Widespread Unquantified Conversion of Old Boreal Forests to Plantations

Across the boreal biome, clear‐cutting of old, previously non clear‐cut forests with high naturalness followed by tree planting or seeding is a major land use change. However, how much previously uncut forest has been converted to plantations remains unquantified. We combine Swedish national databases on clear‐cuts and forest inventories to show that at least 19% of all clear‐cuts since 2003 have occurred in old forests that were most likely not previously cut and planted or seeded. Old forests have been cut and lost at a steady rate of ∼1.4% per year for the same period, and at this rate they will disappear by the 2070s. There is further evidence that this type of unreported forest conversion is occurring across much of the world's boreal forest.

tropical forests to oil palm plantation (Chong et al., 2017). Partly due to these difficulties, countries with boreal forest do not systematically map or track such conversions (Harris et al., 2019;Naturvårdsverket, 2020).
As a first insight into the potential extent, rate and spatial pattern of conversion from older uncut forest to planted and seeded forest in the boreal zone we use a uniquely detailed set of databases on forest structure and clear-cutting available in one major boreal forest nation: Sweden. Sweden has the largest forest cover in Europe (excluding Russian Federation) and is a major exporter of sawnwood and paper (8% of world exports in 2019) (FAO, 2021). Forestry in Sweden is currently almost exclusively rotation-forestry, where large areas of forests are clear-cut, planted, thinned and then cut again over a time scale of 50-100 years. This management follows government regulations designed to promote wood production, and lately, to limit negative impacts on biodiversity (Lindahl et al., 2017). Examples of an uncut forest and a planted forest are shown in Figure 1.

Methods
We cross-referenced the stand ages of a large database of forest inventory plots (n = 74,288) surveyed from 1996 to 2019 with a national clear-cut database to determine the rate and pattern of clear-cutting in differently aged productive forests. The clear-cut database comprises spatially explicit geographical polygons delineating close to a million compulsory reported individual cuts from 2003 until 2019, verified by the Swedish forest agency (Skogstyrelsen, 2021a). The inventory plots were established by the Swedish forest inventory on five regional grids spanning the whole country, designed to provide an unbiased sample of unprotected productive forest land nationally (Fridman et al., 2014). A basal area weighted stand age is calculated in each inventory plot, based on age determination of tree cores from a subset of cored trees (3-5 trees). In plots with uneven tree age, this implies that the oldest tree in the plots will be older than the stand age.
To produce estimates of forests and clear-cuts for all of Sweden from the NFI plot level data we averaged regional estimates using the productive forest area as weights and estimated confidence bands using bootstrapping. Further details on the methods with associated figures and an uncertainty assessment are available in Supporting Information S1.

Using Stand Age to Infer Naturalness
The only available map of potentially uncut forest does not include southern Sweden and is based on aerial photography and satellite images where the earliest photos were taken 1955 to 1970, depending on location (Ahlkrona et al., 2017), implying that it is unlikely that this map will capture early clear-cuts.
Since there is no national map indicating whether a forest has been clear-cut and planted or seeded we relied on historical records to relate plot-level stand age with the likelihood of past clear-cutting and planting or seeding. Analysis of the discussion and articles in the Journal of the Forest Association of northern Sweden between 1883 and 1960 (Lundmark et al., 2013) and a regional study using aerial photography from the 1940s indicates that the widespread adoption of rotation-forestry in Sweden occurred between 1900 and 1950, with ∼10% of the forest area being clear-cut by the early 1900s (Lundmark et al., 2021). Forests with a current mean plot-level stand age substantially predating this transitional period (predating 1880 for the purpose of our analysis) have therefore likely never been clear-cut. We analyzed forest plots based on the year their stand age predated instead of an approach based solely on stand age which changes over time and where the loss of old uncut forests can be compensated by an increasing share of planted and seeded forests when they age (Skogstyrelsen, 2021b). Because also young forests may not have experienced clear-cutting but regenerated following natural disturbances or more intensive selective logging, an unknown fraction of younger forests not predating 1880 may also be previously non clear-cut forests. For all age classes of potentially uncut forests, their share of clear-cuts is ∼2 times higher than their relative abundance across the country, implying that old uncut forests are being inadvertently or otherwise targeted (Figure 2a).

Results
By tracing the fate of all unprotected forest plots with stand age predating 1880 and inventoried between 1996 and 2002, we find that forest land predating 1880 have been cut at a remarkably stable rate of 1.40% (95% CI: [1.33, 1.46]) per year since 2003 (Figure 2b). The clear-cutting of these older forests occurs across the country but is more common in the north, in large part because many of the older forests in southern Sweden were already clear-cut by 1996 (Figure 2c). At this rate, all unprotected older forests in Sweden will be converted to planted or seeded forest within decades (year ∼2073, 95% CI: [2070,2076] (2021)). The unprotected, old and non clear-cut forests may provide a richer set of ecosystem values than similar forests in currently protected areas because the unprotected forests are generally found in warmer, more productive regions.

Discussion and Conclusion
The conversion of old natural forest to planted and seeded forests has potentially important implications for ecosystem functions and local communities but remains poorly understood in part because the pattern and rate of the forest conversion has not been systematically monitored. The extent to which clear-cut and planted and seeded forests achieve the same ecosystem functions and values as uncut, natural and freely developing forests, and over what time scale, remains poorly known. Given this lack of knowledge and the rate at which the old forests are cut, further work addressing these knowledge gaps is an urgent priority.
Beyond Sweden, large scale analysis and modeling of high-resolution satellite remote sensing data suggests that forestry is the major driver of widespread harvest (around 10% gross tree cover loss between 2001 and 2021 in Canada and the Russian federation (World Resource Institute, 2014)) of more southern and more productive forests throughout the boreal biome (Curtis et al., 2018). A regional study from eastern Canada suggests that the majority of this harvest is associated with clear-cuts of the most productive old boreal forest stands (Martin et al., 2020). These regional and remote sensing-based assessments echo our results for Sweden, which together with other supporting scientific literature (Bradshaw et al., 2009;Cyr et al., 2009;Potapov et al., 2012) and some anecdotal evidence from media and NGO reports (Eremenko, 2014;Lindwall, 2019;Vinyard & Skene, 2020) strongly suggest that this process is not limited to Sweden but widespread across the boreal forest zone.
We argue that it is crucial that these land conversions are mapped and monitored across the boreal zone in a coordinated and systematic way, and that targeted studies investigate the implications of the conversions for carbon storage, biodiversity, recreational and cultural values. Given the widespread disapproval about similar conversions in tropical countries we call upon northern governments to urgently investigate these ongoing large-scale conversions in their own territories, to map their scope, analyze their implications and, where feasible, to protect the forests that have not yet been clear-cut.
Data Availability Statement the data that forms the basis for this paper. AA acknowledge funding from the Crafoord Foundation, Ebbe-Kocks Stiftelse, Längmanska Kulturfonden, and the Swedish Research Council (2021-05344). The study is a contribution to the research environment Biodiversity and Ecosystem Services in a Changing Climate (BECC) at Lund University. The authors thank two anonymous reviewers for their thoughtful feedback.