Latitudinal decline in stand biomass and productivity at the elevational treeline in the Ural mountains despite a common thermal growth limit

To quantify tree biomass and stand productivity of treeline ecotones and identify driving factors.


| INTRODUC TI ON
Climate warming during the past century has repeatedly been shown to have profound effects on the productivity, distribution and diversity of vegetation throughout the world (Walther et al., 2002). The most pronounced temperature increases have occurred at high elevations and high latitudes (Pepin et al., 2015;Stocker et al., 2013), and ecosystems in these cold regions have experienced particularly striking changes (e.g. Devi et al., 2008;Steinbauer et al., 2018).
Responses of vegetation to changes in climate are expected to be rapid and extreme in ecotones, boundary ecosystems where plant life-forms and soil conditions change rapidly over relatively short distances (e.g. Allen & Breshears, 1998).
The treeline ecotone, where closed forest transitions to alpine or arctic tundra, is both a boundary and a cold ecosystem and is thus particularly valuable as a bio-indicator of climate change (Holtmeier, 2003). Indeed, treelines in many regions of the world have shifted to higher elevations or higher latitudes over the last century (Hagedorn et al., 2014;Harsch, Hulme, McGlone, & Duncan, 2009;Kullman & Öberg, 2009;Lloyd, 2005;Shiyatov, Terent'ev, Fomin, & Zimmermann, 2007). The expansion of forests into alpine and arctic tundra can impact plant productivity and diversity (Gazol, Moiseev, & Camarero, 2017) and has important implications for carbon and nutrient cycling (Hagedorn, Gavazov, & Alexander, 2019;Kammer et al., 2009;Speed et al., 2015), as well as snow accumulation and albedo (de Wit et al., 2014;Schwaab et al., 2015), which can in turn affect ecosystem functioning and result in further vegetation change.
It is therefore highly relevant to understand how environmental conditions influence current treeline ecosystems and contribute to future shifts in treeline position and characteristics.
The positions of high elevational treelines throughout the world follow an isotherm of a mean growing season temperature of 6-8°C (Körner, 1998;Körner & Paulsen, 2004;Müller et al., 2016), strongly suggesting that the high-elevation limit of tree growth at the global scale is primarily driven by low-temperature constraints on growth processes (Holtmeier, 2003;Körner, 2012). The importance of temperature in treeline formation has been supported by studies documenting treeline shifts in association with rising regional temperatures, although warmer winter conditions with more snow seem to have contributed to treeline changes as well (Hagedorn et al., 2014;Harsch et al., 2009). Other factors have also been found to determine treeline position and characteristics at the local level and may drive tree growth rates once a low-temperature limitation is released. For example, low nutrient availability has been linked to low rates of photosynthesis and growth of trees at the alpine treeline in Alaska (Sullivan, Ellison, McNown, Brownlee, & Sveinbjörnsson, 2015), and increased growth of trees with nutrient addition has been observed in treeline ecotones in the subarctic (Susiluoto, Hilasvuori, & Berninger, 2010;Sveinbjornsson, Nordell & Kauhanen, 1992) and in the Swiss Alps (Möhl et al., 2019).
Most treeline studies have focused on understanding drivers and limitations of tree aboveground growth. In comparison, knowledge of tree biomass and productivity, especially regarding the belowground compartment, and the environmental factors influencing biomass production is severely limited, probably because of the enormous field effort and logistical challenges involved in measuring biomass at remote treeline locations. Investigations including estimates of biomass pools and productivity have mainly come from a single or small number of sites (Bernoulli & Körner, 1999;Liu, Nie, Kong, & Luo, 2016;Moiseev, Bubnov, Devi, & Nagimov, 2016) and/ or have been confounded with experimental manipulation (Dawes et al., 2015;Speed et al., 2015). At larger scales, tree biomass and vegetation productivity have increasingly been estimated by remote sensing techniques (e.g. Park et al., 2016). The resolution provided by satellite images is, however, too coarse to assess treeline dynamics, and the use of aerial photographs or Light Detection and Ranging (LiDAR) requires ground truthing (Coops, Mörsdorf, Schaepman, & Zimmermann, 2013) through labour-intense tree allometry determination. Therefore, larger-scale investigations including detailed measurements of stand structures, tree biomass and productivity within natural treeline ecotones are valuable for improving our understanding of C dynamics and the overall climate balance including albedo effects through vegetation changes in current treeline ecosystems, as well as for predicting how these systems will be altered by global change (De Wit et al., 2014;Schwaab et al., 2015).
In this study, we investigated treeline patterns along a northsouth gradient in the Ural mountains of Russia. The 13.3° latitudinal range of the study sites comprises treelines with different tree species and varying aspects, climatic conditions and soil properties, making it possible to differentiate between local and larger-scale factors shaping treelines. In contrast to other European mountain ranges, the Urals have never been disturbed by extended human activities or exposed to considerable regional air pollution (Hagedorn et al., 2014). Comparisons of historical and recent photographs coupled with detailed analyses of tree demography have demonstrated a clear upward shift in treeline, by 4 to 8 m per decade, across treeline ecotones. The observed latitudinal decline in stand productivity is important for above-and belowground diversity and functioning.

K E Y W O R D S
allometry, climate, growing season length, nitrogen, permafrost, phosphorus, soil, treeline the north-south extent of the Urals (Devi et al., 2008;Hagedorn et al., 2014;Moiseev et al., 2016;Shiyatov & Mazepa, 2015). Here, we combine demographic information with detailed biomass measurements of three key treeline tree species to estimate stand-level biomass and productivity within the treeline ecotone across the entire Ural latitudinal gradient. Our objectives were: (a) to quantify the biomass and productivity of trees along elevational gradients spanning the treeline ecotone, across the north-south extent of the Ural mountains; and (b) to identify the key factors influencing treeline position and productivity of treeline trees along the latitudinal gradient by relating tree data to regional climate records and site properties, including multi-year data on temperature and soil conditions. We hypothesized that treeline position is primarily related to growing season temperature, whereas tree productivity is driven by other factors, such as soil fertility, whose importance increases with increasing distance downward from the treeline.

| Study sites
Our treeline study was conducted at 18 sites across a 1500-km latitudinal gradient in the Ural mountains (54.5 to 67.9°N), span-  Table S1). The elevation of the treeline ecotone decreases from 1225-1375 m a.s.l. in the South Urals to 150-300 m a.s.l. in the Polar Urals ( Figure 1, Table S1). The dominant tree species are as follows: Siberian spruce (Picea obovata Ledeb.) and white birch (Betula pubescens Ehrh. ssp. tortuosa (Ledeb.) Nyman) in the South Urals; spruce, birch and Siberian larch (Larix sibirica Ledeb.) in the North Urals; and larch and birch in the Sub-Polar and Polar Urals ( Figure 1). Although tree species co-exist in each region, each transect was clearly dominated by one species. In the South Urals, the tundra is dominated by dwarf shrubs (Vaccinium uliginosum L., Vaccinium vitis-idaea L.), grasses (e.g. Festuca igoschiniae Tzvelev) and sedges (e.g. Carex vaginata Tausch  Table S1). The transects were located on shallow, evenly inclined slopes (4-15°) with different aspects (Table S1). Each transect consisted of three elevation levels: the tree species line (tree individuals or islands of multi-and single-stemmed trees with heights of more than 2 m, distances between trees from 20 to 60 m, and a total crown cover of 5%-10%); the open forest line (distances between trees from 7 to 30 m and a total crown cover of 20%-30%); and the closed forest line (continuous forest with distances between trees < 7 m and a total crown cover greater than 50% (as defined by Shiyatov et al. (2007) and Hagedorn et al. (2014)

| Climate data
Climate data were obtained from a combination of direct monitoring in or near the transects and external sources. On three

| Stand characteristics
In each plot, all saplings taller than 20 cm and all trunks of single-or multi-stemmed trees were recorded (N = 20,600). We mapped the location of each stem and measured its height, diameter at the base and breast height, and projected ground area covered by the crown.
The age structure of all plots was determined by dendrochronological methods, following protocols used by Hagedorn et al. (2014) and Moiseev et al. (2016). From trees with a diameter ≥ 3 cm at the stem base, we took a single tree core at a height between 0 and 30 cm from every second single-stemmed living tree and from every fourth stem of every multi-stemmed tree. From every second tree taller than 0.2 m, but < 3 cm in basal diameter, we sampled stem discs at the root collar.
All cores were mounted on wooden strips. Cores and stem discs were cleaned with both a paper knife and a shaving blade.
After enhancing ring boundary contrasts with white powder, samples with narrow annual rings were measured on the linear table LINTAB-V (F. Rinn S.A., Heidelberg, Germany) to a precision of 0.01 mm and were cross-dated using the computer programs TSAP-3.0 (Rinn, 1998) and Cofecha (Holmes, 1995). Samples with wide rings were visually cross-dated, paying special attention to frost and light rings. The dates of tree germination (for single-stemmed trees) or the start of upright growth of individual trunks (for multi-stemmed trees) were estimated by correcting for the number of years required to grow to the height of sampling and for the number of years to the pith when the core missed the inner ring. For cores hitting the pith, the distance to the centre of the tree was estimated by fitting a circular template to the innermost curved ring (Braeker, 1981). The number of years it took for a stem to grow to the core height was determined from a regression of tree age with height established for all seedlings and saplings at each study site. At all sites, tree age and height were significantly related to each other with an exponential relationship (R 2 > 0.6, P < 0.001).  Table 1).

| Tree and stand biomass
The aboveground biomass of model trees was determined by separating felled trees into stem wood and bark, branch wood and bark, needles or leaves, and dead branches, as described by Moiseev et al. (2016). The fresh mass of stems, including bark, was determined in the field by cutting them into 1-m sections and weighing them to a precision of 50 g. The percentage of dry matter in this wood and the bark biomass fraction was determined for cross-cut samples from the butt end of the sections. These sections were weighed in the field to a precision of 0.1 g and then transported to the laboratory to measure dry weight. To determine the biomass of the tree crown and its structural components, all branches (including leaves) were cut off and divided into three groups with respect to their location in the upper, middle or lower section of the crown. The total fresh mass of each group was measured in the field, and then the crown was divided into foliage-bearing and foliage-less parts and weighed separately. For a sample from the foliage-bearing part (20%-30% of total crown fresh mass), the foliage was removed from the branches and weighed. A sample from the foliage-less part (5%-10% of total crown fresh mass) was used to determine the proportions of wood and bark in the branches. The fraction of dry matter in foliage was determined for 20-g samples from each section of the crown.
In the South, North and Polar Urals, we estimated belowground biomass by excavating the coarse root system (threshold > 5-cm root diameter) for a subset of the model trees, covering the full elevation range of the transects (53 trees total; Table 1). The weight of excavated roots was estimated directly in the field to a precision of 50 g, and roots were then transported to the laboratory for dry weight a) Aboveground Note.: Parameters were determined for different tree species, for trees with a single-stemmed (single) or multi-stemmed (multi) growth form (South Urals only), and for different elevation levels within the treeline ecotone. The number of model trees used to establish the relationship (n) and the coefficient of determination (R 2 ) for the relationship based on these model trees are given. a Square of cross section of stem base used instead of stem basal diameter.
TA B L E 1 Parameters of exponential equations (y = ax b ) modelling the dependence of (a) total aboveground biomass (kg tree −1 ) and (b)  As forest stands of the treeline ecotone represent the first tree generation established after the Little Ice Age in 1850, we can estimate the 'apparent' stand productivity by dividing tree biomass by tree age. Aboveground stand productivity was calculated by dividing stand biomass by the average age of the stand. In addition, the change in stand aboveground biomass and in aboveground stand productivity per metre of elevation change were calculated by dividing the difference in biomass or productivity between the tree species line and the closed forest line by the total elevation difference in the transect (37-126 m).

| Soil measurements
In the South Urals (Mali Iremel' and Bolschoi Iremel', 2 transects), North Urals (Konzhakovskii Kamen', 1 transect) and Polar Urals (Tchernaya, 2 transects), soil properties were measured at the tree species line in mid growing season (July-August). In three plots from each transect, we collected the L-and F-layer from a 20 × 20 cm area using a frame at locations under the tree canopy and in open areas.
In the South Urals, samples from the deeper soil were taken using a corer with an inner diameter of 2 cm. Specifically, eight soil cores were taken from 0-to 5-cm depth, six cores from 5-to 10-cm depth, and four cores from 10-to 20-cm depth and from 20-cm depth down to the bedrock, which occurred at an average soil depth of 29 cm.
In the stone-rich North and Polar Urals, we sampled soils using two quantitative soil pits per sub-plot, where soils were excavated down to the bedrock (usually less than 20-cm depth) from an area of approximately 20 × 20 cm. The exact soil and stone volume for each depth layer was determined by measuring the pit's dimensions with a ruler or by filling the pit with a known volume of sand. At the field station, we carefully removed root biomass and gravel/stones from the soil samples using a 4 and a 2 mm sieve. An aliquot of the soil was transported to the laboratory, where a portion of the soil was freeze-dried for the determination of water content and texture and for chemical analysis. Clay, silt and sand contents were measured by the sedimentation method according to Gee and Bauder (1986).
Inorganic nitrogen (N) concentrations were extracted with 1 M KCl, using a 1:10 ratio for 1 hr for soil from 0-to 5-cm depth and 1:5 for 1 hr for soil from 5-to 10-cm depth and 10-to 20-cm depth.
In the extracts, NH 4 + concentration was measured by automatic flow injection analysis (PE FIAS-300, Perkin-Elmer, Waltham, MA, USA) and NO 3 concentration by ion chromatography (DX-120, Dionex, Sunnyvale, CA, USA). Soil pH was measured potentiometrically in the same KCl extracts. Extractable phosphorus was determined using the P(Bray I) method, extracting soils with NH 4 F for 1 min (Irving & McLaughlin, 1990). For soil C analysis, subsamples were dried at 40°C for 24 hr and ground with a ball mill. Carbon concentrations were measured using a CN-analyser (Euro EA 3,000, HEKAtech GmbH, Wegberg, Germany) interfaced with a continuous flow isotope ratio mass spectrometer (Delta-V Advanced IRMS, Thermo GmbH, Bremen, Germany). Carbon and nutrient pools were estimated by multiplying the element concentration by the mass of fine earth per area and depth increment. Inc., Shelton, CT, USA) and then measuring total P and K concentrations using ICP-OES (Optima 7300 DV, Perkin Elmer).

| Statistical analysis
We assessed the effect of latitude on tree variables at the stand level (aboveground, belowground and total biomass, belowground to aboveground biomass ratio, aboveground productivity, stand density, projected crown area) and individual tree level (mean age, aboveground biomass, height, basal diameter) with linear mixed-effects models fitted with the restricted maximum likelihood method (Pinheiro et al., 2016). The random effects structure of the statisti- Response variables were square-root transformed (stand aboveground and belowground biomass and productivity, stand density) or log transformed (projected crown area, tree age, tree aboveground biomass, tree height, tree basal diameter, larch foliar nutrient concentrations, soil properties) when necessary to meet assumptions of normality and homoscedasticity of the residuals. We considered fixed effects significant at P < 0.05. All analyses were performed using the nlme package of R version 3.3.3 (R Development Core Team, 2017).

| Climate conditions
In the South and Polar Urals, data from loggers placed at 10-cm soil growing season radiation and winter air temperature were significant (Pearson correlation r > 0.9, P < 0.05; Table S2). Soil temperature Air temperature (Table 1), we calculated that stand aboveground biomass averaged across all 18 transects in the Urals was 27 ± 3 t/ha (mean ± 1 SE, averaged across the three elevation levels). Estimates of long-term annual stand aboveground productivity averaged 0.32 ± 0.04 ha -1 y -1 ( Figure 4). Averaged across the three elevation levels and across the 12 transects where data were available, stand belowground (coarse root) biomass was 13 ± 1 t/ha, yielding a total stand biomass of 43 ± 4 t/ha and an average belowground to aboveground biomass ratio of 0.42 ± 0.03 (Table 2) (Table 3). Stand aboveground biomass averaged across all transects increased from 2.7 ± 0.5 t/ ha at the tree species line to 46.8 ± 5.5 t/ha at the closed forest line. Including slope aspect and associated interactions did not improve the model fit for any of the measured variables, and these explanatory variables were therefore excluded from final models. With increasing latitude, there was a significant decrease by a factor 2.5-3.3 in stand total biomass (modelled slope with latitude, averaged across the three elevation levels = −2.5 t/ha°N −1 ) and aboveground biomass (−2.3 t/ha°N −1 ), as well as in stem density (−73 stems/ha°N −1 ; Figure 4, Table 3). Long-term aboveground net stand productivity, estimated by dividing biomass by mean tree age, exhibited the same latitudinal pattern as stand biomass, showing a decline by a factor of 5.2 along the latitudinal gradient ( Figure 4,   Figure S1). Overall, mean tree age was 78 ± 5 years. The estimates of mean tree ages were not biased by systematic differences in forest age structures with latitude ( Figure S1; for tree spe-  Table 3). This result was supported by a significantly smaller change in stand productivity (but not biomass) per height metre, by a factor of three, along the latitudinal gradient towards the Polar Urals ( Figure 4). Results regarding stand productivity did not differ qualitatively if values were calculated using stand total biomass (aboveground compartment plus coarse roots; data not shown). Stand coarse root biomass and, to a lesser extent, projected crown area also tended to become smaller at higher latitudes, but this effect was not significant (Table 3). A larger reduction in aboveground biomass than in belowground (coarse root) biomass with increasing latitude meant that the ratio of belowground to aboveground biomass at the stand level increased significantly moving north along the latitudinal gradient (Table 2).

In contrast to stand-level variables, at the individual tree level
there was no detectable change with latitude in mean aboveground biomass (mean across all transects and the three elevation levels ± 1 SE = 25 ± 3 kg), height (4.4 ± 0.3 m) or basal diameter (9.8 ± 0.7 cm; Table 3). These results demonstrate that the latitudinal pattern observed for stand-level biomass and productivity was primarily driven by a change in stem density.
Stand aboveground biomass and productivity correlated significantly with growing season length, growing season radiation and winter air temperature, which all correlated with latitude (Pearson correlation r > 0.9, P < 0.05; Table S2). In contrast, stand biomass and productivity did not correlate with summer temperatures or GDD 5 .

F I G U R E 3
Latitudinal patterns of climatic data  reconstructed from meteorological stations along the Ural mountain range using in situ data.

| Soil properties and nutrient pools
Soil properties showed significant differences among the three regions where data were collected and indicated less developed soil in the northern regions (Table 4). In the Polar Urals, the surface was patterned and soils showed signs of cryoturbation, indicating that soils were affected by permafrost. Active layer depth was greater than 20 cm. No signs of permafrost were visible in the South or North Urals, and soil depth decreased from the South to the Polar Urals. Consistent with these results, stone and sand contents were higher in the Polar and North Urals than in the South Urals (Table 4).
In addition, total soil organic C stocks (soil surface down to bedrock) declined from the South to the North and Polar Urals, and extractable N and P pools (0-to 20-cm soil depth) were approximately 20 F I G U R E 4 Stand above-and belowground biomass, stem density, long-term aboveground stand productivity, average biomass of individual trees, and increase in stand productivity with elevation from species to closed forest line. Linear relationships of stand characteristics with latitude of the Ural mountain range are shown when significant at P < 0.05. Stem density includes all stems, with multistemmed trees having several stems. Stand productivity was calculated by dividing tree biomass by mean tree age and represents the longterm net production in biomass.

TA B L E 2
Ratios of belowground to aboveground tree biomass. and 30 times smaller, respectively, in the Polar than in the South Urals (Table 4).

| Larch foliar nutrient concentrations
Larch foliar concentrations of N, P and K were higher at the closed forest line than at the tree species line (Table 5; all P < 0.02) and higher on Mali Iremel' in the South Urals than on Tchernaya in the Polar Urals (all P < 0.02). The difference between regions occurred at both elevation levels for N and K, but was greater at the closed forest line than at the tree species line for P (significant region x elevation level interaction; F 1,12 = 11.07, P = 0.02).

| Consistent growing season temperatures at treeline along the Urals
Our study along a 1500-km north-south transect in the Ural mountains shows that the elevational position of the treeline ecotone TA B L E 3 Results of the linear mixedeffects model testing the effect of latitude on response variables at the stand and individual tree levels (for data see Figure 4)  to those from an Alaskan treeline on permafrost soil, where diminishing soil N availability, rather than low-temperature constraints, was proposed to be primarily responsible for reduced aboveground growth of white spruce towards the elevational treeline. Instead, our findings lend strong support to the concept that the elevational position of treelines is primarily linked to growing season temperatures (Hoch, 2013;Körner & Paulsen, 2004), most likely through thermal constraints on xylogenesis and the metabolization of assimilates in the roots, where temperatures are generally lower than in the canopy (Ferrari, Hagedorn, & Niklaus, 2016;Hoch, 2013;Hoch & Körner, 2012;Rossi, Deslauriers, Anfodillo, & Carraro, 2007).

| Tree biomass and productivity in the treeline ecotone
In agreement with the few other existing biomass estimates for treeline ecotones, calculated for sites in the Alps, the Scandes and Tibet (Leuzinger, Manusch, Bugmann, & Wolf, 2013;Liu et al., 2016;Speed et al., 2015), stand biomass and productivity in the Urals increases dramatically, by a factor of three to nine, as one moves downward TA B L E 4 Soil properties at the tree species line in the South, North and Polar Urals (mean values ± 1 SE). approximately 100 m in elevation from the tree species line, corresponding to a 0.5-0.65°C increase in air temperature. Hence, stand productivity clearly follows a threshold function, with an abrupt increase occurring once the thermal limitation of tree growth is released below the tree species line. In addition to the expected elevational pattern, our study revealed a clear decrease in total biomass and stand productivity with increasing latitude along the Urals (Figure 4).
At first glance, this pattern seems to follow the general decline in forest growth with increasing latitude observed at low elevation, primarily related to declining growing season length and growing degree days (e.g. Jung et al., 2011;Salminen & Jalkanen, 2007). However, treeline trees grow near their thermal growth limit along the whole north-south gradient in the Ural mountains and hence at a very similar temperature regime during the growing season, indicating that factors other than growing season temperature must have been responsible for the decline in productivity towards higher latitudes.
One reason could be different tree species growing at treeline, with spruce dominating in the South Urals and larch becoming increasingly dominant from the North to the Polar Urals, which impedes a clear differentiation between species differences and latitudinal effects.
However, in the southern part of the North Urals, where spruce and larch coexist on slopes with different aspects, north-exposed 'colder' treeline stands dominated by larch had only a 20% smaller stand biomass than south-exposed slopes dominated by spruce. In a 30-yearlong afforestation experiment in central Siberia, Larix sibirica grew larger than Picea obovata and stand productivity did not differ significantly between the two tree species (Kuzmichev, Pshenichnikova, & Tretyakova, 2004). Likewise, in mixed forest stands in the forest-tundra ecotone of central Siberia, tree-ring widths were about 60% greater for L. sibirica than for P. obovata (Knorre, Kirdyanov, & Vaganov, 2005). All these results indicate that the fivefold decrease in stand productivity in the treeline ecotone from the South to Polar Urals is clearly too large to be explained by an inherent difference in biomass productivity among tree species, which actually tends to be greater for larch than for spruce. This conclusion is supported by the latitudinal decline in biomass from the North to the Sub-Polar and Polar Urals when larch-dominated stands were considered separately ( Figure 4). The lower production and biomass may partly result from a lower reproduction and recruitment at higher latitudes (Koshkina, Moiseev & Goryacheva, 2008;Wieczorek et al., 2017), but we regard this as an integral part of treeline stand dynamics. Thus, we consider the observed latitudinal gradient in productivity to apply not only to trees but also to the entire ecosystem.
What drives the latitudinal decline in stand biomass and productivity at treeline? The climatic factors correlating significantly with stand biomass and productivity were growing season length, radiation summed over the growing season and winter temperature.
These variables are all correlated with each other, however, making it difficult to disentangle their individual impacts (Table S2).
Growing season length (GSL) has been observed to correlate more closely with tree growth than average growing season temperatures at treelines in the Alps and in Tibet (Jochner et al., 2018;Li et al., 2017), very likely through its effect on the timespan above critical temperatures when xylogenesis can take place. In our study, the length of the growing season decreased slightly towards the north, with 20% shorter summers, but longer days during summer in the Polar Urals (Figure 3). This finding is consistent with the declining GSL at treeline on the global scale reported by Körner and Paulsen (2004). However, this decrease did not translate to fewer growing degree days (GDD) above 5°C in the Polar Urals, which is generally considered an important control for tree growth at high latitudes and elevations because it quantifies the total energy available for trees to grow (Jochner et al., 2018;Leuzinger et al., 2013;Salminen & Jalkanen, 2007). The decreasing length of the growing season towards the north in our study was also linked to a decline in incoming short-wave radiation per growing season ( Figure 3). Moreover, the low sun angle at high latitudes leads to stronger absorption of UV-B light in the atmosphere than at locations further south (Olsen & Lee, 2011). Both growth models and experimental work, predominantly focusing on agricultural crops, indicate that radiation and growth are linearly coupled (Russel, Jarvis, & Monteith, 1989).
However, given that radiation and biomass production scale linearly, it seems unlikely that diminishing radiation is the primary driver of the latitudinal decline in productivity, as growing season radiation decreases by only 40% along the latitudinal gradient (Figure 3), whereas the observed decline in stand productivity was manifold ( Figure 4). Moreover, tree canopies adapt to lower sun angles through the formation of vertically extending crowns (Kuuluvainen, 1992).
There is also growing evidence that tree growth is predominantly limited by C sink strength, that is the metabolization of assimilated C, rather than by C uptake, in particular for trees growing at high elevations (e.g. Hoch & Körner, 2012;Li et al., 2018). Nonetheless, C metabolization might be indirectly related to sun angle, as a lower solar angle at higher latitudes leads to greater shading by trees and associated cooling of the rooting zone (Aakala, Shimatani, Abe, Kubota, & Kuuluvainen, 2016;Bonan & Shugart, 1989). In our study, the occurrence of such a shading effect was supported by the decline in stand density, corresponding to larger spacing among trees, at higher latitudes. The latitudinal increase in the shading effect might be counterbalanced by the change in tree species from spruce, with a dense canopy, in the South Urals to larch, with a transparent canopy, in the Polar Urals. Systematic assessments of the spatial structures involved in microclimatic conditions would help quantify the importance of this potential mechanism.
Latitude may also influence tree growth through its effects on the circadian rhythm of tree growth, as cell expansion and structural growth occur mainly during nighttime, in concert with improved water status and thus more favourable turgor (see review by Steppe, Sterck, & Deslauriers, 2015). Consequently, shorter nights towards the north during the growing season may impair growth processes. However, in quantitative terms the importance of these ecophysiological processes are still poorly understood. For the latitudinal gradient along the Urals, we consider the effect of diel growth dynamics rather limited because growth at the individual tree level was independent from latitude.

| Winter temperatures and their link to soil nutrient availability
Among all estimated climatic variables, winter air temperature showed the strongest change with latitude, decreasing by about 4°C along the 1500-km gradient (Figure 3). Although winter temperatures do not directly impact tree growth because trees adapt to very low temperatures in the dormant season (Holtmeier, 2003;Körner, 2012), concentrations of non-structural carbohydrates in treeline trees have been found to decrease during the winter season (Li et al., 2018). However, a complete depletion of carbon reserves over winter has not been reported, indicating that the C balance of trees seems rather to be driven by the extent to which C stores in trees are replenished during the growing season.
In comparison to the limited direct effects on tree physiology, winter temperature and conditions may exert a considerable indirect influence on forest growth through impacts on nutrient mineralization (Sturm et al., 2005) and the depth and duration of permafrost. Permafrost has been identified as one of the key determinants of tree growth and treeline advances in Alaska and northern Canada, as permafrost restricts rooting depth and nutrient availability (Lloyd, 2005;Sullivan et al., 2015). In the Urals, permafrost has been reported to reach as far south as the North Urals (Westermann, Østby, Gisnås, Schuler, & Etzelmüller, 2015), but the active layer depth exceeds the rooting zone in these mountain soils.
However, periglacial processes associated with permafrost in the deeper soil (e.g. stone sorting) impede weathering and soil development (Dymov, Zhangurov, & Hagedorn, 2015), which in turn results in reduced soil fertility. In support of these indirect effects of winter conditions, our soil survey indicated a strong decrease in soil development and fertility at treeline with increasing latitude.
Stone contents tripled towards the Polar Urals and soil organic matter stocks were almost seven times smaller in the Polar than in the South Urals (Table 4). Furthermore, mineral N and available P in the soil were 20 and 30 times smaller, respectively, in the Polar than in the South Urals (Table 4), a pattern that was also reflected in foliar nutrient concentrations (Table 5). Fertilization experiments in treeline ecosystems have demonstrated a positive growth response to nutrient additions (Möhl et al., 2019;Sveinbjornsson, Nordell, & Kauhanen, 1992), strongly suggesting that once thermal growth limitation is relieved, nutrient availability may become a key determinant of stand productivity (Hoch, 2013). Along elevation gradients, nitrogen availability has been found to increase from the tundra to the closed forest, due to enhanced N mineralization in more favourable microclimates (Kammer et al., 2009;Thébault et al., 2014), which potentially contributes to growth release below the tree species line.
The decreasing nutrient availability towards the Polar Urals might also be mirrored in the latitudinal increase in the ratio of belowground to aboveground biomass (Table 2), which is generally regarded as a tree's adaptation to resource limitation in the soil (e.g. Solly et al., 2017). However, more systematic soil studies at treeline in other mountain ranges are needed to elucidate how winter climate, soil development and nutrient availability are interlinked and influence stand productivity .

| Consequences for C sequestration
The latitudinal decline in treeline stand productivity and biomass observed here implies that the amount of C stored in treeline ecosystems and sequestered during upward forest expansion into former tundra decreases with increasing latitude. As a tree age structure analysis indicated that all treeline stands in our study represent the first tree generation established since the Little Ice Age around 1850 (Hagedorn et al., 2014), we can roughly estimate the amount of carbon sequestered in these Ural treeline ecotones during the past century. The apparent stand productivity in aboveground and belowground biomass combined corresponds to a net C uptake of approximately 0.37 ± 0.06 t C ha −1 y −1 and 0.09 ± 0.05 t C ha −1 y −1 (averaged for all three elevation levels) in the South and Polar Urals, respectively, and thus a reduction by a factor of four along the latitudinal gradient. Although one may expect a strong C sink associated with the striking forest expansion into former treeless tundra, these sequestration rates are rather small compared with those of temperate forests. Despite the continuous biomass removal with harvest in low-elevation forests, net C uptake rates for European forests have been estimated at approximately 0.5 t C ha −1 y −1 (Janssens et al., 2003).

| CON CLUS IONS
Our study is one of the first assessments of stand biomass and productivity at multiple treeline sites, variables that serve as the basis for estimating and modelling various ecosystem functions.
In combination with novel remote sensing techniques, the allometric functions for treeline trees provided here could facilitate large-scale estimations of stand biomass in treeline ecotones. Our results demonstrate that treeline positions are located at similar temperature regimes during the growing season along the entire 1500-km latitudinal gradient in the Ural mountains. Despite the consistent cold limitation of tree growth, stand biomass and productivity showed a manifold decrease from the South to the Polar Urals. Several potential explanatory factors-growing season length and radiation, winter temperature, and soil nutrient availability-showed a corresponding change with latitude.
As they correlated with each other, it was not possible to identify one key driver. We suggest that soil fertility, restricted by permafrost and low soil temperatures during winter, plays a key and yet underexplored role for stand productivity near treeline once the thermal limitation is relieved. The latitudinal patterns in stand biomass and productivity will impact a number of ecosystem functions, such as above-and belowground diversity and C sequestration, and how these are altered by a changing climate.

CO N FLI C T S O F I NTE R E S T
The authors declare no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
All data used in this publication are available at the WSL data portal: