Domesticated Forest Landscapes in Central Scandinavia during the Iron Age: Resource Colonization for Iron and Subsistence Strategies based on Livestock

ABSTRACT This study explores how resource colonization for iron in central Sweden during the early Iron Age may have affected the use of forest landscapes. Slag heap volume at iron production sites was used to estimate the amount of forest resources required for charcoal production. Forest resources required for livestock grazing and fodder were estimated from literature sources. To produce charcoal at iron production sites, forests were harvested, creating conditions suitable for grazing. Production of livestock winter fodder, leaf-hay, became a constraint due to the conflict between grazing grounds and fodder producing areas near main settlements. Although availability of forest was not limiting, a combination of opportunities and constraints is suggested to have promoted a new spatial ordering of land use. This included land closest to the main settlements allocated to fodder production and development of secondary seasonal settlements (shielings) at iron production sites, which could be exploited for livestock grazing.


Introduction
In southern and southeastern Scandinavia, cultural landscapes expanded during the Neolithic, Bronze Age, and Iron Age. From the 4th millennium B.C. onwards, these landscapes were successively transformed by agriculture, crop production, and management of livestock, resulting in an opening of the forest cover for crop fields and extensive areas for grazing (e.g. Berglund 1991;Pedersen and Widgren 2011;Welinder 2011). In contrast, large tracts of land, particularly in central and northern present-day Sweden were only sparsely populated for much of this time period. These forested regions have previously been considered to have remained little affected by people before the early Middle Ages (e.g. Myrdal 2011), i.e. approximately at the time during which the national state of Sweden was formed. However, there is mounting evidence suggesting that the process of transforming forest regions in central Sweden ( Figure 1) actually commenced in the early Iron Age, the first centuries A.D., or even earlier (Hennius 2021). Several studies suggest that people started with livestock herding and occasionally also farming in this region earlier than previously recognized (e.g. Engelmark 1978;Emanuelsson 2001;Emanuelsson et al. 2003;Von Stedingk and Baudou 2006;Karlsson, Emanuelsson, and Segerström 2010;Lindholm, Sandström, and Ekman 2013). Furthermore, iron production expanded into these forest regions, and it has been suggested that they became integrated parts of trade networks ) based on iron (Hyenstrand 1979;Magnusson 1986; National Atlas of Sweden 2011) and other commodities, for example furs (Lindholm and Ljungkvist 2016) and tar (Hennius 2018a). Also, in Norway, farms were established in the forest outland from the early Iron Age (Øye 2005, 2009), often associated with iron production (Solem et al. 2012;Stenvik 2015;Rundberget 2017).
It has been discussed whether this transformation process was initially the result of a colonization in a conventional sense (people moving there) or if it was a transition of management practices by people already inhabiting the forest regions, i.e. cultures based on hunting and fishing. Evidence, for example that based on graves, suggests that it was at least partly a colonization (e.g. Hyenstrand 1979;Welinder 2008;Magnusson and Segerström 2009;Hennius 2020a), although people with different backgrounds had probably co-existed and interacted for a long time (e.g. Welinder 2008). Several authors suggest that these interactions promoted hybrid cultures characterized by cooperation rather than conflict (Welinder 2008;Hansen and Olsen 2014;Amundsen 2017;Hennius 2021).
Three regions have been proposed as "centers of origin" for this colonization (see Figure 1). Hyenstrand (1979) and Magnusson (1986) suggested that the early colonization process emanated from the Mälaren region. Welinder (2008) suggested that the main route of colonization was from the west, originating from the Trøndelag region in Norway (e.g. Stenvik 2015). Another possible route was from the east, i.e. the region along the Bothnian coast (Ramqvist 2012). Baudou (1978) suggested a first phase (to ca. A.D. 500) with colonization from the west, and perhaps also from the Mälaren region, and a second phase from the east. The presumed focal recipients of the trade were "power centers" located in these established agricultural regions. For example, the region around Lake Mälaren was densely populated during the first centuries A.D., developing into what became a regionally dominant power center, Old Uppsala (Beronius Jörpeland et al. 2018).
A basic tenet for this study is that iron production for trade was a key driver of this colonization. A particular focus will be on the province of Jämtland, henceforth considered to include the present-day province of Härjedalen (see Figure 1). Jämtland is appropriate for a closer examination because iron production in this province has been subject to detailed studies (Magnusson 1986). In addition, iron production in Jämtland peaked between the 4th and 7th century A.D., probably earlier than any corresponding peak from other regions in Sweden (Magnusson 2020;Hatlestad, Wehlin, and Lindholm 2021). This was long before the formation of the Scandinavian nation states. When nation states in the Nordic countries appeared later, Jämtland was part of Norway. It was not until peace negotiations in A.D. 1645, after one of the recurring wars in Scandinavia, that Jämtland was transferred to Sweden.
The specific objectives of this study were to examine how iron production and agriculture based on livestock might have influenced the forest landscape in central Sweden (see Figure 1) and how people handled subsistence challenges. The climate in these regions is harsh, and except for some areas, for example along rivers and lakes, most land consists of boreal forest and mires not initially suitable for agriculture. In order to sustain their livelihood, people must have cleared forests to make room for settlements and for livestock grazing, and probably also for crop fields and slash-and-burn cultivation. The study examines how a combination of opportunities and constraints may have promoted a structured and managed forest landscape. The approach estimates the areal extent of the direct and indirect impact of resource colonization and, based on these impacts, derives a hypothetical model for how the early Iron Age landscape of the forest outland was domesticated. The conclusions of this study should be considered tentative, serving as a basis for further exploration of material evidence from the Iron Age in forested regions of central Scandinavia.
Before presenting the quantitative estimates of how the forest landscape may have been affected, it is useful to proceed with a brief overview of background theory and concepts.

Theory and Concepts: Spatial Structuring of Domesticated Forest Landscapes
During the early Iron Age, land use in agricultural regions in Scandinavia was usually organized as infield systems (Eriksson, Arnell, and Lindholm 2021). Infields (Swedish: inägor) were enclosed land (or otherwise protected from livestock grazing) used for settlements, small crop fields, and spatially extensive hay-meadows. Outside the infields lay outland (Swedish: utmark) used for livestock grazing and collection of resources, including leaf-hay and hay from wetlands. Different terms have been used for the outland, e.g. "outlying land" and "outfields," with or without including land used commonly (e.g. Øye 2005). Henceforth, "outland" refers to all land beyond the infields, including land perceived and used communally (e.g. Lindholm, Sandström, and Ekman 2013). Evidence suggests that infield systems with managed meadows were fully established in agricultural regions during a period from the first centuries B.C. to the first centuries A.D., i.e. when iron tools such as scythes became available (Eriksson and Arnell 2017;Eriksson 2020). Iron was thus a key to the establishment of infield systems. The assumption that exploitation of iron was a likely driver behind the resource colonization process in central Scandinavia can be understood in this context.
Several authors have suggested that infield systems (using the term "field-and-meadow system") were introduced with colonization (e.g. Karlsson, Emanuelsson, and Segerström 2010;Lindholm, Sandström, and Ekman 2013). It has also been proposed that the introduction of infield systems was associated with a perception of land ownership related to the infields surrounding settlements (Eriksson and Arnell 2017;Zachrisson 2017). Thus, the colonizers' mind-set may have reflected not only existing early Iron Age agricultural knowledge but also a mentality manifested as an ordered landscape organized around sedentary settlements (Lindholm, Sandström, and Ekman 2013;Hennius 2020a).
Along with colonization, the forest landscapes would have successively become "domesticated" (Terrell et al. 2003;Widgren 2012). Forests were cleared, manipulated, and transformed to satisfy resource demands for iron production and for the subsistence of the iron-producing people. This domestication meant not only a direct transformation of environments but also included how land was organized and perceived by the settlers. It is important to recognize that people did not just respond to harsh environmental conditions; these environmental conditions were also responding to the activities of people. The causation was reciprocal. This is a key feature in the process of niche construction (Odling-Smee, Laland, and Feldman 2003;Eriksson, Arnell, and Lindholm 2021). Niche construction implies that a species (in this case, humans) alters its own niche (environment), which in turn feeds back to the niche-constructing species; causation goes back and forth. This way of interpretation excludes any form of environmental determinism. People and their environment are viewed as interacting agents, acting continuously and reciprocally. Such a reciprocal causation lies behind the concept of domesticated landscapes, ultimately changing the resource base for people (Terrell et al. 2003). In addition, niche construction might also imply that human-environment interactions became more predictable and resilient (Hatlestad, Wehlin, and Lindholm 2021).
In a synthesis of evidence relating to an extension of the scale of outland use, Hennius (2021, 105) suggested that "the intensified outland exploitation could be described as a resource colonization with increased exploitation of a surrounding landscape, aimed at extracting valued components that could be transformed into commodities of crafts and trade." Furthermore, Hennius (2020aHennius ( , 2021 suggested that resource colonization was associated with a fundamental change in outland use, involving the establishment of stable settlements, claimed land beyond the former infields, and regulated seasonal sites. The activities at outland sites show signs of being clustered, like hotspots, including, for example, the cultivation of crops, production of livestock fodder, livestock grazing, iron production, hunting, and tar production (Emanuelsson et al. 2003;Lindholm, Sandström, and Ekman 2013;Lindholm and Ljungkvist 2016;Hennius 2018a). Two of these activities, iron production and livestock management, are the focus of the present study.

Iron Production and Its Impact on Forests: A Background
There are several overviews of early iron production in Sweden (National Atlas of Sweden 2011; Hjärthner-Holdar et al. 2018;Magnusson 2020). The oldest evidence of iron production in Sweden is from the Late Bronze Age, ca. 1200-500 B.C. (Hjärthner-Holdar et al. 2018). The knowledge of how to produce iron is generally believed to have spread from the Near East (Erb-Satullo 2019), reaching Scandinavia from the south via continental Europe (Buchwald 2005). Recent finds of iron production from the first centuries B.C. in northern Fennoscandia (Bennerhag et al. 2021) suggest that there may also have been a northern route for the import of iron production technology. Iron was produced in bloomery furnaces. Lake or bog ore was used as raw material. Bloomery furnaces were used to some extent until the 19th century A.D., i.e. long after blast furnaces had been invented. The oldest known blast furnace in Sweden is from the 12th century A.D. (National Atlas of Sweden 2011).
Charcoal was used for heating the furnaces and for the reduction of iron in the ore. Charcoal was also used for other purposes: roasting of lake and bog ore, refining of iron, and forging. Hennius (2018b) summarized data on the remains of charcoal production in Sweden, and there are records of over 36,000 charcoal kilns (hearths) and almost 10,000 charcoal pits. It is likely that these figures underestimate the true number of charcoal production sites. During the Iron Age, charcoal was mainly produced in pits. It should be noted that early Iron Age bloomery furnaces may have used pine wood as fuel (Solem et al. 2012;Stenvik 2015). Despite its various uses, considering that charcoal pits represent ca. 21% of all recorded charcoal production sites in Sweden (Hennius 2018b) and that conversion factors necessary for the calculations presented below are based on charcoal, it is henceforth assumed that charcoal was used in iron production.
The iron produced from bloomery furnaces is in a solid state, whereas smelting yields slag. Slag heaps surrounded bloomery iron production sites. These slag heaps are useful to estimate the amount of iron produced but will in this study be used to estimate the consumption of charcoal, and with this as a basis, the amount of wood harvested and the resulting impact on the forest landscape ( Figure 2).

Estimating volume of slag
The basic information on iron production in Jämtland comes from the extensive studies by Magnusson (1986). There were ca. 720 iron production sites in the province of Jämtland recorded by Magnusson, but he suggested that the true number is probably around 1000. It was, however, difficult to get access to the full dataset, and, as the majority of iron production sites are included by Magnusson (1986), this was considered sufficient. As the goal here was to suggest the magnitude of the forest area impacted by iron production, and not details for each site, the summary figures 117-119 in Magnusson (1986, 259-262) were used. Magnusson distinguished between two major types of iron production sites: those located close to lakes (shore-bound) and sites not located close to lakes (forest-bound). The shore-bound sites are generally the oldest (the majority are from the 4th-7th century A.D.), while the forestbound sites are younger (the majority being from the 13th-16th century A.D.) (Magnusson 1986, 144-145). This implies that there were two different phases of iron production in Jämtland (Magnusson 2020). Preliminary and unpublished analyses incorporating all data from iron production sites in Jämtland do not alter the conclusion of two different phases with an early peak of iron production (A. Hennius, personal communication 2022), as suggested by Magnusson (2020). The fraction of the earliest iron production sites, the shore-bound, is about 14% of the total. These early shore-bound sites (Figure 3) are the focus of this study, although results from both types are presented below.
At the shore-bound sites, slag heaps are shallow and extend linearly, whereas the forest-bound slag heaps are best described as ordinary heaps, i.e. approximately rounded, with a clear height and diameter, although sometimes extended in one dimension. To estimate volume, different calculations were needed for the two types.
For shore-bound sites, the breath and height of slag heaps were only given ranges: the breadth range is 5-15 m, and the height range is 0.4-0.8 m (Magnusson 1986, 259). Most slag heaps are not longer than 50 m, although single ones may be up to 200 m long. The average length (based on Magnusson 1986, fig. 119) is 43 m. The size distributions of slag heaps are typically skewed (Magnusson 1986, figs. 117-119), so using the midpoint of the breadth and height ranges given above might be misleading. Therefore, the upper limit of the lower quartile of the range was used for these measures, i.e. breadth 7.5 m and height 0.5 m. With a length of 43 m and a breadth of 7.5 m, the basal area is 322.5 m 2 . Assuming a triangular shape of a cross-section perpendicular to the length axis, and using a height of 0.5 m, an average slag heap would contain ca. 81 m 3 slag (322.5 × 0.5/2). Magnusson (1986) remarked that these slag heaps have been subject to erosion, indicating that this figure may be an underestimate.
For the forest-bound sites, the average diameter was 6.1 m (based on Magnusson 1986, fig. 117) and the average height 0.64 m (Magnusson 1986, fig . 118). Assuming a volume in the shape of a half-ellipsoid (Rundberget 2017) (only half of the volume contains slag, i.e. above ground), the volume was calculated as (4/3πabc)/2, where a, b, and c are the distance from the center point to the perimeter in three perpendicular directions. This gives an average volume of the slag heaps of ca. 12.5 m 3 (4/3 × π × 3.05 2 × 0.64/2).
According to these calculations, the shore-bound slag heaps, on average, had more than a six-fold larger volume than the slag heaps at the forest-bound sites. Thus, the generally younger forest-bound sites yielded a smaller production of iron than the older shore-bound sites, in turn indicating that the two types of sites may have served different purposes.

Estimating charcoal consumption
The next step was to calculate the amount of charcoal needed to produce the amount of iron resulting in the estimated volume of slag. As mentioned in the background section, the calculations below assumes that charcoal was used for iron production. One should note, however, that Magnusson (1986, 242) reported finds of charcoal pits at only two shorebound iron production sites, but he remarked that the remains of charcoal production are likely often overlooked. One would also need to take into account that charcoal was additionally used for roasting lake and bog ore, refining iron, and forging, but these aspects were not considered in the calculations. Thus, if roasting demanded much charcoal, the calculations below should be regarded as underestimating charcoal requirements.
A 1:1 relationship is often used between the volume of slag and iron production, so that 1 m 3 slag approximately corresponds to 1 ton of produced iron (Magnusson 1986, 272). However, studies have yielded very different results (reviewed in Rundberget 2017, 254-255). Using the relationship between slag and iron weight, Rundberget (2017) estimated that the slag-to-iron ratio is 1:0.7 and that slag weighs 1.175 tons/m 3 . This implies that 1 m 3 slag corresponds to 0.82 tons of iron. In turn, this implies that the shore-bound sites, on average, produced 66.4 tons of iron and forest-bound sites, on average, produced 10.2 tons of iron.
The charcoal-to-iron ratio reported in the literature, i.e. how much charcoal was needed to produce a certain amount of iron, varies considerably. Crew (2012) suggested a ratio for medieval bloomery furnaces of ca. 17.5:1. Groenewoudt and van Nie (1995) presented data suggesting a ratio of 7.6-10.8:1. Based on experiments, Birch and colleagues (2015) suggested a ratio of 4.7:1. Rundberget (2017) suggested a consumption of 29.5-59 dm 3 charcoal per kg of iron (corresponding to the measures m 3 and ton, respectively). The  fig. 149). In this area, there are also the remains of an Iron Age farm/settlement. According to Welinder (2008, 54) this farm was established around A.D. 400 or slightly earlier, and it incorporated both small crop fields and livestock management. Table 1. Estimates of forest area required for charcoal production based on the size of slag heaps from two types of bloomery furnace iron production sites, shore-bound and forest-bound, in the province of Jämtland, Sweden (Magnusson 1986). The forest area was calculated based on two levels of estimated charcoal-to-iron ratio: L = low (marked in bold) and H = high, and for L and H, respectively, the range is based on the wood biomass per ha of forest. See explanations in the text.

Estimated outcomes of iron production
Shore-bound iron production sites The forest area required if all wood harvest took place at the same time.

2
The forest area required per year, assuming that the iron production site was used for 50 years. 3 The forest area required if the yearly demand of wood was only based on forest growth.
weight of charcoal depends on the tree species used to produce charcoal. For charcoal made from oak, Crew (2012) reported 180 kg/m 3 . As oak wood weighs ca. 50% more than the major tree species in Scandinavia, Scots pine (Pinus sylvestris) and Norway spruce (Picea abies) (Skogskunskap 2022a), and in want of exact data, it is assumed here that 1 m 3 charcoal weighs 120 kg. This would imply that the charcoal-to-iron ratio in Rundberget (2017) falls between 3.5:1 (29.5 × 0.12:1) and 7.1:1. (59 × 0.12:1), i.e. with a midpoint of 5.3:1.
To account for this variation, two levels of charcoal consumption are henceforth used in the calculations: a "high" charcoal-to-iron ratio (17.5:1) based on Crew (2012) and a "low" charcoal-to-iron ratio (5.3:1) based on Rundberget (2017).

Estimating the required forest area
Several studies have estimated how much wood was used to produce charcoal (Iles 2016(Iles , 2019. Here, it was considered most reliable to use a Swedish source, based on the dominant species of trees in boreal forests, Scots pine and Norway spruce. For Sweden, Arpi (1951) reported a 7:1 ratio (weight) wood to charcoal. Thus, in order to produce 1 ton of charcoal, 7 tons of wood were needed.
As the amount of wood in forests is usually measured as volume/area, one needs to determine the relationship between wood weight and volume. The density of wood from Scots pine and Norway spruce is, on average, ca. 400 kg/m 3 (SkogsSverige 2020). Thus, 1 ton of wood corresponds to ca. 2.5 m 3 of wood (1/0.4).
The wood volume in forests varies a lot, and one needs to account for the fact that modern forests are not likely to be representative for forests during the Iron Age. The overall forest structure may have been similar though, as the southward expansion of Norway spruce had reached well into southern Sweden by around 2000 B.P. (Seppä et al. 2009). Studies on wood volume in boreal forests in central Scandinavia during the 19th and early 20th centuries A.D. (Arpi 1951;Linder and Östlund 1998;Ericsson, Berglund, and Östlund 2005) suggest a range between 60 and 141 m 3 /ha. Putting this range together with the estimates of slag volume, iron production, the two levels of charcoal requirements (charcoal-to-iron ratios), the requirement of wood to produce the necessary charcoal, and wood volume per area means that the volume of slag can be associated with the total forest area required for iron production (see Figure 2). To translate this to a yearly impact, one also needs to account for how long an iron production site was in use. In the following, it is assumed that a site was in use for, on average, 50 years. This assumption is discussed below in the discussion section.
The results are presented in Table 1. Note that the table presents a range based on wood volume per forest area for each of the two levels of charcoal-to-iron ratio, low and high. For reasons explained in the discussion, the low charcoal-to-iron ratio is considered most reliable, and to stress that, the low ratio results are presented in bold. Note also that these estimates assume that the forest area was clearcut. If not all trees were used (large trees may have been difficult to use), the area needed would have been larger.
The total requirement of wood was between 6130 and 20,335 m 3 for the shore-bound sites and between 945 and 3125 m 3 for the forest-bound sites. The forest area needed yearly at the shore-bound sites was 0.9-6.8 ha. For the forest-bound sites, the forest area needed yearly for charcoal production was smaller: 0.1-1 ha. As mentioned above, the figures in the lower part of the ranges are probably most reliable. Extending over a period of 50 years, the total area of forest affected (i.e. forest in various stages of re-growth) would be ca. 50 times these estimates. It is, however, reasonable that previously harvested areas could be harvested again, after 20-30 years, somewhat reducing the total area required.
If only forest growth was used for harvesting the amount of wood needed yearly, and assuming a forest growth of 1.7 m 3 /ha per year (the average for Jämtland given by Arpi [1951]), this would, for the shore-bound sites, require exploiting between 72 and 239 ha, while, for the forestbound sites, about a sixth of this area would have been required.
Despite the uncertainties regarding the figures and assumptions used in the calculations, the overall conclusion is robust: the forest area needed to supply the iron production sites with wood for charcoal was quite small. For the shore-bound iron production sites, assuming that the low charcoal-to-iron ratio is most realistic (see below), one should imagine that a total area in the magnitude of up to 1 km 2 (100 ha = 1 km 2 ) around sites was directly impacted by charcoal production, partly clear-felled, and in various stages of forest recovery. The size of the present-day province of Jämtland is over 49,000 km 2 , of which ca. 56% is presently productive forest (Arpi 1951). For early iron producers, the total forest resource for charcoal production must have seemed inexhaustible. But even if the requirement of wood may seem small considering the vastness of the forests, the effects were substantial at a local scale, and they may have affected other kinds of land use, for example livestock management, to which we turn next.

Subsistence Strategies Based on Livestock: A Background
In contrast to iron production, which leaves interpretable material traces, the possibility to assess the extent of forest grazing during the Iron Age is limited. Evidence from pollen analyses, while useful for conclusions of the timing of management related to agriculture (e.g. Wallin 1996; Emanuelsson 2001; Karlsson, Emanuelsson, and Segerström 2010;Eddudóttir et al. 2021), does not provide a basis for the spatial extent of effects on vegetation. Other evidence, for example seasonal settlements such as shielings, are also difficult to detect in material remains (Lindholm, Sandström, and Ekman 2013). Thus, to assess the potential effects of forest grazing during the Iron Age, a retrospective approach is necessary, based on the assumption that knowledge obtained today on effects of burning, forest grazing, and hay-making is relevant for understanding the effects such activities had during the Iron Age.
It is well-known that until late 19th and early 20th centuries A.D., vast areas of forest in southern and central Sweden were used for livestock grazing (e.g. Kardell 2008;Westin, Lennartsson, and Ljung 2022). From the last centuries, it is known that fire was used to deliberately improve the conditions for livestock grazing (e.g. Groven and Niklasson 2005; Granström and Niklasson 2008; Kardell 2008), and it is believed that intentional burning of forests to create land to be used for livestock grazing was used also much earlier, at least in southern Scandinavia (e.g. Magnusson and Segerström 2009;Welinder 2011). Therefore, it is reasonable that the colonizers of the forest regions in central Sweden were aware of the use of intentional forest burning and were able to employ the same method to create grazing grounds for livestock. In boreal forests, natural wildfires are major drivers of ecosystem processes (e.g. Nilsson and Wardle 2005). Thus, the same effects achieved by intentional burning could have been capitalized on after natural wildfires.
In addition to clearing forest cover, forest fires promote the growth of many plant species valuable as feed for livestock, for example grasses such as Deschampsia flexuosa and herbs such as Chamaenerion angustifolium, as well as the resprouting of heather (Calluna vulgaris) and various deciduous trees (e.g. Groven and Niklasson 2005;Kardell 2008). Furthermore, fires promote soil microbial activity, decomposition rates, and the availability of soil nitrogen (e.g. Nilsson and Wardle 2005), effects which generally promote plant growth, favoring grasses and herbs in the field layer. In addition, grazing by livestock promotes transformation of soil from podsol to mull (Westin, Lennartsson, and Ljung 2022). Thus, livestock grazing, particularly if combined with burning, constitutes a positive feedback process: the changing conditions due to grazing increase the productivity of feed for livestock.
To maintain livestock throughout the year in the forest regions of central Sweden, grazing grounds were not sufficient. The winter season is harsh, and livestock need fodder to survive. Winter fodder could be harvested from lakeor river-shore meadows and mires (e.g. Elveland 2015) or it could be obtained from harvesting twigs and leaves from trees and shrubs (Slotte 2001). There is little evidence of hay-making on outland mires during the Iron Age (e.g. Emanuelsson et al. 2003). Some authors (e.g. Von Stedingk and Baudou 2006) suggest that hay-making on mires developed from A.D. 1000 onwards, i.e. considerably later than colonization commenced. It is likely that most of the supply of winter fodder was initially from leaf-hay (twigs and leaves) collected in the vicinity of settlements (Eriksson 2020). The harvesting of leaf-hay was massive in Sweden until the 19th century A.D. (Slotte 2001). All kinds of deciduous trees and shrubs were used. For the region considered in this paper, the main species would have been birch (Betula spp.), aspen (Populus tremula), rowan (Sorbus aucuparia), and various species of willow (Salix spp.). Engelmark (1978) remarked that the poor conditions for agriculture in the forests of central Sweden enforced mixed subsistence strategies, including hunting and fishing. For example, Emanuelsson and colleagues (2003) suggested that outland use at a settlement from the 7th century A.D. incorporated crop production, livestock management, iron production, and pitfall traps for elk. Hennius (2020b) found that pitfall hunting had existed in the forest regions of central Sweden for several millennia and co-occurred with the early peak of iron production in Jämtland (4th-7th centuries A.D.), and there was a strong increase in pitfall trap systems during the centuries following this peak.

Estimating Livestock Grazing and Fodder Production in Boreal Forests During the Iron Age
The focus here is on livestock, and it has been proposed that, initially, during the colonization of forests in central Sweden, these were mainly sheep and goats (Von Stedingk and Baudou 2006;Magnusson and Segerström 2009) and that cattle came later. For the calculations here, cattle (henceforth implying "cattle equivalents;" see below) are used, as some of the underlying data are based on cattle. In her thesis on livestock grazing in Sweden during the 17th-19th centuries A.D., Dahlström (2006, 133) estimated that the livestock density in forests used for grazing ranged from 0.1-0.5 cattle per ha. Westin, Lennartsson, and Ljung (2022) reported a wider range, from 0.04 to over 2 cattle per ha. Considering that all these figures come from southern Sweden, and that the forests in focus here are less productive, a figure in the lower part of the range was used: in the following, 0.1 cattle per ha.
The number of cattle at early Iron Age farms in southern Sweden has been estimated at 10-16 (Olausson 1998), 6-12 (Petersson 2006, and, on average, 14.6-15.7 (Pedersen and Widgren 1998). For the calculations here, it was assumed that an average farm during the colonization period had 10 cattle (cattle equivalents). Based on a density of 0.1 cattle per ha, the forest area needed would be about 100 ha.
The herding of livestock was necessary due to the presence of predators such as wolves, lynx, and bears. If the livestock was brought back to the settlement every evening, this would imply that the area used for livestock grazing must have been within a reasonable walking distance from each farm settlement. This would limit the forest resource that could actually be used. Field studies on the behavior of cattle grazing in forests in the province of Jämtland (Kardell 2008) suggest that cattle move up to ca. 8 km per day. Allowing for variation in habitat qualities, and considering that cattle do not move in a straight line, the herd would probably not be more than ca. 2 km from the farm (i.e. a quarter of the walking distance per day for cattle). This means that the available area for grazing would be ca. 12.6 km 2 (corresponding to a circular area with a radius of 2 km). Accounting for topographic obstacles such as water bodies hindering movement in some directions, for example if the farm was located along a river or a large lake, it could be that no more than half this area was available, i.e. ca. 6.3 km 2 (630 ha). Based on the figures from Dahlström (2006) above, the requirements for herding 10 cattle would be met. This would also allow for re-growth of the most important feed for the livestock if some areas were avoided each year. This leads to a conclusion similar to the one obtained for charcoal production: for the early colonizers, the forest resource in itself would have seemed inexhaustible.
However, there was a constraint. Harvesting of winter fodder and transporting the fodder to the farm was necessary for keeping livestock over the winter. Wallin (1996) proposed that occurrences of natural hay-meadows (such as wetlands) were so important that their presence was the main factor for settlement location during the Iron Age. As mentioned above, hay from wetlands may not initially have been the most important winter fodder used. In his extensive studies of the harvest of leaf-hay (twigs and leaves), Slotte (2000Slotte ( , 2001 suggested that this form of fodder dominated in forested regions, such as in central and northern Sweden. Scots pine and Norway spruce are useless as leaf-hay, so in order to have a cover of re-sprouting deciduous trees and shrubs, this presumes that the area was managed by clearing or burning (Slotte 2000).
Leaf-hay was collected as leaf-sheaves, i.e. a pile of twigs with leaves collected into a bundle. For highly productive wooded meadows in southeastern Sweden, with tree species such as ash (Fraxinus excelsior), elm (Ulmus glabra), lime (Tilia cordata), and maple (Acer platanoides), Slotte (2000) estimated that about 1000 leaf-sheaves per ha could be harvested in intervals of five years. Thus, 5 ha would be required for a yearly demand of 1000 leaf-sheaves. It is difficult to translate these figures to the much less productive boreal forest, where the deciduous species only occur as early successional species, and the re-growth is slower. Exact calculations are not possible. As a guideline, one may consider that forest growth in Jämtland (1.7 m 3 /ha, Arpi 1951) is ca. 20% of what can be obtained on productive soil in southeastern Sweden in the regions examined by Slotte (2000) (based on Skogskunskap 2022b). This implies that 1000 leaf-sheaves/year would require at least five times as large an area, i.e. 25 ha, assuming that the area was managed by clearing or burning. Slotte (2000) estimated the fodder requirement for sheep, not cattle, and one sheep required from 200 to over 1000 leaf sheaves/year, with the high end of this range indicating small leaf-sheaves. Using a conservative estimate, the low end of the range (200 sheaves/year) implies that feeding a single sheep for one season would require 5 ha of forest managed by clearing or burning. The feeding requirement for sheep needs to be translated to cattle (since the livestock densities referred to above were for cattle). Dahlström (2006, 88) estimated that the relationship for feeding requirements for cattle and sheep is 1:0.25, i.e. in terms of feeding, 10 cattle is equivalent to 40 sheep. Using this 4:1 ratio, 10 cattle equivalents would require 200 ha (10 × 4 × 5) (2 km 2 ) of forest to satisfy the demands for winter fodder. This represents a circular area with a ca. 0.8 km radius. In practice, accounting for impediments, lakes, or any other obstacles for producing fodder, the actual distance from the farm to the outer border of the whole fodder-producing area may have been considerably larger, also affecting the distance harvested fodder needed to be transported. For example, if the farm was located on a river or lake, the distance would be at least doubled. Even accounting for some fodder production at nearby shore meadows, this implies that the distance may exceed the distance a herd of livestock could have been from the farm if herded back to the farm daily.
This leads to the conclusion that the forest area closest to the settlements must have been allocated to the production of winter fodder. These areas could not be used for summer grazing by livestock: that would have destroyed the production of leaf-hay (or crops, if we consider crop fields as being located close to farms). As a consequence, the land used for livestock grazing must have been located further away from the farm.

Discussion
Before interpreting the results of the calculations, some comparisons with other studies are useful. The Swedish forest historian Lars Kardell (2003, 45) made an attempt to estimate the requirement of forest resources for iron production during the Iron Age and concluded that a clear-cut of 5-6 ha, or a forest growth of ca. 500 ha, would suffice to satisfy the average yearly demand for the whole of Sweden. Obviously, this estimate is very low compared with the results in Table 1, considering that this table provides estimates for single average iron production sites (and there were many iron production sites). The calculations in Kardell (2003) are not described in detail, but from the information given, some inferences can be made, which suggest that he underestimated the impact on forests. Firstly, an average for the whole Iron Age is misleading. Production varied considerably, and the early Iron Age in Jämtland represented a peak far above the average for the Iron Age (Magnusson 2020); the results in Table 1 for the shore-bound sites represent production during such a peak period. Secondly, in comparison to the calculations here, Kardell (2003) assumed a higher slag-to-iron ratio and a much higher forest production (Kardell uses figures more relevant for modern production forests in southern Sweden).
The results in Table 1 can also be compared with Stenvik (2015), who reported total wood consumption from an early Iron Age production site in Norway: ca. 7000 m 3 wood. The site produced an estimated 100 tons of iron, ca. 50% higher than the average shore-bound site in Jämtland. In the calculations made in the present study, the total estimated wood consumption for an average shore-bound site was in the range of 6160-20,335 m 3 (for the two levels of charcoal consumption, respectively; see Table 1). The lower figure, based on the lower charcoal-to-iron-ratio, is most compatible with Stenvik (2015), suggesting that using this charcoal-to-ironratio provides the most reliable estimate of the forest area impacted by iron production (and thus marked in bold in Table 1). This is also in line with the experimental results presented by Birch and colleagues (2015) suggesting a charcoal-to-iron ratio of 4.7:1, close to the low ratio used in the calculations.
Another critical factor to consider is the length of time an iron production site was active. This affects the yearly consumption of charcoal. Here, it was assumed that an average site was active for 50 years. We do not know whether this assumption is reasonable and can only indirectly assess its validity. Magnusson (2020) suggested that a bloomery furnace could produce ca. 20 kg of solid iron per day. Rundberget (2017) reached a similar result: 19 kg of iron per day. Using the results in Table 1, the yearly average production for shore-bound production sites would be ca. 1.3 tons of iron (66.4 tons of iron, distributed over 50 years). This amount could then be produced from one furnace active over 65-68 days. Shore-bound iron production sites generally had several furnaces: Magnusson (1986, 262) describes a model of shore-bound sites with six furnaces. Rundberget (2017) and Stenvik (2015) suggested that early Iron Age production sites typically had four furnaces active simultaneously. If that was the case, the yearly average of iron (1.3 tons) could be produced in 16-17 days. Rundberget (2017) considered 20 working days per year as reasonable for the iron production sites. This indicates that using 50 years as the time period of the iron production sites' use is an acceptable approximation.
One can only speculate on the temporal distribution of the different iron production sites. The dating of the iron production sites suggests that the shore-bound sites existed mainly between the 4th and 7th centuries A.D. (Magnusson 2020), i.e. for 3-4 centuries, with a maximum number of sites around A.D. 500. Based on the estimate (Magnusson 1986) that the shore-bound sites amounted to 14% of a total of ca. 720 iron production sites in Jämtland (i.e. around 100 sites), and using an estimate of 50 years for the operation of single sites, a very rough estimate can be made. This suggests that, during the peak of activity (around A.D. 500), there could have been in the neighborhood of a maximum of 20-40 sites in operation simultaneously, with considerably fewer at the beginning and end of this time period.
A general conclusion is that there was plenty of forest area available for the colonizers, so forests per se would not have been a constraint for producing charcoal, for establishing grazing grounds for livestock, or for fodder production. But the forest area required for leaf-hay would have imposed constraints on livestock grazing in the vicinity of the main farm. It is easy to imagine that a strategy emerged of organizing summer grazing at some distance from the main farm, thus relieving the herders of moving livestock daily between the farm and the grazing grounds and ensuring that livestock did not destroy the production of leaf-hay. Magnusson (1986) remarked that shore-bound iron production sites, which were not located close to permanent settlements, were often located at what seemed like suitable places for shielings, i.e. secondary seasonal settlements used in summertime as pasture and for other agricultural activities and which are known historically from a much later time (16th century A.D. onwards). Did the iron production sites during the early Iron Age provide the initial impetus for the development of shielings? Due to the harvesting of wood needed for producing charcoal, iron production sites had the qualities necessary for livestock grazing ground.
The issue of the origin of shielings in Sweden has been controversial. Larsson (2009) made an extensive study (using the term "summer farm") based on a 19th century A.D. model and argued that shielings are no older than the late Middle Ages. The definition of shielings used by Larsson (2009, 421) focused on non-perishable milk products and stressed that it was a specialized female worksite. One may argue that a focus on milk products reflects the market for such products from the 16th century A.D. onwards, and the female aspect may have been a combined result of milk production (probably from early on a female activity) and a generally unbalanced gender ratio (there were few men) resulting from the more-or-less constant wars Sweden was involved in between the A.D. 1500s and early 1800s. Livestock herding was probably a male activity before the Middle Age, as it is in southern Europe (Westin, Lennartsson, and Ljung 2022). Thus, using a definition constructed on a 19th century A.D. model implies that the conclusion with regard to the timing of its origin is built into the definition, meaning there were no shielings before the 16th century A.D.
Other authors have suggested that shielings were established much earlier, during the Iron Age (Svensson 1998;Emanuelsson 2001;Emanuelsson et al. 2003;Von Stedingk and Baudou 2006;Magnusson and Segerström 2009;Rundberget 2017;Hennius 2020a). The results presented here are in line with these authors. The origin of shielings may reflect an adaptation to a resource-poor landscape in regions with a harsh climate, to secure exploitation of fodder (both grazing and winter fodder) by moving livestock seasonally beyond the areas close to the main farm. Due to clearing and burning, the conditions at iron production sites would then have been suitable places for establishing seasonal pastures, perhaps after iron production at the site had ceased and re-growth of wood for future production of charcoal was no longer needed. The total area affected by clearing at iron production sites (44-103 ha, assuming the low charcoal-to-iron ratio; see Table 1) is on the same magnitude, or somewhat smaller, as what is required for grazing by 10 cattle. The same kind of reasoning for the iron production sites may lie behind the suggested hotspots of outland activities (e.g. Lindholm, Sandström, and Ekman 2013;Hennius 2020aHennius , 2021; a spatial concentration of activities was rational from the point of view of workload, due to the associated mutual benefits obtained from the different activities (such as charcoal production and livestock grazing).
Remains of early shielings are archaeologically obscure (Lindholm, Sandström, and Ekman 2013), and grazing effects detected in the vegetation after 1500 years are elusive. Therefore, an approach for further investigations of the suggested spatial association between shore-bound iron production sites and summer pastures for livestock would be to analyze sediments in lakes adjacent to these sites for pollen and various chemical compounds indicative of the past presence of livestock (e.g. Karlsson et al. 2016;Ter Schure et al. 2021).
In conclusion, forests per se, seen as a resource, were not limiting for the colonization of forest regions in central Sweden during the early Iron Age. There was enough wood to satisfy the needs of producing charcoal, and there was enough room for livestock grazing and fodder production. However, livestock cannot graze the same area that was used for fodder production (they will consume the plants useful as fodder), and there is a limit for people herding livestock (they cannot walk daily back and forth over too-great distances). Thus, livestock grazing had to be conducted elsewhere than at the main settlement. While the forest areas cleared for charcoal production, located at iron production sites, initially may not have been intended for livestock grazing, an innovative idea to use them for that purpose would have reduced these constraints, even after the iron production at the site had ceased. This opened areas for fodder production closer to the main settlement, provided that these were properly managed by clearing and burning to promote the secondary succession of deciduous species such as birch, aspen, and rowan. Figure 4 provides an illustration of these suggestions.
The utilized forest landscape eventually extended spatially as the forest areas between the main farm, the iron production sites, and the shielings were influenced by the movement of people and livestock and became part of a domesticated landscape. The spatial connection between farms and shielings may also have demonstrated the extent of the farm's territory, preventing other colonizers from establishing themselves there (Emanuelsson 2001;Karlsson, Emanuelsson, and Segerström 2010). The main conclusion from this study is a hypothesis that such a domesticated landscape developed in the forest regions of central Sweden during the Iron Age: sparsely populated and with a different spatial structure than the landscapes in the agricultural plains in southern Scandinavia. Further research on material evidence of a domesticated Iron Age forest landscape is needed to examine the validity of this hypothesis.