**2. Study sites, sampling design, and analyses**

The two glacier forelands studied are located in the Central Eastern Alps. In summer 2005, permanent plots were established immediately in front of Goldbergkees (Hohe Tauern, Austria) and Lenksteinferner (Rieserferner, Italy), which were revisited every second year thereafter. In 2010, chronosequences were established within the same glacier forelands extending from the permanent plots down to moraines of the Little Ice Age (LIA) maximum which occurred around 1850. **Table 1** summarizes basic information on both study areas.

Sample sites are 10 m<sup>2</sup> (2 × 5 m) for both the permanent plot and the chronosequence studies and represent "mean" site conditions (i.e., no hollows with above-average snow cover duration or wind-exposed knolls with drier conditions). The permanent plots are arranged in two different sets. The A-sites were installed close to the glacier terminus (2–8 m in front of the ice margin in 2005), the B-sites 7–12 m away from the A-sites. Most sites became deglaciated in the year of the initial survey, and some of the B-sites probably already the year before. In total, 26 sample sites at Goldbergkees (13 A, 13 B) and 22 at Lenksteinferner (11 A, 11 B) are studied. Sites are GPS-recorded, flagged, and photo-documented for a precise match at resurveys. Vegetation sampling records groundcover and individual numbers of all vascular plants as well as structural features such as life form composition [12] and dispersal biology types [13] of the species. Mosses are sampled as undifferentiated species group. Taxonomy of vascular plant species follows [14]. Sampling occurs m2 -wise with the smallest unit being 0.01% ground cover (i.e., 1 cm × 1 cm on a 1 m<sup>2</sup> subplot). Raw data are subsequently converted to mean ground cover values as well as total number of species and individuals per 10 m<sup>2</sup> sample site.


**Table 1.** Basic data on the permanent plot and chronosequence studies within the glacier forelands of Goldbergkees and Lenksteinferner.

The chronosequences extend from the permanent plots down to the LIA terminal moraines. For both glacier forelands detailed chronologies about glacier retreat since LIA maximum exists [15, 16], allowing for a quite accurate age determination of sample sites. At Goldbergkees, the total horizontal distance between the youngest and oldest sites is 1.3 km and 200 m elevational difference, on Lenksteinferner 1.25 km and 280 m, respectively. On Goldbergkees, eight different stages of time since melt-out were studied (2 years, 4 years, 15 years, 25–30 years, 55 years, 85 years, 120 years, 155 years) and on Lenksteinferner, nine different stages (2 years, 4 years, 20 years, 35 years, 55 years, 75 years, 90 years, 120 years, 155 years). Each stage is represented by three 10 m<sup>2</sup> plots. For the youngest stages (<10 year-ice-free), three of the permanent plot samples were used (three A-sites of 2007, three B-sites of 2009). Vegetation sampling is primarily the same as on the permanent plots, with the only modification that individual numbers are not counted. Environmental variables collected in situ for all sample sites of both approaches are elevation (by altimeter), exposure (by compass), slope angle (by clinometer), and rockiness of the ground (by visual estimation of coarse rocks > 6 cm in %).

**2. Study sites, sampling design, and analyses**

of vascular plant species follows [14]. Sampling occurs m2

0.01% ground cover (i.e., 1 cm × 1 cm on a 1 m<sup>2</sup>

Sample sites are 10 m<sup>2</sup>

128 Glacier Evolution in a Changing World

sample site.

around 1850. **Table 1** summarizes basic information on both study areas.

The two glacier forelands studied are located in the Central Eastern Alps. In summer 2005, permanent plots were established immediately in front of Goldbergkees (Hohe Tauern, Austria) and Lenksteinferner (Rieserferner, Italy), which were revisited every second year thereafter. In 2010, chronosequences were established within the same glacier forelands extending from the permanent plots down to moraines of the Little Ice Age (LIA) maximum which occurred

**Figure 2.** Hypothetical successional trajectories carrying linear or both positive and negative logarithmic/exponential

characters. At a particular stage also reverse trends of decreasing species or individual numbers might occur.

and represent "mean" site conditions (i.e., no hollows with above-average snow cover duration or wind-exposed knolls with drier conditions). The permanent plots are arranged in two different sets. The A-sites were installed close to the glacier terminus (2–8 m in front of the ice margin in 2005), the B-sites 7–12 m away from the A-sites. Most sites became deglaciated in the year of the initial survey, and some of the B-sites probably already the year before. In total, 26 sample sites at Goldbergkees (13 A, 13 B) and 22 at Lenksteinferner (11 A, 11 B) are studied. Sites are GPS-recorded, flagged, and photo-documented for a precise match at resurveys. Vegetation sampling records groundcover and individual numbers of all vascular plants as well as structural features such as life form composition [12] and dispersal biology types [13] of the species. Mosses are sampled as undifferentiated species group. Taxonomy

to mean ground cover values as well as total number of species and individuals per 10 m<sup>2</sup>

(2 × 5 m) for both the permanent plot and the chronosequence studies


subplot). Raw data are subsequently converted

For data analyses, univariate and multivariate statistical procedures were employed. To assess the successional development in glacier forelands quantitatively, documented changes between different samples are crucial—temporally different in the case of the permanent plots and spatially different in the case of the chronosequences. Primary succession in glacier forelands commonly starts with simple agglomerations of plants and subsequently becomes more and more complex. The increasing complexity during succession becomes obvious in change measures such as mean ground cover values (of singular species and total) as well as species and individual numbers. Temporal trends are derived by linear and/or nonlinear regressions. The range of variation in data sets (species numbers, individual numbers, groundcover, etc.) is depicted by box and whisker plots.

For detecting gradual changes in species composition within large data sets, ordination procedures assuming underlying gradients within data sets are appropriate tools. By means of similarity relationships, gradual changes of samples concerning species composition are calculated in a multidimensional ordination space. The aim of ordinations is to reduce the number of dimensions and to make complex datasets with many species and samples interpretable. For the chronosequence data, unconstrained linear principal component analyses (PCAs) are employed. The graphic presentation of ordination analyses is by two-dimensional scatter plots displaying samples and/or species; explanatory environmental variables (if available) are displayed as arrows. The arrows point from the origin of ordinates in the direction of samples with above average values of the particular variable; the length represents the relevance of the variable. Gradient analyses were performed with Canoco 4.5.
