Controls on and consequences of specific leaf area variation with permafrost depth in a boreal forest

Specific leaf area (SLA, leaf area per unit dry mass) is a key canopy structural characteristic, a measure of photosynthetic capacity, and an important input into many terrestrial process models. Although many studies have examined SLA variation, relatively few data exist from high latitude, climate-sensitive permafrost regions. We measured SLA and soil and topographic properties across a boreal forest permafrost transition, in which forest composition changed as permafrost deepened from 54 to >150 cm over 75 m hillslope transects in Caribou-Poker Creeks Research Watershed, Alaska. We characterized both linear and threshold relationships between topographic and edaphic variables and SLA and developed a conceptual model of these relationships. We found that the depth of the soil active layer above permafrost was significantly and positively correlated with SLA for both coniferous and deciduous boreal tree species. Intraspecific SLA variation was associated with a fivefold increase in net primary production, suggesting that changes in active layer depth due to permafrost thaw could strongly influence ecosystem productivity. While this is an exploratory study to begin understanding SLA variation in a non-contiguous permafrost system, our results indicate the need for more extensive evaluation across larger spatial domains. These empirical relationships and associated uncertainty can be incorporated into ecosystem models that use dynamic traits, improving our ability to predict ecosystem-level carbon cycling responses to ongoing climate change.


Introduction 34
The boreal forest is changing rapidly with climate change [1]. Permafrost soil underlaying the 35 boreal is currently degrading and in many places predicted to disappear entirely, by the end of 36 the 21 st century in Alaska [2] and other circumpolar regions [3]. Permafrost thaw affects 37 ecosystem carbon, water, and nutrient cycling [4][5][6], which are expected to, in turn, produce 38 shifts in tree cover and canopy physiology [7]. Moreover, permafrost thaw has been shown to be 39 a threshold for these environmental shifts [8,9]. 40 Phenotypic plasticity allows plants to adapt to environmental shifts, resulting in intraspecific 41 trait variation. A particularly variable trait is specific leaf area (SLA). SLA-leaf area per unit 42 dry mass-is a trait that corresponds with differences in leaf structure associated with 43 photosynthesis [10] and importantly ecosystem carbon gain [11]. SLA has been used in 44 numerous meta-analyses to predict leaf physiology and other functional traits [12]. 45 Understanding the consequences of permafrost thaw on SLA variation and ecosystem 46 productivity is particularly important, as shifting environmental gradients may impact 47 intraspecific trait variation, with potentially large consequences on carbon accumulation across 48 the landscape. In environments with optimal soil resources (e.g., water, nutrients), for example, 49 plants can produce more leaf biomass with high SLA, maximizing carbon gain per unit leaf mass 50 [13,14]. Conversely, in suboptimal resource environments, small and thick leaves (i.e., with low 51 SLA) allow plants to maximize leaf longevity. The relatively thick (low SLA) leaves of black 52 spruce, a typical boreal evergreen conifer, last an average of 50-60 months, compared to the 4-6 53 month life span of the relatively thinner (high SLA) leaves of boreal deciduous species (Reich,54 Tjoelker, Walters, Vanderklein, & Buschena, 1998).
Global analyses reveal that specific leaf area varies with climatic and edaphic gradients 56 [16,17]. In contrast, SLA variation within a species is less well understood. Intraspecific SLA 57 variation has been shown to contribute significantly to total trait variability [18][19][20][21], and used to 58 understand local and regional community assembly processes and explain the coexistence of 59 multiple species across environmental gradients [22][23][24]. 60 In boreal systems underlain by permafrost, the thaw depth of the seasonally-thawed active 61 layer [9] is coupled to soil moisture and nutrient availability, and is hypothesized to govern leaf specific leaf area as a function of climate or species, these fixed trait-based approaches miss the 72 variation in a trait within a given plant functional type. Without data on intraspecific trait 73 variation, it is not clear whether these models can be successful.

74
In this study we examined intraspecific SLA variation and associated topographic and 75 edaphic factors across a permafrost and vegetation transition within an Alaskan boreal forest. We 76 hypothesized that (i) SLA would be significantly correlated with active layer depth, which 77 governs the availability of soil resources such as water and nutrients, and (ii) intraspecific SLA 78 variation across the permafrost transition would in turn positively correlate with aboveground net 79 primary production. This is an exploratory study to begin understanding SLA variation in a 80 permafrost forest ecosystem. Our results indicate significant influences of environmental features 81 across the permafrost transition zone, indicating the need for more extensive evaluation of SLA 82 and forest production across larger spatial domains in forested ecosystems.   However, we recognize the importance of species-specific data, and thus provide the full (raw) 124 data with putative species tags (see below for data availability).  Net primary production (NPP) was determined using tree cores taken from a representative 147 sample of trees (birch, alder, black spruce, and white spruce) every 10 m along each transect. At 148 each sample point, we cored 3-5 trees per species; sample discs were taken from trees too small 149 to core. Cores were embedded into larger boards for protection, sanded, and scanned at 800 dpi  with leaf nested in tree, and with separate analyses for each species (S1 Appendix).
Because we have continuous data, we used linear regressions (with SLA averaged by tree 170 individual for each species) to test the hypothesized relationships between SLA and topographic 171 and edaphic properties (Fig 1). Spearman's rank correlations were used to determine the 172 significance and strength of relationships between ALD and slope and between ALD and SLA, 173 using R function cor.test in the R 'stats' package, version 3.2.4. Where we saw relationships in 174 the linear regressions, we used T-and F-tests to examine shifts in means and variances, 175 respectively. When univariate regressions showed nonlinear relationships, we used the function 176 segmented in the R 'segmented' package to test for breakpoints and tested for significance of 177 these breakpoints using the davies.test function. If segmented regression did not change our 178 inferences, then a quadratic fit was tested. In this case, AICc (small sample size-corrected AIC) 179 was compared to determine best model fit between linear and quadratic models, using the 180 function AICc in the R 'AICcmodavg' package.

187
Inter-and intraspecific SLA variation 188 Specific leaf area (SLA) varied by over a factor of three between alder and spruce, with high 189 levels of intraspecific variation as well (Table 1). SLA for alder ranged from 121 to 364 cm 2 g -1

191
Partitioning the variance of SLA into the leaf and tree scales revealed that most of the variation 192 in SLA is between rather than within individual trees, for both alder and spruce (   Fig 2a). Relationships between (a) spruce-specific SLA (cm 2 g -1 ) and net primary production 211 (NPP, g C m -2 yr -1 ), and (b) total NPP and soil C:N.

213
We hypothesized that SLA has a direct effect on NPP (Fig 1). However, it has been shown 214 that soil nutrient status (e.g., soil C:N) may have a direct influence on NPP [45]. We found that Across the entire field site, ALD ranged from 54 cm to above 150 cm (median 137.2 cm). Soil 237 moisture at 6 cm ranged from 0.18 -0.93 g g -1 (mean = 0.68 g g -1 ); we used soil data from 6 cm 238 depth for comparisons with landscape and SLA data, as this depth is more relevant than surface 239 soil to rhizosphere processes. Soil was drier at deeper ALD (for 6 cm, p < 0.05, Fig 4). For 240 subsequent analyses, we used the ALD-moisture relationship to divide the data into two ALD  Mean alder SLA for deep ALD was 266 cm 2 g -1 , and for shallow ALD was 158 cm 2 g -1 .

251
Mean spruce SLA for deep ALD was 62.0 cm 2 g -1 , and for shallow ALD was 52.1 cm 2 g -1 . SLA for both alder and spruce increased with thicker ALD (p < 0.001, R 2 = 0.664, rho = 0.75 for 253 alder, Fig 5a; p < 0.001, R 2 = 0.31, rho = 0.62 for spruce, Fig 5b)  Across the field site, slope ranged from 13-56% (mean = 23.7%). ALD was positively correlated 264 with landscape slope (p < 0.001, rho = 0.63, Fig 2). Above 23% slope, ALD was consistently 265 deeper than the maximum probe depth (150 cm); below this value, there was no significant 266 correlation between ALD and slope (p = 0.157, rho = -0.31, Fig 6).  The empirical data presented here suggest that in permafrost-affected systems, the depth of the 296 active layer can function as a threshold for various soil parameters, influencing plant traits such 297 as SLA (Fig 5). SLA is directly associated with leaf-level photosynthesis, and has been shown to have a direct positive correlation with photosynthesis and productivity [14]. Here, we show that 299 across a permafrost transition, a two-fold increase in black spruce SLA corresponds to a five-fold 300 increase in NPP (Fig 2a). Further, our data suggest a direct connection between SLA and NPP 301 (Fig 2a) and an indirect and weaker influence of soil C:N on NPP (Fig 2b). Because the control 302 on maximum SLA is mediated by ALD (Fig 5a,b) [48,49]. The balance between respiration and productivity warrants further exploration in 315 permafrost-affected ecosystems sensitive to climate change.

317
Influence of active layer depth on SLA variation 318 We found that soil moisture does not a have a direct, linear relationship with SLA (Fig 3a, The relatively large range of soil moisture included our study site (0.18 -0.93 g g -1 ) may 327 therefore explain the apparent quadratic relationship between SLA and soil moisture seen in our 328 study, suggesting an optimal moisture condition for maximizing SLA. However, much more data 329 is needed to unambiguously distinguish between a negative linear relationship (more soil 330 moisture always means lower SLA) versus a quadratic one (implying an optimum).

331
Significant variation in SLA was also explained by soil C:N (Fig 3c,d).

341
For both species, we found that ALD is highly correlated with SLA, with an approximate 342 two-fold increase in species-specific SLA values from shallow to deep ALD locations (Fig 5).

343
Further, ALD has a thresholding effect on intraspecific SLA, whereby a relatively small change in ALD corresponds to a large change in SLA (Fig 5). While SLA at the landscape scale is 345 controlled by soil resources (moisture and nutrients) [51], in this boreal system this control is 346 mediated by ALD, which constrains SLA (Fig 5). While other studies have examined SLA 347 dynamics in permafrost systems [28,30,52], our data extend this knowledge by providing a broad 348 range of continuous ALD across a small-scale and critical soil transition zone, while also 349 examining a network of soil and landscape influences to understand controls on SLA (Fig 9). Given the structure of our hypothesized relationships between SLA and environmental 359 correlates (Fig 1), we considered path analysis (a type of structural equation modeling, SEM).

360
However, we have insufficient data given the number of variables that would be contained 361 within an SEM. For example, separating the soils dataset by tree species, we have 18 data points, 362 and the number of parameters is 7 (excluding NPP, Fig 1); our number of samples per parameter 363 is 2.6, which is on the low end of sample adequacy [53]. We instead performed univariate linear 364 regressions as detailed above, using our hypothesized relationships (Fig 1) to construct a 365 conceptual diagram (Fig 9). Further studies should consider increasing sample size to allow for 366 SEM analysis. Because we evaluated relationships based on a priori hypothesized relationships 367 (Fig 1), we did not perform a global selection model to explain variation in SLA. As a caveat, 368 our data describe one spatial domain (75m x 75m), and further study is needed to confirm if the 369 relationships between SLA and active layer depth and soil properties apply at other spatial 370 scales. 371 We assume here that the high degree of SLA variation among individuals in a given species 372 (Table 1) is due to phenotypic plasticity in response to environmental gradients, rather than in this study (all transects are east-facing), allowing us to isolate the effect of slope on ALD. We 386 found a threshold in the slope-ALD relationship, with permafrost not encountered at slopes 387 greater than 23% (Fig 6). This is not unexpected, as deeper ALD is predicted on steeper slopes 388 due both to increased drainage and to high solar radiation inputs in high latitude systems [27].
Because permafrost is a physical barrier to water drainage, the shallower active layers 390 generally maintain high soil moisture conditions [27]. We found that shallow active layers (< 391 140 cm) constrain soil moisture values to near saturation, and lowered their variance (Fig 4).

392
Together, high moisture and low temperatures of permafrost-associated soils limit 393 decomposition, thus maintaining high C:N ratios in the thick moss layer and limiting nutrient 394 availability [5] (Fig 8). Past the observed ALD threshold of 140 cm, soil moisture and thermal 395 conditions shift rapidly (Figs 4, 7), favoring aerobic decomposition of soil organic matter and 396 increasing available nutrients for plant uptake.

397
In boreal systems, the surface moss and organic soil layer play important roles in soil 398 thermal dynamics, nutrient cycling, and ecosystem carbon accumulation [57]. Our data show 399 expected relationships between moss depth, temperature, and ALD. Specifically, the moss-400 organic layer acts as insulation to maintain low soil temperatures and shallow ALD (Fig 7)

401
[9,27]. In turn, low soil temperatures maintained by the moss-organic layer can limit 402 decomposition and maintain high soil C:N (Fig 8) The empirically-derived relationships presented in this exploratory study (Fig 9)