Seed traits, landscape and environmental parameters as predictors of species occurrence in fragmented urban railway habitats

https://doi.org/10.1016/j.baae.2010.11.006Get rights and content

Abstract

Urbanization particularly promotes habitat fragmentation, which in turn strongly affects biodiversity patterns. A major driver of species loss in isolated habitat patches is dispersal limitation. The relative importance of dispersal ability of species, spatial habitat configuration and local environmental conditions for predicting species composition is still unclear though. Addressing urban abandoned railway areas as study areas, we analyzed the relative importance of environmental versus landscape predictors (e.g. proportions of sealed, built-up and ruderal areas) using variation partitioning methods. To add the perspective of individual species dispersal ability, we characterized the effect of species traits on explained variation in species occurrence with a regression tree. The difference in explained variation in the occurrence of individual species (ΔCfit) between a CCA with environmental predictors and environmental and landscape predictors together was analyzed.

The results revealed that environmental predictors explained a slightly larger amount of variation than landscape predictors. Adding landscape predictors to the analysis with environmental predictors resulted in a sizeable increase in explained variation. The most important predictors in the CCA were photosynthetically active radiation, C/N-ratio in the soil and the proportion of ruderal habitats in the surroundings of the plots. The regression tree model showed higher ΔCfit values for species with a long-term persistent soil seed bank. The lowest ΔCfit values were found for species with a transient seed bank and long seeds.

Linking dispersal-related traits to the predictability of species occurrence is a promising approach to reveal the interdependencies between environmental conditions, landscape configuration and species-specific dispersal abilities. Our results suggest that in fragmented urban habitats, a persistent seed bank is advantageous because it allows for stable populations once habitat patches have been colonized.

Zusammenfassung

Das Wachstum von Städten fördert die Fragmentierung von Lebensräumen. Ungeklärt ist bislang, zu welchen Anteilen einerseits Umwelteigenschaften (Boden und Mikroklima) und andererseits Landschaftseigenschaften (z. B. Anteile versiegelter, bebauter und ruderaler Flächen) die Artenzusammensetzung fragmentierter urbaner Lebensräume steuern und wie ausbreitungsrelevante Arteigenschaften das Vorkommen von Arten in diesen Lebensräumen beeinflussen. Am Beispiel räumlich fragmentierter Bahnbrachen analysierten wir die relative Bedeutung von Umwelt- und Landschaftseigenschaften mit Varianzpartitionierungen. Um die Perspektive der Ausbreitungsfähigkeit einzelner Arten zu ergänzen, charakterisierten wir den Effekt ausbreitungsrelevanter Arteigenschaften auf die erklärte Varianz in den Artvorkommen mit Hilfe von Regressionsbäumen. Dabei wurde die Differenz der erklärten Varianz in den Vorkommen einzelner Arten zwischen einer CCA mit Umweltprädiktoren und einer CCA mit Umwelt- und Landschaftsprädiktoren (ΔCfit) analysiert.

Die Ergebnisse zeigten, dass Umweltprädiktoren einen geringfügig größeren Teil der Varianz in den Artvorkommen erklären als Landschaftsprädiktoren. Das Hinzunehmen von Landschaftsprädiktoren in die Analyse mit Umweltprädiktoren ergab eine deutliche Zunahme der erklärten Varianz. Die wichtigsten Prädiktoren der CCA waren die photosynthetisch aktive Strahlung, das C/N-Verhältnis und der Anteil ruderaler Habitate. Der Regressionsbaum zeigte hohe ΔCfit Werte für Arten mit persistenter Samenbank. Die geringsten ΔCfit Werte wurden für Arten mit kurzlebiger Samenbank und langen Samen gefunden.

Die Verbindung von Ausbreitungsmerkmalen mit der Vorhersagbarkeit von Artvorkommen in multivariaten Verfahren ist ein viel versprechender Ansatz, um die Wechselbeziehungen von artspezifischer Ausbreitungsfähigkeit, Umwelt- und Landschaftsprädiktoren zu zeigen. Unsere Ergebnisse deuten darauf hin, dass in fragmentierten urbanen Habitaten die Ausbildung einer persistenten Samenbank begünstigend ist, weil sie in einmal besiedelten Habitaten stabile Populationen ermöglicht.

Introduction

Habitat fragmentation and the modification of landscapes strongly affect biodiversity patterns (Fischer & Lindenmayer, 2007). Habitat fragmentation has been studied for a long time in landscape ecology but has seldom been applied to urban settings (Bierwagen, 2007). Urbanization however promotes habitat fragmentation (Robinson, Newell, & Marzluff, 2005), directly by modifying the landscape and indirectly by changing the biophysical structure of the habitats (Alberti, 2005). Special characteristics of urban environments include a higher level of fragmentation, smaller patch sizes compared to rural or natural landscapes, and an increased matrix heterogeneity (Irwin & Bockstael, 2007). Muratet, Machon, Jiguet, Moret, and Porcher (2007) argue that the urban matrix may isolate wastelands situated in built urban areas from each other, preventing the exchange of species and resulting in a reduction of the high numbers of native and introduced plant species cities may harbor (Kühn et al., 2004, Angold et al., 2006).

Besides the degree of fragmentation of habitat patches, the colonization of patches is strongly governed by species-specific dispersal abilities. The relative importance of dispersal-related predictors and of local environmental predictors for local plant species composition is still not completely understood despite modeling approaches integrating both predictor groups (Ozinga et al., 2005). To our knowledge, no study thus far addresses the predictability of individual species in multivariate models with both environmental and landscape predictors while taking dispersal-related plant traits into account.

When local species composition is predicted based on environmental conditions, species with limited dispersal ability have a lower predictability in multivariate analysis (Ozinga et al., 2005). The missing link between the predictability of species occurrences and the spatial configuration of the matrix is therefore the dispersal ability of individual species. The latter is highly dependent on seed traits and certain life-history traits like seed production. The present study therefore aims at linking plant traits and local environmental and landscape predictors.

We thus compared the relative importance of landscape predictors such as the proportions of different habitats and sealed areas around the plots to that of environmental predictors (microclimate and soil). Then we analyzed, to what extent landscape predictors improve the predictability of species occurrence and how this change in predictability correlates to dispersal-related plant traits.

We anticipate that the increase in explained variation due to the inclusion of landscape predictors would mainly occur in species with low dispersal ability because these species are expected to be most affected by habitat isolation and by having a patchy distribution around their seed sources. In contrast, highly dispersible species are presumably distributed more evenly and are thus more able to reach suitable habitat patches. Using abandoned railway areas as study areas, we thus address the following hypotheses: (1) both local environmental and landscape predictors are important for explaining species composition in abandoned railway areas. (2) The importance of landscape predictors for explaining species occurrence in urban abandoned railway areas depends on dispersal-related plant traits. (3) The increase in predictability due to the inclusion of landscape predictors mainly occurs in species with low dispersal ability.

Section snippets

Study area and vegetation sampling

The six study areas are located in the city of Berlin, Germany. The climate is transitional between continental and atlantic, and Berlin experiences a pronounced urban heat island effect (Berlin Department for Urban Development, 2001). We studied one special type of wastelands, abandoned railway areas, because they have a high floristic diversity, are fragmented and frequently occurring in cities. Furthermore, they share a similar history of utilization and a constrained set of substrates,

Importance of environmental and landscape predictors

Variation partitioning of the x- and y-coordinates of the plots versus landscape and environmental predictors (total variation explained 35.2%) revealed that only 3.5% (p < 0.001) of the variation in the plant species composition was strictly due to spatial autocorrelation. Then, six dummy variables, which coded the membership of the plots to the six study areas, versus landscape and environmental predictors were analyzed with variation partitioning. Only 7.2% (p < 0.001) of the variation explained

Relative importance of environmental predictors

A composition of microclimatic, soil, and landscape variables explained the species set in the studied system. PAR and temperature maxima exerted a strong and significant influence on species composition. This is likely due to the differences in vegetation density and height during succession. The incoming radiation at the soil surface determines the surface temperature and thereby the air temperature near the surface (Stoutjesdijk & Barkman, 1992). Increasing vegetation density during

Conclusions

Linking species traits to the predictability of species occurrence is a promising approach to reveal the interdependencies between environmental conditions, landscape configuration and species-specific dispersal abilities. Different from analyses on the dominance of traits in certain habitats, our results allow for conclusions about how dispersal related functional groups of plant species respond to the spatial configuration of the urban matrix. We conclude that the inclusion of landscape

Acknowledgements

This study was supported by the postgraduate research and training group ‘Perspectives on Urban Ecology II – Shrinking Cities’ (GRK 780) of the German Research Foundation. We thank B. Seitz for help with species determination, A. Okujeni and U. Bangert for help with GIS analysis, A. Chamgoulova and S. Trinks for soil analysis and six anonymous referees for their comments, which greatly helped to improve this manuscript.

References (38)

  • B.G. Bierwagen

    Connectivity in urbanizing landscapes: The importance of habitat configuration, urban area size, and dispersal

    Urban Ecosystems

    (2007)
  • D. Borcard et al.

    Partialling out the spatial component of ecological variation

    Ecology

    (1992)
  • L. Breiman et al.

    Classification and regression trees

    (1984)
  • S. Brunzel et al.

    Neo- and archaeophytes respond more strongly than natives to socio-economic mobility and disturbance patterns along an urban-rural gradient

    Journal of Biogeography

    (2009)
  • G. De’ath et al.

    Classification and regression trees: A powerful yet simple technique for ecological data analysis

    Ecology

    (2000)
  • R. Erfanzadeh et al.

    Factors affecting the success of early salt-marsh colonizers: Seed availability rather than site suitability and dispersal traits

    Plant Ecology

    (2010)
  • J. Fischer et al.

    Landscape modification and habitat fragmentation: A synthesis

    Global Ecology and Biogeography

    (2007)
  • S. Godefroid et al.

    Urban plant species patterns are highly driven by density and function of built-up areas

    Landscape Ecology

    (2007)
  • C. Grashof Bokdam

    Forest species in an agricultural landscape in the Netherlands: Effects of habitat fragmentation

    Journal of Vegetation Science

    (1997)
  • Cited by (0)

    View full text