Forest Fire and Indigenous Sami Land Use: Place Names, Fire Dynamics, and Ecosystem Change in Northern Scandinavia

During the last two centuries, fire suppression has critically modified boreal ecosystems in northern Scandinavia and has undoubtedly affected indigenous Sami land use. We inventoried Sami toponyms referring to fire in a municipality located in Swedish Sápmi, and investigated their past and present meanings by analyzing Sami dictionaries and conducting semi-structured interviews with Sami reindeer herders. We use toponyms based on the Sami word ‘roavve’ - a lichen-rich pine-heath that has burned - as a description of past ecosystems to inventory understory and tree vegetation and date the last occurrence of fire in 15 ‘roavve’ places. The inventories showed that some ‘roaave’ places have developed a late succession vegetation type, reducing their suitability for reindeer grazing. We argue both that fire suppression strongly influences the ecological trajectory of these sites and that one must take into account ethnoecological considerations when using toponyms as ecological markers to fully understand their meanings and avoid misinterpretation.


Introduction
Although fire suppression has been a standard management practice in boreal forests for over a century, fire is today seen as an important ecological disturbance factor that structures the functioning and dynamics of boreal forest ecosystems (Niklasson and Granström 2000;Nilsson and Wardle 2005). It is also acknowledged that fire regimes constitute intertwined natural and cultural phenomena resulting from interactions between humans and the living world (Bowman et al. 2011;Coughlan and Petty 2012;Pyne 1998). In Northern Sweden, fire suppression measures implemented since the late nineteenth century have greatly reduced the number of fires in boreal forests and the extent of burned areas, increasing the mean interval between fire events from 80 years to centuries (Carcaillet et al. 2007; Granström and Niklasson 2008;Wallenius 2011;Zackrisson 1977). Conversely, the boreal landscapes of pre-industrial Sweden were characterized by a mosaic of habitats heavily structured by fire regimes that maintained diverse, open, multi-storied stands (Nilsson and Wardle 2005;Östlund et al. 1997;Zackrisson 1977). Because fire's beneficial long-term effects on commercial forest productivity (Nilsson and Wardle 2005;Wallenius 2002) and biodiversity conservation are increasingly acknowledged (Halme et al. 2013;Linder 1998;Östlund et al. 1997), fire restoration is now commonly performed in the Swedish boreal forest during conservation and prescribed burnings.
Fire management measures and the associated changes in fire regimes also affect indigenous peoples' livelihoods (Natcher 2004). In northern Fennoscandia, the indigenous Sami people practice reindeer husbandry, relying on lichen-rich forestlands for winter grazing, and are therefore significantly impacted by fire regime shifts. Sandström et al. (2016) showed that the area of lichen-rich forests in Northern Sweden has declined by 71% over the last six decades. Some studies have suggested that fire suppression during the twentieth century reduced the abundance of ground reindeer lichen (Cladonia spp.) (Berg et al. 2008;Moen and Keskitalo 2010) by reducing the abundance of old, open, low productivity pine forests (Esseen et al. 1997).
Although fire initially destroys ground lichens, it can promote their long-term persistence by suppressing competing vegetation and maintaining suitable conditions for their establishment. The post-fire chronosequence, which describes the vegetation dynamics of ecosystems on broad geographic and time scales, predicts that ground lichens become established after an average of 30-40 years following a fire, and reach maximum coverage after 40 to 120 years depending on geographical conditions (Coxson and Marsh 2001;Crittenden 2000;Kumpula et al. 2000). Organic matter subsequently accumulates, creating moister soil conditions that initially favor the establishment of ericaceous dwarf shrubs such as lingonberry (Vaccinium vitisidaea) and bilberry (Vaccinium myrtillus), followed by crowberry (Empetrum nigrum), heather (Calluna vulgaris), and feather mosses. After a period of 80-100 years on average, these species are expected to become dominant over lichens unless a new perturbation occurs (Ahti and Oksanen 1990;Miller 1996;Nilsson and Wardle 2005).
Indigenous peoples around the world, including those in boreal regions, manage fire for various purposes (Kimmerer and Lake 2001;Lewis and Ferguson 1988;Miller and Davidson-Hunt 2010). However, little is known about the historical relationships between Sami livelihoods and fire. Consequently, there is a need to better understand how the Sami historically made use of entire forest landscapes and how central resources were used and managed (Norstedt 2018). Studies of Sami toponyms can provide critical information on their livelihoods and relationship with fire that can complement ecological investigations in reindeer grazing areas. A few studies have used place names as historical landmarks to study ecological change in other geographical areas (Conedera et al. 2007;Henshaw 2006;Jones 2016;Sousa and García-Murillo 2001;Sousa et al. 2010). Sami toponyms often describe the places they refer to, such as the configuration of the terrain, or the possibility of finding specific plants and animals, and are frequently linked to reindeer husbandry (Korhonen and Anderson 2010). Some toponyms refer to forest fires or their consequencesan important example is place names incorporating the Sami category roavve, which refers to a burned forest suitable for reindeer grazing.
Our overall objective was to understand how places named roavve have evolved over time, their current status, and how fire and fire suppression have impacted the vegetation. We conducted an interdisciplinary study encompassing forest history and ethnoecology to (i) clarify the current and historical meanings of roavve to Sami reindeer herders; (ii) determine how places named roavve have evolved in the last ca. 100 years, using the indigenous definition of roavve as an ecological reference; and (iii) evaluate the role of fire and fire suppression in this evolution by linking vegetation characteristics to time since the last fire (TSLF). More broadly, we aim to determine how connecting place names, ethnoecological studies, and ecological investigations can shed light on ecological changes and trends in indigenous people's land use.

Material and Methods
The study area was limited to the borders of Jokkmokk municipality, Norrbotten County, Sweden. The municipality covers an area of 19,334 km 2 centered at approximately 66°37′0°N; 19°50′0°E, and includes commercial forestlands, protected forest areas, and mountain lands. The forests have two main coniferous species: Scots Pine (Pinus sylvestris) and Norway Spruce (Picea abies). The most common deciduous trees are birches (Betula pendula, B. pubescens). Less abundant deciduous trees include willows (Salix caprea), aspen (Populus tremula), and alder (Alnus incana). Indigenous Sami people practice reindeer husbandry over the whole area. The municipality is home to five of the 51 Swedish herding communities: three mountain-based communities (Sirges, Jåhkågasska, and Tuorpon), and two forest-based communities (Udtja and Slakka). Mountain-based communities conduct yearly migrations with their reindeer between the winter pastures in the lowland forests, and the summer pastures in the high mountains near the Norwegian border. The forest communities keep their reindeer in the lowland forests all year round, although they also undertake migrations between winter and summer pastures. The Sami reindeer herding communities have a right of use over their herding territories. However, most of these territories is owned by public and private forestry companies, with a smaller proportion belonging to private smallholders. Reindeer herders are facing increasing encroachment of their winter grazing areas because of the activities of various industries in northern Sweden, including forestry, mining, wind power, and hydropower. These encroachments and the fact that Sami reindeer herders do not own the land they use for grazing, are imposing severe pressures on their communities. The area of Jokkmokk corresponds to the historic linguistic area of Lule Sami, one of the 10 Sami languages still spoken today. Despite restrictions imposed by early twentieth century colonial policies and practices, Lule Sami remains in active use, especially among reindeer herders' communities. However, the North Sami language is also widely spoken in the area since the forced displacement of Sami families from the North in the early twentieth century.

Identification of Sami fire Toponyms
Our first step was to identify Sami terms referring to forest fires that occur in place names in the study area. Sami place names, like those in many other languages, are often compound words consisting of a generic term that designates a topographic feature and an additional term that describes its physical properties (often metaphorically) or refers to the presence of plants, animals, resources, or associated socio-cultural elements (Rankama 1993). Sources used for vocabulary identification were dictionaries for North Sami (Nielsen and Nesheim 1979[1932-1962; Svonni 2013) and Lule Sami (Grundström 1946(Grundström -1954Korhonen 1979), and records of place names (Collinder 1964;Grundström 1927Grundström -1934Korhonen and Anderson 2010). Three words related to fire and their derivatives were found in place names of the area: buollem or buollám, guorbba, and roavve. Buollem and buollám derive from the verb buollet, to burn, and mean burned, or burned forest area (Grundström 1946(Grundström -1954Korhonen 1979); guorbba means burned place, or terrain that has been destroyed by forest fire (Grundström 1946(Grundström -1954Korhonen 1979); and roavve refers to the type of vegetation found after a forest fire (Nielsen and Nesheim 1979[1932-1962).
To find place names containing these words and their derivatives, we used the Swedish National Land Survey's (Lantmäteriet (2017)) online map of place names and the place name records of the Swedish Institute for Language and Folklore (Institutet för Språk och Folkminnen (2017)). We identified 39 place names incorporating these three words (30 for roavve, 5 for buollem or buollám, and 4 for guorbba) (Table A, Annex). We found one roavve place name, Tjájároavve, which was absent from place names records or maps, in Ruong (1964). We then used place name registers and the first ordnance maps of the area (published in 1890 and 1893) to locate the identified place names (Grundström 1927(Grundström -1934Pellijeff 1992Pellijeff [1936). We follow the official Lule Sami orthography, in accordance with the 2001 Swedish Heritage Conservation Act, for the spellings of the place names (Nyström et al. 2007). We chose to focus on roavve place names because they were the most numerous in the study area and because it allowed us to maintain a degree of homogeneity in the sites we inventoried.

Semi-Structured Interviews
Semi-structured interviews were conducted with reindeer herders to investigate how Sami native speakers who interact closely with their environment understand the word roavve, and to deepen the definitions given by the dictionaries. An additional goal was to clarify the role and importance of roavve places in reindeer herding practices. The interviews were based on a set of themes addressed through openended questions, enabling interviewees to explore specific topics or introduce new ones. The main themes addressed were the meaning of roavve, the role of fire in its definition, and the role of these places in reindeer herding today as well as in the past. Ten reindeer herders from five reindeer herding communities around Jokkmokk municipality were interviewed (nine men and one woman aged from 53 years old to over 80): seven Lule Sami speakers, one who spoke a dialect of Lule Sami from the Flakaberg area (about 140 km East of Jokkmokk), and two North Sami speakers. An interview was also conducted with a consultant in Lule Sami at the Sami parliament in Jokkmokk. The 11 interviews were conducted in Swedish, recorded, and fully transcribed.

Sampling Strategy
Since the purpose of the ecological inventories was to characterize the vegetation composition at roavve sites, we based our sampling strategy on place names and their locations. Study sites were defined as the surfaces encompassed by the topographic feature designated in the place names. Sampling areas were then selected within each study site. The sampling areas were 200x200m squares with their centers located as close as possible to the coordinates assigned to the place name. In cases where the study site was a hill or a mountain, this position usually corresponded to the top. In a few cases, the coordinates provided by Lantmäteriet did not correspond to the feature referred to in the place name. In such cases, the sampling area and its center coordinates were determined using aerial photographs and topographic maps. In some cases, the place names recorded designated non-forested land features, such as bodies of water and mires. Because of the ambiguity of the locality of the forest stand designated by the roavve place name in such cases, they were excluded from the potential inventoried sites. The remaining sites were prioritized according to their accessibility (distance from a road). In total, 15 sites were inventoried ( Fig. 1).

Estimation of the Time Since Last Fire from Fire Scars
To determine the time since the last fire (TSLF) at the study sites, we used dendrochronology (Arno and Sneck 1977;Niklasson and Granström 2000;Stokes and Smiley 1996;Tirén 1937;Zackrisson 1977). We conducted searches at each study site along transects covering the sampling areas, separated by 100 m intervals, to find fire scars on living trees. If no fire scars were encountered, we extended the search to a 400x400m square. The presence of other signs of fire such as charred stumps, branches, or roots was also recorded. One or more tree cores were taken with an increment borer from the first fire scar encountered on a living tree inside the sampling area. In three sites where modern clear-cutting had deeply modified the stand structure, we found no living trees with fire scars and so took cross-sections from dead trees or snags with fire scars, to be cross-dated using a pre-established master-chronology from the Torneträsk area (Arno and Sneck 1977;Grudd et al. 2002). Tree cores were measured using a LINTAB™ tree-ring measuring station with a resolution of 1/100 mm, and the measurements were analyzed using the TSAP-Win™ software package (Rinntech technologies).

Ground and Tree Vegetation Inventories
We made a vegetation inventory in each study site at three randomly selected points where we took a tree core from the presumed oldest tree to estimate the stand age in even-aged stands. If the stand contained several age-classes, we took a core for each age-class. The ground vegetation was characterized according to Hägglund and Lundmark (2004). We evaluated the percentages of cover for reindeer lichen, bilberry, lingonberry, and crowberry inside a circle of 10 m radius centered on the sampling point. In addition, the characteristics of the tree layer (i.e., tree density and basal area for each tree species) were measured in a circle of 11.28 m radius (400 m 2 ) at each sampling point.

Statistical Analysis
To characterize the variability of the study sites' vegetation and establish a typology, we performed a principal component analysis (PCA) in conjunction with hierarchical clustering using the ade4 package (Chessel et al. 2004) with the R 3.4.0 software (R Core Team 2013). The variables included in the PCA were vegetation variables structuring the definition of roavve. The bilberry, lingonberry, and crowberry covers were also included in the analysis because these species are the dominant dwarf shrub species in Fennoscandian boreal forests and their abundance can be indicative of edaphic conditions. The variables used to construct the PCA were therefore lichen cover (% of inventory area), pine basal area (m 2 / ha), spruce basal area (m 2 /ha), bilberry cover (% of inventory area), lingonberry cover (% of inventory area), and crowberry cover (% of inventory area). We constructed a biplot in which we plotted the position of each study site with respect to the first two principal components. To represent the vegetation composition at each site within the scores plot, radar charts showing the values of the vegetation variables for each site were generated using the fmsb package. We tested the correlation between the PCA scores and two measured variables supposed to be explanatory, namely time since last fire and tree density, to interpret the distribution of the sites along the principal components' axes. We also performed linear regressions to determine the extent to which TSLF could explain the variability of the vegetation variables among the sites.

Definitions of Roavve
Every dictionary we consulted other than the most recent (Svonni 2013) gives a first definition of roavve as a place where there has been a forest fire (Table 1). Three dictionaries also give a second meaning of roavve as a morphological feature (a camber) that may be associated with the curvature of a ski (Grundström 1946(Grundström -1954Korhonen 1979) and can designate a topographical feature (Nielsen and Nesheim 1979[1932-1962). However, Svonni (2013) defines this word only as a type of foresta heath.
Seven of the herders were able to clearly define the term roavve, two were uncertain of its meaning, and one did not know it. The definitions given by those who knew the term included several criteria and collectively revealed all the dimensions that can be included in the meaning of roavve. All of the herders stated that roavve always referred specifically to a pine heath; one of them noted in passing that he had Bnever heard anyone talk about a guossaroavve [a spruce-roavve].M oreover, they all associated roavve with richness in reindeer lichen and thus good winter pasture for reindeer. Four of them regarded fire as a defining element of roavve, always designating a pine heath that had burned. These herders considered the fire responsible for the richness in reindeer lichen and high pasture suitability of roavve places, suggesting that it  (Table 1), who was over 80 years old and mindful of knowledge transmission issues, as he used to map and teach place names to his children, provided a very precise definition of roavve and its relationship to fire. To him, it was not merely a pine heath that had burned: it designated a stage of the forest after the fire, when the vegetation had started to return and the ground reindeer lichen had become established. His definition increases the term's complexity because it implies a temporal dimension: as he understood it, a place could not be named roavve immediately after a forest fire, but only once the vegetation had reached a specific state: BIt is the result, they do not call it just when it has burned, but when it has been a few years, and vegetation begins to come back in the area. And reindeer lichen grows, and a young forest takes over, I think it is then one calls such a place roavve.Ĉ onversely, herders 3 and 4 provided definitions of roavve that did not include fire as a defining criterion ( Table 1). One of them stated that roavve had Bnothing to do with fire.^For them, it was the topography that defined roavve: it designated a pine heath on a long ridge. One specified how roavve were usually situated on southeastern mountain  Grundström 1946Grundström -1954 Lule Sami R åvvē: place where forest fire has gone through, burn. råvvē-tnam: terrain where forest fire has gone through; Flakaberg area: also, ridge. Råvvē: the even upwards bending in the middle of a ski Korhonen 1979 Lule Sami Roavve: camber, the even upwards bending in the middle of a ski Roave: burn, burned area, burned place (where there has been a forest fire) Korhonen and Anderson 2010 Lule Sami Roavve, gen. Roave: burn, burned aera, burned place (where there has been forest fire). The name element refers to mountains or heaths with coniferous forest, where forest fires can come up through lightning strikes. The name element roavve is therefore missing from the mountain area naming register.
[…] Nielsen and Nesheim 1979[1932-1962 North BRoavve, it is a heath where there has been forest fire, and… it is an old forest-fire place, which, after many years can become very good for reindeer pasture, but it takes many years before it becomes a reindeer pasture at such a place. So, my interpretation of roavve is that it has been a forest heath, with an old forest, which has burned away. And then, the lichen and everything has grown and reindeer pasture begins to grow, lichen for example, and wavy hair-grass [Deschampsia flexuosa] as well. It is a grass that often grows at burned places. slopes. Thus, two meanings of roavve emerged from the reindeer herders' definitions: a main meaning as a lichenrich pine heath that has burned, and a second meaning as a lichen-rich pine heath on a long ridge having no particular association with fire. Some dictionaries include both meanings, which they associate with specific dialects in certain cases (Grundström 1946(Grundström -1954Korhonen 1979;Nielsen and Nesheim 1979[1932-1962). The languages spoken by the interviewees, namely Lule Sami or North Sami, did not seem to influence their preference for one definition of roavve or the other.
The herders also noted that places called roavve were changing, causing the typical roavve vegetation to disappear, and losing their potential as good reindeer pastures. One herder, who was very interested in observing and understanding vegetation dynamics, said: BThey do not need to be lichen-rich today, but, in the favorable period in the cycle, then they were very good reindeer pasturelands, I believe. (…) They have surely been lichen-rich. Then the years have gone by and they have become richer in nutrients, it has changed the vegetation, so that the reindeer lichen is gone.T o him, the vegetation in the vast majority of lichenproductive lands, including roavve places, was changing and becoming unsuitable for lichen, for example by becoming bilberry dominated. Others noted that in many areas the forest was becoming denser and darker, preventing lichen persistence and enabling the establishment of competitor plants. They attributed this to noncommercial forests growing older, and dense planting practices in commercial forests.

Time Since Last Fire
The last fire was accurately dated using dendrochronology at seven of the 15 inventoried sites (Table 2). At three sites, fires were dated (to 374, 272, and 427 years ago) using the master-chronology; these dates are probably earlier than the actual dates of the last fires at these sites. Indeed, many signs of past fires, such as charred stumps, branches, or roots were observed at these sites, indicating more recent fires. Moreover, all three sites contained young stands (the stands at sites 2, 8, and 13 were 53, 35, and 77 years old, respectively) under intensive forestry management that had recently been clear-cut and regenerated naturally or by planting. The felling of older trees during forestry operations would explain the absence of living trees bearing fire scars. In three other sites, the trees with fire scars encountered were rotten inside, and we could only determine a minimum time since the last fire, corresponding to the age of the overlapping part of the fire scar. The time since last fire at these sites must thus be greater than these estimates. Consequently, these estimates are henceforth referred to as minimum estimates. For the two remaining sites, no fire scars on living trees were detected, and cross-dating with masterchronologies gave non-significant results. We therefore used the age of the spruce trees in the stand as a proxy for the TSLF in these cases. The most recent fire across all sites was in 1971, and the oldest in 1643. The mean TSLF, excluding the fires dated with the master-chronology, was 134.4 years (SD = 70.4). The mean TSLF for the seven fires dated accurately by dendrochronology was 124 years (SD = 50.8).

Vegetation Composition and Its Relationship to Time Since Last Fire
At the time of inventory, the 15 sites exhibited high variability in terms of their vegetation composition (Table B Annex), particularly with respect to the two variables defining roavve, i.e., reindeer lichen cover and tree composition. Some sites were covered with an extensive reindeer lichen mat with scattered pine trees (Fig. 2a). Others had less or no lichen cover but were dominated by ericaceous dwarf shrubs with a denser tree layer consisting of spruces and a few old largediameter pines (Fig. 2b). There were also sites dominated by ericaceous dwarf shrubs with the tree vegetation dominated by pines (Fig. 2c). Finally, some sites had young pine stands that had been clear-cut and regenerated, resulting in low ground reindeer lichen cover (Fig. 2d). Across all sites, the lichen cover ranged from 1% to 70% with a mean of 24% (SD = 20.5), while the basal areas of pine and spruce ranged from 0.5 to 13.9 m 2 /ha and 0 to 6.5 m 2 /ha, respectively, with means of 6.7 m 2 /ha (SD = 4.1) and 1.68 m 2 /ha (SD = 2.2), respectively.
The PCA structured the heterogeneity in the sites' vegetation; the first and second dimensions of the PCA explained 42.3% and 30.2% of the variation in the data, respectively. Hierarchical clustering revealed some common trajectories, enabling the determination of a typology that separated the inventoried sites into four groups (Fig. 3). The vertical axis of the PCA distinguished sites characterized by an association between spruce and bilberry (group 1) from sites displaying an association between lichen, pine, and lingonberry (groups 2 and 4). Within the lichen-pine-lingonberry association, the horizontal axis distinguished sites with lichen-rich vegetation (group 4) from sites rich in crowberry and lingonberry at the expense of lichen (group 2). This axis also distinguished sites within group 3 that have intermediate levels of crowberry, lingonberry, and lichen cover.
The two explanatory variables used to interpret the PCA dimensions, TSLF and tree density, were uncorrelated (Pearson correlation = 0.01; p value = 0.96). However, the first axis of the PCA was explained by TSLF (simple regression: coefficient = −0.02; F-Test: p value = 5.21 × 10 −2 ). This relationship became statistically significant (p = 8.65 × 10 −4 ) if the sites for which only minimum estimates of the TSLF were available (i.e., sites 1, 7, and 9) were excluded from the analysis. Additionally, the tree density (after being logarithmically transformed to establish a normal distribution of residuals) explained the second axis of the PCA (Simple regression: coefficient = 1.32; F-Test: p value = 0.03) (Fig. 4). Finally, the model relating TLSF to lichen cover showed a tendency to significance (Fig. 5) and became significant (coefficient = −0.28; p = 4.65 × 10 −2 ) upon exclusion of data corresponding to the three sites for which only minimum estimates of the TSLF were available.

Discussion
The Category Roavve and Its Meaning for Reindeer Herders All our interviewees interpreted the term roavve as referring specifically to a pine heath. Prior lexicological analyses lack this precision: with the exception of Svonni (2013), the dictionaries refer more broadly to a Bplace^or a Bforest^. The herders also considered the richness of ground reindeer lichen at the site (and thus the site's potential quality as a winter pasture for reindeer) to be an important aspect of the definition of roavve, notably absent from both old and recent dictionaries.
Most of the herders also considered forest fire to be a defining criterion of roavve. This was reflected by all of the dictionaries other than the most recent North Sami dictionary (Svonni 2013). However, some herders defined roavve without reference to fire, focusing instead on topography as a defining feature, a definition also reported by some dictionaries (Grundström 1946(Grundström -1954Korhonen 1979;Nielsen and Nesheim 1979[1932-1962). The presence of both meanings in dictionaries from the early twentieth century together with the use of both meanings by the interviewees suggest that the two definitions have coexisted for at least a century. Grundström (1946Grundström ( -1954 notes that the meaning of roavve as a topographic feature is associated with the Flakaberg Lule Sami dialect, which was confirmed by the definition given by our interviewee (herder 3). However, it should be noted that the two definitions are not mutually exclusive and can describe the same type of features. Indeed, studies on fire ecology have shown that forest stands on ridges are more likely to burn than those on flat land because lightning strikes are more likely on convex surfaces (Engelmark 1987;Zackrisson 1977). Because such environments are usually drier than lower sites, it is perhaps unsurprising that the term roavve incorporates this double meaning: a pine heath that has burned is most likely to be found on a ridge, and vice versa. This is well known by Sami herders, who are keen observers of the forest. However, without further linguistic analysis, it is impossible to establish whether both meanings were forming only one in the past or to determine why they have diverged into two different interpretations. Finally, some herders' definitions included a dynamic element: they considered roavve to refer to vegetation at a burned site after a certain  Nielsen and Nesheim (1979[1932-1962): roavve designates an Barea where there was once a forest fire and which is now for the most part covered with young trees.â To summarize, herders' understandings of the term roavve are structured by several criteria based on Sami modalities of categorizing the environment linked to potential reindeer grazing. A roavve is not simply a lichen-rich pine heath (referred to as gielas). It is also an environment generated through fire that may be found on ridges and is associated with good winter grazing for reindeer. Roavve is thus a semantic category that connects multiple domains: botanical, ecological, topographical, and land use. As such, it is a Bcomplex category^in the terminology of ethnoscience, the study of indigenous classification systems (Friedberg 1999;Roturier and Roué 2009). The definition of roavve implies that the Sami who named these places had a detailed understanding of dynamic ecological processes and were well aware of forest fires' long-term effects on vegetation and reindeer pasture. Furthermore, the many locations within the study area that have names incorporating the word roavve suggest that such environments were historically important for Sami livelihoods.

Vegetation Changes Compared to the Roavve Reference
It is impossible to know precisely when roavve places were named. However, of the 15 we considered in this research, five are recorded in ordnance maps from the 1890s, and five are listed in place name records from the 1920s and 1930s, indicating that the names were assigned before fire suppression became common in forest management. Therefore, based on the assumption that the 15 study sites were named because they once exhibited the characteristics of roavve, we used its Sami multi-criteria definition as an ecological reference to estimate vegetation changes during the fire suppression period. Our results show that some of the sites no longer exhibit key roavve characteristics according to the Sami definition.
While all of the sites presented signs of past fire, the results displayed in the radar charts (Fig. 3) reveal their heterogeneity with respect to understory vegetation and forest structure. Ten sites still contain pine-dominated forest, but five had become mixed-coniferous or spruce-dominated forests. In addition, the reindeer lichen cover was above 25% at six sites but below this level at the remaining nine. The PCA we constructed on the basis of the roavve definition structured the sites' heterogeneity and revealed a typology that divided them into four groups (Fig. 3). Sites in Group 4 display the key vegetation characteristics that the Sami use to define roavve, i.e., they contain pine-dominated forests with extensive lichen cover. Conversely, sites in Groups 1, 2, and 3 have understory vegetation and forest structures that are deviating or have deviated from these. On-site observations were particularly important in this respect. Sites 5 and 9 (Group 1) contained a few oldgrowth, large-diameter pine trees (diameter > 40 cm) with large crowns growing among a younger spruce population, suggesting that the forest composition was very different in the recent past, when it was probably dominated by lowdensity pines and thus would have had high lichen cover. In addition, no signs of more recent fires (e.g., fire scars on young trees) were detected, but many signs of historical fires such as burned stumps or roots were visible, indicating that fire suppression has played a role in changing vegetation composition, and its deviation from the roavve definition.

Time Since the Last Fire at Roavve Sites
It should be taken into account that the TLSF dataset generated during our research is incomplete. For instance, the TSLF values for sites 2, 8, and 13 were estimated to be over 250 years based on cross-dating of fire scars on dead trees using the master-chronology. However, these dates are unlikely to correspond to the most recent fires in these stands. Before the advent of fire suppression, the average fire interval in the Jokkmokk region was around 80 years, with the lowest being 46 years (for lichen-Calluna type pine forests on flat, sandy river terraces), and the highest 122 years (for mixed Vaccinium myrtillus-type coniferous forests on north-facing slopes with moraine soils (Zackrisson 1977). In cases where stand regeneration has been driven by forestry operations, it is often impossible to date the last fire based on fire scars because all live trees with fire scars have been felled. For some sites, we used spruce age as a proxy of TSLF. Zackrisson (1977) argues that the inability of spruce to survive forest fires has been overstated because topographic conditions may allow some to survive fire events. However, an abundance of spruce trees in a stand covering a wide area is strong evidence that the stand has not burned during its lifespan.
It is also worth noting that historical documents can complement TSLF measurements. For example, Grundström (1934) indicated in his place name register that Tjiednekroavve (site 11) had Bburned 65 years ago,^i.e., ca. 100 years before the fire we dated at the site. Field observations confirmed the occurrence of successive fires at the inventory sites: multiple fire scars were identified on dead stumps or living trees. Notably, our results indicate that at least seven of the 15 sites have not burned since the advent of fire suppression, which was progressively applied in Jokkmokk from the end of the nineteenth century. Another three sites burned only once during the last century (1918,1940,1976).

Fire Suppression as a Driver of Roavve Trajectories?
In this study we used a Sami toponym referring to fire to select study sites and investigate the role of fire suppression in the variability of the understory vegetation composition at 15 selected sites. A principal component analysis indicated that TSLF values and tree density could explain the variation in vegetation cover between sites. The PCA-derived group with the lowest lichen cover, the highest spruce basal area, and the highest bilberry cover (Group 1, Fig. 4) had a mean TSLF of 151 ± 66 (the second highest of the four groups even though the TSLF values for two sites in this group were minimum estimates). Conversely, Group 4, with the highest lichen cover, had a TSLF of 108.3 ± 37 years. Groups 2 and 3 had intermediate levels of lichen cover and mean TSLF values of 99.5 ± 79 and 159 ± 84 years respectively. This is consistent with the   (Esseen et al. 1997;Steijlen and Zackrisson 1987). Our results also indicate a negative correlation between the lichen cover and TSLF (Fig. 5a). While there are few data on sites with low TSLF values, it appears that the lichen cover peaks around 100 years after burning and then decreases as the TSLF increases. This trend is strengthened by the fact that the TSLF values for three sites are known to be underestimated (Fig. 5), and is consistent with the expected pattern of lichen recovery after burning (Ahti and Oksanen 1990;Kumpula et al. 2000). There is a striking convergence between this pattern established by ecological studies and the detailed descriptions of vegetation dynamics after forest fires provided by some of our Sami informants. Differences in tree density may explain the different vegetation compositions of the lichen-rich group (4) and the group rich in crowberry and lingonberry (2), which have similar TSLF values (Fig. 4). Sites belonging to Group 2 were subject to active forest management, which may account for their high tree density compared to Group 4. This greater tree density would be expected to reduce lichen cover because it would create light conditions that are not optimal for lichen growth (Čabrajič et al. 2010). Group 1 sites, characterized by an association between spruce and bilberry, also exhibited relatively high tree densities despite not having recently been actively managed. In unmanaged forests, increasing tree densities have been linked to the abandonment of selective cutting practices together with fire suppression (Esseen et al. 1997;Hedwall and Mikusiński 2014).
The variability of understory vegetation between the sites was also certainly influenced by soil conditions, which we did not consider in this research. For instance, sites 3, 6 and 14 had relatively high TSLF values but also had quite extensive lichen cover. The persistence of high lichen cover in the absence of major natural disturbances such as fire can be explained by sustained nutrient-poor edaphic conditions that prevent vascular plants from outcompeting ground lichen (Nilsson and Wardle 2005;Taylor and Chen 2011).
In conclusion, our data and our sampling design based on fire toponyms did not allow us to demonstrate fully the role of fire on understory vegetation and roavve trajectories. However, there were converging results and field observations showing that, with few exceptions, roavve sites exhibit the Bsprucification^seen in Scandinavian protected forest areas, which has been linked to fire suppression and non-active management (Hedwall and Mikusiński 2014;Linder et al. 1997;Linder 1998;Zackrisson 1977). Simultaneously, they can undergo densification, a phenomenon that reindeer herders have identified as a cause of ground lichen loss in both unmanaged forest stands and commercial forests. Our results thus indicate that most of our study sites no longer exhibit the defining characteristics of roavve.

Place Names as Markers of Ecological Change
Place names with social and cultural significance for a community can persist over time and remain in use even after the environment referred to has disappeared or changed (see Ruotsala 2004 for examples from the Finnish Sami area). They can thus become markers of environmental change (Sousa and García-Murillo 2001).
The use of place names as ecological markers is not without difficulties and has raised many debates in the scientific community. First, it is generally impossible to know what motivated the original assignation and interpretations of a name's meaning in a contemporary context may be little more than guesswork (Sousa and García-Murillo 2001). Moreover, place names are often assigned according to a few salient features (Chessex 1945). It is then impossible to know whether a place was named because it exhibited all or merely some of these features. Additionally, a particular name may have originally referred to a circumscribed area and later been extended over a wider area. Some names might also be missed from the place names' inventory, because within societies with oral traditions, such as those of many indigenous peoples, all the names might not have been recorded on maps or otherwise documented, and might have faded from memory (Cogos et al. 2017). Finally, Kadmon (2000) also suggests that toponyms can lose the transparency of their meaning due to changes in the features that inspired them.
Our interviews with reindeer herders suggest that such effects could be occurring with place names based on roavve. Indeed, two of our 10 interviewed herders did not know the meaning of the term or gave somewhat simplified definitions, and two noted that it was an old word not in regular use. To account for this, one herder suggested: BThey have disappeared, these expressions… Because then there were surely names for different stages after forest fires for example. (…) But we have been extinguishing forest fires for so long, so that they have simply disappeared. (…) And surely there were some expressions in Sami for the smoke from fire, I am certain about that. Because it had a strong signification for people who lived at that time.Â s suggested by Korhonen and Anderson (2010), the absence of fire may also explain shifts in the meaning of the word. Contemporary definitions focus on the vegetation resulting from the fire, i.e., pine heaths, omitting the burning process itself since it no longer occurs; this is consistent with the most recent definition given by Svonni (2013) in North Sami.
Ultimately, toponyms are created in a socio-cultural context and carry a set of social and cultural values (Basso 1988;Hunn 1996;Jett 2011;Jones 2016). As such, they represent far more than ecological markers: they are testimonies of past livelihoods and their links to the environment (Jett 2011). Indigenous toponyms are based on alternative classifications of the environment to those developed by ecological scientists (Johnson 2000). A central aim of ethnoscience is to holistically study indigenous classifications, thus providing access to the indigenous way of understanding, representing and experiencing the environment. The interdisciplinary approach we adopted here followed this path, and aimed to combine the use of place names as ecological markers with a deeper analysis of their meaning based on herders' own definitions and interpretations to better understand the dimensions embedded in the associated places in modern and historical Sami livelihoods. Places names incorporating the term roavve are thus testimonies of types of environment that are important for Sami reindeer husbandry and are associated with fire, which remains a key factor governing the dynamics of the boreal forest.

Conclusion
The definitions of the category roavve provided by reindeer herders indicate that such areas were especially important for reindeer herding because they provide good winter pasture for reindeer, and that fire played a central role in establishing these characteristics. Our vegetation inventories at the study sites show that some no longer exhibit their original characteristics, probably because of long-term fire suppression. Some are currently dominated by spruce and ericaceous dwarf shrubs, making them unsuitable for reindeer lichen. We argue that Sami place names, since they reflect the relationship between land use and the environment, could serve as a basis for implementing fire management practices that ensure the sustainability of reindeer husbandry. In broader terms, studies on indigenous place names related to forest fires can provide a basis for understanding the effects of fire regime shifts on indigenous livelihoods.