Urban areas are often found to have high levels of biodiversity, but little is known on how fine-scale land use is structuring species diversity in cities. We used species occurrence records from GBIF and official land cover classifications to determine how habitat affects total species richness, and the number of threatened and alien species. We did so by constructing spatially correlated Generalised Linear Mixed Effects Models based on habitat, habitat heterogeneity, aspect and taxonomic group within 500 m × 500 m grid cells across the Trondheim municipality, selecting the best models based on ∆AIC. The best models varied for overall-, threatened and alien species richness, with total species richness depending on all predictors and their interaction with taxon, whereas threatened species richness depended on habitat, aspect and taxon, and alien species richness only depended on taxon. The relationship between species richness in general are highly complex and dependent on multiple factors and interactions (Table 2, Figs. 3–6). Threatened, native species are associated with non-anthropogenic habitats (Table 3, Figs. 4 and 6), whereas alien species are mainly affected by spatial correlations on the investigated spatial scale (Table 4). The key findings of this study advance our understanding of the field by confirming the association of threatened, native species with more natural habitats, and the potential for establishment of alien species across all habitats on a management-relevant spatial scale.
The retention of all predictors and interactions in the model of overall species richness illustrate the complex relationships between environmental variables and different taxonomic groups. Nevertheless, the overall negative effect of northness reflect the low species richness of north-facing slopes, compared to south-facing ones42 (Fig. 4). The different taxa responded differently to increasing habitat heterogeneity, the only unidirectional response being for plants (positive) (Fig. 5). This supports the results of Matthies et al. (2017)30 and Beninde et al.(2015)56, in which respectively habitat heterogeneity and habitat richness were positively associated with species richness in urban areas. However, other studies have found a positive correlation for restricted taxonomic groups, such as arthropods29,40, birds and mammals30, which was not observed here.
Non-surprisingly, the different taxa responded differently to various habitats (Fig. 6). Interestingly, whereas plants, fungi and non-avian animals generally responded negatively to urban areas (differences not necessarily significantly different from other habitats however), the effect was less pronounced for birds. This could reflect their high mobility, and potential for an “urban adapter/exploiter”-status of some bird species13. In contrast, the habitat with the highest predicted number of both plant-, non-avian animal-, and fungi species, had the lowest predicted number of bird species.
Threatened species richness generally responded positively to increasing northness, in contrast to what would be expected (Fig. 4). This could potentially be an artefact of the habitat associations; Coastal areas had higher northness values (mean = 0.758, S.E. = 0.022 compared to the overall mean = 5.35, S.E. = 0.005). The effect of taxon reflects the differences in the number of species within each taxonomic group being classified as threatened; (50 bird species, 26 plant species, 12 non-avian animal species, 33 fungi species included in the study). For all taxa, the lowest species numbers are predicted for all variants of coniferous forest, contrary to the initial expectations, and urban areas (Fig. 6). The negative effect of urban areas on threatened species richness mirrors the findings of Aronson et al. (2014)8, and emphasises how vulnerable native species are not pre-adapted to the changed environments of the city. Contrary to expectations, the effect of the various forest habitats on threatened species is lower than for most other habitats (Table 3). The low number of threatened species in forests can be due to the lack of sampling, showing a spatial bias in the data rather than an effect. This should however be accounted for by using the number of records as an offset in the models. Rather, large parts of the forested areas in Trondheim are srongly affected by previous afforestation for timber production, where mainly coniferous species (both native and alien) were planted33. These forests might not provide the needed conditions for native species57. Plantations and secondary vegetation have been shown to harbour fewer species than primary forests58,59. The lack of association between forested areas and threatened species calls for a nuanced perspective on what forest types constitute suitable habitats for species of interest, as indicated by Ingram et al. (2015)58 and Horák et al. (2019)59. The highest species numbers are predicted for Open marsh and coniferous forest and Coastal areas (Fig. 6); the former is likely the habitat category reflecting the lowest human impact. The high number of threatened species in coastal habitats can likely be ascribed to these habitats being ecotones, providing varied habitat conditions. Ecotones have been suggested to have an increased species richness36. Lloyd et al. (2000)38 found ecotonal species to mainly be natives, which is supported by the findings here.
Interestingly, in the model of alien species richness, only taxon was retained as a significant predictor, reflecting the differences in the number of species within each taxonomic group being classified as alien (5 bird species, 156 plant species, 10 non-avian animal species, 6 fungi species included in the study). The lack of response to either of the other investigated variables stands in stark contrast to the expectations and previous findings, but can be attributed to alien species often being generalist opportunists; the spatial scale investigated does not reflect the fine-scale conditions affecting the individual species. This result highlights that on this spatial scale, all parts of the municipality are open for potential invasion by alien species. Given the spatial correlations (Supplementary material 4, Fig. S.2), it is evident that founder events and subsequent spread of alien species are of crucial importance: on the investigated scale, even more important than the configuration of environment. As many alien species are introduced through urban areas mainly due trade and traffic12,15,31, emphasis must be put on the importance of avoiding unintentional introduction of potential invasive species. As an example, the review by Kowarik (2011)5 found cities to be hotspots of alien plant species. In addition, port cities have been suggested as even greater hotspots of introductions, leaving Trondheim even more vulnerable60,61.
As the explanatory variables used in these models are “indirect” (sensu Guisan and Zimmermann (2000)62), the habitats are proxies for underlying environmental (direct) drivers. Therefore, a direct extrapolation to other geographical areas should be cautious62. However, the general methods are highly applicable elsewhere.
Of the 1,509 grid terrestrial cells, 485 qualified for analyses; species occurrence data was sparse in the rest. Those used in the analyses were biased towards urban areas (Table 1), supporting the general trend in citizen science data; concentrated around inhabited or areas otherwise accessible to the public63,64. For example, areas within Trondheim municipality relatively far from human activity are under-sampled, with two habitats not being represented in the analyses at all (Open firm ground and forest –and cultivated land). This bias was accounted for in the models, but the differentiated sampling effort nevertheless leaves varying degrees of uncertainty for each habitat and taxon. The sample sizes differed among species groups, with many more observations of threatened than alien species. The differences thus might reflect sampling strategy rather than reality.
As the models are by nature rather crude, they inevitably lack predictor variables, which could have increased model accuracy. However, including highly detailed variables was not the aim of this study. Since the data set included a wide array of species, these will not respond in similar ways to variation in the included variables, or to missing variables65. The more species included in the models, the more opposing mechanisms are attempted to be fitted within a single modelling framework, giving a poorer fit, compared to models with a narrower scope.
The number of GBIF records have increased in recent years (see Speed et al. (2018)64). Of all species recorded in Trondheim, approximately 1/3 have been recorded within the municipality from year 2013 to 2018. Of the 6,020 species from the downloaded data set not included in the analyses, 33.9% (2,039) have only been recorded once, and 85.5% (5,150) have been recorded <10 times. Most of these infrequent species are insects. This taxonomic skew is likely due to this species group being poorly sampled or requiring expert knowledge to identify to species level.
Different correlations with environmental variables are expected at different spatial scales for different organisms19,23,66. Taxa and species with opposing responses to the included variables could mask each other, thus not revealing the underlying mechanisms56. Simultaneously, the mechanisms underlying species distributions vary with spatial scale, not necessarily in the same direction for different taxa19,23,67. As multiple different taxonomic groups were included in this study, the used spatial scale is potentially inappropriate for all taxa.
According to Pautasso (2007)19, a negative correlation between urbanisation and species richness is expected when the study grain is smaller than 1 km, as in this study, but positive at larger scales. This is ascribed to the larger scale reflecting human settlement in productive areas, competing for space with other species, whereas the small scales reflect more detailed environmental- and land cover effects.
Our results indicate that if the Trondheim municipality is to be managed to favour biodiversity, favouring threatened species and excluding alien species, the following actions can be recommended:
Habitat heterogeneity on a relatively small spatial scale should be ensured, favouring overall species richness. This should however not be confused with fragmentation of natural habitat.
To favour threatened species, non-anthropogenic- and coastal areas should be monitored and protected, potentially expanding the actions to ecotones in general.
To limit the spread of alien species, initial introduction and establishment should be avoided. Thus, urban- and other anthropogenic areas should be closely monitored and managed12,68.
Protection of important and heterogeneous habitat should be accounted for in unison with ensuring large habitat patches, rather than multiple smaller ones; a metastudy by Beninde et al. (2015)56 showed patch area to have the largest positive effect on urban biodiversity.
Source: Ecology - nature.com
