Richness of vascular flora in Lublin city, east Poland, under different urban pressure

Lublin is the principle city of southeastern Poland and capital of the province. In its present administrative boundaries, it covers the area of 148 km² situated between 51°08′–51°18′N and 22°27′–22°41′E and has a population of 300,000 permanent residents. The city is located in the central-northern part of the macroregion of the Lublin Upland on the border of four subregions (Fig. 1): Nałęczów Plateau, Bełżyce Plain, Świdnik Plateau, and Giełczewska Elevation (Kondracki 2009). The city of Lublin is characterized by fairly specific climatic properties. The average growing season lasts 209 days (Kaszewski 2008). The basic Quaternary substratum is composed of loess predominating in the western part and loess-like and clay covers predominating in the southern part (Chałubińska and Wilgat 1954). Most often, these are urban and industrial soils with different degrees of pollution and contamination (Kukier 1985; Turski et al. 2008). The oldest records concerning the origins of Lublin date back to the ninth century. The city began as a settlement on the three Old Town’s Hills (Rozwałka 1997). At the beginning of the twentieth century, the growing population numbered 50,000 (Sochacka 1997).

Research data was collected along a transect consisting of 20 1-km² study squares. The established squares correspond to the arrangement of the squares used in the ATPOL grid system (Zając 1978). The transect crosses the area of Lublin from NE to SW and passes through diverse types of habitats that are representative for the city area (Fig. 1). Each study square was assigned a number of traits classified into two categories—flora and urban indicators—described in Tables 1 and 2.

Habitat analyses were performed taking into account selected urban indicators (Table 1). PU was determined based on historical data (Rozwałka 1997). This period in the study squares distinguishes between study areas that have been used for >600 years and those that were included in the city limits in the 1980s (Fig. 1; Table 1). The landscape structure (caused by land-use and land-cover variability), communication network, and HH was assessed on the basis of Landsat digital data available at the European Environmental Agency (EEA). Those urban areas are defined from classes of land cover contributing to the urban tissue and function and laying less than 200 m apart. Land-use and land-cover data contain three main data sets—from general in label 1 to details in label 3. In this study, three land-cover and land-use classes from label 1 were taken into account: agricultural areas (A), artificial surfaces in work-named urban areas (U), and forests and seminatural (F) areas distinguished on the basis of the dominant major form of land use and land cover in the study square (Fig. 1). Differentiation of particular land-use and land-cover labels are given in Table 1 (labels 1 and 3) and their proportions in Table 3 and “Appendix”. We mapped the distribution of land use and land cover and calculated their proportion using a geographic information system (ESRI 2003) for each transect square. The length of roads (Ro) and railways (Ra) were calculated in the same way. The communication network included paved roads and railway tracks (“Appendix”). Landscape or HH is a complex phenomenon involving size, shape, and composition of different landscape areas and the spatial relations between them. Land-use and land-cover units have been used to compare differences in heterogeneity of study patches within city landscapes (Cale and Hobbs 1994). On the basis of number of land-use and land-cover types in a study square, for the purposes of this paper, HH indicator was evaluated. If in a study square only one form of land use and land cover exists (label 1), HH = 1, regardless of the variation (label 3); for example, the proportion of A in square A1 is 100% and HH = 1; in square F16, HH = 3, which is caused by three types of land-use and land-cover units (Table 3).

A slightly modified 4-grade scale developed by Sukopp (1972) was used to assess HL given in Table 1. Levels of species hemeroby were estimated during the field research ascribed in accordance with the increasing intensity of anthropopression. Our data set was taken from the BiolFlor database, according to which hemeroby is a measure of departure from naturalness. Habitats and vegetation types are classified along the hemeroby scale from ahemerob (natural) to polyhemerobic (nonnatural). Sites without plant life are metahemerobic. BiolFlor indicates the amplitudes of hemeroby so that all levels of hemeroby are indicated in which a plant species can occur. Particular species may vary in scope of HL. There are strongly transformed habitats appearing simultaneously in natural and seminatural environs. The HL was calculated by assigning a value to each hemeroby degree (Table 1). The HL for species is the sum of hemeroby degrees divided by their number, e.g., Adoxa moschatelina: 1 = oligohemerobic, 2 = mesohemerobic, which gives 1 + 2 = 3/2 and HL = 1.5. HL for the square is the sum of the average degree of species hemeroby divided by the number of species in a square (Table 3).

Floristic data had been collected during field studies, i.e., by mapping the Lublin vascular flora in 2002–2008 (Rysiak 2009), which was repeated in vegetation seasons 2010–2011 in all habitats along a study transect. Distribution of species was noted in the grid composed of squares 1 × 1 km. Occurrence of a species in a square was regarded as its locality (binary variables, presence or absence of species in square). Analysis of flora was performed at two levels: general flora in the entire transect, and flora of each study square. The description of each taxon comprises affiliation to geographical-historical (geohistorical) elements on the basis of papers by Zając (1979) and Tokarska-Guzik (2005), syntaxonomic (synecological) affiliation of species after Matuszkiewicz (2008) and endangered species, both according to the Regional (Kucharczyk 2003) and National (Mirek et al. 2006) Red Lists, and legally protected species listed in the Regulation of the Minister of Environment (2012). Based on the number of native and AS, synanthropization indicators were calculated for each plot (Kornaś 1977; Jackowiak 1990). This data was defined as flora indicators. Latin names of species are based on nomenclature proposed by Mirek et al. (2002).

To analyze the relationship between flora and urban indicators in Lublin, we calculated Spearman’s rank coefficient. Species responses to experimental treatments were evaluated with multivariate methods in Canoco version 4.5 (ter Braak and Šmilauer 2002), with flora indicators in squares as response variables and values of urban indicators as explanatory variables, i.e., predictors (Tables 1, 2). According to gradient length from a preliminary detrended correspondence analysis (DCA), a linear method, redundancy analysis (RDA) was used for flora and urban indicators. Relationships between flora and urban indicators were analyzed by three series of RDA: SR/urban indicators, geohistorical elements/urban indicators, and synecological groups of species/urban indicators. The first RDA analysis was based on SR as compositional data matrix and urban indicators as explanatory variables. In the second and third RDA analyses, based on SR, number of geohistorical elements and synecological groups of species in each study square were calculated. RDA models are free from problems caused by multicollinearity of variables. Variance of the inflation factor (VIF) of each explanatory variable was measured in the extent of multiple correlation with other predictors. If VIF variables are large (>20), they are almost perfectly correlated with each other. High VIFs indicate multicollinearity among explanatory variables. If an explanatory variable is completely multicollinear, its VIF is set to 0 (−0.000) and its regression coefficient to 0. Normally, VIFs are usually >1.0. (ter Braak and Šmilauer 2002). During the fitting of explanatory variables, collinearity of U, VIF_U = 27.13 and multicollinearity among forests (F) VIF_F = −0.000, variables were detected; in all RDA analyses, these factors were removed.

RDA analysis was combined with Monte Carlo permutation tests (499 permutations). The test has the ability to evaluate the significance of constrained ordination models and relate to the general null hypothesis, stating the independence of the primary (species) data on values of explanatory variables. Consequently, values of environmental variables are randomly assigned to individual samples of species composition, ordination analysis is done with this permuted (shuffled) data set, and the value of the test statistic is calculated. In this way, both response variable distributions and explanatory variable correlation structures remain the same in the real data and in the null-hypothesis simulated data. Stepwise selection of the model testing the usefulness of each potential predictor (environmental) variable for extending the subset of explanatory variables was used in the ordination model. The relationship is characterized by the F value of the analysis of variance (ANOVA) of the regression model. Unconstrained correspondence analysis (CA) was run to quantify the unique contribution of each of the 20 groups of plants in transect squares. In this way, data set homogeneity was tested (ter Braak and Šmilauer 2002; Lepš and Šmilauer 2002).

The study transect contained three major types of habitats (Fig. 1): Agricultural (A1–A5 and A20), urban (U6–U14), and forest (F16–F19) areas differ in terms of particular land-use and land-cover, proportion of communication network, and urban indicators. A and F areas are less diverse in terms of habitat types than are U areas (Fig. 1; Table 3). HL increases significantly in areas that have been under urban pressure over a longer period, particularly when it is associated with compact development and a dense communication network. The highest Ro density is in square F16, located along an F complex (Table 3). Soils naturally occurring in the study transect are fairly varied; however, since they are covered with artificial forms and highly transformed, they do not have a significant impact on SR and flora quality.

The RDA of SR and main urban factors shows grouping of the environmental variables (Fig. 2). Three of the seven urban indicators are statistically significant (Table 4). HH is positively correlated with the two axes of the ordination diagram. A areas are negatively correlated with the first axis and strong positively with the second. Road (Ro) density is negatively correlated with both axes and weaker with the second. Statistically significant urban factors (HH, A, and Ro) are correlated with other factors, and HL is positively correlated with Ro and PU. The proportion of A is negatively correlated with the other groups of variables. HH and Ra density are positively correlated with the first axis and each other. HH is positively correlated with the second axis, in contrast to the proportion of Ro density. HH is not correlated with the PU or with HL but depends on the form of land use (A, Ra) (Fig. 2). HH is a statistically significant factor in relation to flora of the entire transect but is, however, important for each species group. It is positively correlated with the number of spontaneophytes (N _SP) and negatively correlated with anthropophytization (I _AN) and kenophytization (I _KN) indicators (Table 5).

The communication network is an important factor in flora quality. Ro and Ra density is varied in the study area. Ro density has a significant impact on AS distribution (N _AS) and is positively correlated with the number of archeophytes (N _AR), which is reflected in the high proportion of segetal (N _S) and ruderal (N _R) vegetation on these sites (Table 5). There are no roads in three squares (F17–F19), whereas square U7 has the highest road density. There are railway tracks in all major habitat types, which are related to the presence of the Ra network.

The PU is a not statistically significant factor (Table 4) but is strongly negatively correlated with the first and second axis of the ordination diagram (Fig. 2). It is positively correlated with the squares located in the city center, which was longest under urban pressure.

In total, 795 vascular plant species occurred in the study squares. The greatest SR was in U, where 698 species were recorded which represents ~88% of the total number of species in the transect, while in A, 474 species were observed (~60%) and 388 were found in F (~49%). The greatest number of species, both native and alien, was recorded in squares U13 and U15. The lowest level of SR was displayed by the two most peripheral squares: A1 and A20. The number of species therein reached ~100, while the number of AS did not exceed 50. Squares A2, A4, U14, U15, and F17 were the richest in endangered/protected taxa (Table 6).

The quantitative ratio of native (N _NS) and alien (N _AS) species is similar in A and F areas. In U, the mean number of AS increases, whereas the mean number of native species remains unchanged in comparison with the other habitat types. This exerts a positive effect on SR in these squares (Table 6).

In terms of geohistorical elements, the study area is dominated by native (540) rather that alien (255) species. Among anthropophytes, kenophytes (127) predominate over archaeophytes (81) and diaphythes (47). Vegetation of the study area is classified into eight synecological groups (Table 2). All these groups, with various abundance, are represented in transect squares. The highest proportion was recorded for ruderal (N _R), forest (N _F), meadow (N _M), xerothermic (N _X), and segetal (N _S) species, which considerably contribute to SR in the squares and the established transect (Table 6).

Indicators of anthropogenic flora transformation for individual study squares are highly diverse (Table 6). High anthropophytization indicator values (I _AN) were recorded in the city center—squares U6, U8, U10, and U12—whereas the lowest were in F16 and F17 and grassland areas at the city limits (A3). The kenophytization indicator value (I _KN) shows that the particularly distinguished city-center habitats are characterized by compact development, dense transport networks, and industrial areas with a high HL—from ~20% (U10 and U12) to 22% (U8). Spatial distribution of the archaeophytization indicator (I _AR) is not as concentrated. A relatively high proportion of archaeophytes can be observed in areas that are distant from the center and that have retained their agricultural character (A1). The flora modernization indicator (I _M) exhibits the highest values in densely developed residential and Ra infrastructure areas: U8, U13, U15, and F16. Its minimum values have been recorded in areas dominated by arable land (A1), forest (F18), and urban areas (U14) in close vicinity of the forest. The highest mean values of most flora indicators were noted in squares U6–U15 (Table 6), while the lowest were characteristic of squares A1–A5. Notably, the mean number of native species and spontaneophytes—with its highest values recorded in squares dominated by forest and seminatural areas—is an exception.

Four peripherally located squares are atypical (Fig. 3). Square U12 on the right, localized in a housing development district, is characterized by moderate SR, with a substantial proportion of alien elements compared with native elements. This is accompanied by high HL values (Table 3). Opposite is square F16, comprising a seminatural habitat, which exhibits different characteristics: small number of species, dominance of native elements, and low HL. The upper peripheral part of the diagram shows square A2 situated in a suburban area with an agricultural character and a high proportion of forest vegetation, namely, midfield woodlots. There, flora is characterized by remarkable richness, with a mixture of natural, seminatural, segetal, and ruderal vegetation (Table 6). The counterweight for this plot is square U8 in the city center, where SR declines considerably and the proportion of AS increases. HH for both squares is the same, but those areas differ in terms of spatial management, PU, and HL (cf. Table 1; Fig. 1). Squares A1–A3 are a newly incorporated area, while U8 and U9 are in the Lublin settlement.

The diagram clearly shows four aggregations of the remaining squares, which display similar characteristics to those mentioned above (Fig. 3). The first cluster (I) comprises two study squares, which are the richest areas located in prefabricated concrete residential districts, similar to square U12. The second (II) cluster, composed of four squares, has similar characteristics to those of square U8 and is associated with the city center. The third (III) covers areas in the NE part of Lublin (A1, A3–A5), which exhibit similarity to square A2 and have a dual characteristic of seminatural associations of xerothermic vegetation and the area of a railway junction. Covering a fragment of Ra infrastructure, loess ravines, and plateaus, and with its less dense residential developments, the district located in square U15 appeared similar. The last cluster (IV) is clearly similar to square F16, as it comprises squares associated with the seminatural habitats and the neighboring residential district (U14). Spearman rank correlation coefficients (Table 7) show the gradient according to which study squares are ordered (Fig. 3). The first axis is strongly correlated with urban indicators, such as U proportion, HL, PU, and Ro density. The value of anthropopressure increases along this axis (Axis 1), which expresses those U indicators. The second axis is strongly positively correlated with the proportion A areas, indicating their increasing proportion in the studied transect.

The greatest SR, >300 in each square, was reported in the city center where residential areas border railway infrastructure and extensively cultivated A. The lowest SR, <180 per square, was found in the peripheral transect squares. Those areas were overgrown by seminatural vegetation, i.e., forests and fragments of xerothermic grasslands (Table 6). Low habitat diversity and the low HL result in species composition stability and low susceptibility to disturbances.

The proportion of U areas in the study transect was the only indicator that exhibited a significant correlation with SR (Table 5). The correlation between the number of native species (N _S) in study squares and U indicators was not statistically significant. Native species occur in all transect squares and predominate over AS. The number of AS (N _AS), both keno- and archaeophyte, is positively correlated with the typically urban features of the habitat (PU, HL, U, Ro). In this plant group, a negative correlation is found between the U and A areas. This is additionally confirmed by indicator values of anthropogenic flora changes. A similar correlation exists between individual ecological groups. Xerothermic, meadow, aquatic, segetal, and ruderal species prefer typically U habitats.

Analysis of statistical significance of the correlation between sample species composition and habitat indicators shows that HL, proportion of A areas, and Ro density are the most significant (Table 4). These indicators account for 21% of the total variability of occurrence of a given set of species. The other variables are not statistically significant at P < 0.05.

RDA analysis (Fig. 2) confirmed the great significance of HH for SR in transects. This indicator is positively correlated with Axis 1; likewise the proportion of U areas and Ro density. These two indicators are the most typical for U habitats and exhibit a strong correlation. Nearly all squares dominated by U habitats are concentrated around these indicators. The proportion of F and A are equally important for flora richness. The first indicator is positively correlated with the first ordination axis and the other with the second axis. These indicators are linked to squares with an A and F characteristic, respectively. The center of the diagram is occupied by U squares U7, U13, and U15, which do not exhibit typical features of a U habitat. They border seminatural areas. F and A squares (F17, F19), characterized by the increasing gradient of Ra in the square, are noteworthy. The Ra network across F areas influenced flora features (U14, F16, F17). The number of AS increases, which is reflected by the higher values of the following coefficients: archaeophytization, kenophytization, and flora modernization (Table 6).

Stepwise selection of the variables of the impact of U indicators on the proportion of geohistorical elements of the study area (Table 8) demonstrates that the proportion of the communication network Ro and Ra is a significant factor, which accounts about 34% of the total SR variability in the transect.

RDA analysis (Fig. 4) shows that the most significant indicators for number and distribution of geohistorical elements include a proportion of U areas, HL, and Ro density; the latter in the case of archaeophytes (N _AR) and apophytes (N _AP). They are correlated with the first axis, which accounts for 33% of the total sample variability. The number of diaphytes (N _DIA) is negatively correlated with this axis. The second axis has a high positive correlation with the proportion of F areas in transect. HH, Ra network, and proportion of A has a lower significance.

Anthropophytes compete efficiently with native plant species in habitats with heavy urban pressure, which is evident in the central part of the transect and less visible in the peripheral fragments of the study area (Tables 3, 6). An interesting phenomenon can be observed in the case of archaeophytes: they are concentrated in the center of the transect, in compact residential and industrial areas, whereas their number is lower in A. The distribution of the proportion of kenophytes exhibits greater variability: the increase in their proportion (N _KN) is particularly visible in the industrial areas and along the transport network. AS occur spontaneously there and push out native plants. Nonpermanent flora elements (N _DIA) are concentrated in areas where the HL reaches the highest values, i.e., in various anthropogenic habitats with compact development and spontaneous ruderal vegetation and intensively cultivated greenery—city squares and lawns. Ornamental and crop plants occur in these habitats ephemerally (Table 6).

Areas with high HH and HL values proved statistically significant for the distribution of synecological groups (Table 9). They account about 24% of variability at P < 0.05. The other variables were not statistically significant. RDA analysis for synecological groups (Fig. 5) demonstrated a distinct correlation between the number of segetal (N _S), ruderal (N _R), and taxonomically unidentified species (N _UN) and the typically U habitat indicators (PU, Ro). HH has great significance for the other groups, including F species (N _F). All these indicators are positively correlated with the first axis, which accounts for 41% of total variability of the sample. The proportion of A areas exhibits a negative correlation with the two axes, but it is not correlated with any of the synecological species group.

Ruderal (N _R) and segetal (N _S) vegetation is concentrated in the highly transformed habitats of the central part of the transect; its lowest proportion was observed in peripheral areas. The distribution of segetal vegetation (N _S) partly corresponds to that of archaeophytes (N _AR) (Table 6). The proportion of segetal taxa, which is lower than that of ruderal taxa, is associated with changes in A management and subsequent transformation into residential areas. The occurrence of ruderal (N _R) and segetal (N _S) vegetation exhibits a high positive correlation with the typically U habitat indicators (PU, HL, U) and Ro (Table 5), in contrast to A habitats. The number of F species (N _F) clearly declines toward the NE. This synecological group is the most abundant in the southern peripheral areas, which comprise F squares. In more transformed parts of the city, F species colonize replacement habitats, e.g., parks, cemeteries, tree buffer strips. Meadow vegetation (N _M) exhibits a mosaic distribution. The number of meadow species is homogenous in all types of habitats (Table 6). They prefer fresh and wet habitats and occur alternately with xerothermic grass species, which predominate on loess plateaus. The number of meadow species (N _M) exhibits a significant positive correlation with the proportion of U habitats and a negative correlation with the proportion of the A habitats (Table 5). Additionally, xerothermic species (N _X) occur on high Ra and Ro embankments and along roadsides. This is promoted by loess substrate, southern or western exposition, and artificial enrichment of the substrate with calcium carbonate. The increasing proportion of U habitats clearly contributes to the occurrence of xerothermic species (Table 5).