The research was conducted on an eroded area with a chronosequence of reclamation sites at the Holar farm, which is situated in southern Iceland (63°59′N, 19°57′W; c. 150 m a.s.l.). The climate is maritime, with cool summers and mild winters (Takhtajan 1986). The closest synoptic weather station, Hæll (64°04′N, 20°15′W; 121 m a.s.l.), is about 17 km from the study area. The mean temperature ranged from a low of −1·2 °C in January to a high of 11·7 °C in July 2000 and annual precipitation was 1150 mm (Icelandic Meteorological Office 2001).
Anthropogenic and natural disturbances have contributed to land degradation and soil erosion at the Holar farm and the surrounding area. The nearby Hekla volcano is very active. It has erupted 20 times in historical times, producing both tephra and andesitic lava flows (Gudmundsson 1996) and has occasionally strewn tephra over the study area. The history of the Holar probably resembles that of large areas in Iceland where wood gathering and heavy grazing has destroyed native birch woodlands (Thorsteinsson 1986). Loss of woodlands was often followed by soil erosion and barren land (Thorarinsson 1961; Arnalds et al. 2001). The resulting soils are poorly developed, low in nitrogen and carbon content (Elmarsdottir 2001) and the surface is sandy-gravel, as frost-heave lifts stones to the surface. Furthermore, the soils (Typic or Lithic Vitricryands) have low allophanic content and rather limited water holding capacity (Arnalds & Kimble 2001).
The study area was approximately 50 ha in size and the topography nearly level (some 1–5% slopes). Four study sites were selected within the area, one unreclaimed site (will be referred to as control) and three reclamation sites that had undergone 2, 5 or 11 years of reclamation effort. The reclamation objective was to increase vegetation cover on eroded areas and improve the land for grazing animals. Reclamation activities began with spreading a 1- to 2-cm thick manure layer (organic fertilizer), followed by application of about 150 kg ha−1 of inorganic NPK (20 : 12 : 8) during the second year. Inorganic fertilizer was then applied to each site every second or third year thereafter. Approximately 350 sheep graze the study area and vicinity during the early growing season (mid-June to mid-July), 60 during the summer months and 200 sheep during the autumn (late September to early November).
The vegetation at the control site consisted mainly of forbs (Thymus praecox Opiz, Silene acaulis (L.) Jacq., Rumex acetosa L., Galium normanii O. Dahl), and grasses (Festuca richardsonii Hooker, Agrostis stolonifera L.) (Elmarsdottir 2001). Racomitrium ericoides (Brid) Brid. was the most common moss, but others (e.g. Bryoerythrophyllum recurvirostrum (Hedw.) Chen, Pogonatum urnigerum (Hedw.) P. Beauv. and Schistidium papillosum Culm.) were also common. Lichen species present were Peltigera leucophlebia (Nyl.) Gyeln., Stereocaulon alpinum Laur., S. rivulorum Magn. and P. canina (L.) Willd. Nomenclature for vascular species, mosses and lichens follow Kristinsson (1998), Johannsson (1998) and Kristinsson (1997), respectively.
Cover of vascular species was 10% at the control site and at the reclamation sites that had undergone 2, 5 or 11 years of reclamation effort, the cover was 50%, 77% and 75%, respectively (Elmarsdottir 2001). Cover of moss was less than 10% on the control site and the 2-year-old site and was 55% on the oldest reclamation site, but cover of lichens was less than 3% at all sites. Plant species diversity was greatest on the 5-year-old site and least on the control site, and grasses were more abundant on the reclaimed sites (Elmarsdottir 2001). Above-ground biomass, measured in adjacent plots in 1999, was 0·3 and 8·4 t ha−1 for the control and 11-year-old reclamation sites, respectively (Aradottir et al. 2000).
sampling and data analysis
Within each of the four study sites, four plots (10 × 10 m) were chosen subjectively to represent vegetation that was as homogeneous as possible within each plot, but encompassed what appeared to be differences in vegetation composition and aspect at that site. Within each plot, 10 quadrats (0·5 × 0·5 m) were placed randomly. All data were collected in July and August 2000.
Cover of each microsite type (Table 1) was assessed by 25 equally distributed point measurements within each quadrat. Seedling density was recorded within a portion (0·5 × 0·3 m) of each quadrat. All seedlings from the current growing season were identified to species or genus, and the relevant microsite type was recorded.
Table 1. Description of the nine microsite types used to derive indicators for safe sites for seedlings in nonreclaimed and reclaimed plots at the Holar study area
|Biological soil crusts||Biological soil crusts* and ≤ 1-cm thick mosses|
|Mosses||Mosses > 1 cm thick|
|Sand||Sand ≤ 0·2 cm in diameter on the surface|
|Small rocks||Rocks 0·2–2 cm in diameter on the surface|
|Large rocks||Rocks > 2 cm in diameter on the surface|
|Litter||Combination of litter, humus and dung|
The data were analysed with sas programs and were treated both as continuous and categorical data (SAS 1999). It should be noted that true replication of sites does not exist (Hurlbert 1984) and anova with subsampling was used to compare data among sites where plots were nested within sites and quadrats were nested within plots. Data were analysed with anova by using the sas proc mixed procedure with site as a fixed effect and plots within a site as the error term. If a significant (P < 0·05) overall F-test was found, individual site means were compared using a pairwise t-test. The Type I error rate was 0·05 per comparison. Variance was examined for homogeneity with residual plots and, when necessary, data were transformed to ensure normal distribution of residuals. Cover data of microsite types and density of all seedlings were transformed using a square-root prior to anova analysis and the data were log (natural) transformed to test for differences in seedling density among sites for each species. All means and values in tables and graphs are presented using nontransformed data.
Chi-square goodness-of-fit statistics were used to determine if observed frequency of seedlings showed a departure from a random occurrence (expected) among microsite types (Conover 1980). The analyses were carried out separately for each reclamation site. The observed frequency was the frequency of seedlings found in each microsite type within a reclamation site. The expected frequency was calculated as the total frequency of seedlings within a reclamation site multiplied by the proportion (cover) of each microsite type within that site. Data for microsite types that had low expected frequencies were combined for analysis to ensure that no more than 20% of microsite types had an expected number < 5 (Conover 1980). As the seedling data was based on numbers within quadrats (cluster sampling), the analysis might fail to meet the assumption of independent distribution, which would increase Type I error rates and results might be overstated (Garson & Moser 1995). To counter this problem, we used a 99% significance level and a conservative test statistic, i.e. the deviation for each microsite type had to be equal to or greater than the critical chi-square value for the full analysis for each reclamation site.