Genetic variation on the rocks – the impact of climbing on the population ecology of a typical cliff plant


Correspondence author. E-mail:


1. Rock climbing enjoys enormous popularity world-wide. As a consequence, the anthropogenic pressure on the vegetation of formerly undisturbed cliff ecosystems is continuously increasing.

2. The impact of rock climbing on population structure and genetic variation of the rare plant species Draba aizoides was investigated representatively for many other typical central European cliff plants. Populations from eight climbed and from eight pristine cliffs were compared through the use of vertical transect analyses and molecular markers.

3. Population structure differed between climbed and pristine cliffs. Individuals of D. aizoides were significantly smaller and less frequent on climbed compared with pristine cliffs. On plateau sites, the species’ occurrence was unaffected by climbing activities; it was significantly less frequent on the faces, but more frequent on the tali of climbed in comparison with pristine cliffs.

4. Genetic variation was greater in populations from climbed compared with pristine cliffs, and genetic differentiation was stronger between subpopulations from pristine cliffs than between subpopulations from climbed cliffs.

5.Synthesis and applications. Rock climbing clearly affects population structure and genetic variation of D. aizoides. Seed dispersal is presumably enhanced by rock climbers but climbers remove and drop plant individuals from cliff faces, which causes a downward shift in population structure. This shift in turn reduces genetic differentiation between upper and lower subpopulations. In mountain regions that attract sport climbing, conservation management plans should therefore always ensure the provision of completely unclimbed cliffs to protect the native vegetation.


Recreational outdoor leisure activities have been increasing since the 1960s (Boyle & Samson 1985), and this trend is expected to continue (Patthey et al. 2008; Törn et al. 2009). As a consequence, the pressure on naturally fragile ecosystems is growing (Patthey et al. 2008). Ecologists have already demonstrated the impact of numerous recreational outdoor activities, such as mountain biking (Thurstone & Reader 2001), skiing (Wipf et al. 1997), mountaineering (Willard & Marr 1970) and hiking (Potito & Beatty 2005), on vegetation structure. Rock climbing, together with many other recreational outdoor activities, is becoming increasingly popular (Perwitzschky 2003), especially in mountain regions at low elevations where there are no seasonal restrictions. In Germany, the Northern Franconian Jura and the Swabian Alb represent important climbing areas of international repute and recreational activity has increased dramatically during recent decades (Herter 1996; Schwertner 2001).

Owing to their inaccessibility, cliffs are amongst the few ecosystems that have been relatively unaffected by humans over the last centuries. However, with increasing climbing activity, this situation is changing. It is of considerable conservation importance that the impact of climbing on these habitats is assessed. They harbour a multitude of rare, endemic and endangered plant species and contribute substantially to regional biodiversity (Larson, Matthes & Kelly 2000).

Studies evaluating the effects of rock climbing have focused principally on changes in the structure and composition of cliff vegetation. Climbing has been shown to significantly reduce total plant cover as well as species richness and diversity (McMillan & Larson 2002; Rusterholz, Müller & Baur 2004). Kuntz & Larson (2006) revealed that pristine cliffs harbour almost twice the species richness as climbed cliffs. Proximity to a climbing route causes a significant decrease in plant cover and species density (Camp & Knight 1998; Müller, Rusterholz & Baur 2004). Correspondence analyses have shown a clear separation of plant community composition between climbed and unclimbed cliffs (Rusterholz, Müller & Baur 2004). Rock climbing leads to an increase in ruderal, alien and disturbance-indicating species and to a decrease in rare and endangered species (Herter 1996; McMillan & Larson 2002; Rusterholz, Müller & Baur 2004).

Although knowledge about the impact of climbing on vegetation cover and community composition is increasing, little is yet known of the effects of climbing on the population structure of individual species. This topic is only sparsely addressed in the literature (Nuzzo 1995; Kelly & Larson 1997). Rusterholz, Kissling & Baur (2009) have recently shown that human trampling decreases the genetic and clonal diversity and changes the spatial genetic structure of the woodland herb Anemone nemorosa. It is possible, therefore, that ‘vertical trampling’ by climbers also affects, and possibly reduces, the genetic variation of plant populations on cliffs.

Cushion plants, such as the rare and endangered Draba aizoides L., are a typical plant life-form in rocky habitats. Species such as D. aizoides are restricted to cliffs and are known to be highly sensitive to rock climbing disturbance (Herter 1996). Trampling has been shown to cause immense harm to this species (Kay & Harrison 1970). Being endangered in Bavaria, D. aizoides has its main central European distribution area in two of the most important climbing areas in Germany: the Northern Franconian Jura and the Swabian Alb (Haeupler & Schönfelder 1988; Herter 1996; Schwertner 2001).

The aim of the study presented here was to investigate the impact of rock climbing on the population structure and genetic variation of the rare and endangered cliff species D. aizoides representatively for many other typical central European cliff plants. We analysed: (i) whether population structure varies between climbed and pristine cliffs as a whole, or whether the situation differs between the components of a cliff: the plateau, face and talus and (ii) whether genetic variation within and among populations or subpopulations differs between climbed and pristine cliffs.

Materials and methods

Species Description

Draba aizoides is a perennial evergreen chamaephyte that grows in dense cushions in crevices and on ledges (Kay & Harrison 1970; Dörr & Lippert 2001). Flowering takes place from March to August (Seybold 2006). Pollination is mainly entomophilous but selfing is possible (Kay & Harrison 1970; Sebald et al. 1998). Seed production is abundant (Wilmanns & Rupp 1966), but vegetative propagation by detached rosettes is also possible (Sebald et al. 1998). In Bavaria, the species is endangered (Seybold 2006). D. aizoides is distributed throughout the Alps and in the mountains of the Pyrenees and Carpathians, and there are, additionally, many locations within the low mountain ranges such as the Jura (Kay & Harrison 1970; Hegi 1986; Seybold 2006).

Study Design

Eight pairs of climbed and pristine cliffs were selected in the Northern Franconian Jura (NJ, five pairs) and on the Swabian Alb (SA, three pairs) to study the impact of climbing on the population structure and genetic variation of D. aizoides (Table 1). We first chose pristine cliffs, representative of the distribution of D. aizoides in these regions. These were paired with climbed cliffs which were located close to the pristine cliffs, easily accessible, used for climbing for at least 50 years and having more than five climbing routes and being, if possible, of similar size and orientation. Cliffs were classified as climbed if they were mentioned in rock climbing literature or if climbing was evident. Cliffs within any pair did not differ greatly in distance, aspect or elevation above sea level (mean differences ± 1 SE: distance: 15·89 ± 7·78 km, aspect: 50·50 ± 16·02, elevation: 40·38 ± 18·92 m). The number of climbing routes varied between 7 and 117, depending on the size of the cliff. Population structure of D. aizoides was analysed on all cliffs using vertical transects.

Table 1.   Climbed (C) and pristine (P) cliffs from the Northern Franconian Jura (NJ) and the Swabian Alb (SA) included in this study, with geographical position, altitude, cliff face area, climbing intensity (CI, number of climbing routes per 100 m2) and the size of the Draba aizoides populations
CodePopulationLa. (N)Lo. (E)Altitude (m)Cliff area (m2)CIPopulation size
NJ01Nankendorfer Block (C)49°52′59″11°20′21″37615000·47489
NJ02Kreuz-Berg (P)49°50′02″11°20′30″402800·0057
NJ03Röthelfels (C)49°44′51″11°14′41″51387501·3411 181
NJ04Hohe Leite (P)49°48′27″11°23′59″528640·00140
NJ05Pfarrfelsen Egloffstein (C)49°42′21″11°15′44″44520000·95811
NJ06Heidelberg (P)49°43′34″11°51′16″4498000·001448
NJ07Düsselbacher Wand (C)49°32′51″11°28′58″45636000·581547
NJ08Eschenbacher Geißkirche (P)49°32′03″11°28′47″51311250·003422
NJ09Mittelbergwand (C)49°33′01″11°31′40″44345001·243954
NJ10Zanklstein (P)49°30′11″11°31′07″45712000·001750
SA01Rossfels (C)48°30′37″09°19′26″7469003·89318
SA02Cliff below Rossfels (P)48°30′17″09°19′23″7701400·00249
SA03Wiesfels (C)48°30′11″09°19′20″76612003·75335
SA04Höllenlochfels (P)48°29′54″09°19′16″7508750·00370
SA05Untere Peilerwand (C)48°23′43″09°46′55″5427502·40382
SA06Öschlesfels (P)48°12′41″09°02′59″91710000·00596

Population Structure Analysis

Vertical transects were spanned over the cliffs and ranged 5 m downwards from the cliff plateau, through the whole face and 5 m deep into the cliff talus (Fig. 1). For safety reasons, the ability to fix the rope for rappelling downwards (e.g. a tree or a bolt) was a prerequisite for the position of the transect. Furthermore, we selected transect positions lacking extreme ledges (to avoid distortions of the transect) and having an orientation comparable to that of the pristine cliffs. Vertical transects had a constant width of 4 m and a varying length because of different heights of the cliff faces. Transects were divided in a 1-m2 grid, and during rappelling downwards, the exact position of each cushion of D. aizoides within the transect was noted, the number of rosettes per cushion was counted, and the length and width of each cushion as well as the diameter of the smallest and largest rosette within each cushion were measured. Subsequently, cushion frequency (CF, number of cushions within a transect/transect surface in square metre), rosette frequency, coverage and cushion surface (CS, length times width of each cushion) were determined. Statistical analyses revealed strong correlation between these parameters. For this reason, only data on the most significant parameters CF and CS are presented.

Figure 1.

 Structure of the vertical transect. With a constant width of 4 m, transects ranged downwards from 5 m deep in the plateau over the whole cliff face into the talus with a depth of 5 m. Transects were subdivided into eight sections, whereas sections 1 and 2 lay in the plateau, 3–6 in the face and 7 and 8 in the talus.

The transect records were divided into the cliff parts ‘plateau’, ‘face’ and ‘talus’ and further subdivided into eight transect sections. To yield these eight sections, the cliff plateau and talus were subdivided into two planes of the same size by dividing the transect line after 2·5 m. Having begun ranging downwards from the plateau, sections 1 and 2 (plateau) and 7 and 8 (talus) all had a length of 2·5 m each. The sections 3–6 were subsequently determined by subdividing the cliff face height into four equal distances. For the subdivided transects, the percentage surface covered by the species was calculated as the surface covered divided by the surface of the transect section of each cliff. Because of heterogeneity in the field, the absolute length of these sections differed between cliff pairs. To remove these inequalities, variables were recalculated on a unit area (m−2) basis and these data were used in subsequent analyses.

Before sampling plant material for molecular analyses, the extent of the total rock area of each cliff and the rocky area covered by D. aizoides was determined by measuring width and height of the respective cliff parts. Using this information, we calculated the total size of the population for each cliff by multiplication of the CF with the size of the cliff area covered by the species. Climbing intensity was calculated as climbing route density per 100 m2 by dividing the total rock area by the number of climbing routes (obtained from the Deutsche Alpenverein, DAV).

Molecular Analysis

Twenty young, green, rosettes were collected from each of the five pairs of climbed and pristine cliffs in the Northern Franconian Jura. For sampling, each cliff was divided into an upper cliff half and a lower cliff half such that the upper cliff half was defined as the whole plateau and 2 m of the topmost cliff face and the lower cliff half as 2 m of the lowermost face and the whole talus, providing two subpopulations per cliff. In each subpopulation, 10 samples (200 samples in total) were collected at approximately the same distances on the total width of the rock area. To avoid sampling from the same individual, a minimum distance of at least 2 m between rosettes was used.

Plant material was dried on silica gel and total cellular DNA was extracted following the CTAB protocol from Rogers & Bendich (1994) in an adaptation by Reisch (2007). Concentrations of the DNA extracts were measured photometrically. Solutions were diluted with water to 7·8 ng μL−1 and used for the analysis of amplified fragment length polymorphisms (AFLPs), which were conducted in accordance with the protocol from Beckmann Coulter (Brea, USA) as described previously (Bylebyl, Poschlod & Reisch 2008; Reisch 2008). DNA adapters were prepared by adding equal volumes of both single strands of EcoRI and MseI adaptors (MWG Biotech, Ebersberg, Germany) following a 5-min heating at 95 °C with a final 10-min step at 25 °C. DNA restriction and adapter ligation were performed in one step by adding a 3·6-μL mixture containing 2·5 U EcoRI (MBI Fermentas, St. Leon-Rot, Germany), 2·5 U MseI (MWG Biotech), 0·1 μM EcoRI and 1 μM MseI adapter pair, 0·5 U T4 Ligase with its corresponding buffer (MBI Fermentas), 0·05 M NaCl and 0·5 μg BSA (New England BioLabs, Ipswich, USA) to 6·4 μL of genomic DNA in a concentration of 7·8 ng μL−1. Following an incubation at 37 °C for 2 h with a final enzyme denaturation step at 70 °C for 15 min, the restriction-ligation products were diluted 10-fold with 1× TE buffer for DNA (20 mM Tris–HCl, pH 8·0; 0·1 mM EDTA, pH 8·0).

For preselective DNA amplification, 1 μL diluted DNA restriction-ligation product, preselective EcoRI and MseI primers (MWG Biotech) were added to an AFLP Core Mix (PeqLab, Erlangen, Germany) containing 1× Buffer S, 0·2 mM dNTP’s and 1·25 U Taq-Polymerase. In a 5-μL reaction volume, polymerase chain reaction (PCR) was performed on an automated thermocycler at 94 °C for 2 min, then 30 cycles of 20-s denaturation at 94 °C, 30-s annealing at 56 °C and 2-min elongation at 72 °C and a final 2-min 72 °C and 30-min 60 °C step for complete extension ending with a final cool down to 4 °C. After PCR, products were diluted 20-fold with 1× TE buffer for DNA.

Three primer combinations were chosen for a subsequent selective PCR. Therefore, PCR was carried out in a total reaction volume of 5 μL containing an AFLP Core Mix (1× Buffer S, 0·2 mM dNTP’s, 1·25 U Taq-Polymerase; PeqLab), 0·05 μM selective EcoRI (Proligo, Paris, France), 0·25 μM MseI (MWG Biotech) primers and 0·75 μL diluted preselective amplification product. For detection, EcoRI primers labelled with different fluorescent dyes (M-CAC/D2-E-AGC, M-CTT/D3-E-AAG, M-CAC/D4-E-ACT; Beckmann Coulter) were used. PCR parameters used were 2 min at 94 °C, 10 cycles of 20-s denaturation at 95 °C, 30-s annealing at 66 °C and 2-min elongation at 72 °C, after which annealing temperature was reduced every subsequent step by 1 °C, additional 25 cycles of 20-s denaturation at 94 °C, 30-s annealing at 56 °C and 2-min elongation at 72 °C and a following 30-min step at 60 °C for complete elongation and a cool down to 4 °C. Selective PCR products were diluted fivefold (D2) and 10-fold (D4) with 1× TE buffer for DNA. D3 products were used without a further dilution.

After pooling 5 μL of each selective PCR product of a given sample and adding them to a mixture of 2 μL sodium acetate (3 M, pH 5·2), 2 μL Na2EDTA (100 mM, pH 8) and 1 μL glycogen (Roche, Manheim, Germany), DNA was precipitated in a 1·5-mL tube by adding 60 μL of 96% ethanol (4 °C) and an immediate shaking. DNA was pelleted by 20-min centrifugation at 14 000 g at 4 °C, the supernatant was poured off, and the pellet was washed once by adding 200 μL of 76% ethanol (4 °C) and centrifuged at the latter conditions and was subsequently vacuum-dried in a concentrator.

After redissolving the pelleted DNA in a mixture of 24·8 μL Sample Loading Solution (SLS; Beckmann Coulter) and 0·2 μL CEQ Size Standard 400 (Beckmann Coulter), selective PCR products were separated by capillary gel electrophoresis on an automated sequencer (CEQ 8000; Beckmann Coulter).

Results were examined using the ceq 8000 software (Beckmann Coulter) and analysed using the software Bionumerics 6.6 (Applied Maths, Kortrijk, Belgium). From the computed gels, only those fragments were taken into account for further analyses that showed intense and articulate bands. Samples yielding no clear banding pattern or obviously representing PCR artefacts were repeated or finally excluded from further analyses. Reproducibility of molecular analyses was investigated with 10% of all analysed samples by means of estimating the genotyping error rate (Bonin et al. 2004), which was 2·3%.

Statistical Analysis

In the case of normal distribution and variance homogeneity, population structure was analysed using dependent t-tests; otherwise paired Wilcoxon and U-tests were conducted. Correlation analyses of geographical characteristics and population structure parameters were based on Spearman’s rank correlation coefficient. All statistical analyses were conducted with pasw Statistics 17 (IBM, Munich, Germany) for Windows.

From the AFLP bands, a binary (0/1) matrix was created wherein the presence of a fragment of a given length was counted as 1 and the absence as 0. As one individual was omitted from the study, the final matrix and all further calculations comprised 199 samples. Employing the software PopGene 1.32 (Yeh et al. 1997) genetic variation within populations was computed as the percentage of polymorphic bands (PB), as Nei’s Gene Diversity H (H = 1 − Σ (pi)2) and as Shannon’s Information Index SI (SI = Σ (pi) ln (pi); pi = allele frequency). Wilcoxon tests were used to compare genetic variation within populations because the data were not normally distributed (KS test < 0·200). Correlation between genetic variation within populations and population size was tested using Spearman’s rank correlation coefficient. Both analyses were conducted with pasw Statistics 17 (SPSS) for Windows. The apportionment of genetic variation within and between populations and subpopulations was assessed by hierarchical amova with the software GenAlEx 6.3 (Peakall & Smouse 2001).


Populations from Climbed and Pristine Cliffs

Within the 16 transects, a total of 1727 D. aizoides cushions were recorded. A highly significant decrease in CF and CS was observed on climbed cliffs (Table 2). For cliff faces, the majority of cushions were found on pristine cliffs, which supported 83·77% of all cushions recorded from this part of the transect. In contrast, the tali of climbed cliffs showed an immense increase in the proportion of cushions found, (97·01%) compared with that of pristine cliffs. Six of eight pristine cliffs did not harbour any cushions or rosettes on this part of the cliff, whereas cushions were found on the tali of all climbed cliffs. Although CF did not differ significantly between the plateaux of climbed and pristine cliff, conspicuous differences for the cliff faces and tali were obtained. Climbed cliff faces showed a significant reduction in CF, but this fact was significantly inverted in climbed cliff tali.

Table 2.   Number of cushions recorded (N), cushion frequency (CF in m−2) and cushion surface (CS in mm2) of plants from climbed and pristine cliffs
 Whole CliffPlateauFaceTalus
  1. Values indicate means ± 1 SE. Significant differences are indicated by *< 0·05, **< 0·01 and ***< 0·001.

CFclimbed0·56 ± 0·121·76 ± 0·770·20 ± 0·041·01 ± 0·27
CFpristine1·50 ± 0·322·04 ± 0·931·93 ± 0·370·03 ± 0·02
CSclimbed1523 ± 1771359 ± 2641927 ± 3431324 ± 484
CSpristine2374 ± 5171992 ± 3122269 ± 557470 ± 465

Subdivision into eight transect sections revealed all pristine cliffs to harbour considerably more cushions in the face sections. In contrast, seven of eight climbed cliffs exhibited higher cushion numbers on the tali (Fig. 2). Significant decreases in percentage surface covered on climbed cliffs were obtained for the sections 3 (Z = −2·521, = 0·012) and 5 (= −2·201, = 0·028), and a significant increase was obtained for the climbed section 7 (= −2·240, = 0·025). Values for section 4 were marginally reduced on climbed cliffs (= −1·820, = 0·069).

Figure 2.

 Population density distribution of Draba aizoides on climbed and pristine cliffs. Transect sections 1 and 2 lay within the cliff plateau, 3–6 within the cliff face and 7 and 8 within the cliff talus. Error bars indicate one standard error of the mean. Asterisks show significant differences at < 0·05.

Spearman’s rank correlation did not reveal any relationship between geographical differences between cliffs and population structure. The CS correlated negatively with climbing route density (= −0·857, = 0·007).

Genetic Variation Among and Within Populations from Climbed and Pristine Cliffs

Amplified fragment length polymorphism analysis based on 214 fragments was conducted with a total of 199 individuals, comprising five pairs of populations from climbed (n = 100) and pristine (n = 99) cliffs. In the AFLP analysis, 90·19% of all fragments were polymorphic. Shannon’s Information Index (SI) was significantly higher in populations from climbed cliffs. The percentage of PB and Nei’s Gene Diversity (H) were also larger in populations from climbed cliffs but these differences were not significant (Table 3). Population size of D. aizoides did not differ significantly between climbed and pristine cliffs and was not correlated with genetic variation (rPB = 0·462, pPB = 0·179; rH = 0·309, pH = 0·385; rSI = 0·552, pSI = 0·098). Moreover, genetic variation did not differ significantly between subpopulations from climbed upper and lower cliff halves or between pristine upper and lower cliff halves. An overall analysis of molecular variance (amova) showed no significant genetic differentiation between climbed and pristine cliffs. Further analyses revealed that both climbed and pristine cliffs have about 50% of genetic variance within and between subpopulations from upper and lower cliff halves (Table 4). However, genetic differentiation between upper and lower subpopulations from pristine cliffs was stronger (ΦPT = 0·50) than that between upper and lower cliff halves from climbed cliffs (ΦPT = 0·43). Genetic variation did not correlate with climbing intensity (PB, SI, H all: r = 0·100, = 0·873).

Table 3.   Genetic variation within populations from climbed (C) and pristine (P) cliffs measured as the percentage of polymorphic bands (PB), Nei’s Gene Diversity index (H) and Shannon’s Information Index (SI). Population sizes are given by PS
  1. Values indicate means ± 1 SE. Significant differences are indicated by *< 0·05.

C10024·952 ± 0·7180·0792 ± 0·00350·1197 ± 0·00473596 ± 1991
P9922·338 ± 0·9500·0714 ± 0·00270·1071 ± 0·00381363 ± 616
 199P = 0·080P = 0·225P = 0·043*P = 0·500
Table 4.   Results of the analyses of molecular variance (amova) between populations from climbed and pristine cliffs and between upper and lower subpopulations from climbed and pristine cliffs
Level of variationd.f.SSMS%P
  1. SS, the sum of squares; MS, mean squares and % the proportion of genetic variation. Significances are indicated by ***< 0·001.

Populations from climbed vs. populations from pristine cliffs
 Among climbed/pristine cliffs1118·26118·260·00n.s.
 Among populations81058·56132·3246·260·001***
 Within populations1891379·627·3053·740·001***
Individuals of the upper half vs. individuals from lower half of climbed cliffs
 Among groups8470·4058·8043·000·001***
 Within groups90627·806·9857·000·001***
Individuals of the upper half vs. individuals from lower half of pristine cliffs
 Among groups8625·6878·2150·000·001***
 Within groups89644·607·2450·000·001***


Rock Climbing Affects Population Structure

Generally, cliffs harbour ancient, largely undisturbed and unproductive ecosystems with a large number of rare species (Haig, Matthes & Larson 2000; Larson, Matthes & Kelly 2000; Baur, Fröberg & Müller 2007). D. aizoides, a cushion plant of low competitive ability, is highly adapted to limestone cliff habitats and very sensitive to external disturbances such as trampling or sideways shearing forces (Wilmanns & Rupp 1966; Kay & Harrison 1970; Herter 1996).

The results presented here provide clear evidence for the impact of rock climbing on the population structure of D. aizoides. Individual frequency and size were conspicuously reduced on climbed cliffs as compared with pristine cliffs. This is in line with changes in the population structure of D. aizoides because of climbing reported in a previous study (Wezel 2007). Because pairs of climbed and pristine cliffs differed only marginally in distance, aspect or elevation and because population structure was not correlated significantly with distance, elevation, aspect or population size, the differences observed in population structure between climbed and pristine cliffs can clearly be attributed to rock climbing activities. Our study revealed, however, not only a general decline of D. aizoides on climbed cliffs, but also a shift in population structure as a result of climbing. The species occurred at almost the same frequency on the plateaux of both climbed and pristine cliffs but was significantly less frequent on cliff faces and more frequent on cliff tali of climbed than of pristine cliffs. Herter (1996) and Wezel (2007) pointed out that rock species such as D. aizoides and Saxifraga paniculata sometimes occur unexpectedly on tali of climbed cliffs. Furthermore, cliff face vegetation often exhibits lower species densities on climbed cliffs (Camp & Knight 1998; McMillan & Larson 2002), and the proximity to a climbing route can cause a significant decrease in plant cover and species density on the rock face (Müller, Rusterholz & Baur 2004).

Climbing adversely affects the populations of D. aizoides in a direct way. Abrasion by climbing ropes and the use of pockets, cracks, holes and ledges as hand- and footholds obviously lead to a decline in the species’ abundance. Herter (1996) found 23% of D. aizoides individuals studied to be threatened by climbing and trampling. In the worst case, climbers directly remove D. aizoides individuals growing near climbing routes to clear the surface. Climbing bolts are often placed within the upper cliff face for fixed rope installations or top rope climbing, a set-up where the rope is fixed on the wall for belaying climbers from the ground to reduce the risk of accidents (Perwitzschky 2003). Thus, heavy abrasions and removals can explain the significantly decreased number of D. aizoides cushions found and the CF, especially in the transect sections 3–5. In this context, it should be noted that in our study the extent of damage was negatively correlated with climbing intensity, as already demonstrated in a previous study (Camp & Knight 1998). On the wall-proximate cliff talus, climbers start ascents, store ropes and belay partners and thus cause a decrease in total plant cover whilst creating new gaps (McMillan & Larson 2002; Müller, Rusterholz & Baur 2004). These gaps obviously allow the establishment of new individuals both from rosettes and from the seeds of D. aizoides falling down during climbing, which results in the observed shift in population structure.

Rock Climbing Shifts Genetic Variation

Our study revealed stronger genetic variation within populations of D. aizoides from climbed than from pristine cliffs. This could theoretically be attributed to higher levels of disturbance on climbed cliffs, following the intermediate disturbance hypothesis (Grime 1973; Connell 1978; Reusch 2006). Hence, variation initially increases with increasing disturbance because of the removal of superior competitors and then declines as a result of the survival of only those individuals able to cope with these disturbances. Rock climbing could have produced gaps in cliff habitats, and the reduction of monopolizing genotypes might have allowed the establishment of more new genotypes. Barrett & Silander (1992) and Kudoh et al. (1999) previously demonstrated that physical disturbance provides gaps allowing the successful recruitment of sexual progeny.

As described earlier, we observed a downward shift in population structure because of climbing activity. This shift had no measurable impact on genetic variation within lower subpopulations; we observed no significant differences in the genetic variation of lower subpopulations from climbed and pristine cliffs. Genetic differentiation between upper and lower subpopulations was, however, stronger on pristine as compared with climbed cliffs. This finding corroborates the results of Rusterholz, Kissling and Baur (2009), who also observed an impact of human recreational activities on genetic differentiation. Lower levels of genetic differentiation between upper and lower subpopulations on climbed cliffs can indeed be explained by the fact that climbers remove rosettes of D. aizoides from the cliff face which fall down and establish in the cliff talus. This process is credible, as D. aizoides has a high potential (c. 86%) to produce adventitious roots (Wilmanns & Rupp 1966). Wezel (2007) has shown that D. aizoides cushions, cut off by rock climbers, can re-established in the talus of climbed cliffs. Moreover, the downward shift of individuals might have redirected much seed dispersal from the upper to the lower subpopulation because of climbing activities. Enhanced seed dispersal increases gene flow, which in turn reduces the levels of genetic differentiation (Slatkin 1987). Seed dispersal by climbers could therefore also have contributed to the observed differences in genetic differentiation between climbed and pristine cliffs.

From a long-term perspective, it seems quite possible that climbing-induced gene flow will further diminish genetic differentiation between upper and lower subpopulations. It is, however, difficult to assess whether in the future variation will increase because of disturbance or decrease because of reduction in population size.

Conclusion and Perspectives

The results of our study give us a clear picture of the impact of rock climbing on population structure and genetic variation of a rare and endangered cliff plant, D. aizoides. However, detailed analyses concerning the small-scale genetic variation within upper and lower subpopulations would be of interest, because stronger dynamics within lower subpopulations owing to the population shift should have an impact on spatial genetic structure. Furthermore, long-term monitoring of genetic variation would be helpful to assess the future impact of climbing on genetic variation. In summary, we conclude that in mountain regions that attract sport climbing, conservation management plans should always provide for the retention of completely unclimbed cliffs to protect natural population structure and genetic variation of the native vegetation.


The authors thank P. Gerstberger, J. Wagenknecht and the ‘Verein zur Erforschung der Flora des Regnitzgebietes’ for their help with plant localities and P. Poschlod for his generous support. Furthermore, we thank M. Bartelheimer and O. Vogler for their helpful comments on an earlier version of the manuscript, Hilary Wallace and Mike Prosser for the improvement of our English and two anonymous referees for their inspiring reviews.