The Icelandic government approved activities including revegetation and afforestation in the 1990s to increase the terrestrial carbon sequestration from the atmosphere (Sigurdsson and Snorrason, 2000; Aradottir and Arnalds, 2001; Ministry for the Environment, 2007). In effect, land reclamation has been carried out for over 100 years in order to halt land degradation and soil erosion events (Crofts, 2011) caused by human activities since the island's settlement about 1100 years ago (Aradottir and Arnalds, 2001), as well as natural stress factors such as volcanic eruptions or the harsh climate.

Woodlands and species-rich heathlands form the undisturbed ecosystem type on drylands at lower elevation (<400 m a.s.l.) (Aradòttir et al., 1992). Fertile Brown Andosol is the typical soil type of these ecosystems and it is found across 13 360 km2 (Óskarsson et al., 2004). Andosols have a tendency to accumulate higher quantities of SOC than other soil types, due to the cover of soil organic carbon (SOC) enriched surface horizons by volcanic ejecta and the andic properties resulting from the formation of organo-mineral complexes (Dahlgren et al., 2004; McDaniel et al., 2012; Delmelle et al., 2015; Arnalds, 2015a). Hence, the average SOC stock of the Brown Andosols is estimated at 227 t C ha-1 (Óskarsson et al., 2004). During the last centuries, about 43 000 km2 of Iceland (∼40 %) has been affected by severe extreme soil erosion (Arnalds et al., 2016). Consequently, about 120–500 Mt of SOC has been lost in the past (Óskarsson et al., 2004). Presently, approximately 45 000 km2 (∼45 %) of the land area is covered by sparsely vegetated areas which range to barren deserts, in addition to disturbed areas with reduced carbon levels (Arnalds, 2015b). These landscapes are characterized by limited vegetation cover on vitric soil types (Arnalds et al., 2013) with low biomass production and low SOC stocks (Óskarsson et al., 2004). Vitrisols (Vitric Andosols and Leptosols), which are the typical soil types of the deserts, contain less than 45 t C ha-1 on average (Óskarsson et al., 2004).

Based on these differences, the potential of the severely degraded soils to sequester high amounts of carbon has been demonstrated (Arnalds et al., 2000; Ágústsdóttir, 2004). An important aspect of reclaiming degraded land is the recovery of ecosystem services including rehabilitation of farm land, protection against soil erosion or public recreation (Aradóttir et al., 2013). For example, the large-scale project called Hekluskógar was established in southern Iceland in 2007 with the aim of restoring resilient birch woodlands on about 900 km2 in the vicinity of Mount Hekla in order to reduce the effect of volcanic hazards (Aradóttir, 2007).

The Icelandic carbon stocks have been reported in a national inventory for the UNFCCC (Hellsing et al., 2016). The Icelandic National Inventory Report uses a country-specific soil carbon sequestration factor of 0.51 t C ha-1 yr-1 for soils during the conversion of severely degraded land (“Other Land”) to forest land or grassland (Hellsing et al., 2016). This is based on Icelandic field studies which found an increase in the SOC stock and therefore assigned a positive soil carbon sequestration effect to the reclamation (Aradóttir et al., 2000; Snorrason et al., 2002; Ritter, 2007; Bjarnadottir, 2009; Kolka-Jónsson, 2011; Arnalds et al., 2013). In addition, the Icelandic Soil Conservation Service continuously reviews this value by ongoing C sequestration monitoring (Hellsing et al., 2016). The establishment of a vegetation community passes through different development stages; consequently, the sequestration rate, as a function of SOC change over time, is not linear until the new SOC stock equilibrium is reached (Smith et al., 1997; Six et al., 2002; Stewart et al., 2007). Hence, the development of the SOC stock and the SOC sequestration rates need to be recorded with a high temporal resolution instead of using the data of only two inventories (e.g. t0 and t1 or initial and developed vegetation type).

Monitoring the total C stock is not sufficient to characterize the overall potential for removal of atmospheric C by afforestation since soil organic matter (SOM) consists of a heterogeneous mixture with respect its physical protection and chemical structure (Schmidt et al., 2011). This leads to dynamic patterns of SOC stocks, composition of SOC, decomposability and turnover rates of SOC during land-use changes (von Lützow et al., 2008; Poeplau and Don, 2013). Recent studies show that the labile SOC pool, which is composed mainly of particulate organic matter (POM), increases simultaneously with the total SOC in the mineral soil during the establishment of vegetation systems with a higher net primary production rate (Guidi et al., 2014; Gabarrón-Galeote et al., 2015; Trigalet et al., 2016; Hunziker et al., 2017). To date, however, soil studies in Iceland have not focused on such changes in SOC fractions during the establishment of woody vegetation systems.

In Iceland, the conversion of vegetation cover is currently driven by revegetation of severely degraded land (Aradóttir et al., 2000; Arnalds et al., 2013), natural succession following glacier retreat (Vilmundardóttir et al., 2015) and afforestation of different types of tree species on heathland (Ritter, 2007), as well as on grazed land (Snorrason et al., 2002). However, there is limited information concerning carbon sequestration in soils associated with the afforestation of severely degraded landscapes by the only native forest tree species, Betula pubescens Ehrh. ssp. czerepanovii.

The present study was part of the CarbBirch project (Halldórsson et al., 2011) which was launched in 2008 and involved two of the five CarbBirch study areas. The main goal of CarbBirch was to study the ecological impact of the restoration activities in Hekluskógar. The present study aims at characterizing the long-term carbon sequestration potential of afforestation efforts with mountain birch on severely degraded soils. For this, we compared the SOC patterns in mountain birch stands of different ages to those of severely degraded and barren areas, reclaimed grasslands and natural old growth birch woodlands. The article first introduces the commonly used SOC parameters, SOC concentration and the SOC stocks (0–30 cm), and then discusses the vertical distribution of SOC, SOC quality and the interaction between the SOC and the volcanic clay minerals.

The study area is in the vicinity of the Mount Hekla volcano (Fig. 1a). Due to the unsustainable land use and volcanic activity, most of this area has been affected by erosion. The resulting landscape is characterized by sandy deserts (Arnalds et al., 2016), which often leads to the formation of important sandstorms (Crofts, 2011; Arnalds et al., 2016), and, in consequence, reclamation activities have been carried out over the last decades (Halldórsson et al., 2011). The soil parent material generally consists of lava field material, glacial till, aeolian deposits or buried soil materials (Dugmore et al., 2009; Thorarinsdottir and Arnalds, 2012).

The afforested woodland area “Gunnlaugsskógur” is located approximately 1 km north of the Icelandic Soil Conservation Service Headquarters at Gunnarsholt (Fig. 1c). In 1926, the eroded area was excluded from sheep grazing by fencing. After the stabilization of the ground surface and the fertilization of the soil, birch was seeded on small plots in 1939 and 1945. In 1945, birch seedlings resulting from the activity in 1939 were transplanted on a nearby lava field, although most of the present birch area at Gunnlaugsskógur has naturally regenerated through seed production of the previously planted birches (Aradóttir, 1991; Aradottir and Arnalds, 2001). The age of the afforested birch sites was determined by dendrochronology as part of the CarbBirch project. The mean ages of the sampled afforested birch plots were 15, 20, 25, and 50 years (Birch15, Birch20, Birch25, and Birch50), respectively.

In addition to the birch plots, soil samples were taken from three severely degraded sites with barren surfaces and from three revegetated sites with grass vegetation north of Gunnlaugsskógur (Fig. 1d). In the present study, the severely degraded and eroded sites (Barren Land) represent the stage before any restoration activity has begun. The Barren Land sites were selected at 4 km distance from Gunnlaugsskógur, as barren areas were not available near the afforested birch sites, and it was assumed that the geologic and pedologic characteristics were comparable to those at the birch sites. The grassland sites (Grass50) were located next to the severely degraded sites; these were protected against sheep grazing by fencing and then revegetated by using fertilizers and grass seeds about 50 years ago and at present are not used for hay production. The topsoil at these sites has been found to be degraded, while horizons buried by wind deposits may contain some carbon (Arnalds, 2010; Arnalds et al., 2013). This has to be considered when assessing carbon sequestration by actual restoration programmes because parts of the found SOC may result from earlier vegetation. Due to the same age of Birch50 and Grass50, the two different reclamation types can be compared directly.

The differently aged birch sites were further compared to a naturally growing birch woodland located at Hraunteigur (Fig. 1b). This area was protected against sand encroachment by two streams, but was subjected to deposition of large amounts of dust and periodic tephra fallout. Thus, it represents the original mountain birch woodlands (Birchnat) which covered large areas in the vicinity of Mount Hekla in the past (Árnason, 1958). As the vegetation cover is subject to large-scale sediment deposition, the area has accumulated soils with depths of more than 2 m (Kolka-Jónsson, 2011).

Field sampling was carried out in summer 2011. Each of the tested categories (e.g. Barren Land, Birch15, Birchnat) described above was represented by three test sites (three replicates) (Fig. 1e). After removing the litter layer, the top 30 cm of the mineral soil was sampled. This sampling depth interval represents the common depth for SOC stock inventories (Aalde et al., 2006; Snorrason, 2010), and in addition, the top 30 cm of the mineral soil contains most of the belowground living root biomass at grassland and birch sites (Snorrason et al., 2002; Bjarnadottir et al., 2007; Hunziker et al., 2014). Thus, the dominant belowground organic carbon source deriving from plant growth is located between 0 and 30 cm soil depth.

At each site, five soil pits were randomly placed. At the woody sites, sampling occurred within one-half of the crown diameter of a dominant mountain birch (Betula pubescens Ehrh. ssp. czerepanovii) tree. The soil was sampled with a cylindric metal core (Eijkelkamp Soil & Water, Giesbeek) of 100 cm3 volume and 5 cm in diameter at given soil intervals (0–5, 5–10, 10–20 and 20–30 cm). The five sub-samples per depth interval were immediately mixed in order to form one composite sample. Thus, each depth interval per category was represented by 3 composite samples (3 replicates per depth interval) (Fig. 1), resulting in a dataset of a total of 84 composite samples.

The amount of soil organic carbon that is stored in a given soil profile is defined as the SOC stock and is given in tons per hectare. According to Ellert et al. (2008) and Rodeghiero et al. (2009), the SOC stock (SOCstock; t C ha-1) is a function of the soil's carbon content (SOCconc; mg g-1), the bulk density (BD<2mm; g cm-3) of the fine soil fraction (<2 mm) and the investigated soil depth (d; cm). The conversion factor between the units is 100. The study calculated the amount of soil organic carbon which was stored in the fine soil fraction (<2 mm) within the top 30 cm according to the given formula: SOCstock=SOCconc×BD<2mm×d×100. SOC stocks stored in the SOC fractions were calculated after Poeplau and Don (2013) and Guidi et al. (2014).

The soil material sampled at all depth intervals was andic ((Al+1/2Fe)ox>2 %) and therefore classified as Andosols (Table 2) (IUSS Working Group WRB, 2014). According to the Icelandic soil classification (Arnalds, 2008), freely drained soils under vegetation are termed Andosols and desert soils are classified as Vitrisols. The results of the present study confirmed that the sampled soils of the birch and grass stands are classified as Brown Andosols (1 %–12 % C and >6 % allophane) (Tables 1, 2). The calculated bulk densities of the fine earth material fractions were within the range of 0.3 to 0.8 g cm-3 of typical Icelandic Andosols (Arnalds, 2008). The severely degraded and unvegetated soils of Barren Land were classified as Vitrisols (Arnalds, 2015c) due to the relatively high pH values (>7.0; H2O). The pH values of the top 20 cm of these soils were significantly (p≤0.05) higher compared to those of the other tested categories. However, the soils of Barren Land contained more organic carbon and clay minerals than usually found in Vitrisols of Icelandic deserts (Table 1) (Óskarsson et al., 2004). At Barren Land, we found the highest concentrations of allophane and ferrihydrite clay minerals (Table 2). These high concentrations stand in contrast to the typically low concentration (2 %–5 %) found in desert Vitrisols (Arnalds, 2015d). Lilienfein et al. (2003) found an increase in allophane and ferrihydrite concentrations with increasing age of mudflow soils at Mt Shasta. Based on these findings, our results indicate that the soils at Barren Land are pedogenetically developed and the high carbon and clay contents found on Barren Land are more representative of severely degraded soils than Icelandic desert soils.

SOC concentrations varied between 0.6 % and 9.8 % within the whole dataset of the 84 samples (Table 1). Surprisingly, the lowest C contents were not found at the Barren Land sites as was expected; this is most likely due to the absent vegetation cover at the time of sampling. This finding supports the hypothesis that organic carbon was sequestered in the soil before the onset of soil erosion. Further, at the Barren Land, Grass50 and Birch15 sites, the SOC concentrations were higher in the deeper sampling intervals (5–30 cm) than at the shallow depths of 0–5 cm (Table 1). At the Barren Land, Grass50 and afforested Birch sites, the C:N ratios of the soil mostly varied between 10 and 14 (–). However, the ratio was considerably higher in the top 5 cm at Grass50 and significantly (p=0.05) higher at all tested birch categories compared to the soil at Barren Land. In deeper soil layers the differences were only significant (p=0.05) between Barren Land and Birchnat in sampling layer “5–10 cm” (Table 1). Hence, the soils at Birchnat also showed the highest C:N ratios with a maximum value of 19.2 (Table 1). These findings can be attributed to the presence of freshly deposited and less decomposed organic matter close to the surface ground at the vegetated sites. In contrast, the C:N ratio was slightly higher at deeper sampling intervals at Barren Land, which gives evidence that the carbon originates from past vegetation cover.

Based on the analysis of the C concentrations of the soil samples, the study showed that the unvegetated, severely degraded volcanic soils contained appreciable amounts of SOC. And further, afforestation with mountain birch increased the soil C concentration during the first 50 years of shrub establishment, predominantly in the top 10 cm. Further, the values of bulk density and SOC concentration are inversely proportional. Consequently, to further discuss the SOC sequestration potential of the soils studied, further detailed information on SOC stocks and SOC quality are needed, in addition to measurements of SOC concentration. Lastly, the unknown influence of the sampling depth also needs to be accounted for.

In the present study, the initial state before afforestation starts is represented by the sites of Barren Land. The SOC stock (0–30 cm) at Barren Land (median value: 39 t C ha-1) is higher (p>0.05) than the SOC stocks at Birch15, Birch20 and Birch25. The present study found a continuous increase in the median SOC stock (0–30 cm) with birch stand age (Birch15: 31; Birch20: 33; Birch25: 36; Birch50: 46 t C ha-1) (Fig. 2). After 50 years of birch growth the SOC stock (0–30 cm) of Birch50 is significantly (p=0.05) higher than the SOC stocks of younger birch stands (Birch15, Birch20) or severely degraded soil (Barren Land) (Fig. 2). The given results of the SOC stocks (0–30 cm) might lead to the assumption that the soil acts as a C source during the first 25 years of the establishment of birch and that there is a carbon sink between after 25 years until 50 years of birch growth. This would be in accordance with Hunziker et al. (2017), who found a decline in the SOC stock (0–30 cm) during the first 40 years of green alder encroachment on former subalpine pastures. Another finding of the present study is that after 50 years of birch growth, the SOC stock was still significantly (p=0.05) lower than that of the old-growth woodlands of Birchnat (Δ15 t ha-1) (Fig. 2). This means that the soils at Birch50 can sequester additional organic carbon during the succession towards mature woodlands which reflects the equilibrium state. Overall, the results indicate that afforestation by mountain birch, and the establishment of birch woodlands, can significantly increase the SOC stock (0–30 cm) (Birch15–Birch50), which is in accordance with Icelandic studies given in the literature.

During the period between Birch15 and Birch50 (35 years), the sequestration rate is 0.42 t C ha-1 a-1 on average, without taking the SOC stock of Barren Land (as the status before afforestation begins) as a reference for calculation. The reason for this is given in the assumption that Barren Land contains a lot of SOC which does not originate from the revegetation process. The rate is lower than the given removal factor of 0.51 t C ha-1 a-1 for afforestation activities (Hellsing et al., 2016).

A literature review revealed that the succession of already vegetated heathland to birch woodland in eastern Iceland shows no change in C stocks (Ritter, 2007). The SOC stocks were about 40 t C ha-1 (0–20 cm) for 26- and 97-year old birch stands. Snorrason et al. (2002) found a higher SOC stock (0–30 cm) in a 54-year old birch stand (65 t C ha-1) compared to that of grassland (54 t C ha-1) at Gunnarsholt, which leads to the assumption that the effect of afforestation is more effective than that of revegetation concerning SOC sequestration. However, Snorrason et al. (2002) and Ritter (2007) did not report the initial SOC stock before the ecosystem change began. Soil development and natural vegetation succession on moraine till after glacial retreat are other typical processes of land-cover change in Iceland. Vilmundardóttir et al. (2015) found a SOC accumulation within the top 20 cm from 0.9 (initial status) and 13.5 t C ha-1 at sites with a maximum age of 120 years, thereby demonstrating that the process of vegetation succession on moraine till leads to an increase in soil carbon stock. Our results indicate that the change in SOC stocks during afforestation with mountain birch on severely degraded soils (Fig. 2) is comparable with those given for shrub encroachment in the cited literature.

Restoration by revegetation is another process of land-cover change in Iceland (Aradóttir et al., 2000; Arnalds et al., 2013). The present study compared the effects of afforestation and revegetation. Within 50 years, the revegetated sites (Grass50), which were restored by fertilizer and grass seeds, showed a median SOC stock 9 t C ha-1 higher compared to the soils of Birch50. This leads to the assumption that revegetation of severely degraded soils enhances the SOC stock (0–30 cm) more effectively than afforestation (Fig. 2). Aradóttir et al. (2000) and Snorrason et al. (2002) also studied revegetated grassland sites near Gunnarsholt showing a site history comparable to the grassland sites of the present study. Accordingly, Snorrason et al. (2002) reported a SOC stock (0–30 cm) of 54 t C ha-1, which is comparable with the SOC stocks of the present study. However, Aradóttir et al. (2000) found 28 t C ha-1 (0–20 cm) for a 46-year old grassland site, compared to the median value of 34 t C ha-1 (0–20 cm) for the present grassland sites.

The results of the SOC stocks within the commonly used soil depths of 30 cm of the tested categories indicate that the SOC pool decreases between Barren Land and 15-year old birch stands. It then increases during birch establishment to reach the level of naturally grown birch woodlands (Fig. 2). This pattern is comparable with other field studies (e.g. Goulden et al., 2011; Hunziker et al., 2017). However, the analysis of mineral SOC dynamics in such a temporally dynamic landscape, which results in unequal SOC and volcanic clay concentration patterns across the tested categories, calls for more detailed and alternative methods (Tables 1, 2). Thus, the present study further focused on the vertical distribution of the SOC and its quality to verify whether afforestation results in the soil becoming a C source and whether more C is sequestered during revegetation than afforestation.

Clay minerals found in volcanic soils, such as those found in Iceland, may play a key role in stabilizing soil organic carbon due to their amorphism, high degree of hydration, extensive specific surface area (200–1500 m2 g-1), and pH-dependent charge and the high reactivity (Torn et al., 1997; Basile-Doelsch et al., 2007; McDaniel et al., 2012; Arnalds, 2015a). The major stabilization mechanisms are either the formation of allophane– or ferrihydrite–humus complexes, which is favoured at pH >5.0, or the building of metal–humus complexes which are more effective at pH values lower than 5.0 (Arnalds, 2015a). The Alpyr:Alox and Fepyr:Feox ratios (Table 1) are used as an indicator of the occurrence of metal–humus complexes. The higher the ratio, the more clay minerals are bound to organic compounds, which suggests an increase in the SOC stabilization. In order to discuss stabilization processes of SOC found in the “<63 µm” fraction' with mineral clays (Sect. 3.3), we therefore considered the SOC concentration of the “<63 µm” fraction in Fig. 5.

In general, the results showed a decline of the Alpyr:Alox and Fepyr:Alox ratios with soil depth for all tested categories (Table 2). Further, Birchnat and Birch50 showed the highest Alpyr:Alox and Fepyr:Feox ratios, while at Barren Land and Grass50, the lowest ratios were found (Table 2). The present study found a strong positive (r=0.68, p value < 0.001) correlation between the allophane concentration and the pH value and a strong negative (r=-0.77, p value < 0.001) correlation between the (Alpyr+Fepyr):(Alox+Feox) ratio and the pH value, respectively (Fig. 5a, b). These findings are in accordance with Arnalds (2015c). The allophane concentrations are highest in the soils which were unvegetated (Barren Land). On the other hand, the concentrations of Al and Fe bound to metal–humus complexes were highest in the top sampling intervals of sites with the longest vegetation covers (Birchnat, Birch50 and Birch25; dotted circle). Both correlations indicate a possible influence of the different stages of vegetation cover on the amounts of allophane and the ratio (Alpyr+Fepyr):(Alox+Feox), respectively. This can be explained by the increase in protons resulting from vegetation processes in the soil, which leads to acidification and simultaneously a lowering of the pH value. Hence, the establishment of vegetation favours the formation of metal–humus complexes (Arnalds, 2008, 2015c).

The scatterplots comparing the allophane concentrations with the unfractionated SOC concentrations, as well as the “<63 µm” SOC concentrations, show no clear trends as most of the samples contained <4 % of unfractionated SOC or “<63 µm” SOC. The highest SOC concentrations were found in the upper sampling intervals at Birch25, Birch50 and Birchnat (dotted circles). However, the allophane content is lowest in these cases (dotted circle), which may be attributed to the fact that soil weathering and the formation of clay minerals takes longer than the allocation of soil organic carbon during birch growth. Regarding SOC sequestration during the reclamation of severely degraded land and soils, soil material of eroded and capped soil profiles most likely already passed through weathering processes and therefore contained a high amount of clay minerals. The fresh SOC originating from reclamation activities can be stabilized by the already existing clay minerals like allophane (Table 2, Fig. 5) and the carbon sequestration potential of these eroded soils may be relatively high (Arnalds et al., 2000; Ágústsdóttir, 2004).

The stabilization of the SOC in the form of metal–humus complexes seems to be hampered due to the relatively high measured pH values (Table 1), which were higher than the upper threshold value of 5.0 for the building of metal–humus complexes given in the literature (Fig. 5d, f). The pattern that the uppermost sampling intervals of the vegetated sites (dotted circle) are decoupled from the nested scatters was also observed in the relationship between the selected SOC pools (SOC concentration, “<63 µm” SOC concentration) and the organo-mineral complexes (Fig. 5d, f). The decoupled nested scatter shows an almost strong positive relationship (r=0.69, r2=0.48, p<0.05) between the “<63 µm” SOC concentration and the (Fepyr+Alpyr):(Feox+Alox) molar ratio (Fig. 5f). This observation of relatively high “<63 µm” SOC concentrations as well as relatively high (Fepyr+Alpyr):(Feox+Alox) molar ratio values indicates that the SOC in the “<63 µm” fraction might be sequestered as organo-mineral complexes in the upper sampling intervals at Birch25, Birch50 and Birchnat. In such cases, the formation of metal–humus complexes might comprise a reasonable stabilization process of the SOC in the “<63 µm” fraction (Fig. 5f). The regression analysis between the SOC and volcanic minerals indicates that during afforestation, the organic carbon is preferably stabilized in metal–humus complexes. It implies that this process starts to be an effective stabilization process for total SOC and “<63 µm” SOC after 20 years of birch growth and can occur in deeper sampling intervals in older birch stands.