenging terrain) using equal-weighted simple addition. We threat parameters were determined through expert consultation then rescaled each indicator map and the resulting map to a 0–1 range (following Kuhnert et al., 2010). This information was gathered from a using Eq. (1) following Carver et al. (2013) and Müller et al. (2018): purposedly designed structured survey administered to 21 international experts, on urban ecology, biodiversity conservation, and urban envi­ (Xij − OmV)*(NMV − NmV) Sij = (1) ronmental management who suggested values to parametrize our urban (OMV − OmV + NMV) landscape model (see Questionnaire S1, in Supplementary material) (Tables S5 and S6). where Sij refers to the standardized value of cell j in map i, Xij to the Since the list of 18 naturalness classes was too extensive for an expert current value of cell j in the map I, OmV to the old minimum value, OMV survey, we grouped these classes into 8 categories of urban habitats to the old maximum value, NmV to the new minimum value, and NMV to based on urban green infrastructure typology according to Hansen et al. the new maximum value. (2017) (Table S4, Supplementary material). During the administering of We performed a Jack-knife analysis to assess the sensitivity - the survey, we did not share any results or feedback among the group of robustness and weigh the uncertainty of the final UW model by checking experts as in Terrado et al. (2016), so our method relied on experts the effect that leaving one input variable out of the model would have on having a good understanding of the questions. To estimate the model the results (Quenouille, 1956). We calculated three alternative maps by uncertainty and to determine parametrization scores from each excluding one input map, respectively. Then we calculated the corre­ descriptor variable, we first calculated the mean (µ), standard deviation lation coefficients of the three resulting maps when compared to the (SD), and coefficient of variation (CV). We then identified and excluded original (full) UW map by using the r.regression.line module in GRASS from the analysis extreme values in the valuation of any descriptors from 7.8.5 (QGIS.org, 2021). All the spatial analysis was performed in QGIS the expert survey by z-score deviation. The HQ score uncertainty was 3.18.2 (QGIS.org, 2021) and ArcGIS 10.7.1 (ESRI, 2019). calculated with the Zonal Statistic tool (QGIS) as the average HQ and their coefficient of variation for each habitat typology class (Di Febbraro 2.3. Habitat quality spatial model et al., 2018). The final values used as input parameters to build the HQM are reported in Tables S5 and S6 from the supplementary material. We calculated the indicator for habitat quality (HQ) using the The total threat level value in each raster cell x of habitat typology j spatially-explicit InVEST Habitat Quality Model (v.3.8.9; Stanford Uni­ was given by Dxj (Eq. (2)), where y indexes all grid cells on r’s map and Yr versity, Stanford, CA, USA) which estimates the relative degradation corresponds to the set of raster cells of r’s map. If Sjr (relative sensitivity extent and status of different habitat types in a given region. This model of each habitat type to the threat) equals 0, then Dxj is not a function of is based on the relative sensitivity to different threats, distance to these threat r. Threats were normalized so the sum of all considered threats potential threats, and location of protected areas. The HQ approach equals 1 (Sharp et al., 2020): assumed that habitats with higher quality will support higher biodi­ ∑R ∑Yr Wr versity and is a general estimator of biodiversity suitability (Sharp et al., Dxj = r=1 y=1 (∑R )ry rxi βx Sjr (2) r=1 Wr 2020). Therefore, we estimated HQ as a function of i) the suitability (Hj) of each land use to provide habitat for biodiversity (as local species The raster cell’s degradation score was then translated to habitat richness), ii) anthropogenic threats which might impair HQ, and iii) the quality (Qxj) scores along with Hj using scaling parameters (z and k) in sensitivity of each land use type to each threat. HQ models can be used Eq. (3) (Sharp et al., 2020). Qxj is equal to 0 if Hj is equal to 0. Qxj can for a particular species or general patterns of biodiversity associated to never be greater than 1. As raster cells degradation score increases, habitat quality, regardless any particular taxonomic or functional group, habitat quality decreases and vice versa: which makes it a suitable rapid tool for general biodiversity assessment ( ( )) studies (Sharp et al., 2020). Since not all habitats are affected by Dzxj Qxj = Hj 1 − (3) different threats, in the same way, we characterised the sensitivity of Dzxj + kz habitat types to various threats. We considered habitat suitability to be affected by: i) the relative impact of each threat (Wr, weighted relative importance of each threat), ii) the distance between habitat and the 2.4. Spatial correlation with urban biodiversity data source of the threat (including the maximum threshold distance, Max. D), and iii) the relative sensitivity of each habitat type to the threat (Sjr). We evaluated the spatial correlation of the UW and HQ models with Further details on the InVEST HQ model and parameterisation can be observed biodiversity data considering the correlation coefficient (R), found in Sharp et al. (2020), Sallustio et al. (2017), and Terrado et al. the adjusted coefficient of determination (R2), and the regression of the (2016). residuals, with the ‘r.regression.line’ and ‘r.regression.multi’ modules in Since our work was based on an urban context, the HQ models were GRASS 7.8.5 (QGIS.org, 2021). Biodiversity data relates to observed adapted from the InVEST approach by considering the spatial distribu­ species richness (as a proxy of alpha biodiversity, i.e. mean species di­ tion of habitats including a broad range of managed land use classes as versity at a local scale (Whittaker, 1972)) from occurrence records of potential habitat providers to urban biodiversity rather than merely mammals, birds and butterflies obtained from the ‘Ornitho.eus’ data­ threats (Sharp et al., 2020). To define the potential habitats, we used the base (www.ornitho.eus) for the period January 1994 to December 2020. categorization of the 18 naturalness classes described in section 2.2. Birds and butterflies are common core indicators of biodiversity ac­ (Table S2, supplementary data). Instead of a binary (i.e. ‘natural’ vs cording to the City Biodiversity Index (SCBD, 2012), and all groups 4 C. Aznarez et al. Landscape and Urban Planning 228 (2022) 104570 selected are well represented across citizen science and monitoring centre (Fig. 2). Areas with higher UW corresponded to challenging programs in the city (Albaina et al., 2020; De la Hera, 2019). Each record terrain zones or with barriers to access, particularly wetlands, streams, included a species occurrence at a specified geographic location along foothill forests, and hills mostly located at the edges of the city and with the date and number of individuals. We first cleaned the species distant from the urban centre (i.e. Ramsaŕs Salburua wetlands in the East occurrence records retrieved from the ‘ornitho’ dataset by removing and mountains to the South). Several large-sized UGBS, like large parks geospatial errors, duplicated, and museum registers. To correlate and freshwater ecosystems (i.e. rivers and wetlands), also showed observed species richness with HQ and UW values, we first joined the relatively high UW even when located near the city centre (Fig. 2). cleaned species occurrences to a grid cell polygon layer of 100 × 100 m Build-up infrastructure from residential and industrial areas was map­ covering the whole study area. We then extracted the total count of ped as least wild, corresponding to ‘completely sealed’ or ‘mostly sealed species occurrences, species richness, and the mean values of our areas’ according to Müller et al. (2018) classification. UW decreased explanatory variables (i.e. UW and HQ) matching this grid cell scale. The towards higher impervious coverage and was particularly low in large responses of species richness to UW and HQ were analysed separately for industrialized with few or absent UGBS. each taxonomic group by Gaussian generalized linear models (GLM) The sensitivity analysis of the UW model, comparing the differences first, for the whole city and then spatially constrained to UGBS. We in UW between the original (full inputs) and three alternative maps after selected the best models based on their statistical significance (F and p excluding one of the three input parameters (perceived naturalness, values <0.05) and goodness of fit (adjusted R2), checking for normality remoteness and challenging terrain), showed a stable mapping trend and and homoscedasticity through visual inspection of the residuals. To high correlation (correlation when excluding: Naturalness R = 0.75; achieve a better fit for the whole city scale model, species richness was Remoteness R = 0.97; Ruggedness R = 0.97) (Appendix A). log10-transformed. To estimate the strength of the relationship between species richness, UW and HQ, we performed non-parametric Spearman’s 3.2. Spatial patterns of urban habitat quality correlation tests. To test how different or complimentary were UW and HQ at predicting urban species richness, i.e. if the model had a detect­ There was a clear urban–rural gradient in relation to the HQ spatial able effect on richness, we regressed that variable (e.g. UW) against the pattern, where most urban core areas were associated with low HQ residuals of related variables (i.e. richness as a function of habitat values, while peripheral and large-sized UGBS showed the highest HQ to quality) and vice versa. All the statistical analyses were conducted in support biodiversity (Fig. 3). These areas mostly corresponded to the RStudio 1.4.1717 (R Core Team, 2021) and GRASS 7.8.5 (QGIS.org, Ramsar wetland Salburua (to the East, Fig. 3), foothill forests and hills 2021). (to the South and West, Fig. 3). Streams showed high HQ values but were mostly restricted to the watercourse and not in the surrounding envi­ 3. Results ronment (i.e. floodplain). Furthermore, these ecosystems were particu­ larly affected towards the city centre depicting very low HQ (Fig. 3). We found that UGBS with higher UW and HQ values had the highest Vitoria-Gasteiz showed a large coverage of UGBS, mostly related to biodiversity values for all considered taxonomic groups. These UGBS green corridors or parks near the city centre (e.g. Zabalgana and Park were generally represented by large-sized peripheral areas, with high Arriaga, Fig. 1) that were however associated with low HQ values, some perceived naturalness, remoteness, challenging terrain (for UW), and of them ranging between 0.12 and 0.15 (Fig. 3). high habitat suitability (for HQ). Both UW and HQ estimators were The coefficients of variation (CV) that expressed the uncertainty for positively correlated to each other and associated with biodiversity. the expert-based parametrization of habitat suitability for HQ, showed that freshwater ecosystems, natural, semi-natural areas and feral areas 3.1. Spatial patterns of urban wildness were generally perceived as more natural and similarly valuated (CV freshwater ecosystems = 14 %, CV natural, semi-natural areas and feral UW mapping showed that peripheral areas were associated with areas = 11 %, Appendix S5). Whereas, the habitat suitability valuation higher wildness values, and in general lower values towards the city uncertainty was much higher for more anthropized areas, and the Fig. 2. Urban wildness map: relative wildness for Vitoria-Gasteiz city on a 10 m resolution. 5 C. Aznarez et al. Landscape and Urban Planning 228 (2022) 104570 Fig. 3. Urban habitat quality (HQ) map for Vitoria-Gasteiz city: Landscape-scale relative HQ mapping on a 10 m resolution. Habitat quality is based on a continuous scale from 0 to 1, where the lowest values indicate lower habitat quality (warm colours) and the higher values indicate higher habitat quality (cold colours). highest level of uncertainty corresponded to predominantly sealed areas UW depicted higher values than HQ in large-sized green spaces and (CV = 238 %) (Appendix S5). Other land uses such as agricultural land, peripheral areas including forests, hills, wetlands and rivers but also in parks or grasslands were associated with intermediate habitat suitability centrally-located areas such as urban parks had higher UW values CV values of 30 %, 24 % and 30 % respectively (Appendix Table S5). compared to HQ (in red Fig. 4). Particularly, freshwater ecosystems Furthermore, the coefficient of variation concerning the threats for HQ (river and wetlands) and the surrounding areas depicted higher UW was highly variable depending on the land use class and type of threat compared to HQ (Fig. 4 in red). While in the case of HQ, peripheral and (Table S6). industrial areas showed higher values than UW (Fig. 4 in blue). 3.3. Compared wildness and habitat quality spatial patterns 3.4. Wildness and habitat quality as proxies for biodiversity support Even if UW and HQ were positively correlated (Adjusted R2 = 65 %) The ‘ornitho’ occurrence registers for Vitoria-Gasteiz consisted of and explained a similar amount of biodiversity variation, they showed 50,168 ind of 250 species of birds, 2721 ind of 48 species of mammals spatial differences for biodiversity estimations in particular areas, with and 2063 ind of 97 species of butterflies. The most frequently observed some areas with a significantly higher valuation for UW than for HQ (in bird species were mostly generalists and songbirds like Pica pica red Fig. 4) and vice versa (in blue Fig. 4). (Eurasian Magpie) or Turdus merula (Eurasian Blackbird), and the Fig. 4. Comparative mapping of urban wildness (UW) and habitat quality (HQ) explaining urban biodiver­ sity patterns. Red areas indicate higher UW values compared to HQ, while blue areas indicate higher HQ values compared to UW. The higher the colour in­ tensity the higher the differences between UW and HQ. Residual map based on the regression of UW as a function of habitat quality (HQ). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) 6 C. Aznarez et al. Landscape and Urban Planning 228 (2022) 104570 worldwide dominant Passer domesticus (House sparrow) (Fig. 5). The adapted to urban environments to capture an area’s capacity to support most common butterflies were Pararge aegeria (Speckled Wood), Pieris biodiversity. UW and HQ were significantly positively correlated with rapae (Small White) and Colias crocea (Clouded yellow). Among the most the biodiversity for all the considered taxonomic groups, and thus common mammals, we found Sus scrofa (Wild Boar), Oryctolagus cuni­ correlated to each other. The selection of these taxonomic groups was culus (European Rabbit) and Vulpes vulpes (Red Fox), mostly mid-sized appropriate since they are common in urban environments and easily generalists and omnivores. However, other mammals observed in the identifiable, making them suitable for citizen science programs. The case study area include the endangered Arvicola sapidus (Southern Water most frequent species within each group, particularly in the case of Vole) and the critically endangered Mustrela lutreola (European mink) generalists birds and small to medium-sized mammals, are in general according to the IUCN red list (IUCN, 2021). highly recorded in crowdsourced databases due to their visibility and Species richness of the different considered taxonomic groups (birds, widespread presence in UGBS (Sultana & Storch, 2021). However, the butterflies and mammals) was positively correlated to both, UW and HQ biodiversity in UGBS is largely conditioned by the spatial and vegetal values for the whole study area (Fig. 6 and Table S7 – Supplementary conformation of the UGBS (Reyes-Riveros et al., 2021; Lepczyk et al., material). The Spearman’s correlation for the strength of the relation­ 2017; Turner & Gardner, 2001). Some of the most frequent bird species ship between species richness and UW, showed a strong positive cor­ found in the city such as magpies and blackbirds are commonly asso­ relation for birds (p < 0.001, R = 0.26), butterflies (p < 0.01, R = 0.33) ciated with weedy patches and native shrubs in central areas within the and mammals (p < 0.001, R = 0.34), respectively. Similarly, the city and for which the amount of vegetation is a major determinant of Spearman’s correlation for richness and HQ, was also strongly and density, diversity and distribution (Forman, 2014). Butterflies are usu­ positively correlated for birds (p < 0.001, R2 = 0.32), butterflies’ (p < ally the least species-rich in areas with intensive management and 0.01, R2 = 0.29) and mammals (p < 0.001, R2 = 0.37). design such as the case of urban parks (Aguilera et al., 2019; Forman, The goodness of fit increased substantially when focusing our anal­ 2014). The list of frequent butterfly species observed in our case study, ysis on UGBS instead of the whole study area (Fig. 6). Here, the rela­ coincided partially with previous research showing the decline of spe­ tionship between species richness and both modelled predictors UW and cies richness in butterflies related to management intensity (Aguilera HQ tested positive for all the considered taxonomic groups showing et al., 2019). In the case of large wild mammals, despite being less likely considerably higher correlation values and significance (Fig. 6). to be observed on an urban scale compared to birds and butterflies, the The Spearman’s correlation of richness against UW was positively presence of mid-sized generalists within the city may be indicating that correlated for birds (p < 0.001, R2 = 0.39), butterflies (p < 0.001, R2 = part of the UGBS may act as wildlife corridors, entering the city (For­ 0.79), and mammals (p < 0.01, R2 = 0.61). Likewise, species richness man, 2014). and HQ were positively correlated for birds (p < 0.05, R = 0.2), but­ We found that UGBS located in the periphery of the city, generally terflies (p < 0.001, R = 0.65) and mammals (p < 0.01, R = 0.62). large-sized, were associated with similar high range values for both, UW Overall, the richness of all considered taxonomic groups showed a and HQ and therefore the highest potential to support urban biodiver­ consistent positive response to UW and HQ. sity. These areas were mostly represented by, freshwater ecosystems, Residual regressions analysis showed that UW and HQ were not large-sized parks, and foothills, where human access and threats were influencing richness independently, but rather capturing the same re­ lower, and they are currently protected as community interest sites in lations (Table S8, Supplementary material). All the different taxonomic the Natura 2000 network, such as the Ramsar wetland Salburua and the groups used to test the relationship between HQ, UW and richness, Zadorra river (CEA, 2014). Parks and other residential UGBS, patchier, showed a clear trend to increase as our indicators increased. small-sized and centric (Fig. 1) were characterized by low values of both proxies of biodiversity support UW and HQ, being particularly low when 4. Discussion considering HQ. We found contrasting patterns of UW and HQ values between these areas, suggesting that not only the location but the size 4.1 Habitat quality and wildness to biodiversity and shape of UGBS may have an important role in influencing biodi­ versity (Figs. 2 and 3). These findings are consistent with previous Our results suggest that both UW and HQ models can be adequately research highlighting the positive relation between UGBS characteristics Fig. 5. Ordered list of the 10 most common species of birds, butterflies and mammals, in the ‘ornitho’ biodiversity dataset as occurring in Vitoria-Gasteiz (year: 1994–2020). 7 C. Aznarez et al. Landscape and Urban Planning 228 (2022) 104570 Fig. 6. Upper panel: Whole-city patterns, regression analysis of richness against UW (left) and HQ (right), and fitted linear model results for each taxonomic group. Lower panel: Regression analysis of richness against UW and HQ constrained to UGBS and fitted linear model results for each taxonomic group (R2 and significance values of p < 0.05 and confidence intervals (shaded area). on species richness, as size and habitat quality of UGBS in cities are key account not only for the size but for the shape and connectedness of local factors to biodiversity (Aronson et al., 2017; Lepczyk et al., 2017; UGBS (Turner & Gardner, 2001). As increasing the size of UGBS in Goddard et al., 2010). Larger UGBS with a lower perimeter:area ratio consolidated cities can be challenging, due to land availability con­ have less border effect, i.e. are less exposed to threats and effects of the straints (Stott et al., 2015), previous research suggests that increasing surrounding cover types (Turner & Gardner, 2001). This may explain the tree coverage and vegetation stratum complexity is a good alterna­ why larger areas had both higher UW and HQ: the larger the UGBS were, tive to improve UGBS functionality (Garizábal-Carmona & Mancera- the more distance from the central areas to the edge, i.e. the higher the Rodríguez, 2021; Müller et al., 2018; Sandström et al., 2006). However, distance to potential threats. Contrastingly, smaller-sized or edged such measures should be carefully considered as management for one shapes with a higher perimeter:area ratio, UGBS are potentially more taxonomic group may be detrimental to others (e.g. increasing tree exposed to threats and the interference cover of human activities (higher coverage would generally benefit birds but not butterflies, as these tend edge effect) (Turner & Gardner, 2001). Studies in large-sized cities to prefer semi-open rather than shaded spaces) (Warren et al., 2021; evidenced that the perimeter:area ratio and indicators related to shape Sandström et al., 2006). and connectedness to surrounding areas (edge effect) are better pre­ Both proxies, UW and HQ were in general higher in the periphery of dictors of bird biodiversity than merely the size of UGBS (Garizábal- the urban areas indicating a higher potential to support biodiversity. Carmona & Mancera-Rodríguez, 2021; Shih, 2018). Despite the positive Some centric areas (e.g. centric parks Fig. 2) were identified with high effect of UGBS size on species richness, it remains poorly understood UW but low HQ, and in general, HQ values were high exclusively in how size thresholds on individual patch size may influence biodiversity areas located in the periphery of the city. These differences may be due conservation in urbanized landscapes (Lepczyk et al., 2017). There are to the inputs used to estimate both indicators. While UW takes into ac­ further current discussions about how land sharing (extensive sprawling count structural descriptors of land use and terrain (perceived natural­ urbanization) vs land sparing (intensive and compact urbanization) ness, remoteness and challenging terrain) (Müller et al., 2018, 2015), urbanisation initiatives along a landscape fragmentation continuum may HQ includes assessing the expert perception of the habitat suitability of influence biodiversity in urban contexts (Stott et al., 2015). Therefore, different land uses and their susceptibility to different anthropogenic effective strategies to increase biodiversity in urban contexts should impacts. Thus, HQ provides not only structural descriptors that can be 8 C. Aznarez et al. Landscape and Urban Planning 228 (2022) 104570 obtained from land use maps and digital elevation models but incor­ assumptions on biodiversity support, data quality, and potential porate functional descriptors of habitat suitability and vulnerability to modelling uncertainties, as happens with other biophysical models (e.g. threats to spatially describe the capacity to support biodiversity (Sharp Aznarez et al., 2021; Bagstad et al., 2013). UW was in general a robust et al., 2020; Di Febbraro et al., 2018; Sallustio et al., 2017). HQ seems to indicator of biodiversity, and this was supported by the results of the better consider potential threats (i.e. roads, night light, population Jack-knife test, where it remained with stable values even when density, railways and noise), their distribution and potential impact on excluding one of its three basic components (perceived naturalness, the capacity of the different land use classes to harbour biodiversity. remoteness and challenging terrain) (Appendix A). However, from these Contrastingly, other areas were identified with higher UW than HQ. For three indicators, the maps excluding perceived naturalness had lower instance, industrial areas with low UW values showed medium HQ performances, indicating that perceived naturalness is a key input values since the spatial distribution and addition of threats had a lower contributing the most to UW character. Yet, other studies suggest that influence in these areas as they are far from the city centre (e.g. night­ remoteness was more important in influencing UW (Müller et al., 2018), light sources were fewer and population density was considerably lower which indicates that the potential role of each indicator may vary than towards the city centre). depending on the habitat distribution of the study area. In general, we found high UW and HQ values in freshwater ecosys­ Expert consultation was effective when lacking threat data to enable tems (e.g. Zadorra river) indicating an overall high capacity to support spatial comparisons and use relative scores to rank habitat suitability biodiversity. Preserved wetlands, streams and surrounding areas and quality indicators. However, different expertise backgrounds from (floodplains) typically support high biodiversity and are ecological the consulted experts, considering the suitability of the urban habitats to corridors that may connect UGBS to larger ecosystems (e.g. peripheral different types of biodiversity, along with the different anthropogenic lakes or mountains) (Mansur et al., 2022). However, these ecosystems threats to habitat suitability may influence the results (Di Febbraro were highly vulnerable in urban areas (Stroud et al., 2022) and were in et al., 2018). Despite the subjectivity and variability in expert consid­ our study identified as with higher UW than HQ due to the high erations (Sallustio et al., 2017) introducing moderate uncertainty in our vulnerability of these ecosystems to potential impacts. This vulnerability HQ predictions, the overall assessment proved to be consistent with the seemed to be particularly high for rivers possibly due to their higher relative UW outcomes and coherent with the species richness indicators. edge effect which lowers the distance to threats (Sharp et al., 2020). The parametrization scores obtained from the international expert Mapping of HQ allows identifying how these potential threats can affect consultation can be further tested to any other urban context, as the the habitat suitability to support biodiversity as well as the connectivity survey was designed for urban biodiversity and land uses in general. The of these ecological corridors. crowd-sourced data of the occurrence of taxonomic groups used to assess potential associations with UW and HQ indicators may have 4.2. Methodological considerations different biases that hinder their suitability as predictions of species richness as overrepresented data in easily accessible UGBS, highly Both UW and HQ are simple indicators, yet robust in describing the populated areas or circulation ways (Sultana & Storch, 2021; Petersen spatial patterns of biodiversity in the urban context, though they have et al., 2021; Callaghan et al., 2020). For instance, bird richness was different potential uses. UW seems to be more suitable when functional higher due to an artefact in the greater number of observations (Fig. 5) data is not available as it incorporates indicators related to the landscape since it is a more charismatic group, widely used in citizen science and structure and the input requirement is relatively low (i.e. digital eleva­ reported in ‘ornitho.eus’. However, these data allow studying large tion model, land use, mechanized access). Besides, both models may spatial scales based on large databases from multiple observations which partially address some limitations associated with NDVI/NDWI in­ overcome at least part of these biases and made these data suitable for dicators, which are widely used in urban environmental studies to urban biomonitoring programs (Callaghan et al., 2020). quantify exposure to UGBS but are less adequate for measuring their UW and HQ were positively correlated and robust proxies to predict quality. Such indicators are sensitive to seasonal variability and do not biodiversity for all considered taxa in our study. This suggests that at differentiate between green space typologies as considered in this study, least with the data availability we had for biodiversity, land use, habitat thus limiting their applicability in the context of urban biodiversity suitability and potential threats, in a middle-sized city such as Vitoria conservation (Trethewey & Reynolds, 2021). Gasteiz, these databases were reliable to estimate biodiversity at urban On the other hand, HQ is a more robust model to evaluate impacts scales. However, there is a substantial unexplained variation between affecting the functionality of the landscape, since it includes information our proxies and biodiversity that is attributable at least partially to other on threats and their relative impact on biodiversity conservation ob­ non-contemplated factors, e.g. seasonal variability, unassessed threats jectives. These differences in UW and HQ, provide more accurate in­ or habitat conditions, biogeographical filters, the effect of invasive formation to managers by identifying areas that can be categorized with species. Yet, further research considering small (e.g. micro-scales) and high UW that are not so likely to support high biodiversity due to the unconventional habitats with aggregated distribution patterns in the influence of threats. Furthermore, our outcomes allow identifying key environment should consider possible limitations in the use of these areas for urban biodiversity protection, (un)suitable places for new models developed for larger areas (Knapp et al., 2021; Soanes et al., urban development or where efforts need to focus on reducing threats 2019). This approach should allow to improve the currently limited impacts. However, both models do not integrate size or connectivity availability of biodiversity data and inventory completeness with an descriptors that are highly associated to habitat quality for biodiversity, adequate aggregation scale for different taxonomic groups across the which would be a good improvement to further model development. urban landscape, ultimately assessing spatial patterns at fine scale res­ Acquisition of extensive information on (urban) biodiversity status is a olution. Recognizing and measuring the value of small spaces and un­ time and resource-consuming endeavour (Jalkanen et al., 2020) and conventional habitats to urban biodiversity (i.e. from brownfields and since the input requirements to assess general biodiversity proxies in our cemeteries to cavities within buildings and infrastructure) calls for spatial models were relatively low, they are particularly suitable for exploration along with novel conservation opportunities for manage­ places where biodiversity information is deficient or monitoring efforts ment (Knapp et al., 2021; Soanes et al., 2019). Our results suggest that are scarce. The combination with expert consultation, a widely-used even if the obtained outcomes may be sensitive to biases, which is often method to obtain information when data is limited (Terrado et al., the case in ecological modelling, they provide valuable and robust 2016; Kuhnert et al., 2010), makes this approach cost-effective to insight into identifying urban areas with high potential to support evaluate potential urban biodiversity support while incorporating biodiversity and to pinpoint conservation areas. habitat suitability and sensitivity to threats to the different habitats. These models have however some limitations based on their different 9 C. Aznarez et al. Landscape and Urban Planning 228 (2022) 104570 4.3. Implications for urban biodiversity management areas with very low UW and HQ values coincided with highly imper­ vious areas. Further research should address how urban biodiversity In our study, UGBS close to the city centre showed low and very low patterns are distributed among vulnerable urban dwellers (Mansur et al., UW and HQ values, suggesting that cities such as Vitoria-Gasteiz, with 2022). However, understanding the way that urban dwellers value and 20 m2 UGBS per inhabitant, do not necessarily encompass high HQ to experience nature among contrasting perspectives is key to identifying biodiversity as these UGBS may be ecologically very poor, although still socio-ecological trade-offs and feedback (Mansur et al., 2022; Hill et al., providing nature contributions to people. Coinciding with other studies 2021). For instance, urban dwellers often perceive rewilding actions (i.e. (Mansur et al., 2022; Soanes et al., 2019; Turner & Gardner, 2001; low maintenance regime) negatively in the city, seeing such areas as Forman, 2014) our results suggest that although small and fragmented abandoned, particularly when applied to urban contexts (Kowarik, UGBS, are important for species persistence, biodiversity would benefit 2018; Botzat et al., 2016). On the other hand, some species present in from decreasing perimeter:area ratio, along with increasing the size, urban surroundings can be seen as pests (e.g. wild boar, seagulls, pi­ quality and connectivity of UGBS from peri-urban areas to core areas. As geons), leading to new human-wildlife conflicts or reinforcing existing the Vitoria-Gasteiz municipal Biodiversity Conservation Strategy points ones (Forman, 2014; Dunn et al., 2006). Management actions should be out, the overall biological diversity of the flora and fauna in urban parks carefully considered and designed to successfully communicate that has decreased significantly due to intensive management practices (i.e. unmanaged areas are intended by planners and not the product of the frequency with which the grass is being cut or pruned and dead trees neglect (Kowarik, 2018; Müller et al., 2018; Botzat et al., 2016). Since removed) (CEA, 2014). Another reason adding to the low HQ for most of the world population including decision-makers and financial biodiversity support is the specific and structural simplicity of UGBS, resources reside in urban areas, urban biodiversity conservation actions which mostly consists of the vertical structure of green spaces (i.e. lawn will ultimately rely on the ability of the population to experience and and trees) (Garizábal-Carmona Garizábal Mancera-Rodríguez, 2021; maintain the connection with urban nature (Callaghan et al., 2020; Soga Aronson et al., 2017; CEA, 2014). Previous studies suggest that classical & Gaston, 2016; Dunn et al., 2006). parks with lawns and tall deciduous trees usually harbour few breeding species (Forman, 2014). Increasing HQ for biodiversity in urban socio- 5. Conclusion ecosystems is critically needed (Soanes et al., 2019), which is also correlated to UW according to our findings. This includes actions aiming UW and HQ indicators have been found to be appropriate for pre­ to improve the conditions for human-wildlife coexistence, by accom­ dicting the potential for biodiversity support in an urban context and modating dynamic natural processes (Mansur et al., 2022), providing therefore be integrated as key layers into urban planning. However, it resources such as refuge sites (e.g. holes in tree chunks, adding shrub should be noted that their use and individual suitability will depend on cover), feeding sources or reducing the impact of different threats the data availability and study focus. UW is more suitable when focused impact to habitat suitability, such as night light or fragmentation on landscape structure indicators if functional data is not available, (Soanes et al., 2019; Forman, 2014). In other words, to increase biodi­ while HQ adds to this spatial information the effects of anthropogenic versity, cities should bring some wildness back. Reducing management impacts on biodiversity and landscape functionality. Overall, this (through rewilding sensu Perino et al., 2019), without affecting people’s research has developed easily adaptable and replicable indicators that accessibility to UGBS could be a cost-effective way to enhance HQ in the can provide information at a fine urban scale, accounting for different city and therefore general biodiversity (Mansur et al., 2022; Müller categories of UGBS and considering the anthropogenic effects on the et al., 2018). A variety of wildness-friendly actions such as reducing quality of UGBS. The methodology used was based on open-access pruning or mowing would be suitable to increase the environmental software, public data and crowdsourced data for all the phases of the quality of UGBS (Kowarik, 2018) while reducing maintenance costs, modelling approach. This approach can provide a detailed level of in­ ultimately benefitting urban dwellers through encouraging interaction formation on urban biodiversity conservation in cities with limited with nature. Vacant land and brownfields may likewise be an opportu­ monitoring capacities. Our study suggests that strategies aiming to in­ nity for increasing the network of green infrastructure in urban land­ crease biodiversity in urban landscapes should be based on improving scapes (Kabisch & Haase, 2013). Due to the scattered and low density of wildness by enhancing structural dimensions of UGBS (i.e. perimeter: urban residential developments, Vitoria-Gasteiz has a considerable area ratio, size and connectivity), reducing management frequency and share of vacant and abandoned land, which has been included in the intensity along with evaluating potential anthropogenic pressures on municipal Green Strategy. Yet, previous research warns that the repur­ different habitats. posing of such areas may have undesired consequences for biodiversity Urban dwellers would benefit from enhanced contact with nature by (Macgregor et al., 2022; Broughton et al., 2020). Therefore, the increasing these drivers of urban biodiversity, but this would rely on the ecological value of vacant land and brownfields along with its potential capacity to adequately communicate the aim and intentions behind for urban wildness for increasing both, local biodiversity and recrea­ rewilding actions. We, highlight that UGBS planning should focus not tional purposes, should be carefully considered (Macgregor et al., 2022; only on access and quantity but also on the ecological quality and Kabisch & Haase, 2013). particularly on the support for biodiversity that will ultimately enhance As the size and connectivity of UGBS influence biodiversity, it is also urban dwellers’ experiences in their nearby nature. closely related to public access to recreation in natural environments (Senetra et al., 2018). Experiencing nature in cities is a necessary con­ CRediT authorship contribution statement dition for conservation since people with more exposure and access to nature are more interested in its conservation (Callaghan et al., 2020). Celina Aznarez: Conceptualization, Methodology, Data curation, As an increasing majority of the human population lives in cities, urban Validation, Formal analysis, Investigation, Visualization, Writing – biodiversity is becoming the main people’s interaction with nature original draft, Funding acquisition. Jens-Christian Svenning: Concep­ (Dunn et al., 2006). However, considering the link between contact with tualization, Methodology, Resources, Supervision, Writing – review & nature and conservation measures, it is important to address the ineq­ editing. German Taveira: Conceptualization, Methodology, Formal uitable distribution of urban nature, and specifically, biodiversity, as it analysis, Writing – review & editing. Francesc Baró: Conceptualization, may contribute to lower levels of participation from minorities in Methodology, Resources, Supervision, Writing – review & editing. Unai environmental leadership initiatives (Mansur et al., 2022; Dunn et al., Pascual: Conceptualization, Methodology, Resources, Supervision, 2006). Here, our results may be also used to indicate areas within the Writing – review & editing, Funding acquisition. city with little or no UGBS, and relate it to the demand side for such UGBS (i.e. through population density at the neighbourhood scale) as 10 C. Aznarez et al. Landscape and Urban Planning 228 (2022) 104570 Declaration of Competing Interest Callaghan, C. T., Ozeroff, I., Hitchcock, C., & Chandler, M. (2020). Capitalizing on opportunistic citizen science data to monitor urban biodiversity: A multi-taxa framework. Biological Conservation, 251, Article 108753. The authors declare that they have no known competing financial Carver, S., Comber, A., McMorran, R., & Nutter, S. (2012). A GIS model for mapping interests or personal relationships that could have appeared to influence spatial patterns and distribution of wild land in Scotland. Landscape and Urban the work reported in this paper. Planning, 104(3–4), 395–409. Carver, S., Tricker, J., & Landres, P. (2013). Keeping it wild: Mapping wilderness character in the United States. Journal of Environmental Management, 131, 239–255. Data availability Carver, S. J., & Fritz, S. (2016). Mapping wilderness. Netherlands, Dordrecht: Springer. Cao, Y., Carver, S., & Yang, R. (2019). Mapping wilderness in China: Comparing and integrating Boolean and WLC approaches. Landscape and Urban Planning, 192, Article Data will be made available on request. 103636. Centro de Estudios Ambientales (CEA). (2014). The Urban Green Infrastructure of Acknowledgements Vitoria-Gasteiz. Retrieved from: https://www.vitoriagasteiz.org/docs/wb021/ contenidosEstaticos/adjuntos/eu/34/21/53421.pdf. Centro de Estudios Ambientales (CEA). (2017). Strategic noise maps for Vitoria-Gasteiz. We are grateful to the Environmental Studies Centre (CEA) from the Retrieved from: https://www.vitoriagasteiz.org/wb021/was/contenidoAction.do? Vitoria-Gasteiz council, especially to Asier Sarasua and Aitor Albaina for idioma=es&uid=u1183c8cf_13967725e8d__7ffdu. Centro de Estudios Ambientales (CEA). (2017). Digital elevation moel for Vitoria-Gasteiz. their collaboration and assistance with data collection and for providing Retrieved from: https://www.vitoria-gasteiz.org/wb021/was/contenidoAction.do? valuable information from the city and greenbelt. We are also grateful to idioma=es&uid=app_j34_0119. all the experts who kindly participated in the survey design and filled De la Hera, I. (2019). Seasonality affects avian species distribution but not diversity and nestedness patterns in the urban parks of Vitoria-Gasteiz (Spain). Animal Biodiversity out the questionnaire for model parametrization, along with providing and Conservation, 42(2), 279–292. interesting discussion points. We thank Juan Pablo Pacheco for his Di Febbraro, M., Sallustio, L., Vizzarri, M., De Rosa, D., De Lisio, L., Loy, A., et al. (2018). feedback on the manuscript. CA was supported by the Doctoral INPhI­ Expert-based and correlative models to map habitat quality: Which gives better NIT–INCOMING program, fellowship code (LCF/BQ/DI20/11780004), support to conservation planning? Global Ecology and Conservation, 16, e00513. Díaz, S., Pascual, U., Stenseke, M., Martín-López, B., Watson, R. T., Molnár, Z., et al. from “La Caixa” Foundation (ID 100010434). JCS considers this work a (2018). Assessing nature’s contributions to people. Science, 359(6373), 270–272. contribution to his VILLUM Investigator project “Biodiversity Dynamics Dunn, R. R., Gavin, M. C., Sanchez, M. C., & Solomon, J. N. (2006). The pigeon paradox: in a Changing World” funded by VILLUM FONDEN (grant 16549). This Dependence of global conservation on urban nature. Conservation Biology, 1814–1816. work was supported by the Basque Centre for Climate Change ‘‘Unit of Dymond, C., Carver, S., & Phillips, O. (2003). Investigating the environmental cause of Excellence’’ (Spanish Ministry of Economy and Competitiveness; MDM- global wilderness and species richness distributions. In Seventh World Wilderness 2017-0714). Congress Symposium: Science and Stewardship to Protect and Sustain Wilderness Values. USDA Forest Service, Rocky Mountain Research Station, Ogden, Utah, USA. This research contributes to ICTA-UAB “María de Maeztu” Pro­ Engemann, K., Pedersen, C. B., Arge, L., Tsirogiannis, C., Mortensen, P. B., & gramme for Units of Excellence of the Spanish Ministry of Science and Svenning, J. C. (2019). Residential green space in childhood is associated with lower Innovation (CEX2019-000940-M). risk of psychiatric disorders from adolescence into adulthood. Proceedings of the National Academy of Sciences, 116(11), 5188–5193. Eustat. (2020). Euskal Estatistika Erakundea. Basque Institute of Statistics. Municipal Appendix A. Supplementary data statistics for inhabitants. Retrieved from: https://www.eustat.eus/estadisticas/tema_ 268/opt_1/tipo_1/ti_estadistica-municipal-de-habitantes/temas.html#el. Fuller, R. A., Irvine, K. N., Devine-Wright, P., Warren, P. H., & Gaston, K. J. (2007). Supplementary data to this article can be found online at https://doi. Psychological benefits of greenspace increase with biodiversity. Biology Letters, 3(4), org/10.1016/j.landurbplan.2022.104570. 390–394. Forman, R. T. (2014). Urban ecology: Science of cities. Cambridge University Press. References Garizábal-Carmona, J. A., & Mancera-Rodríguez, N. J. (2021). Bird species richness across a Northern Andean city: Effects of size, shape, land cover, and vegetation of urban green spaces. Urban Forestry & Urban Greening, 64, Article 127243. Aguilera, G., Ekroos, J., Persson, A. S., Pettersson, L. B., & Öckinger, E. (2019). Intensive Goddard, M. A., Dougill, A. J., & Benton, T. G. (2010). Scaling up from gardens: management reduces butterfly diversity over time in urban green spaces. Urban Biodiversity conservation in urban environments. Trends in Ecology & Evolution, 25 Ecosystems, 22(2), 335–344. (2), 90–98. Albaina, A., Sarasua, A., Lobo, L., Belamendia, G., Vilela, J., Agut, A., & Itziar-Velasco, A. Gómez-Baggethun, E., & Barton, D. N. (2013). Classifying and valuing ecosystem services (2020). Aplicación de la nueva versión del índice de biodiversidad urbana (índice de for urban planning. Ecological Economics, 86, 235–245. Singapur) en la ciudad de Vitoria-Gasteiz. CONAMA. Hall, L. S., Krausman, P. R., & Morrison, M. L. (1997). The habitat concept and a plea for Aronson, M. F., Lepczyk, C. A., Evans, K. L., Goddard, M. A., Lerman, S. B., MacIvor, J. S., standard terminology. Wildlife Society Bulletin, 173–182. et al. (2017). Biodiversity in the city: Key challenges for urban green space Han, Y., Kang, W., Thorne, J., & Song, Y. (2019). Modeling the effects of landscape management. Frontiers in Ecology and the Environment, 15(4), 189–196. patterns of current forests on the habitat quality of historical remnants in a highly Aznarez, C., Jimeno-Sáez, P., López-Ballesteros, A., Pacheco, J. P., & Senent-Aparicio, J. urbanized area. Urban Forestry & Urban Greening, 41, 354–363. (2021). Analysing the Impact of Climate Change on Hydrological Ecosystem Services Hansen, R., Rall, E., Chapman, E., Rolf, W., Pauleit, S. (eds., 2017). Urban Green in Laguna del Sauce (Uruguay) Using the SWAT Model and Remote Sensing Data. Infrastructure Planning: A Guide for Practitioners. GREEN SURGE. Retrieved from Remote Sensing, 13(10), 2014. http://greensurge.eu/ working-packages/wp5/. Bagstad, K. J., Semmens, D. J., & Winthrop, R. (2013). Comparing approaches to Hill, R., Díaz, S., Pascual, U., Stenseke, M., Molnár, Z., & Van Velden, J. (2021). Nature’s spatially explicit ecosystem service modeling: A case study from the San Pedro River, contributions to people: Weaving plural perspectives. One Earth, 4(7), 910–915. Arizona. Ecosystem Services, 5, 40–50. Houlden, V., Jani, A., & Hong, A. (2021). Is biodiversity of greenspace important for Baró, F., Calderón-Argelich, A., Langemeyer, J., & Connolly, J. J. (2019). Under one human health and wellbeing? A bibliometric analysis and systematic literature canopy? Assessing the distributional environmental justice implications of street tree review. Urban Forestry & Urban Greening, 66, Article 127385. benefits in Barcelona. Environmental Science & Policy, 102, 54–64. IUCN. 2021. The IUCN Red List of Threatened Species. Version 2021-3. https://www. Bhagabati, N. K., Ricketts, T., Sulistyawan, T. B. S., Conte, M., Ennaanay, D., Hadian, O., iucnredlist.org. Accessed on [22/08/2021]. et al. (2014). Ecosystem services reinforce Sumatran tiger conservation in land use Jalkanen, J., Fabritius, H., Vierikko, K., Moilanen, A., & Toivonen, T. (2020). Analyzing plans. Biological Conservation, 169, 147–156. fair access to urban green areas using multimodal accessibility measures and spatial Borsekova, K., Koróny, S., Vaňová, A., & Vitálišová, K. (2018). Functionality between the prioritization. Applied Geography. size and indicators of smart cities: A research challenge with policy implications. Kabisch, N., & Haase, D. (2013). Green spaces of European cities revisited for Cities, 78, 17–26. 1990–2006. Landscape and Urban Planning, 110, 113–122. Botzat, A., Fischer, L. K., & Kowarik, I. (2016). Unexploited opportunities in Knapp, S., Aronson, M. F., Carpenter, E., Herrera-Montes, A., Jung, K., Kotze, D. J., et al. understanding liveable and biodiverse cities. A review on urban biodiversity (2021). A research agenda for urban biodiversity in the global extinction crisis. perception and valuation. Global Environmental Change, 39, 220–233. BioScience, 71(3), 268–279. Boulton, C., Dedekorkut-Howes, A., & Byrne, J. (2018). Factors shaping urban Kowarik, I. (2011). Novel urban ecosystems, biodiversity, and conservation. greenspace provision: A systematic review of the literature. Landscape and Urban Environmental Pollution, 159(8–9), 1974–1983. Planning, 178, 82–101. Kowarik, I. (2018). Urban wilderness: Supply, demand, and access. Urban Forestry & Broughton, R. K., Parry, W., & Maziarz, M. (2020). Wilding of a post-industrial site Urban Greening, 29, 336–347. provides a habitat refuge for an endangered woodland songbird, the British Willow Kuhnert, P. M., Martin, T. G., & Griffiths, S. P. (2010). A guide to eliciting and using Tit Poecile montanus kleinschmidti. Bird Study, 67(3), 269–278. expert knowledge in Bayesian ecological models. Ecology letters, 13(7), 900–914. Lepczyk, C. A., Aronson, M. F., Evans, K. L., Goddard, M. A., Lerman, S. B., & MacIvor, J. S. (2017). Biodiversity in the city: Fundamental questions for 11 C. Aznarez et al. Landscape and Urban Planning 228 (2022) 104570 understanding the ecology of urban green spaces for biodiversity conservation. Sandström, U. G., Angelstam, P., & Mikusiński, G. (2006). Ecological diversity of birds in BioScience, 67(9), 799–807. relation to the structure of urban green space. Landscape and Urban Planning, 77 Macgregor, C. J., Bunting, M. J., Deutz, P., Bourn, N. A., Roy, D. B., & Mayes, W. M. (1–2), 39–53. (2022). Brownfield sites promote biodiversity at a landscape scale. Science of the SCBD (Secretariat of the Convention on Biological Diversity). (2012). Cities and Total Environment, 804, Article 150162. Biodiversity Outlook (CBO). Available online: https://www.cbd.int/doc/health/cbo- Mansur, A. V., McDonald, R. I., Güneralp, B., Kim, H., de Oliveira, J. A. P., action-policy-en.pdf (accessed 03.09.2021). Callaghan, C. T., et al. (2022). Nature futures for the urban century: Integrating Schwarz, N., Moretti, M., Bugalho, M. N., Davies, Z. G., Haase, D., Hack, J., et al. (2017). multiple values into urban management. Environmental Science & Policy, 131, 46–56. Understanding biodiversity-ecosystem service relationships in urban areas: A Marselle, M. R., Martens, D., Dallimer, M., & Irvine, K. N. (2019). Review of the mental comprehensive literature review. Ecosystem Services, 27, 161–171. health and well-being benefits of biodiversity. In Biodiversity and health in the face of Scottish Natural Heritage (2014). Mapping of Scotland’s wildness and wild land: Non- climate change (pp. 175–211). Cham: Springer. tech- nical description of the methodology. Retrieved from: https://www.nature.sco Marselle, M. R., Hartig, T., Cox, D. T., de Bell, S., Knapp, S., Lindley, S., et al. (2021). t/doc/guidance-mapping-scotlands-wildness-and-wild-land-non-technical-desc Pathways linking biodiversity to human health: A conceptual framework. ription-methodology. Environment International, 150, Article 106420. Senetra, A., Krzywnicka, I., & Mielke, M. (2018). An analysis of the spatial distribution, Moreira, M., Fonseca, C., Vergílio, M., Calado, H., & Gil, A. (2018). Spatial assessment of influence and quality of urban green space–a case study of the Polish city of Tczew. habitat conservation status in a Macaronesian island based on the InVEST model: A Bulletin of Geography. Socio-economic Series, 42(42), 129–149. case study of Pico Island (Azores, Portugal). Land use policy, 78, 637–649. Shanahan, D. F., Fuller, R. A., Bush, R., Lin, B. B., & Gaston, K. J. (2015). The health Monclús, J. (2018) From Park Systems and Green Belts to Green Infrastructures. In: Díez benefits of urban nature: How much do we need? BioScience, 65(5), 476–485. Medina C., & Monclús, J. (eds) Urban Visions. Springer, Cham. https://doi.org/ Sharp, R., Tallis, H. T., Ricketts, T., Guerry, A.D., Wood, S.A., Chaplin-Kramer, R., 10.1007/978-3-319-59047-9_26. Nelson, E., et al., (2020). InVEST 3.8.9 User’s Guide. Müller, A., Bøcher, P. K., & Svenning, J. C. (2015). Where are the wilder parts of Shih, W. Y. (2018). Bird diversity of greenspaces in the densely developed city centre of anthropogenic landscapes? A mapping case study for Denmark. Landscape and Urban Taipei. Urban ecosystems, 21(2), 379–393. Planning, 144, 90–102. Soanes, K., Sievers, M., Chee, Y. E., Williams, N. S., Bhardwaj, M., Marshall, A. J., et al. Müller, A., Bøcher, P. K., Fischer, C., & Svenning, J. C. (2018). ‘Wild’ in the city context: (2019). Correcting common misconceptions to inspire conservation action in urban Do relative wild areas offer opportunities for urban biodiversity? Landscape and environments. Conservation Biology, 33(2), 300–306. Urban Planning, 170, 256–265. Soga, M., & Gaston, K. J. (2016). Extinction of experience: The loss of human–nature Newman, P., & Cabanek, A. (2020). Bioregional Planning and Biophilic Urbanism. In interactions. Frontiers in Ecology and the Environment, 14(2), 94–101. Bioregional Planning and Design (Volume I, pp. 113–128). Cham: Springer. Stott, I., Soga, M., Inger, R., & Gaston, K. J. (2015). Land sparing is crucial for urban Noss, R. F. (2020). The spectrum of wildness and rewilding: Justice for all. ecosystem services. Frontiers in Ecology and the Environment, 13(7), 387–393. In Conservation (pp. 167-182). Springer, Cham. Stroud, S., Peacock, J., & Hassall, C. (2022). Vegetation-based ecosystem service delivery Orive, L. A., & Lema, R. D. (2012). Vitoria-Gasteiz, Spain: From urban greenbelt to in urban landscapes: A systematic review. Basic and Applied Ecology. regional green infrastructure. In Green Cities of Europe (pp. 155–180). Island Press/ Sultana, M., & Storch, I. (2021). Suitability of open digital species records for assessing Center for Resource Economics. biodiversity patterns in cities: A case study using avian records. Journal of Urban Perino, A., Pereira, H. M., Navarro, L. M., Fernández, N., Bullock, J. M., Ceaușu, S., et al. Ecology, 7(1), juab014. (2019). Rewilding complex ecosystems. Science, 364, eaav5570. Terrado, M., Sabater, S., Chaplin-Kramer, B., Mandle, L., Ziv, G., & Acuña, V. (2016). Petersen, T. K., Speed, J. D., Grøtan, V., & Austrheim, G. (2021). Species data for Model development for the assessment of terrestrial and aquatic habitat quality in understanding biodiversity dynamics: The what, where and when of species conservation planning. Science of the Total Environment, 540, 63–70. occurrence data collection. Ecological Solutions and Evidence, 2(1), e12048. Trethewey, S. P., & Reynolds, E. K. (2021). Exposure to green spaces and all-cause QGIS: A Free and Open Source Geographic Information System. https://qgis.org/en/site/ mortality: Limitations in measurement and definitions of exposure. The Lancet index.html. Published (2021). Accessed. Planetary Health, 5(8), e501. Quenouille, M. (1956). Notes on Bias and Estimation. Biometrika, 43, 353–360. Triguero-Mas, M., Donaire-Gonzalez, D., Seto, E., Valentín, A., Martínez, D., Smith, G., R Core Team. (2021). R: A language and environment for statistical computing. Vienna, et al. (2017). Natural outdoor environments and mental health: Stress as a possible Austria: R Foundation for Statistical Computing. https://www.R-project.org/. mechanism. Environmental Research, 159, 629–638. Radford, S. L., Senn, J., & Kienast, F. (2019). Indicator-based assessment of wilderness Turner, M. G., Gardner, R. H., O’neill, R. V., & O’Neill, R. V. (2001). Landscape ecology in quality in mountain landscapes. Ecological Indicators, 97, 438–446. theory and practice (Vol. 401). New York: Springer. Reyes-Riveros, R., Altamirano, A., De La Barrera, F., Rozas, D., Vieli, L., & Meli, P. Twohig-Bennett, C., & Jones, A. (2018). The health benefits of the great outdoors: A (2021). Linking public urban green spaces and human well-being: A systematic systematic review and meta-analysis of greenspace exposure and health outcomes. review. Urban Forestry & Urban Greening, 127105. Environmental Research, 166, 628–637. Ribeiro, A. I., Triguero-Mas, M., Santos, C. J., Gómez-Nieto, A., Cole, H., Anguelovski, I., Warren, M. S., Maes, D., van Swaay, C. A., Goffart, P., Van Dyck, H., Bourn, N. A., et al. et al. (2021). Exposure to nature and mental health outcomes during COVID-19 (2021). The decline of butterflies in Europe: Problems, significance, and possible lockdown. A comparison between Portugal and Spain (p. 106664). Environment solutions. Proceedings of the National Academy of Sciences, 118(2). international. Whittaker, R. H. (1972). Evolution and measurement of species diversity. Taxon, 21 Russo, A., & Holzer, K. A. (2021). Biodiverse Cities: Exploring multifunctional green (2–3), 213–251. infrastructure for ecosystem services and human well-being. In Urban Services to Wu, Y., Tao, Y., Yang, G., Ou, W., Pueppke, S., Sun, X., et al. (2019). Impact of land use Ecosystems (pp. 491–507). Cham: Springer. change on multiple ecosystem services in the rapidly urbanizing Kunshan City of Sallustio, L., De Toni, A., Strollo, A., Di Febbraro, M., Gissi, E., Casella, L., et al. (2017). China: Past trajectories and future projections. Land Use Policy, 85, 419–427. Assessing habitat quality in relation to the spatial distribution of protected areas in Zoderer, B. M., Carver, S., Tappeiner, U., & Tasser, E. (2020). Ordering’wilderness’: Italy. Journal of Environmental Management, 201, 129–137. Variations in public representations of wilderness and their spatial distributions. Landscape and Urban Planning, 202, Article 103875. 12