The study was conducted within a 15 × 75 km area (approximately at 62.5°–63.1°N, 17.9°–18.7°E, DD) in the High Coast Region situated in Västernorrland County located in the southern boreal zone of Sweden (Fig. 1). The High Coast Region on the east coast of Sweden along the Baltic Sea is characterized by a rugged and steep terrain. The area was covered with ice during the latest ice age and after the ice had melted the area was under sea level for millennia; only the highest hill tops > 285 m.a.s.l. could be seen as small islands (Lundkvist 1986). Most of the plots are situated on hilltops and the altitude ranges from approximately 50 to 230 m, i.e. below the highest coastline. Mean annual temperature in the region is around 3 ° C, mean annual precipitation is 800 mm, the length of the growing season (number of days with temperature > 5° C) is approximately 150 days. The average maximum snow depth is 70 cm and the snow covers the ground 150–175 days every year (SMHI 2016). Norway spruce (Picea abies (L.) Karst.) and Scots pine (Pinus sylvestris) are the two dominant tree species in the study area with Scots pine dominating on dry and nutrient poor sites. A general inventory of 26 sites with rocky pine forests was conducted in 2011 in the study area by the County Administration in Västernorrland with the aim to map core areas with high conservation values, e.g. presence of big and old trees, diverse forest structure and abundant dead wood (Salomonsson and Bader 2015). Eight of these 26 study sites with the highest ranking (high conservation values) were selected for this study.

In August 2014 we established three band transects at each site, resulting in total in 24 transects. We used band transects since they sample spatial heterogeneity better than circular plots. The studied forests are highly heterogeneous and exposed bedrock occurs mixed with parts of more closed forest. Transects were 10 × 100 m for living trees, but for dead wood sampling we expanded plots to 20 × 100 m to ensure a sufficient number of sampled dead wood units.

To avoid subjectivity in the placement of transects we randomized their placement by using a numbered grid that was placed over a map of the core area at each site and starting points as well as the direction (N, NE, E, SE, S, SW, W or NW) were randomly assigned, but with a minimum distance of 100 m. Within each transect, we recorded × and Y coordinates, species and diameter at breast height (DBH) for all living (stems ≥ 1.3 m high) trees. To describe the spatial distribution pattern, we calculated the distance from each tree to the nearest tree in the × direction and then computed the variance to mean ratio of these distance, i.e. a one-dimensional spatial analysis (Horne and Schneider 1995). We then used a χ²-test to test whether the distribution of trees where random or significantly aggregated. We extracted two increment cores at 30 cm height from each living tree (DBH ≥ 10 cm) using a 6 mm increment borer (Haglöfs, Mora, Sweden). We measured the height of the tallest tree along the transect with a Laser Height Meter (Nikon Forestry 550) at 0 m, 50 m and 100 m.

All snags and logs with a maximum diameter of ≥ 10cm and with their base inside a transect were included. We recorded both base- and top diameters, DBH, length and decay class. We used a four class system for decaying wood: 1 Hard dead wood – The volume of the stem consisting of > 90% hard wood, hard surface, very little impact from wood-decaying organisms; 2 Partly decayed wood – 10–25% of the stems’ volume consists of soft wood, a sheath knife goes through the surface but not through the whole sapwood; 3 Decayed wood – 26–75% of the stems volume consists of soft to very soft wood; 4 Substantially decayed wood – 75–100% of the stems volume consists of soft-very soft wood, a sheath knife can penetrate the whole stem, but a hard core can occur (Esseen et al. 2003). We collected samples for dead wood for dating with a chainsaw, taking only small samples (cookies) rather than whole cross-sections when possible. The dead wood samples were used for age determination and calculating time since death. Naturally created high stumps were recorded as snags. We used the conic-paraboloid formula (Fraver et al. 2007) to calculate the volume of dead wood, a formula that has a greater precision and lower bias than more commonly used formula. However, due to the shape of the trees with many branches and bent, crooked stems, the volume calculations are rough estimates rather than exact values. Cut stumps were not included in dead wood volume calculations but we noted diameter and species and took samples for dating to quantify past harvesting.

In autumn (September-November) 2015 we carried out a comprehensive search of each site for fire scars in stumps, living and dead trees. Two persons visited each site one whole day, resulting in approximately 10 search hours at each site covering the core area and the adjacent forest. Every tree with a fire scar was mapped and samples were later collected with a chainsaw. We only took partial sections whenever possible to avoid unnecessary damage. A total of 52 fire-scarred wood samples were collected and dated from eight sites. An additional six samples from dead wood sampling in 2014 and 3 samples from Skuleskogen, sampled in 2010, also had fire scars and were included, resulting in a total of 61 dated samples. More than half of the samples (33) contained scars from repeated fires and the maximum number of fires in one sample was four.

We mounted cores from living trees and cross-sections from dead trees and sanded them with increasingly fine grain size until a fine polish was achieved (down to grain size 400 for all samples and to 600 for some samples) using standard methods (Speer 2010). Ring widths and number of rings (age) were measured using a scanner and the image analyzing software WINDENDRO (version 2014). We used Applequist (1958) pith locator in samples that had missing rings in the core. Age was determined by counting every year ring to the pith. A few samples had a rotten core and we then used the mean growth rate for each site and the DBH to estimate the age. As a measure of growth at each site, we used the average ring width from each tree and its variability for all trees at each site. Samples from living trees with high age and several samples from dead wood that were visually dated with high accuracy were selected to create a master chronology. We used the software program COFECHA (Holmes 1983; Grissino-Mayer 2001) to cross-date our samples and suggested years were always visually double-checked to make sure that distinct marker years (with typical late-wood features) were dated correctly. The program ARSTAN was then used to create a standardized master chronology and tree ring series with questionable dating were eliminated before standardization (Cook and Krusic 2005). The master chronology from living and some dead wood samples resulted in 130 samples covering 818 years (1197–2014) with a mean sensitivity of 0.30 and a relatively strong inter-correlation of 0.515. This master chronology was used to date samples from dead wood and as dead wood samples were dated, they were added to the master chronology. We were able to date 70% of all dead wood samples; 30% of the logs were too decomposed to allow sampling or difficult to date. The dated samples belonged to all decay classes but only a few samples in decay class four were possible to sample and date. The master chronology was also used to date fire scars and because the growth is related to precipitation during the main fire season in June, we noted the average ring width for all the fire years. The final chronology (see Appendix 1) with living as well as dead wood samples with and without fire scars resulted in 248 samples covering 819 years (1197–2015) with a mean sensitivity of 0.29 and an inter-correlation of 0.505 and with a minimum of ten samples, beginning with the year 1431. We also calculated time since death for the dead wood; however, we cannot rule out the possibility that we have overestimated this value as in some cases the outermost parts might have eroded. The computer program FHAES (Fire History Analysis and Exploration System, version 2.0) was used for structuring and analyzing fire data. FHAES was also used to produce fire history graphs.

Very little written information is available that is relevant for the specific historical use of the rocky pine forests. The general history of the region is, however, relatively well documented (e.g. Lundkvist 1986; Baudou 1995) . Based on findings of old building structures in stone and grave cairns, it is clear that some places in the area have been populated by humans for at least 3,000 years (Baudou 1995). The few early settlers based their livelihood mostly on fishing and hunting (e.g. seal along the coastline) and fishing has been the main income in the region for a long time (Lundkvist 1986). The possibility for mining was explored during the 1700´s at several places along the High Coast Region, e.g. at southern Ulvön and Hemsön (sites SU and PB), and, although it never became an industrial use, the attempts might have impacted the nearby forests to some extent (Andersson 1975; Lindh 1991). When the first large-scale wave of industrial usage of the forests started during the 1800´s, it is unlikely that the nutrient-poor pine forests with rather small trees on the hill-tops along the High Coast also were exploited. In a map of Vårdkallberget from 1851 it is possible to read (in Swedish) that the area close to the hill-top where the rocky pine forests occur is named “Näs by skogsmark” and “Myre by skogsmark” (Renström and Hedström 1985). This means that this area was classified as a common forest resource, free to use for the village members, but to what extent it was actually used is not known. Close to one (VB) of the eight sites, remnants of two old tar pits have been found (Fig. 2). The pits are located just along the coastline and it is unclear whether the wood for the tar was collected nearby or from the pine forest in the hill-top (the hill-top is located approximately 500 m from the tar pit remnants). The area around VB was not populated before 1780 but nearby settlers used the area mainly for fishing even before the settlements (Renström and Hedström 1985; Lundkvist 1986) . We only found a few harvested stumps on our plots, at one plot at GA and in two neighboring plots at the SU site, indicating very limited forestry activities. We found nine stumps at GA and ten stumps at SU, most of them heavily eroded, and we were only able to date six stumps. The outermost year ring from the stumps at SU was dated to 1736, 1827 and 1886 and for GA 1752, 1813 and 1828, corresponding to 200 years since harvest on average.

Figure 2.

Old map from one rocky pine site. Map of Vårdkallberget (VB) from 1851, which shows the classification of the nutrient poor rocky pine forests around the hill-top (marked with a triangle) as “general forest land” and the nearby settlement (marked with a plus sign). Two remnants of tar pits are marked with a star. The widest distance from east to west is approximately 2 km.

Scots pine was the dominating tree species in rocky pine forests and stands were very scattered and open; they had very little mineral soil and the ground was interspersed with bare rock and with a vegetation dominated by lichens and dwarf shrubs (Fig. 3). A general pattern was that the forest became even less dense and increasingly rocky higher up on the mountain. Trees had different sizes and they grow upright but also strongly leaning, some even grow horizontally despite being healthy and alive.

Figure 3.

General description of rocky forests in the High Coast Area. Photos from Fanön (FA), Gårdberget (GA) and S. Ulvön (SU), which show the topography, typical characteristics in rocky pine forests with dispersed trees and dead wood in various sizes in rocky terrain, spruce interspersed at concave surfaces and a fire-scarred pine.

The general stand structure at the eight sites was characterized by sparsely distributed trees with a mean basal area of approximately 9 m² ha^-1 and a density of 640 trees ha^-1 on average (Table 1). Variance to mean ratio showed that trees both had an aggregated and a random distribution pattern, depending on site (Table 1). The forests mainly consisted of Scots pine, which constituted approximately 90% of all the trees. The second most common tree species was Norway spruce, which constituted 7% on average. Deciduous trees were usually present but in very low density (Table 1). Less than 4% of the trees were deciduous with birch (Betula spp.) as the most common species, while aspen (Populus tremula L.) and rowan (Sorbus aucuparia L.) occurred only sporadically. The average maximum tree height for pine was 7.5 m and varied between 5.5 and 10.1 m at the different sites (Table 1). The average age for living pines varied between 147 and 194 with a mean of 167.5 years and the maximum age was even more varied with the oldest living trees ranging from 276 to 418 years (Table 1) at the eight different sites. The average ring width of living pines varied from 0.52 to 0.68 mm yr^-1 between sites (x‒ = 0.55), indicating very low growth rate (Table 2). In addition, there was a considerable within-site variability in tree-ring width, indicating a large variation in growth at the stand level (Table 2).

Table 1.Download as CSV 

General stand structure. Stand structure data for eight rocky Pinus sylvestris dominated forests in the High Coast Region in Northern Sweden. Values are means with SE in parenthesis and based on three band transects per site. Site acronyms as in Figure 1.

Study site	GB	PB	VB	FA	GA	SU	SK	HU
Dbh (cm), Trees ≥ 1.3 m height	11.4 (0.6)	9.8 (1.2)	11.2 (0.8)	11.3 (2.8)	11.1 (1.0)	12.2 (0.7)	12.6 (0.3)	9.3 (0.6)
Basal area (m2 ha-1), Trees ≥ 1.3 m height	11.0 (1.9)	8.8 (0.9)	10.2 (1.1)	8.1 (2.4)	10.2 (1.9)	6.3 (0.8)	10.1 (2.1)	8.8 (0.9)
Maximum height, Pine (m), average	10.1 (0.67)	7.0 (0.27)	6.6 (0.23)	8.4 (0.42)	8.4 (0.45)	6.2 (0.05)	8.0 (0.67)	5.5 (0.25)
Age, Pine (yr), average ≥ 10 cm DBH	194 (16)	157 (10)	182 (10)	147 (4)	157 (15)	189 (21)	156 (30)	158 (24)
Age, Pine (yr), maximum	334	299	372	276	418	403	442	376
No. of living trees ha-1 ≥ 1.3 m height	647 (94)	713 (91)	636 (76)	563 (39)	753 (234)	363 (19)	593 (118)	783 (72)
No. of living trees ha-1 ≥ 10 cm DBH	280 (70)	300 (15)	297 (32)	300 (60)	323 (33)	220 (6)	303 (61)	363 (35)
Spatial distribution (variance to mean ratio, VMR)	1.9*	1.1	1.2	3.3*	1.2	1.7*	1.7*	1.0
Pine, share (%)	77 (15)	95 (3)	95 (1.4)	78 (6)	88 (11)	92.5 (4)	94 (1.3)	98 (2.3)
Spruce, share (%)	20 (13)	4 (2.2)	5 (1.4)	4 (0.2)	9 (7)	6 (4)	4 (0.8)	0 (0)
Deciduous, share (%)	3 (1.9)	1 (1.1)	0 (0)	18 (5.8)	3 (3.4)	1.5 (1.5)	2 (1.1)	2 (2.2)
Dead wood volume (m3 ha-1) Standing	4.8 (1.09)	1.0 (0.49)	1.7 (0.31)	1.9 (0.52)	2.6 (1.30)	1.9 (0.89)	1.5 (0.76)	0.6 (0.21)
Downed	3.3 (1.56)	1.7 (0.29)	1.6 (0.30)	3.3 (1.14)	2.0 (0.15)	2.0 (0.99)	4.0 (1.44)	1.2 (0.52)
Proportion dead wood, basal area (%)	18.9 (1.91)	13.5 (1.69)	16.7 (2.88)	21.7 (5.64)	16.7 (2.64)	26.1 (6.00)	20.3 (8.47)	9.9 (2.61)

*Significantly aggregated

Table 2.Download as CSV 

Tree-ring width based growth. The growth of living Scots pines measured as average annual tree-ring width in eight rocky Pinus sylvestris dominated forests in the High Coast Region in Northern Sweden. Values are means and SD from all trees > 10 cm DBH at the sites. Site acronyms as in Figure 1.

Study site	GB	PB	VB	FA	GA	SU	SK	HU
Growth (mm yr-1)	0.53	0.53	0.58	0.68	0.56	0.48	0.55	0.52
SD	0.21	0.20	0.29	0.38	0.26	0.25	0.24	0.19
n	82	85	84	70	83	59	49	89

The tree diameter distribution was positively skewed with small trees (DBH ≤ 10 cm), dominating and with decreasing frequency of trees with increasing size (Fig. 4a). However, even though the large trees were rare, they did exist and there were on average one pine ha^-1 larger than 40 cm in DBH and 15 pines larger than 30 cm ha^-1 in the studied forests. For trees larger than 10 cm in DBH, medium age classes (75–200 yr) were most common for both pine and spruce (Fig. 4b). Pines were on average 165 years (SD = 75, n = 624) and spruce 144 years (SD = 47, n = 52).

Figure 4.

Categories of tree size and tree age. Summary of the distributions of diameter at breast height (DBH) (a) and age (b) (> 10 cm DBH) for living trees of pine and spruce at the eight sites combined (with mean and SD, n = 8).

All ages are present for pine but really old spruces were lacking. The oldest spruces were a maximum of 275 years old whereas pines were older; approximately 1 tree ha^-1 was above 400 years old and 13 trees ha^-1 were more than 300 years old. The correlation between age and DBH was generally low (R² = 0.31 for pine and R² = 0.11 for spruce) but due to the large sample size significant for both tree species (p < 0.001 and p < 0.05 respectively; Fig. 5). The oldest pine (418 years) was also the largest (48 cm DBH), while for spruce the oldest tree (259 years) was only 11 cm at breast height (Fig. 5). The age span was wide across all diameter classes; for example, the age span of trees with a DBH of 10 cm varied from 28 to 282 years and from 57 to 216 years for pine and spruce, respectively.

Figure 5.

Relationship between size and diameter at breast height (DBH) for pine (a, n = 624) and spruce (b, n = 52). Only trees with a DBH larger than 10 cm are included.

The amount of dead pine wood was low (average of 4.5 m³ ha^-1, range: 2–8, SD = 2.0, n=8) and varied between sites, but both logs and snags were present at all sites (Table 1, Fig. 6a). On average, there were 53 pine DW units ha^-1 (SD = 12.8, n=8). DW in all decay stages were present and with decay stage 2 as the most common. However, at three sites dead trees in decay stage 1 were more common (FA, GA, HU; Fig. 6b). The average decay class for all dead wood was 2.45 and for dated dead wood samples 1.92. The difference stems from lack of datable samples from dead wood in decay class four. The basal area of DW in relation to total basal area, including living trees, varied between 10–26% at the different sites with an average of 18%. Most dead wood had died fairly recently; almost half of the dated dead wood samples had died during the last 50 years (Fig. 7).

A majority of the DW, 88%, had died during the last 200 years, which corresponds to a DW addition of approximately 2 trees per ha^-1 and decade since the beginning of the 1800s. However, there were clear signs that dead wood can remain for several hundred years. The DW did not totally decay even in cases when the logs had been dead more than 300 years and we even found DW that had been dead more than 500 years (Fig. 7). Snags constituted a higher proportion of newly dead wood whereas logs were the most common DW type for units older than 350 years, but important to note is that there were also really old snags present. For DW that had been dead between 100–300 years the two types are equally common (Fig. 7). The average time since death for all dated dead wood units was 106 years (SD = 129, n=220), for snags 92 years (SD = 151, n=109) and for logs 120 years (SD = 129, n=111). The average age for the dated DW was 192 (SD = 82, n=220).

Figure 6.

Number of logs and snags from pine in rocky pine forests at the eight different sites (SE, n = 3) as well as the total mean (SD, n = 8). The abundance of DW is classified as standing (Snag) or horizontal (Log) (a) and also classified in decay stages (b).

Figure 7.

Time since death for dead wood for snags and logs.

Most signs of fire were found in FA where 18 fire years were detected from 12 collected fire-scarred samples (Fig. 8a). FA was also the only site with individual trees that had experienced at least four fires. On average, 1.8 fires per sample from all sites was detected but on FA where most fires occurred, we detected 2.4 fires per sample. We did not find any fire-scarred trees at GB and only 3 fires could be detected at PB. On average, eight fire events per site were detected covering a time period from 1235 to 1923 but the number of detected fires varied between sites (Fig. 8a and Table 2). The period with most fires was during the 1600s and the 1800s when 13 fires occurred across all eight sites (Fig. 8b). We found signs of 47 fires between 1500 and 1900 and we detected very few fires earlier than 1500 and later than 1900. The earliest detected fire occurred in 1235 but as with most dated fires, this fire was only detected at one site (Fig. 8c). However, one of the fire years, 1693, was found at five sites. The fire in 1693 was found in PB, FA, SU, SK and HU but not in GA and VA which are located close to FA, SK and HU (Table 2). GA and FA are situated very close together but the only fire year that is common for the two sites is 1563. Approximately half of the fire-scarred trees (45%) had only one fire scar and trees with most scars (repeated fires) had four scars and was found at FA (see Appendix 2 for detailed information on samples with fire scars). In addition, most of the dated fire years was only detected in one sample (64%). The average fire interval varied between sites (shortest: FA 20.5 years and longest: VB: 65.5 years), but with an average fire interval of 42 years (excluding GB, where no signs of fire were detected) (Appendix 3). We also noted that many fire years had smaller year rings (indicating a dry fire season) than the average sizes in the master chronology. Twenty-eight fires had significantly smaller year rings than average, 25 had average size while only four had unusually large rings (Appendix 4).

Figure 8.

Fires in rocky pine forests. The total number of fires at the eight sites and the average from the sites (a). All detected fires are included, regardless of the number of samples the fire was detected in. Total number of fires are also separated into centuries (b) and individual fire years and their occurrence in the number of areas, with fire years occurring at more than one site highlighted (c).

The studied rocky pine forests are characterized by a very open structure. The low density of trees can partly be explained by the scarcity of water and nutrients, as low productivity is generally connected with low basal area and openness (Liira et al. 2007). The rocky pine forests also have an uneven distribution of tree species, size and age. This creates a heterogenic and diverse environment and a wide age distribution typical for old pine forests (Kuuluvainen et al. 2002). Attributes that usually are associated with old-growth forests are present, such as several canopy layers, high variation in tree sizes and ages and high spatial heterogeneity of tree distribution (irregular distribution of gaps) (Bauhus et al. 2009).

Although the diameter distribution pattern in our study is descending with size, the pattern is not a typical reversed J-shaped curve, common in many undisturbed forest stands (e.g. Parker and Peet 1984; Linder et al. 1997; Kuuluvainen et al. 2002; Rouvinen and Kuuluvainen 2005). In a study by Lilja and Kuuluvainen (2005), three different semi-natural (old-growth) pine stands had all a descending pattern, one that showed a reversed J-shaped pattern and two stands that were more similar to our findings. Rouvinen and Kuuluvainen (2005) found a bimodal pattern in a managed pine stand but a descending pattern, but no sharp reversed J-curve, in a natural pine stand. In a sub-xeric old-growth stand, dominated by pine, Uotila et al. (2001) found a pattern with most trees in intermediate diameter classes and not a clear descending pattern. Since we have no data on regeneration (trees < 1.3 m), we have no clear explanation to why we did not find a clear reversed J-shape on the diameter distribution in our study. It could be that the small trees are less frequent because there are only a few places/spots where trees can establish or it could be influenced by moose (Alces alces), which browses on small trees in winter. Neither did we find a bimodal diameter distribution pattern, often seen in forests with repeated disturbances, such as fires (e.g. Zenner 2005). This could indicate that the fires in this type of forests have been low-intensity fires of small sizes, as discussed below. In addition, the lack of bimodal diameter distribution pattern also could be an effect of lack of fires during the last 150 years.

As well as being a sign of naturalness, the heterogeneity and patchiness of tree distribution may also be influenced by an uneven water and nutrient availability for the trees. A tree that grows where the ground consists of bare rock has a very different potential for growth than trees in a concave patch covered with mineral soil, which probably can explain why the trees grew spatially irregularly at four of the sites. The trees generally have a low growth and high variability between trees at all of the sites, indicating also heterogeneity in growth. The high variability in growth between trees clearly contributes to the limited correlation between age and DBH. Kuuluvainen et al. (2002) found a stronger correlation between age and size; their R² was 0.58 for pine. However, they had a wider range of stand types included in their study and not only nutrient poor pine forest, which could explain the higher predictability. Nevertheless, they also found a lower correlation for spruce (R²=0.36) than for pine, similar to our findings.

The average maximum tree height is generally low in rocky pine forests. The maximum tree height is related to height growth when the tree is young (Ryan and Yoder 1997), which in turn is influenced by genetic and environmental factors (Junttila 1986; Kaya et al. 1999; Pensa et al. 2005). Tree height seems to depend on a combination of temperature, precipitation and nutrient availability (Jansons et al. 2015) and the environment in the rocky pine forests is characterized by low water and nutrient availability and can probably partly explain the low mean maximum tree height. Other factors that might influence tree height in the rocky pine forests are wind-exposure and openness (Tomczak et al. 2014).

The presence of old trees is one of the key features of natural forests (Östlund et al. 1997; Andersson and Östlund 2004). Old trees are relatively common in the studied rocky pine forests and we found approximately 13 pine trees ha^-1 older than 300 years and as many as 70 pines ha^-1 older than 200 years. As a comparison, Edwards and Mason (2006) found approximately 20 and 60 trees ha^-1 older than 200 years in two different old-growth pine stands in Scotland. In another Scottish old (and open) forest, only about 6 trees ha^-1 older than 250 years were found (Summers et al. 2008). This is in line with the findings of Kuuluvainen et al. (2002); they found approximately 5–10 trees ha^-1 older than 250 years in an old-growth pine forest. The oldest living pines in our plots were 418 years but we know that living trees older than 600 years (data not shown) existed in the studied area. In line with our findings, in old-growth forests in Finland, ages of more than 500 years for pine and almost 300 years for spruce were found (Kuuluvainen et al. 2002). In a study from a similar area, Hornslandet, approximately 200 km south from our study area, the authors actively searched for very old pine trees and found the oldest recorded pine in Sweden with an age of more than 750 years (Andersson and Niklasson 2004).

There are not many deciduous trees in the rocky pine forests and although birches (the most common deciduous species in the area) can tolerate rather dry conditions (Sutinen et al. 2002), regular drought may well limit the abundance of deciduous trees. In addition, it is likely that the density of deciduous trees is limited by the presence of moose (Alces alces L.), which can reduce the abundance of several deciduous species (e.g. rowan and aspen) when present (Edenius et al. 2002). The abundance of moose has since the 1970s massively increased in northern Sweden (Cederlund and Bergström 1996) and moose is today abundant throughout the whole of Sweden with the highest densities in the world (Skogforsk 2016). Moose is highly abundant also in the rocky pine forests in the High Coast Region and likely have a negative impact on the abundance of deciduous trees.

The CWD volume is very low and corresponding to levels in managed forest (Jonsson et al. 2016; Ylisirniö et al. 2012) and at a much lower level than what usually is common in natural forests with no or little human impact (e.g. Rouvinen and Kuuluvainen 2001; Karjalainen and Kuuluvainen 2002; Koster et al. 2005; Shorohova and Kapitsa 2015). Given the low tree density it is, however, more relevant to compare the proportion of dead trees to living trees. In the rocky pine forest the CWD share of the total basal area (living and all dead wood combined) is 18% on average, which is in line with several other pine-dominated forests with old-growth characteristics (e.g. Sippola et al. 1998; Siitonen et al. 2000; Rouvinen and Kuuluvainen 2001). By contrast, Karjalainen and Kuuluvainen (2002) found a higher proportion (32%) of the volume of CWD in forest stands on dry soils. The variation in CWD:live wood volumes is greater in pine dominated forests than in spruce forests (Shorohova and Kapitsa 2015). We found a slightly lower proportion of dead wood in this study, but it should be noted that the ratio between dead and living trees is driven primarily by tree life span (mortality rate) and decay rates. Even though decay rates are slow in the studied rocky pine forests, it is to be noted that tree life span is likewise very long.

Charcoal and tar production has been common in the area a long time and the CWD quality in rocky pine forests is particularly suitable for tar production. Situated < 1 km from our sampling sites, two old remnants of tar pits have been found at one site, but there were no reported signs of old tar pits at the other seven sites. Hence, we cannot totally rule out the possibility that some CWD has been used for char and/or tar production in the studied area because these activities tend not to leave any visible traces behind.

The most common decay stage in this study is stage 2, which is in line with other pine forests with a high degree of naturalness (Karjalainen and Kuuluvainen 2002). The late decay stages were not as common in the study area as the most decayed wood (decay stage 4) constituted approximately only 12% of the total amount of CWD. In other studies of pine forests with high degree of naturalness the distribution of CWD in decay classes was more even (Rouvinen and Kuuluvainen 2001) or later decay classes were also more common (Sippola et al. 1998). One explanation for the relatively low presence of CWD in late decay stage found in this study could be reoccurring small fires because fire consumes logs in later decay stages to a higher extent than logs in earlier stages (Eriksson et al. 2013). It seems that CWD is accumulating slowly and can remain for very long periods in dry conditions. This makes every CWD unit very valuable and even a small outtake can have a large impact on the CWD-depending organisms. The slow decay rate also explains why there is really old CWD; some has been dead for 300–500 years, in the rocky pine forest. In addition, we were not able to sample and date the most decayed dead wood, so the result is probably an underestimation of the amount of very old CWD.

Only two of the sites had signs of past cutting. A low frequency of manmade stumps indicates a high degree of naturalness (Uotila et al. 2001; Rouvinen et al. 2005). The outermost year rings on the six dated stumps varied between 1736 and 1886. This suggests that harvesting took place a very long time ago, although not necessarily giving the exact year of harvest since all stumps lacked bark and some erosion of the outer parts cannot be excluded. Consequently, the years should be regarded as the earliest possible cutting year but the actual cutting could have happened later. Nevertheless, there is evidence that the limited cuttings that did occur happened during the 1800s, or even earlier, and with no recent harvesting during the last 100 years. It is unlikely that rocky pine forests were targeted by industrial forestry since trees with large diameter are rare and the accessibility is very low. However, it is possible that settlers used the pine forests at small-scale for certain construction details, e.g. window frames, for which dense high quality wood was preferred. Such limited harvest/extraction is unlikely to leave any long-lasting visible traces.

Surprisingly, many fires were detected in the rocky pine forests. There is not much fuel on the ground in the studied rocky pine forests, bare rock is common and trees are scattered. On the other hand, the ground surface easily becomes very dry and vegetated areas are mainly composed of reindeer lichens and dry mosses, which together with scattered dwarf shrubs potentially could carry a ground fire during dry years. Not many fires were detected before 1500. This could partly be a sampling artefact due to the limited number of fire scarred old trees and snags, but different climate and lower human population size cannot be ruled out as explanations for lower fire frequency in the beginning of our time series. Many fires happened during the 1600s, a pattern also observed by others (Zackrisson 1977; Niklasson and Granström 2000; Wallenius et al. 2004). The year 1693 stands out as a major fire year; a fire year that has also been documented by others (Drobyshev et al. 2014).

There was a population increase after the Black Death plague during the 1600s and several “slash and burn” immigrants from Finland also settled down in northern Sweden during this period, which can be an explanation for more frequent fires. However, most of the ”slash and burn” farmers did not settle down along the coast area where fishing was the main livelihood but rather in the inland areas where the land was more suitable for the ”slash and burn” cultivation technique (Lundkvist 1971). One fire happened in 1721 in Hummelvik nature reserve, a year when Russians made raids where they burned hundreds of farms along the coastline in the studied area (Lundkvist 1971; Lindh 1991). A village just outside Hummelvik nature reserve is mentioned as being burned (Lundkvist 1971), which could potentially explain the fire 1721 since the year ring has an average size. Unusually small year rings have been shown to be closely related to low precipitation in June in the Rocky pine forests in the High Coast Region and fire years are influenced by summer temperature and precipitation to a large extent in the northern region (Drobyshev et al. 2012; Drobyshev et al. 2014).

Most fire years were detected only in a few samples, which indicates that the fires in this area and forest type might have been small in size, but not necessarily rare events. A plausible explanation is that there are plenty of small-scale dispersal barriers for fires in this heterogenic landscape. Both bare rock and wet hollows often occur. Moist depressions, swamps and Picea abies patches often do not burn even when nearby dry patches do (Wallenius et al. 2004). The mean fire interval varied between 20.5 and 65.5 years at the different sites (except of one site with no fires found at all). These intervals included fire years that only were detected in one sample, and hence do not necessarily indicate that the whole forest area (site) burned. It is more likely that the fires were small in this heterogenic environment which creates a small-scale spatial variation influencing fire frequency and size (Wallenius et al. 2004). Many trees also survived the fires and many age classes were present. Taken together, this suggests that the fires were not stand-replacing, had low intensity but happened on a regular basis.