The study area is located in the Lower Saint Lawrence Region (LSL) in the south-east of the province of Quebec, Canada (47° 92′ N to 48° 91′ N and 66° 84′ W to 68° 86′ W). It is delineated to the north by the St. Lawrence River and to the south by the province of New Brunswick and lies within the Appalachian geological formation, characterized by sedimentary bedrock. The topography generally consists of low elevation hills with moderate slopes with mean and maximum elevations of 350 m and 910 m, respectively. Surface deposits are mainly from glacial and in situ weathering origins. The meteorological data (1981–2010) of Rimouski (20 m a.s.l.), Trinité-des-Monts (260 m a.s.l.), and Saint-Jean-de-Cherbourg (320 m a.s.l.) (Figure 1) show mean annual temperatures of 4.4°C, 2.5°C, and 1.9°C and annual total precipitations of 959 mm, 1100 mm, and 1138 mm, respectively, 30% of which falls as snow (Environment Canada, 2019).

This area is situated at the northern limit of the Great Lakes-St. Lawrence mixed forests at the transition between the temperate deciduous and the boreal coniferous forests (Rowe, 1972). Up to 75% of the study area corresponds to public forest, mostly in the backcountry, whereas 25% is private land, mostly along the St. Lawrence River. The study area is composed of two former forest concessions of the Price Brothers & Company: the Rimouski (2279 km²) and Matane (938 km²) sectors. Price Brothers & Company operated in the LSL region since the 19th century until the first half of the 20th century. The two sectors belong to the balsam fir-yellow birch bioclimatic domain (temperate mixedwoods), although the south-eastern part of the Matane sector, which is higher in elevation with steeper slopes, is located in the balsam fir-white birch domain (southern boreal forest) (Robitaille and Saucier, 1998; Grondin et al., 1999). Balsam fir (Abies balsamea), white spruce, and yellow birch (Betula alleghaniensis) are abundant in mesic sites in the balsam fir-yellow birch bioclimatic domain, while balsam fir, white spruce and white birch are abundant in mesic sites in the balsam fir-white birch domain. In both domains, trembling aspen (Populus tremuloides) and balsam poplar (Populus balsamifera) are also present. Sugar maple and red maple generally occupy hill tops and reach their northern range limit in the study area. Black spruce (Picea mariana) and northern white cedar generally occur on organic deposits along water courses (Robitaille and Saucier, 1998).

The study area has been subject to extensive anthropogenic fires, logging practices of increasing severity and spruce plantations during the 19th and 20th centuries. Fires were frequent at the margins of inhabited areas during the early 20th century due to slash burning for land clearing and agriculture (Fortin et al., 1993; Terrail et al., 2019). In the second half of the 20th century, population stabilization and effective fire suppression (Blanchet, 2003) decreased the frequency of fires to very low values. Logging practices during the 19th century and until the 1930s were mainly selective winter harvesting of large pine and spruce trees (Fortin et al., 1993; Boucher et al., 2009a). Logs were then transported to saw mills using water courses (Proulx, 1985). Forest operations slowed down during the great economic depression of 1929–1939 and vigorously recovered following the Second World War due to the increasing demand for residential construction material. In the 1940s, the rarefaction of large diameter trees prompted a progressive transition to the pulp and paper industry exploiting smaller diameters of spruce and balsam fir (>3 inches DBH). This phase was also marked by the progressive introduction of chainsaws and log transport by trucks (Fortin et al., 1993). In the 1970s, dramatic changes occurred over the landscape, due to mechanization and the increased cutting rate in the snow-free season (Boucher et al., 2009a). Meanwhile, a severe spruce budworm outbreak (SBW) (Choristoneura fumiferana) between 1975 and1992 (Boulanger and Arseneault, 2004) triggered extensive salvage logging and consecutive large scale spruce plantations (Fortin et al., 1993). To our knowledge, there is no documentation of pre-European anthropogenic disturbances in the LSL region. Nevertheless, Aboriginal influence in modifying the natural fire regime has been documented elsewhere in eastern North America (Williams, 2000; Blarquez et al., 2018), and could have participated in influencing pre-industrial forest composition.

We digitized and georeferenced maps from several sources (Table 1). In their study, Boucher et al. (2009b) digitized and analyzed maps produced during an exhaustive 1930 forest inventory by the Price Brother’s & Company. Along with these data, in the present study we discovered maps of a significant extension of the 1930 inventory and digitized stand age as well as all disturbance polygons. We also georeferenced and digitized 61 maps of the Price archive fund showing logged areas, logging previsions and fire polygons for different sections of the Rimouski and Matane sectors. These historical maps were georeferenced using lakes and confined rivers from Quebec ministry of natural resources’ modern maps (MFWP). In addition, we used a map of the LSL region made from a 1938 aerial survey showing fires of the ∼1900–1938 period (Hébert, 1938). This map was previously analyzed by Terrail (2013). Our database also included more recent fire polygons obtained as shapefiles from the fire protection agency in Quebec (SOPFEU). Finally, we used the four decadal forest inventory maps produced by the Quebec’s ministry of Forests, Wildlife and Parks (MFWP) from ground surveys and aerial photographs taken in 1973–1974, 1985–1986, 1993, and 2004, respectively. We digitized disturbance polygons for the first and second decadal inventory maps, while the third and fourth decadal maps were directly obtained as digitized polygons by the MFWP.

Creating a homogenous database of all disturbance polygons was somewhat challenging, mostly due to the variable resolution of maps and the occasional repetition of some polygons in different data sources, sometimes with different dates. When a same polygon was repeated on two maps with a spatial mismatch, both polygons were merged. When dates were inconsistent for a same polygon on two different maps, the oldest date was kept, as it indicates the first mention of the occurrence of the disturbance polygon. On the maps, logging polygons between 1895 and 1965 were identified as Cut Over zones, without a precise logging type. However, these cut over zones were most likely diameter limit selective logging operations, based on written accounts of logging practices at that time (Price Brothers and Company Limited, 1944; Fortin et al., 1993) as well as an analysis of forest diameter structure in survey plots in 1930–1931 of logged areas against unlogged forest (see Appendix 1 and 2). Logging polygons for the period 1965–2005 were classified as partial cuts (25–75% basal area removal) or total cuts (more than 75% basal area removal) based on Quebec’s department of forests decadal inventories. The 1970s SBW severity was equally categorized by Quebec’s department of forests into moderate outbreaks (between 25 and 75% defoliation) and severe outbreaks (more than 75% defoliation).

The four main disturbance categories that have been considered in our analyses are (1) fire (1895–2005), (2) logging (1895–2005), (3) plantations (1955–2005), and (4) SBW (1975–2005). Although the natural or anthropogenic origin of fires is not directly known, their spatial pattern indicates a predominant anthropogenic origin (Terrail, 2013). In addition, even if a plantation is not a disturbance per se, plantations were included in our analysis as a human-imposed forest recovery pattern that caused important compositional difference compared to naturally regenerated stands.

The SBW of 1975–1992 and windthrows are considered natural disturbances in our database for which we have spatial information. Two additional outbreaks occurred in the region during the 20th century in 1947–1958 and 1914–1923 (Boulanger and Arseneault, 2004) for which we have no spatial information. Hence, the impact of SBWs during the 20th century might be under-estimated in our database. It is also important to note that SBWs might be indirectly influenced by human activity through climate change (Gray, 2008). Yet, we included the spatial data we have of the 1970s SBW to compare its extent and rotation period and to examine its interactions with previous and subsequent anthropogenic disturbances. Areas impacted by windthrows are negligible in extent compared to other disturbances and are only included for descriptive purposes of the disturbance regime. For each disturbance type, we estimated affected area (km² and % of the study area), rate (% of the study area /year) and rotation period (i.e., the time necessary for a disturbance to cover an area equivalent to the study area; Frelich, 2002). We also calculated the affected area disturbed per decade and the cumulative area disturbed for the four main disturbance types.

We assessed tree composition in 1930–1931 and in 1985–2005. For the 1930–1931 period, we used a detailed forest inventory conducted by the Price Brothers & Company in a total of 16 345 rectangular plots (10 × 100 m) of 0.1 ha each, systematically distributed across the Rimouski and Matane sectors (Figure 1). Tree taxa (>3 inches; ∼ 7.6 cm) were then tallied by 2 inches DBH classes. We georeferenced plots from their location on a 1930 map at the scale of 1: ∼ 16000. Some taxa in the historical forest inventory were grouped at the genus level: Maples (Acer spp.), Spruces (Picea spp.), Pines (Pinus spp.), and Poplars. Other taxa were identified at the species level: balsam fir, yellow birch, white birch, northern white cedar, black ash (Fraxinus nigra), American larch (Larix laricina), and American elm (Ulmus Americana). Similarly, present-day forest composition (1985–2005) was described from the sampling plots of the second, third and fourth decadal forest inventories (1985–1986, 1993, and 2004) conducted by the MFWP. The first inventory was excluded because plot location is not georeferenced. During these inventories, circular sampling plots of 0.04 ha were randomly stratified according to forest stand type, excluding non-forested, unproductive and inaccessible (slope > 40%) stands. Within plots, tree taxa were tallied by 2 cm DBH classes, although DBH classes <8 cm were excluded to match the 1930 plot database. We grouped the diametral structure (Density of stems / 10 cm diameter class) for each time period. Plots of the 1985–2005 time period were separated into two categories: naturally regenerated stands and plantations. We divided the study area into 333 grid cells of 3 km by 3 km. The cells contained at least four (4) inventory plots per period. We calculated the mean basal area per taxa for redundancy analysis (RDA). The study area division in 3 km by 3 km grid cells should attenuate the bias associated with the uncertainty of historical inventory sampling plots position and the variable resolution of maps used to reconstruct disturbance history. The historical inventory sampling plots position uncertainty range from 0 to 340 m and the resolution of maps used range from 1: 16 000 to 1: 190 000 (Table 1).

Two redundancy analyses (RDA) were performed. First, we used RDA in order to examine the effect of environment and vegetation composition in 1930–1931 on subsequent occurrence of disturbances. Second, we performed a partial RDA after controlling for environment and pre-industrial vegetation in order to examine the influence of disturbances on present-day forest composition. A grid of 333 cells of 3 km by 3 km covering the study area was generated to calculate, within each cell, disturbance cover percentage, environmental variables and basal area of tree taxa in 1930–1931 and 1985–2005. We calculated the cover percentage of Fire (1895–2005), Cut Over (1895–1935), Cut Over (1935–1965), Partial Cut (1965–2005), Total Cut (1965–2005), Plantations (1955–2005), and SBW (1975–2005). For environmental variables (Table 2), we used the mean value of 10 random points per grid cell to calculate altitude (m), slope (°), aspect (°), closest distance from main rivers (m), closest distance from secondary rivers and lakes (m), and closest distance from private lands (m). Cells located inside private lands have a distance of 0 m. We also calculated soil deposit types and drainage classes cover per cell. For the two RDAs, we selected only significant variables using forward selection with the package vegan in R (Tables 4, 5). Disturbances, taxa basal area of 1930–1931 and 1985–2005 variables were Hellinger transformed while environmental variables were standardized to zero mean and unit variance. The variation explained was reported using the adjusted R2, which takes the number of predictor variables and sample size into account to prevent the inflation of R2 values. A permutation test was applied to test for the significance of the models and axes. Partial RDAs were also performed to calculate the individual adjusted R2 for each variable in environmental and forest composition of 1930–1931 and 1985–2005 datasets. These analyses were performed using the R 3.0.3. software (R Development Core Team, 2016).

During the 20th century, logging was the most widespread disturbance type, followed by fire, the 1970s SBW and plantations. Windthrows were negligible compared to other considered disturbances, with a rotation period of 3000 years (Table 3 and Figures 2, 3). The cumulative area logged between 1895 and 2005 attained 144% of the study area (Figure 4B). This indicates the some zones were logged more than once over the last 110 years. In total, 84% of the study area has been subjected to partial and/or total cut: 39% has been logged once, 34% twice, 10% thrice, and 1% logged four times. 16% of the study area has not been logged in our database. Approximately a third of the non-logged area has burned during the last century (Figure 2D). The unlogged, unburned area might have been subjected to non-documented anthropogenic disturbances absent from our data base, is located in inaccessible areas for logging, or reflects the bias associated to maps variable scales in our database. Logging rotation period in the study area between 1895 and 2005 corresponds to 76 years. However, logging rotation period has continuously shortened throughout the 20th century, due to increasing industrial capacity, from 152 years in 1895–1935, to 79 years in 1935–1965, and 47 years in 1965–2005 (Table 3). Fire events were clustered around 1925–1955 (Figure 4A), in conjunction with the population increased in the study region (Figure 4E). Between 1895 and 2005, fires burned 19% of the study area (Figure 4A), corresponding to a rotation period of 594 years. The fire rotation period decreased from 1668 years in 1895–1925 to 200 years in 1925–1955. Following the peak of settlement in the late 1950s, the fire rotation period increased to 2925 years (Table 3). Large scale plantations are relatively recent in the region and have covered 17.8% of the study area between 1955 and 2005 with a peak of 362 km² between 1985 and 1995 (Figure 4C). SBW data are available only for the recent time (1965–2005) and show a peak of 486 km² area affected between 1974 and 1986. The total area affected by the outbreak is 31.7% (Figure 4D) corresponding to a rotation period of 95 years (Table 3).

The RDA model indicates that the anthropogenic disturbances regime of the 20th century has been spatially structured by environmental variables and by the 1930–1931 tree composition (Adjusted R2 = 32%; Figure 5). This structure reflects human settlement and land-use history during the 20th century. The first RDA axis (20% of constrained variance) mostly expresses increasing distance from main rivers to the left (Individual R2adj = 0.084; Table 4) as well as the basal area of spruces in 1930 (Pic30; Individual R2adj = 0.076). The second axis (6% of constrained variance) shows a gradient from high Pines basal area in 1930 (Pin30, Individual R2adj = 0.017), steep slope (Individual R2adj = 0.009) and weathering deposit (Individual R2adj = 0.013) up to Glacial deposit (Individual R2adj = 0.012) and high birch basal area in 1930 (Bea30 and Bep30, respectively, with Individual R2adj = 0.030 and 0.014). Distance from private lands (PR; Individual R2adj = 0.067) and altitude (alt, Individual R2adj = 0.007) are explained by both the first and second axis of the RDA. Cut over zones between 1895 and 1935 (CO35) are situated next to main rivers. Between 1935 and 1965, logging activities (CO65) moved away from main rivers (situated in the opposite direction of CO35 in the multivariate space) toward areas where spruces were still abundant. Between 1965 and 2005 logged areas (CP05 and CT05) became diffuse across the study region and showed no specific spatial location in the multivariate space. Plantations (1955–2005; PL) occurred in flat zones, while SBW is observed in steep zones. Finally, fires (1895–2005) occurred in low altitude areas close to private lands.

According to the partial RDA, the 20th century disturbance regime explained 6% of the modern basal area (1985–2005) in the landscape (Figure 6). The first RDA axis mostly express a gradient of partial cut (1965–2005) (CP05; Individual R2adj = 0.021; Table 5) while the second axis shows a gradient from fire (FIRE; Individual R2adj = 0.014) to spruce budworm outbreak (SBW; Individual R2adj = 0.008). The third axis shows a gradient of plantation (PL; Individual R2adj = 0.011). Partial cuts between 1965 and 2005, fires and plantations explain 4.6% out of the 6% of the variance among disturbance variables. Trembling aspen (Pot) is correlated to burned areas. Sugar maple (Acs) and yellow birch (Bea) are correlated to partial cuts between 1965 and 2005. Black spruce (Pim) is associated with plantations. Northern white cedar (Tho) is not explained by any studied disturbance type as it is perpendicular to all disturbances in the multivariate space.

Between historical (1930–1931) and present-day (1985–2005) periods, we observed a change in basal area and diameter structure of tree taxa. Total basal area increased noticeably by 5.1 m²/ha in naturally regenerating stands, while it decreased by 2.8 m²/ha in plantations (Table 6). Most plantations were likely thinned at the time of sampling. The increase in basal area in naturally regenerating stands is largely due to a threefold increase of northern white cedar by 2.7 m²/ha and a nine fold increase in maples by 1.5 m²/ha. This increase was accompanied by an increase in the density of northern white cedar in all diameter classes and an increase in maples density for small diameter classes (Figures 7, 8). A slight increase in basal area for spruces, balsam fir and poplars and a decrease in yellow birch and white birch also occurred (Figure 8), although an increased density of stems in small diameter classes occurred for these species in naturally regenerated stands (Figure 7). In plantations, 41% (7.1 m²/ha) of the mean basal area is represented by Spruces (Table 6) with a significant increase in stem density in the smallest diameter class for this taxa compared to historical and modern naturally regenerated stands values (Figure 7). The mean basal area of secondary natural regeneration of balsam fir (fir has not been planted) in plantations represent 44% (7.6 m²/ha). Balsam fir density has slightly increased in the smallest diameter class compared to historical and modern naturally regenerated stands values (Figure 7).

During the 20th century, anthropogenic disturbances were omnipresent in the landscape, as for other regions of the eastern North American mixed forests (Foster, 1992; Whitney, 1994; Friedman and Reich, 2005). Fire activity was concentrated at low elevation during the settlement period, close to private lands suggesting ignitions from anthropogenic origins (Figures 3A, 5). Logging activities expanded from around main rivers during the first half of the 20th century toward inland areas and higher altitude by the second half of the century (Figures 2, 5), marking the transition from log transport by watercourses to the use of forest roads and trucks. Moreover, 18% of these logged zones have been subsequently transformed into plantations, mostly between 1985 and 1995, during the last SBW (1975–1990). Planted zones were chosen in backcountry accessible sectors and mostly planted with insect resistant black spruce (Pim), while the less accessible sectors on steeper slopes were disturbed by the SBW outbreak (Figures 5, 6).

Before European settlement, the pre-industrial disturbance regime of south-eastern Quebec and northern New England was dominated by secondary disturbances like windthrows and insect outbreaks (Blais, 1961; Lorimer, 1977; Bormann and Likens, 1979; Payette et al., 1990; Frelich, 2002; Lorimer and White, 2003). For the LSL region, this can be inferred from the old-growth, uneven aged forest structure of the pre-settlement forest, which was also dominated by fire-sensitive species (balsam fir, white spruce, eastern white cedar) (Etheridge et al., 2005; Boucher et al., 2009b; Dupuis et al., 2011). The very long rotation period we estimated for windthrows in the study area over the 20th century (about 3000 years; less than 1% of the landscape affected; Table 3) is considerably longer than the value of 1150 years estimated by Lorimer (1977) for Northern Maine between 1793 and 1827. This discrepancy may emphasize a relatively low occurrence of large windthrows in the study area. Indeed, wind storms that create large canopy openings are known to be infrequent in this type of mixed forests (Frelich and Lorimer, 1991; Lorimer and White, 2003). Nevertheless, windthrows could be more frequent with high harvesting intensity in partial cut stands (Montoro Girona et al., 2019). Although tree ring evidence indicate a time interval of 40 years between successive SBWs in the LSL region over the past 450 years (Boulanger and Arseneault, 2004), the rotation period for presettlement SBW s is hard to calculate due to the lack of spatial data.

Escaped-settlement fires and extensive logging during the 20th century have strongly altered the type, rotation period, severity and extent of disturbance events (Table 3). Compared to a fire rotation period of 200 years at the peak of human settlement in the region (1925–1955), the value of 1668 years calculated for the 1895–1925 interval is closer to values suggested elsewhere for pre-industrial landscapes of the temperate zone. Natural fires were documented to be rare in the presettlement land surveys in the temperate forests of eastern North America (Siccama, 1971; Lorimer, 1977; Foster et al., 1998). For example, recently burned areas covered 10% of the landscape in Northern Maine shortly before settlement, with a rotation period estimated to be 800 years (Lorimer, 1977). An even longer fire rotation period of 1240 was reported for spruce-fir forests in Maine (Fahey and Reiners, 1981). At the same time logging evolved from partial cutting diameter-limit management, associated to the saw mill industry, at the beginning of the century to total cutting, associated to the pulp and paper industry, by the end of the century (Fortin et al., 1993; Boucher et al., 2009a, c). In the past decades, efforts for forest sustainable management suggested more partial cutting as an alternative to clearcutting in the province of Quebec and elsewhere in eastern North America (Landres et al., 1999; Harvey et al., 2002; Lindenmayer and Franklin, 2002). Overall, logging covered 144% of the study area between 1895 and 2005, with 45% of the study area logged more than once.

Mixed forests in our study area showed resilience in terms of composition, despite the omnipresence of anthropogenic disturbances during the 20th century. There is a correlation between tree taxa basal area in 1930–1931 and present-day taxa basal area (Figure 8). At the beginning of the century, diameter limit selective logging was conducted in winter and with minimal machinery (Fortin et al., 1993). This might have protected coniferous regeneration and limited the expansion of early successional deciduous tree species, contributing to the overall observed resilience of forests to logging operations at the beginning of the 20th century (Table 5). A study of Boucher et al. (2017) have used a similar database and disturbance data, as in this study, in the southern boreal forest of central Quebec, and have concluded that logging have a minor impact on compositional change.

Disturbances of the 20th century explain only 6% of present-day tree composition (Figure 6), suggesting that they have not been an important agent of forest reorganization in our study area. This contrasts with more salient views in the literature which emphasize the impact of anthropogenic disturbances on forest composition (Whitney, 1994; Foster et al., 1998; Fuller et al., 1998; Friedman and Reich, 2005; Dupuis et al., 2011; Boucher et al., 2014; Fortin, 2018; Danneyrolles et al., 2019). Many of these studies use vegetation data from historical land surveys (frequency of occurrence and dominance). The difference between our study area and other studies using historical land surveys might be that our study area is mostly located in public territory, while these studies were mostly located in private lands. More severe and frequent anthropogenic disturbances are likely to occur in private lands, as in the case of anthropogenic fires in our results (Figure 5). Moreover, the response of vegetation to disturbances might be site specific and could have been attenuated by the use of a 9 km² grid cells. The influence of total cuts on vegetation composition might not be fully ceased by our analysis (CT05 adj R2 = 0.005; Table 5). Although naturally regenerated stands originating after total cuts between 1965 and 2005 cover 25% of the study area, many of the trees in these stands are 40 years old or younger and have smaller DBH than the 7 cm minimum DBH for trees measured in our database. Finally, the relatively short period of disturbance data availability (110 years) in this study and lacking data for some disturbance episodes (SBW) might have contributed in diminishing the effect of disturbances in present-day forest composition. Disturbances that have occurred long before our study might as well have a non-measured influence.

Despite the relatively low importance of disturbances to explain the present-day forest composition, some taxa responded to disturbances in an individualistic manner, reflecting their functional traits. The increase of early successional trembling aspen in naturally regenerated stands compared to the 1930–1931 time period (Figures 6–8 and Table 6) is correlated to the important increase in fire events. Several previous studies have linked the increase of trembling aspen with anthropogenic fires in eastern North America (Bergeron, 2000; Cleland et al., 2001; Friedman and Reich, 2005; Boucher et al., 2014, 2017; Terrail et al., 2019). We also found that the fire sensitive balsam-fir (Rowe, 1972) is mostly abundant away from fire polygons (Figure 6). Species usually associated with gap dynamics, such as yellow birch and sugar maple, are correlated to partial cuts (1965–2005). These species might have benefited from the elimination of shade tolerant competitors in partial cuts as well as in areas impacted by SBW (Figure 6; Palik and Pregitzer, 1992; Abrams, 1998; Nie et al., 2018). Moreover, red and sugar maples produce a large quantity of seeds and can grow rapidly when exposed to light (Fei and Steiner, 2008; Nolet et al., 2008). In addition, management practices which aim at preserving yellow birch seed trees in logging operations might have favored the abundance of this species.

Planted areas have also participated to the observed present-day forest composition at the landscape level. They are human made ecosystems which are less diverse than naturally regenerated stands (Table 6), as 41% of plantations basal area is composed of spruces against 19% in natural regenerated stands (Table 6 and Figure 7). Other tree species found in planted stands consist in naturally regenerated balsam fir and early successional white birch.

Surprisingly northern white cedar, a late successional shade tolerant species increased in basal area and was not correlated to disturbances (Table 6 and Figure 6). This result is somewhat contrary to the general tendency of the decreased cedar abundance in eastern Canadian mixed forests, even though variable between regions (Danneyrolles et al., 2017). It’s present-day increase and distribution in the landscape is probably better explained when environmental factors are considered. Northern white cedar is usually found in zones with poor drainage, organic deposit and in low altitude (Robitaille and Saucier, 1998). The study area’s position in public territory might have also made northern white cedar exploitation less frequent. Yet, the interaction with partial natural and human disturbances (insect outbreaks and partial cuts) combined with an abundant advanced regeneration of northern white cedar was suggested to explain the dominance of this species (Heitzman et al., 1997; Ruel et al., 2014; Danneyrolles et al., 2017).

Determining the contribution of anthropogenic disturbances on forest composition should be studied along with other agents and their interactions at various spatial and temporal scales (Krankina et al., 2005; Gimmi et al., 2010; Landhäusser et al., 2010; Fisichelli et al., 2014; Nowacki and Abrams, 2015; Plieninger et al., 2016; Vayreda et al., 2016; Danneyrolles et al., 2019). For example, Boisvert-Marsh et al. (2019) have suggested that disturbance of the last 40 years have been less important than climate change to explain present-day forest composition, whereas Danneyrolles et al. (2019) concluded the contrary at the scale of the last century. Moreover, the characteristics of disturbances themselves (extent, frequency, intensity, type, return interval) and their combination might also have different influences on vegetation composition (Turner, 2010). In the present study, the conjugated analysis of disturbance history and associated compositional change over the transition period encompassing the early settlement of the population and the deployment of industrial forestry suggest moderate change in forest cover composition despite extensive anthropogenic disturbances in this representative eastern Canadian mixed forest.

Our study supports the recommendations of forest ecosystem management practices to use partial cutting to maintain the forest in its natural range of variability (Landres et al., 1999; Harvey et al., 2002; Lindenmayer and Franklin, 2002). On the other hand, further investigation is needed to study the impact of clear cutting as the length and data of our study don’t permit the estimation of the impact of this stand replacing disturbance. Paleoecological studies which could encompass longer time-frames could present precious complementary information on forest composition and natural disturbance history at the millennial scale.