The flora and vegetation of an old Solvay process tip in Jaworzno [Upper Silesia, Poland]

This paper demonstrates the flora, plant communities and substrates of an old solvay process spoil tip in Upper Silesia, Poland. In an area of 15 000 m 2 there are growing 136 vascular plant species. The flora is characterised by the preponderance of Asteraceae - species and long-lived perennial herbs, many of them coming from meadows and grasslands. Ninety-five percent of species are apophytes despite the anthropogenic origin of the site. A majority of species are associated with moderately dry, base-rich soils with low or moderate levels of nitrogen. The site is shown to be an important refuge for some protected species, montane species and other elements uncommon in the local flora. An analysis of a series of samples used a methodology based on the assessment of percentage cover of particular species and multivariate analysis based on TWINSPAN. Both suggested a relatively high overall similarity between the samples with minor variations associated with moister substrates.Elemental analysis and pH determinations of soil samples associated with the releves revealed a narrow range of pH and an absence of any strong concentrations of heavy metals. A redundancy analysis of the soil-plant relationships suggested that the strongest trend of differentiation was most closely associated with a phosphate gradient, and the next strongest was pH and possibly waterlogging. The most species-rich vegetation was associated with low phosphate and high pH levels. The results could be interpreted to suggest that processes of soil development and plant succession are slow but nevertheless perceptible, with implications for future loss of diversity. The vegetation constitutes an assemblage essentially of one type showing only weak relationships with described vegetation types such as Molinio-Arrhenatheretea meadow, Festuco-Brometea grassland and Caricetalia davallianae mire. The results also suggest that the vegetation of the site is of considerable value for nature conservation. The site should be protected and be the subject of further research.


INTRODUCTION
Various types of waste land cover a significant area in in dustrial Upper Silesia. Each site represents an unusual and often unique set of chemical and physical conditions for the establishment of colonising organisms. Man-made habitats such as these present important opportunities for scientific in vestigation. Many have already been the subject of biological and ecological investigation e.g. Ash et al. (1994), Rostański (1998), Shaw (1992Shaw ( , 1998, Tokarska-Guzik et al. (1991), Tumau and Rybka (1991) and Woźniak (1998).
Many authors have considered and assessed the role of an thropogenic habitats in the process of establishing and main taining biodiversity (Buszman et al. 1993;Hind 1956-57;Trzcińska-Tacik 1966;Tokarska-Guzik and Rostański 1996). Many examples show that these waste lands can be covered with interesting vegetation and that even rare plant species can be found there. The importance of many post-industrial sites has already been recognised as refuges for protected and rare plants (Greenwood and Gemmell 1978;Hind 1956Hind -1957Kelcey 1975;Tokarska-Guzik 1991a). This seems to be par ticularly true for lime waste heaps.  commented on the significance of 6 lime waste tips in Poland and England as refugia for protected and regionally rare species. Wilkoh-Michalska and Sokol (1968) reported 249 species of flowering plants from lime spoil mounds in the Notec valley in Poland, including many uncommon species. Lee and Greenwood (1976) described species-rich vegetation characteristic of base-rich habitats which are otherwise absent from the county, on calcareous waste from salt and chemical industries in Cheshire, UK. Vegetation on calcareous waste from the Leblanc process at Nob End in Greater Manchester, UK was given protected site ("Site of Special Scientific Inter est") status in 1988 on account of its unusual flora (Shaw and Halton 1998). The solvay process tip or "soda heap" at Jaworzno, Upper Silesia seems to be another example of this phenomenon (Tokarska-Guzik 1991a. The aim of the present study was to describe, characterise and determine the nature conserva tion value of the vegetation of the Jaworzno site and to relate the floristic composition of the vegetation and its variation to soil characteristics. A range of approaches have been used in the survey and analysis of the vegetation, allowing a compari son of phytosociological methods and multivariate computer analyses. The site at Jaworzno has been the subject of syste matic observation for some time (Tokarska-Guzik 1991a, 1999 with the first botanical records made in the 1980s (Celinski et al. 1982;Tokarska-Guzik 1991b). This period of observation allows the present analysis to be placed in the context of dynamic changes in the vegetation and its persistence.

Site description
The study area is located within a 20 ha site called Wapniowka, which could be translated as "lime place", after the lime heaps which cover 50% of the site with an average height of 4 m. The site, shown in Fig. 1, lies in the Jaworzno Hills in the Silesian Upland (Grid Ref. 50° 18';19° 10') and is an abandoned solvay process slurry tip from a former soda factory.
The Szczakowa window glass factory, formerly an Austrian soda factory, to the west of the site produced soda until 1911 using the Solvay process. The process produced "white seas" of calcium chloride as an aqueous suspension and the present day heaps probably represent the places where it was de posited.
Attempts to assess the potential of the waste material either as a fertiliser or to fill nearby dolomite extraction pits have shown that the material contains too high concentrations re spectively of lead and sulphate ions for these two purposes (unpublished material gathered by the existing window glass factory Szczakowa S. A.). On the other hand, research carried out by the Institute of Fertilisation in Pulawy for the glass factory (unpubl.) showed that the lime waste had the follow ing properties: -very good deacidification potential, particularly for soils poor in lime and magnesium a significant proportion of magnesium higher chemical activity compared with ordinary lime fer tilisers no NaCl addition as an admixture low levels of heavy metals a paste-like consistency throughout the depth of the heaps The study area The part of the site which was the subject of the current in vestigation is a c. 2 ha. 'D'-shaped plateau c. 200 x 100 m. It is isolated on all sides by steep, more or less bare, vertical eroded banks c. 3 m high. The study area as a whole is more or less horizontal with an undulating microtopography, most ly within the range of 20 cm, varying in moisture content but Fig. 1. Localisation of a solvay process tip in Jaworzno. 1 -build-up areas; 2roads; 3 -railways; 4 -forests; 5 -rivers and reservoires; 6 -town border; 7 -solvay process tip. generally moist. The edges are dished as though there has been some shrinkage on the site.
The vegetation of the site consists mainly of fairly short, species-rich grassland, and can be described as healthy and thriving. The whole plateau bears a scattering of about 100 birch trees, on average around 10 m high, with a circum ference around 32-50 cm. There is also a scattering of shrubs in the vegetation, but these seem to be subject to dieback, particularly Frangula alnus.

Vegetation survey methods
A preliminary assessment of the study area was made by walking over it and making a list of all vascular plant species present, scoring them on the DAFOR (Dominant, Abundant, Frequent, Occasional, Rare) scale. Broadly four vegetation types were discernible at this stage. There were a number of patches of vegetation in the central areas which tended to be coarser, less species-rich and dominated by Calamagrostis epigejos or Molinia caerulea. Other facies observed included a variant on the Calamagrostis vegetation with Valeriana of ficinalis, Medicago saliva, Frangula alnus and Rubus caesius, and a fairly tall vegetation with Gymnadenia conopsea. These vegetation types were sampled by recording 2 m x 2 m sam ples representative of homogeneous areas, as shown in Table  1. In each sample, all vascular plant species and bryophytes were identified and their abundance estimated and recorded on two scales as shown in Table 2. Multivariate computer analysis was performed using the Dornin scale, and non-computer tabulation analysis was performed using the other abun-  Table 2. Total percentage cover of the vegetation and bare ground were also recorded together with the mean height of the vegetation. Slope and aspect of each sample was estimated and micro topographical details were re corded. The area represented by each sample was estimated and recorded (Table 4). Vascular plant nomenclature follows Mirek et al. (1995) and bryophyte nomenclature follows Ochyra and Szmajda (1978).

Substrate survey and analysis
Substrate samples were collected from the centre and to wards the four comers of each vegetation sample or relevé. The five samples were taken from 0-5 cm depth following removal of superficial litter and plant material. The samples were combined and thoroughly mixed to provide a single composite sample for each relevé. The substrate samples were then air dried at room temperature for several days. The air dried soil samples were passed through a 2 mm sieve and the fraction >2 mm was discarded. This comprised roots and other plant material only. The fraction <2 mm was then divided into 3 subsamples for pH testing, elemental analysis and storage.
Substrate pH was determined by mixing a subsample of the air-dried material with twice the volume of distilled water. The mixture was stirred for 1 minute and pH recorded using a pH electrode.
Elemental analysis on the soil was carried out by X-Ray Fluorescence Spectrometry. The air-dried sieved material was oven-dried at 4°C overnight. Samples were then ground to a fine powder in a tungsten teemer mill. Hersh wax was added to the milled soils in a vial in the ratio of 1.5 g wax to 8.5 g soil. Two agate balls were added to the vials to aid mixing and the vials placed on a thrapulator for 20 minutes. Pellets of the soil/wax mixture were made by pressing the mixture into plastic cups at a pressure of 15 tonnes per square inch. All equipment was cleaned with ethanol between samples. Concentrations of elements, as listed in Table 6, in the sample pellets were determined against International Standards of soil sample pellets containing a single element of known concen tration using an ARL 8410 X-Ray Fluorescence Spec trometer.

Analysis of the flora
The list of vascular plant species for the site, which in cluded all previous known records as described in the Intro duction, was analysed in terms of their Raunkiaer life forms, ecological group, Ellenberg indicator values (Ellenberg et al. 1992), geographical-historical group (Komas 1981), families and current protection status.

Vegetation analysis
Two different methodologies were used to analyse the samples.
In the first, a tabular arrangement of the samples and species was made based on the identification of widely ac knowledged differential or characteristic species with a high degree of fidelity to particular plant communities. The sam ples were ordered according to similarities in species compo sition and the frequency of occurrence of particular species.
In the second, computer-based methodology, multivariate analysis methods were used to analyse the abundance records. A hierarchical classification method. TWINSPAN (Hill 1979) based on the ordination of the vegetation records, was used to try to identify vegetation types by classifying samples accord ing to their floristic similarities. The floristic relationships be tween samples, the ecological relationships between species

Analysis of the flora
The study site is rich in species considering the small area it covers and its geographical and ecological isolation from other areas with similar physical characteristics. Of the 136 vascular plant species recorded for the site (Table 3), 114 were recorded in the study area, 73 of them in the releves. The best represented families in the flora of the site are: It is interesting to note that orchids also make a large con tribution to the flora. These comprise eight of the nine nation ally protected species present on the site (the ninth is Carlina acaulis). A further three species are partly protected. Ranun culus hulbosus is regionally protected and there are a further seven locally rare species (Ajuga genevensis, Antennaria dioica, Botrychium lunaria, Inula salicina. Parnassia palustris, Polygala amarella, Teucrium botrys).
The results of life form analysis of the flora are shown in Fig.  2. They show a clear preponderance of long-lived perennials (hemicryptophytes, geophytes, chamaephytes) which comprise 87% of the flora, while the annuals only contribute 13%.  Examination of species from the various ecological habitat groups (Fig. 3) show that meadow and grassland species make the largest contributions, while species from forest and anthropogenic habitats are less abundant and wetland species make the smallest contribution. Fig. 4 shows that 95% of the flora are native species (apophytes) and only 5% are alien (anthropophytes).

q Anthropophytes Apophytes mKn ■ Ar nAp
It is possible to assess the regional significance of the site since a Flora of Jaworzno township has recently been com pleted (Tokarska-Guzik 1999). Botrychium lunaria, Polygala amarella and Orchis militaris have their only record in the township on the Solvay Process heap. Teucrium botrys, and Antennaria dioica. both quite uncommon in Upper Silesia, have only one other record in Jaworzno township. Epipactis palusiris occurs in eight other sites but is at its most abundant on the tip. Fig. 10a represents the presence of protected species in the township on a l km grid. The larger the dot the greater the number of protected species. The l km square which includes the Solvay Process tip is the second from the top within the township boundary immediately west of line DD (DF4459). The presence of the tip clearly makes this square one of the most significant for protected species in the township. Fig. 10b shows a similar coincidence map for montane species. These are mostly associated with squares with signi ficant areas of forest. The square containing the Solvay Pro cess tip includes no forest, but clearly the tip acts as a similar refuge for montane species. Similar methodology may be used to demonstrate that the tip is an important refuge for thermophilous species and calcicolous species (Tokarska-Guzik 2000).
The unusualness of this site is evident in the considerable contribution to the vegetation of species such as Carex panicea, Epipactis palusiris. Pamassia palustris and Tofieldia calyculata.

Non-computer analysis of the vegetation
The results of this analysis are shown in Table 4. There are twelve species of high (>80%) frequency in the vegetation. Lotus corniculatus, Rhinanthus minor, Ranunculus acris, Festuca rubra, Trifolium pratense and Achillea millefolium grow in meadows, Carex flacca, Plantago media and Briza media are grassland species. Most of these species are characteristic of neutral to calcareous soils where management or environ mental stress suppress the vigour of more competitive species. The frequent occurrence of Epipactis palustris and Pamassia palustris is of particular interest since these plants grow in rare eutrophic peat bogs. Only Calamagrostis epigejos is a common plant. A further 9 species occurred with a frequency of over 50%. The Table 4 also includes a consider able group of meadow and grassland species.
Many species occurred at only low frequency in the data, 29 with only 2 occurrences, but in combinations which pro duced distinctive facies, such as Ononis spinosa and Galium mollugo in the taller vegetation, and Thymus serpyllum, Hieracium pilosella and Erigeron acris in the shorter vegetation. Plant cover varied from 75% up to 100% but is mostly 95-100%. The number of vascular plant species is quite high in all samples, ranging from 16-34 with a mean and standard error of 24.65 +/-1.06. Six moss species were recorded, Brachythecium salebrosum, B. rutabulum, B. velutinum, Bryum cespiticium, Amblystegium serpens and Plagiomnium rostratum. Mosses were recorded in only 5 samples where their cover was 1-2% except in sample 7 where they reached 25% cover. Table 4 shows the relative homogeneity of the vegetation, but distinguishes 2 major sub-units. The first contains 11, possibly 13, samples and is defined by the presence of species such as Avenida pubescens, Carlina vulgaris and Polvgala vulgaris which grow in dry, basic habitats. These are the more species-rich samples. The second sub-unit is defined by Molinia caerulea and Valeriana officinalis which grow in wet meadows and calcareous mires. It contains 7, possibly 9, samples. The vegetation of this sub-unit is taller and coarser than that of the first one. Samples 7 and 8 appear transitional, having some of the species of both sub-units, but being more species-rich than either and having a higher pro portion of bare ground and mosses.
The table shows a distinctive unit with Tofieldia calyculata, Arabis hirsuta, Potentilla arenaria and Sanguisorba minor represented by the two transitional samples and three others. Three other facies are shown in the table characterised re spectively by Gymnadenia conopsea, Anthyllis vulneraria and Botrychium lunaria. Table 4 also shows two groups of species, one associated with meadow, the other with grassland. However, the overall character of the soda heap vegetation is quite distinctive and is not consistent with these two vegetation types.

Computer analysis of the vegetation
Multivariate vegetation analysis of the sample data using the computer program TWINSPAN confirmed the overall ho mogeneity of the vegetation with low eigenvalues at each di vision. Even at the first division of the samples the eigen value was only 0.204, suggesting that only 20% of the vari ability in the data was explained at that division and that there was relatively little coordinated variation in species dis tribution between samples. The program has identified a num ber of vegetation types shown in Table 5. This table is analo gous to the results shown in Table 4. The numbers in the body of this table are abundance levels on a scale of 1-4, the sample classification is shown horizontally at the top and bot tom of the table, and a simplified species classification is shown vertically at the right hand side. Binary notation is used for both sample and species classifications.
The overall character of the vegetation is defined by the large number of constant species which appear as a large block two-thirds of the way down Table 5, species Group 01. The species above this block characterise the main vegetation type present in the study area (species Group 001), and a single variant (species Group 000). Species Group 001 is clearly the larger group identified in Table 4 using non-computer methods. The species below the main block, species Group 1, characterise a second vegetation type and also a fur ther variant of the main vegetation type.
The first main division of the samples identified a distinc tive vegetation type represented by samples 16, 18 and 19 (sample Group 1) and differentiated by Valeriana officinalis at high levels of abundance, Ononis spinosa, Medicago sali va, Galium mollugo and Molinia caendea. These species are Fig. 11. RDA ordination diagram of releves, species and environmental variables.
12 56 not all exclusive to this vegetation type, but they are strongly preferential.
The next two divisions of the samples differentiated the main group (sample Group 010). The first of these divisions separated off the two samples 7 and 8 which proved "anoma lous" in the non-computer analysis (sample Group 00). The second division separated off three samples, 12, 17 and 20, (sample Group 011) with abundant Molinia caerulea. These were placed alongside the other Molinia samples 16, 18 and 19 (Group 1) as in Table 4. However, these six Molinia samples are divided here on the basis of the strongly preferential species of sample Group 1 and the greater affinity of the companion species in sample Group 011 to sample Group 010.

Comparison of analyses
Both analyses have identified a typical vegetation type con sisting of almost the same samples: 1-4, 6 and 9-15. Only sample 5 was classified differently in the two analyses. Both analyses have distinguished samples 7 and 8. In the non-computer analysis they were shown to span 2 vegetation types and a variant of one of them. In TWINPAN they were separated out at one end of the sample ordination as a distinct variant of the typical vegetation. A tall vegetation type characterised by Molinia caerulea was clearly identified by both methods, but whereas in Table 4 it appears as a homogeneous vegetation type, TWINSPAN has recognised two distinct variants. No farther variants or distinctive facies were apparent in the TWINSPAN analysis, while the non-computer analysis has recognised a Tofieldia calyculata -Arabis hirsuta variant on the typical community and a farther three facies each charac terised by a single species.

Substrate characteristics
The substrates, at least in the upper horizons sampled, were all fine textured with a perceptible crumb structure, a signifi cant penetration by roots, and were dark stained with organic matter. Table 6 shows that the pH across the site does not vary greatly. Elemental concentrations varied across the site with manganese, zinc, lead, aluminium, iron, barium and magnesium tending to be higher in samples 1-4 and 20, while calcium and sodium were higher in other parts of the site.

Redundancy analysis of substrate-plant relationships
A preliminary assessment was made of the entire sample and substrate data using both Canonical Correspondence Ana lysis and Redundancy Analysis. This indicated that Redun dancy Analysis was the more appropriate method, and that few of the substrate variables had a distinct, or coordinated, influence on the vegetation. Despite the relatively low vari ability in total phosphorus shown in Table 6, phosphorus and barium and to a lesser extent pH, sodium and manganese were shown to have the strongest influence, though none were statistically significant. These five plus calcium in view of its relative abundance in the substrate, were the only 6 variables selected for the main Redundancy Analysis, the results of which are shown in Fig. 11.
Ordination of samples along axes 1 and 2 in Fig. 11 show similar groupings to those identified by the TWINSPAN  Variations in the vegetation in relation to bases can be seen on both axes. The differences between the apparent relation ships of barium and sodium probably relates to their different mobilities in the soil, with sodium being more readily leached than barium; calcium was intermediate and less significant. The association of manganese with the Molinia caerulea and Parnassia palustris vegetation at the lower end of axis 2 is suggestive of periodic waterlogging. Fig. 11 shows two main trends in the vegetation, both starting from the upper right hand quadrant. Moving left to the upper left hand quadrant, there is a successional transition from short, dry, open, base rich, nutrient poor vegetation to taller, coarser, more nutrient demanding vegetation.
Moving clockwise from the upper right hand quadrant there is a gradual transition to less base rich more mesic vegetation and then to damp coarser vegetation in the lower left hand quadrant, although the influence of bases is still apparent. The samples clustered in the centre of this diagram show the cur rent status of the vegetation in relation to these trends.

DISCUSSION
The analyses have shown that the vegetation of the soda heap is not strongly differentiated, although there is some clear variation in both the species composition and in the vi gour of the vegetation, some of which can be related to suc cessional trends and water regime. The typical vegetation, which contains the majority of samples, is species rich and contains a number of rare species which are usually only found in uncommon and highly valued vegetation types. These include a number of calcicolous and xerothermal species whose presence is related to the particular charac teristics of the substrate on the soda heap. The general ap pearance of the vegetation is of species rich calcicolous grass land and has some of the species of the classical vegetation types Molinio-Arrhenatheretea meadow, Festuco-Brometea grassland and Caricetalia davalianae mire. This unique com bination of species has arisen in response to the very specific habitat conditions associated with the soda heap. Nevertheless, the vegetation is coherent and not simply a random assemblage of species, as indicated by the overall similarity of the floristic structure of the vegetation in most of the samples, and the identification of successional trends and localised environmental influences in others. In the absence of a comprehensive classifi cation of synanthropic vegetation, a possible name for this com munity could be Rhinantho-Caricetum flaccae.
Both the computer and non-computer analyses have been valuable in characterising the vegetation of the old soda heap. The small differences in the definition of variation in the vegetation between the two methods are related to the differ ent ways in which Tables 4 and 5 are constructed. Patterns and trends in the data and known groupings of species are systematically identified by eye and experience in the con struction of the non-computer table. TWINSPAN on the other hand, takes no account of typical groupings of species, only groupings within the data set. It simultaneously takes account of all species, their abundance and occurrence in all samples so that sample groupings are only defined by the co-ordinated responses of individual species, or groups of species, drawn from the entire data. Thus although the two samples with Gymnadenia conopsea, for example, are placed together in the sample ordination, in the TWINSPAN analysis their overall species composition is not sufficiently different for them to be separated from adjacent samples, even at lower levels of division.
In the analysis of unusual vegetation such as that in the present study which has not previously been described, both methods have merits. The non-computer method (similar to Braun-Blanquet methodology) facilitates the comparison with naturally occurring associations and recognises subtle vari ations related to particular species, while TWINSPAN allows the objective definition of vegetation types and variants.
The development of spontaneous communities on post-in dustrial sites is influenced by aspects of soil chemistry such as alkalinity or acidity, nutrient level and toxicity, and by physical factors such as soil structure and particle size (Brad shaw and Chadwick 1980). Although there were some clear trends in the influence of soil chemistry on the Jaworzno soda heap vegetation, the absence of any statistically significant ef fects suggests that there are also factors other than these in fluencing the composition of the vegetation. Lee and Green wood (1976) found that even in wastes younger than those of Jaworzno, the plant associations observed generally could not be related to soil chemistry and that physical features such as compaction and its effect on hydrology were more important. They observed that the water content of wastes up to 32 years old was always high within the rooting zone, although com paction resulted in the formation of a hard dry crust impene trable to roots with localised standing surface water in places. These observations are consistent with the nature of the vege tation at Jaworzno, where the better drained edges of the pla teau supported a more open, stress-tolerant, calcicolous vege tation, while there were some areas in the centre of the pla teau where the CANOCO analysis indicated periodic water logging. Lee and Greenwood (1976) describe soil development on lime wastes over a period of 32 years as simply tire accumula tion of organic material on the surface of unalterd parent ma terial. The dieback of shrubs such as Frangula alnus and the rarity of other woody species except Betula, suggests that the situation is similar at the Jaworzno soda heap even after 87 years. Birch is a very successful pioneer species on industrial wastes being tolerant of a wide range of toxic compounds (Atkinson 1992;Bialobok 1979). Its success on the Jaworzno soda heap may have implications for the future of the vegeta tion on the site. The establishment and growth to a consider able size of substantial numbers of birch trees may already have contributed to nutrient enrichment on the site. Fig. 11 shows a trend for coarser less species rich vegetation to be as sociated with nutrient enrichment which if continued, could further threaten the diversity of the typical vegetation.
The expansion of industrial, agricultural and urban activities has resulted in massive losses of natural habitats and the na tive species they support. In built-up areas, therefore, special attention should be paid to the spontaneous vegetation appear ing in synanthropic habitats since many vulnerable and rare species of vascular plants have been recorded in them (Adamowski 1998;Shaw and Halton 1998;Tokarska-Guzik 1991a. With its calcicolous and xerothermal species, the soda heap in Jaworzno is an example of the way in which synan thropic habitats can be the mainstay of locally uncommon and even rare plant species in an area, providing a refuge or re placement habitat. It is also important as a reservoir of diver sity for the surrounding industrial landscape. As with the most valued semi-natural plant communities, the vegetation at Jaworzno is important not only as a refuge for locally and nationally uncommon and rare species, but also for its intrinsic beauty and for its unusual vegetation which shares many of the characteristics of the vegetation of other soda heaps (Lee and Greenwood 1976; Wilkon-Michal-ska and Sokół 1968). Whether semi-natural or anthropogenic, plant communities develop in response to particular combina tions of environmental conditions and at Jaworzno the condi tions, and the vegetation they have given rise to, are very un usual. Unusual biological communities such as this, can be so intimately linked with the particular industrial process which created them that they can be considered to be unique and non-recreatable (Box 1993). As such, they have a value even beyond their undoubted ecological value in their contribution to the cultural, industrial and economic heritage of the region to which they belong.