1 Background

Ever since mankind developed permanent settlements, biodiversity patterns of the occupied areas and their surroundings have been modified by human activity. Soil sealing, drainage, and waste production are modern examples of means by which human activities may change environmental conditions (e.g. water supply, nutrient supply, temperature) and, therefore, assemblages of species and structural elements (Ricklefs and Miller 2000). These changes are accompanied by an emergence of spatial patterns of different patches within urban areas. These patterns reflect the type and intensity of human activities in the urban matrix, like housing, business, or recreation. However, the mosaic of different patches created and maintained by humans form a mosaic of habitat types ranging from fairly natural to highly modified ones (Sukopp et al. 1995). Within cities, open spaces (e.g. parks, allotments, urban forests) in particular provide habitat to numerous species of different taxa (Knapp et al. 2008; Kühn et al. 2004; Meffert and Dziock 2012; Melles et al. 2003). They may even fulfill basic nature conservation functions as endangered or rare species colonize urban open spaces (Niemelä 1999). However, little is known about patterns of species richness and key factors that influence total and native species numbers as well as the relation between native and non-native species.

Many factors influence species richness in urban open spaces, like the heterogeneity of vegetation structures (Blair 1996), human disturbance (Guntenspergen and Levenson 1997), age of habitats (Drayton and Primack 1996), and surrounding land uses (Guntenspergen and Levenson 1997). Nevertheless, patch properties like the size of urban open spaces may be of prime importance for development and maintenance of species richness (Mörtberg 2001; Knapp et al. 2008; Meffert and Dziock 2012; Bolger et al. 1997).

In recent years an increasing number of studies attempted to explore the relationship between patterns of the urban matrix and biodiversity. Many of these studies describe the relationship as a linear gradient of decreasing urbanization from city center to fringe areas which is associated with a general trend of increasing species richness (e. g. Guntenspergen and Levenson 1997; Mörtberg 2001; Niemelä 1999). McKinney (2008), who reviewed 105 studies focusing on flora and fauna diversity along urban-rural gradients, points out that diversity was lowest in the most intensively urbanized areas, commonly located at the central districts of the cities. Approximately 70 % of these studies indicated that vertebrate and invertebrate diversity peaked in the least urbanized areas, whereas a similar proportion of the studies found vegetation diversity to be highest in moderate levels of urbanization. As increasing urbanization is correlated with increasing fragmentation (Wickham et al. 2000), the low species diversity in the central districts may be explained by a smaller size and increased isolation of open spaces as compared to areas with a lower degree of urbanization. Populations of such patches may not be viable on the long term as colonization probabilities may be reduced and/or the area may be too small to support a minimal viable population (Snep et al. 2006). The studies reviewed by McKinney (2008), however, assume that the degree of urbanization is positively correlated to the distance to the urban center.

Ramalho and Hobbs (2012) point to the inadequacy of using a linear urban-rural gradient as many studies viewed habitat availability and quality along ‘city center – city edge – peri-urban’ gradients, or ’urban – suburban – natural’ area gradients, whereas modern cities exhibit more complex spatial patterns. Modern mega-cities and urban areas exhibit fractal-like patterns, characterized by multiple hierarchical clusters (Batty 2008) and expand non-linearly by presenting leapfrogging type of dynamics (Benguigui and Czamanski 2004). For a better understanding of the function of urban ecosystems Ramalho and Hobbs (2012) call on expanding beyond the traditional view of city structure and dynamics and propose to consider spatial patterns of the urban matrix and their temporal dynamics. Thus, the distance of a patch from a more urbanized city center may be less relevant than the distance of an open space from the urban edge. However, neighborhood relationships among open spaces, their size, connectivity and evolution through time may all be factors important for determining the properties of the urban-ecological framework.

In the light of the above mentioned discussion we studied species richness patterns in relation to the complex environment of two urban areas that differ in climate, topography, age and structure of the urban landscape. We tested if patch size and distance from the urban edge affects vascular plant species richness in urban open spaces. In particular we focused on the following hypotheses:

  • H1: The total number of vascular plant species in urban open spaces increases with increasing patch size.

  • H2: The number of native vascular plant species in urban open spaces increases with increasing patch size.

  • H3: The proportion of native vascular plant species in urban open spaces increases with increasing patch size.

  • H4: The total number of vascular plant species in urban open spaces decreases with increasing distance from the urban edge.

  • H5: The number of native vascular plant species in urban open spaces decreases with increasing distance from the urban edge.

  • H6: The proportion of native vascular plant species in urban open spaces decreases with increasing distance from the urban edge.

Hypotheses H1, H2, H4 and H5 explicitly assess whether the species respond to either distance from the urban edge or to patch size. Hypotheses H3 and H6, however, implicitly consider the response of the native species to the urban landscape. In spite of the different study areas we explore whether ecological principles can outweigh the impact of geographical position.

2 Methods

2.1 Study Areas

To test our hypotheses the cities of Hannover, Germany and Haifa, Israel were selected. Hannover is located in the north of Germany (city center: 52°23′ N, 9°42′ E) in humid Central Europe. Hannover represents an old city (founded in medieval times) along the banks of a low land river (Leine) in with a human population of more than half million (average of 2,576 inhabitants per km2). The township of Hannover covers an administrative area of 204.1 km2 with 37 % built-up, 16 % traffic, 15 % agriculture, 14 % open space, 11 % forest, 4 % water bodies and 4 % other land uses (Table 1) reaching altitudes from 44 m to 118 m a.s.l.. The city is located in the warm temperate climate zone with a mean annual temperature of 9.6 °C and mean precipitation of 661 mm (DWD 2013a, b).

Table 1 Comparison of the study areas Hannover, Germany and Haifa, Israel

Haifa is located in North-western Israel (32°48′ N, 35° E) on the Carmel Ridge, rising up to an elevation of 480 m above the Mediterranean Sea (Table 1). The jurisdictional area of the city is 63.6 km2, and mean population density is 4,246 inhabitants per km2. The climate is Mediterranean, and mean annual precipitation varies with elevation. At the lower elevations, at sea level, mean annual precipitation is close to 540 mm, and at the higher elevations it is approximately 685 mm (IMS 2011). The majority of the soils are formed on carbonate sedimentary lithology, dominated by limestone and chalk-marls.

2.2 Selection of Study Sites Within the Study Areas

A prerequisite to identification of the open spaces within the city was delineation of the urban area. We defined the urban edge as the border between the urbanized area with a high amount of sealed surface, and the open landscape with a much lower percentage of sealed soil. In Hannover we located the urban edge using detailed land cover classes of an digital landscape model (DLM) (LGN 2007). In Haifa the urban edge was delineated from aerial images obtained during 2008 and 2012. This city borders to the west and southwest large tracts of undeveloped maquis and planted pine forests, which mostly reside in the Carmel National Park. To the east the study area was truncated by an Industrial zone which beyond is followed by large intensively cultivated fallow crop fields. Within the urban edge all urban open spaces were identified by a low amount of sealed surface (generally <; 25 %). Open spaces that had a mutual border were considered a single patch. The areas of each open space and its distance from the centroid of a polygon or, respectively, from the nearest patch border point was measured to the urban edge using GIS.

Study sites were chosen using a stratified random approach (see Underwood 1997) with three size classes (Hannover: >0.5–2 ha, >2–6 and >6–100 ha; Haifa: >0.1–1.6 ha, >1.6–37.5 ha, >37.5 ha) in combination with three distance classes (Hannover: >50–1,000 m, >1,000–2,000 m and >2,000 m; Haifa: ≤300 m; >300–2,000 m, >2,000 m). From each category up to 5 open spaces were randomly chosen as study sites (Figs. 1 and 2). In Hannover 32 study sites were investigated. These contain 4 study sites for each size and distance class combination, but none for medium sized open spaces in the largest distance class. Habitat types found in these patches include forests, parks, cemeteries, allotments, ruderal sites and combinations of these. In Haifa 37 study sites were surveyed in which maquis, garrigue, ruderal site, park and grasslands were found.

Fig. 1
figure 1

Distribution of open spaces appropriate for selection within the studied area in Hannover. Triangles represent the sites chosen for sampling (note the log-scaled y-axis)

Fig. 2
figure 2

Distribution of open spaces within the studied area in Haifa. Triangles represent the sites chosen for sampling (note the log-scaled y-axis)

2.3 Survey of Vascular Plant Species

All vascular plant species growing in each open space were recorded in Hannover from June to August 2011 and March to May 2012 and in Haifa during spring 2011 and spring 2013. Surveys were conducted afoot by a single person (SM for Hannover, DK for Haifa) and each open space was completely covered. Study sites were divided into dominant vegetation structures as they were surveyed (e.g. beech forest, lawn). Every vegetation structure was visited several times during the growing period to record all species. Ornamental species and self-established native and non-native species were recorded whereas submerse plant species were neglected. The nomenclature follows Buttler and Hand (2008) for wild growing species and Zander and Erhardt (2008) for ornamentals in Hannover. In Haifa the nomenclature follows Feinbrun-Dothan and Danin (1991).

2.4 Data Analysis

To test the six hypotheses total species number and the sub-groups into which they were classified – natives and proportion of natives – were designated as dependent variables. The decision whether a species is native or non-native to the region was based on the classification of Garve (2004) for Hannover and Feinbrun-Dothan and Danin (1991) for Haifa. The number of non-native species includes established and not established non-natives as well as ornamental species. The proportion of native species was calculated as the number of natives of all the species found in a patch. Patch size and distance from the urban edge were the independent variables used and their values were calculated with a GIS. To linearize the relationship between the dependent and independent variables data were log-log transformed. Hypotheses H1-H3 were tested by regressing the dependent variables (number of species, number of native species and proportion of native species) against open space size. Hypotheses H4-H6 were tested by regressing the dependent variables against distance of open space from the urban edge.

3 Results

In Hannover 1,372 plant species were found in the open spaces surveyed, of which 577 (58 %) are natives and 795 (42 %) non-natives. In Haifa the total number of 363 species was found in urban open spaces, of which 334 (92 %) are considered natives and 29 (8 %) non-natives.

The patches of open space embedded in the urban matrix can be viewed as islands of suitable habitat, separated by a hostile background. If this is the case indeed, one can speculate that the relationship between patch size and the number of species will follow the commonly observed ‘Species-Area curve’ (Preston 1962). Accordingly, the data was log-log transformed, i.e., both the number of species and patch size were log-transformed, in an attempt to linearize the relationship between the number of species and patch size. Additionally, both within the framework of the Theory of Island Biogeography (MacArthur and Wilson 1967) and modeling of dispersal success (Levin et al. 2003), number of species or individuals is commonly viewed as decaying with distance from source. Consequently, we also log-transformed distance following the same rationale presented above.

The analyses revealed that the total number of species increased with increasing patch size and that this relationship was found to be significant both in Hannover and in Haifa (Table 2, Figs. 3 and 4). Similarly, the numbers of native species increase significantly with increasing patch size in Hannover and Haifa (Table 2, Figs. 5 and 6). In Haifa, however, the explanatory power of patch size, i.e., the percent of the variance associated with it, was lower compared to the Hannover case, as is evident by the lower adjusted R2 values. Additionally, each change in a unit area in Hannover has a greater average effect on the expected number of species, as is evident by the higher b-coefficients obtained. Nevertheless, no significant influences were found between the proportion of natives and patch size (H3).

Table 2 Linear regressions calculated for the dependent factors total species number, number of native species, and proportion of native species in combination with independent factors area and distance from the urban edge for Hannover and Haifa
Fig. 3
figure 3

Hannover – the relationship between the total species number and area (note the log-log transformation)

Fig. 4
figure 4

Haifa – the relationship between the total species number and area (note the log-log transformation)

Fig. 5
figure 5

Hannover – the relationship between the native species number and area (note the log-log transformation)

Fig. 6
figure 6

Haifa – the relationship between the native species number and area (note the log-log transformation)

Additionally the effect of decreasing species numbers with increasing distance to the urban edge was not found to be significant (H4-H6). Results show no relation between the total number, the number of native or the proportion of native species to distance from the urban edge in both study areas.

4 Discussion

As described above, Hannover and Haifa are dissimilar in many geographical properties. In addition, the land use and cultural history left their signature on the studied systems. The proportion of agricultural land within the boundaries of Hannover was close to 15 % whereas in Haifa there are no agricultural areas. When comparing the results of the analyses, however, the broad similarities of the dependent variable responses are striking, as in both urban landscapes the only explanatory variable found to be significant was patch size. This was demonstrated despite fundamental differences between Hannover and Haifa with respect to climate, topography and composition of the flora.

The urban matrix can be viewed within the framework of the Theory of Island Biogeography (MacArthur and Wilson 1967), and therefore it can be expected that the natural and semi-natural areas surrounding the cities provide the “mainland” species pool. Similarly, the open spaces within the city are the “islands” that are subjected to extinction and colonization processes. Therefore, it could be hypothesized that species numbers within open spaces are a function of patch size and distance from the urban edge. Our findings are in only a partial agreement with this theory as was indicated by the fact that distance from the urban edge was not found to be a significant factor. Results which identified the relationship between patch size and number of species were also obtained by others (Knapp et al. 2008; Meffert and Dziock 2012; Mörtberg 2001; Bolger et al. 1997), and these findings correspond with the concept of the species-area relationship (Preston 1962).

Additionally, the hypothesis that increasing patch size will result in increasing proportions of native vascular plant species did not yield any significant relationships. This may be due to the very different types of open spaces selected for investigation. The Hannover study sites included for example forests, parks, and cemeteries. Forests generally contain lower percentages of native species compared to higher values in parks or cemeteries, mostly due to ornamental species. The fact that our third response variable, the proportion of native species in open spaces, did not respond to either size or distance of a patch from the city’s border indicates that the underlying assumption, that the overall pool consists of a pool of native species available from the natural areas surrounding the city, and a pool of non-native species from cultivated areas in the inner parts of the city, may be incorrect. Alternatively, the impact of the cultivated areas may not be strong enough to be detected.

In both cities distance from the urban edge was not a factor explaining species richness. This does not conform to many of the previous studies which investigated species richness patterns along urban to rural gradients. McKinney (2008) reviewed 17 studies related to plant species richness along an urban to rural gradient and 13 of the 17 studies report an association with the gradient, where 9 of the studies indicate peak diversity at the intermediate values of the gradient, one study reported peak values at the rural end of the gradient and in three studies peak values appear at the urban edge of the gradient. Other studies also demonstrated this relationship (Hope et al. 2003; Huste and Boulinier 2007; Bolger et al. 1997). The lack of relationship found in the current study may be explained by the fact that immigration processes may have a minor relevance in cities and that the urban matrix is more permeable to vascular plant dispersal than expected. This may be particularly true for cities which are heterogeneous and which are also characterized by many small patches that may serve as stepping stones for dispersal. For example Wania et al. (2006) studied flora richness in urban and agricultural landscapes of central Germany and concluded that the major driver for species richness is landscape heterogeneity, particularly of small patches associated with different land use types. In spite of this discussion the Theory of Island Biogeography does not work for the influence of distance from the urban edge on total and native vascular plant species numbers and the proportion of native species.

Therefore, we conclude from our preliminary results that city planning should focus on the creation and conservation of large open spaces, and also on providing a mosaic of different habitats within and between open spaces in order allow for a high urban biodiversity. The location of open spaces with respect to their distance from the urban edge seems to be of minor relevance, although the combined effect of patch size and the distance has not been investigated yet. Other factors, not investigated in this study may be of considerable influence on vascular plant species richness as well, for example the mosaic of different habitats within and between open spaces, disturbance characteristics and age of habitat.