Variation of key elements in soils and plant tissues in subalpine forests 1 of the northern Rocky Mountains

Abstract. The essential elements for the structure and function of forest ecosystems are found in relatively predictable proportions in living tissues and soils; however, both the degree of spatial variability in elemental concentrations and their relationship with wildfire history are unclear. Quantifying the association between nutrient concentrations in living plant tissue and surface soils within fire-affected forests can help determine how these elements contribute to biogeochemical resilience. Here, we present elemental concentration data (C, N, P, K, Ca, Mg, S, Fe, Mn, Zn) from 72 foliar and 44 soil samples from a network of 15 sites located in the fire-prone subalpine forests of the northern Rocky Mountains, USA Plant functional type is strongly correlated with carbon (C) and nitrogen (N) – C concentrations are highest in coniferous needles, and N concentrations are highest in broadleaved plant species. The average N / P ratio of foliage among samples is 9.8 ± 0.6 (μ ± 95 % confidence). This suggests that N is the limiting nutrient for these plants, however several factors can complicate the use of N / P ratios to evaluate nutrient status. Average C concentrations in organic soil horizons that were burned in regionally extensive fires in 1910 or 1918 CE are lower than those from sites that burned prior to 1901 CE (p 



Introduction 35 36
Although living plant tissue is primarily composed of carbon (C), there are 37 approximately 20 other elements that are necessary for biochemical reactions and growth.38 These other key elements-including nitrogen (N), phosphorus (P), potassium (K), 39 calcium (Ca), magnesium (Mg), and sulfur (in the form of SO 4 )-play important roles in 40 Table 1).The study sites are located between 1623-2011 m above sea level in the northern 99 Bitterroot Mountains along the border of Idaho and Montana, U.S.A. (Fig. 1).

Patterns in soil and foliar elements 161 162
Soil C and N concentrations were highest on average (34.94% and 1.32%, 163 respectively) within litter material and declined with depth through organic (18.30% and 164 0.84%) and mineral (6.05% and 0.29%) soil horizons (Table S1, Fig. 2).The 165 concentration of other nutrients such as K, Ca, and Mg were highest on average within 166 the litter (0.016%, 0.115%, and 0.012%, respectively) and organic soil horizons (0.019%, 167 0.104%, and 0.013%), and decreased in the mineral soil horizon (0.009%, 0.046%, 168 0.007%).Na concentrations displayed the opposite pattern and increased from litter 169 (<0.001%) to mineral (0.001%) soil horizon.170 The Foliage contained C concentrations between 47-52%, with the exception of Larix 178 occidentalis, Alnus viridis, and Menziesia ferruginea, which had slightly lower 179 concentrations of C between 43-47% (Fig. 3, Table S2).N concentrations were generally 180 low (0.75-1.89%) in coniferous needles, and high in leaves of the broadleaved species 181 Alnus viridis and Menziesia ferruginea (2.5-3.5%).P, K, and Ca concentrations were 182 variable within and between tree and shrub species, with no clear pattern differentiating 183 needle and broadleaf species (Fig. 3).184 There was considerable variability in trace element concentrations among samples 185 of foliar material (Fig. 3).Mg was below 0.20% in all leaf material, except for Alnus 186 viridis (0.35%) and Menziesia ferruginea (0.50%).Fe, Mn, and Zn concentrations were 187 highly variable within and between species.Menziesia ferruginea contained high 188 concentrations of Mn (0.48%) and Zn (0.01%).SO 4 concentrations ranged between 0.01-189 0.15% for all foliar material, except for higher values found in Alnus viridis (0.18%).positively correlated (Fig. S2), indicating that leaf tissues are built in predictable 214 stoichiometric ratios, but the variation in the concentration of those nutrients among 215 sampling sites suggests that the availability of these elements in soils and bedrock 216 likewise varied.Ca was weakly correlated with N, P, K, Mg, and SO 4 in leaves.Ca 217 concentration is strongly affected by foliage age (Turner et al., 1977), which represents 218 another source of variation in the sampled foliage not controlled for in this study.

219
We found substantial differences in the foliar nutrient concentrations between 220 evergreen and deciduous plant types (Fig. 3).N, Mg, and SO 4 were found at higher 221 concentrations in broadleaf material relative to evergreen needles, probably due to higher 222 rates of metabolic activity and photosynthesis in broadleaf material during the summer 223 (Linzon et al., 1979).Leaves from Alnus, a nitrogen-fixing plant, contained greater N 224 (2.55-3.56%)relative to non-fixing species (Taylor et al., 1989) Biogeosciences Discuss., https://doi.org/10.5194/bg-2018-443Manuscript under review for journal Biogeosciences Discussion started: 15 November 2018 c Author(s) 2018.CC BY 4.0 License.regulating terrestrial ecosystem processes.The quantities of these elements are controlled 41 by a number of factors including bedrock parent material, soil composition (Kramer et 42 al., 2017), vegetation type (Hu et al., 2001), and climate (Campbell et al., 2009).It is 43 generally thought that plant tissue and soil nutrient concentrations can provide 44 information about element limitation to growth (Wardle et al., 2004;Boerner, 45 1984;Schreeg et al., 2014).However, relatively few studies have analyzed the 46 concentrations of a large suite of elements in both plant and soil samples from one region.47 Plant traits, such as biochemical, physiological, and anatomical features measured 48 at the individual level (Violle et al., 2007), reflect the outcome of evolutionary processes 49 responding to environmental constraints (Valladares et al., 2007).Traits determine how 50 primary producers respond to abiotic and biotic environmental factors and influence a 51 host of other ecosystem services (Kattge et al., 2011).Plant functional type (PFT) models 52 capture a substantial fraction of the observed trait variation across plant species.A 53 relatively small number of PFTs (5-20) have been used to represent the functional 54 diversity of >300,000 documented plant species on Earth in global vegetation models 55 (Kattge et al., 2011).Growth-limiting elements in foliar material derived from soils, such 56 as phosphorus, make nutrient concentrations among the most plastic of plant traits 57 (Wright et al., 2004) because the availability of these nutrients varies spatially as a result 58 of parent bedrock composition.Nutrient concentrations in plants are influenced by a 59 number of other factors, including the age of the tissue and overall age of the plant 60 (Turner et al., 1977;Reich et al., 1992;Schreeg et al., 2014;Luo et al., 2017).Additionally, 61 sunlight and location on a tree branch can also affect nutrient concentration in leaves 62 (Schreeg et al., 2014), and weather-related factors such as precipitation and temperatures 63 can affect nutrient levels on a seasonal basis (Turner et al., 1977).Therefore, samples of 64 plant tissues taken over a short time period for several species across one region are 65 particularly useful for differentiating factors that influence foliar nutrient concentrations 66 and subsequent bioavailability of nutrients in leaf litter and upper soil horizons (Qualls 67 and Haines, 1991;Taylor et al., 1989).68 High severity fires can profoundly impact the cycling of C, N, and other nutrients 69 in plants and soils (Chen et al., 2017;Fletcher et al., 2014;Certini, 2005;Dunnette et al., 70 2014), and could therefore affect the ability of the forest to regenerate to pre-fire 71 conditions (Smithwick, 2011;McLauchlan et al., 2014).Several mechanisms determine 72 how forests respond to fire, and how cycling of C, N, and other elements subsequently 73 change (Pellegrini et al., 2018;Certini, 2005).For example, combustion of vegetation and 74 soils from frequent, low-intensity burning can lead to the loss of C and N to the 75 atmosphere (Pellegrini et al., 2014;Reich et al., 2001;Deluca and Sala, 2006;Yelenik et 76 al., 2013).Fires also release nutrients from plant tissues as ash, which can potentially 77 increase post-fire vegetation growth through soil nutrient enrichment (Boerner et al., 78 2009;Hurteau and Brooks, 2011).Over longer time scales of centuries to millennia, 79 ecosystem modeling informed with paleofire history records suggest that changing fire 80 Biogeosciences Discuss., https://doi.org/10.5194/bg-2018-443Manuscript under review for journal Biogeosciences Discussion started: 15 November 2018 c Author(s) 2018.CC BY 4.0 License.regimes can have substantial and long-lasting impacts on C and nutrient cycling 81 (Hudiburg et al., 2017;Kelly et al., 2016).82 Here we provide data from 72 leaf and 44 soil samples from a network of 15 sites 83 in the coniferous subalpine forests of the northern Rocky Mountains, U.S.A. (Fig. 1, 84 were collected at 15 sites along a 100-km northwest to southeast transect (46.78°-47.36°114 N; 114.76°-115.57°W) from 12 July to 25 August, 2017 CE (Fig. 1).At each soil core 115 site, except for Hoodoo, associated leaf samples were collected from nearby trees and 116 shrubs, avoiding new growth and juvenile trees.The majority of the foliar samples 117 consist of coniferous needles from eight species (n=68): Abies lasiocarpa, Larix 118 occidentalis, Pinus albicaulis, Pinus contorta, Picea engelmannii, Taxus brevifolia, 119 Thuja plicata, and Tsuga mertensiana.We also collected two samples each of the 120 Biogeosciences Discuss., https://doi.org/10.5194/bg-2018-443Manuscript under review for journal Biogeosciences Discussion started: 15 November 2018 c Author(s) 2018.CC BY 4.0 License.broadleaf species green alder (Alnus viridis; n=2) and false huckleberry (Menziesia 121 ferruginea; n=2).Foliar samples were taken directly from living vegetation located 122 approximately 2 m above the soil surface.Soil cores were sampled by horizons: 'litter' 123 was unconsolidated plant material from the surface; 'organic' soil was identified visually 124 and sampled at a depth between approximately 0-10 cm; the 'mineral' soil layer was 125 sampled between 10-25 cm below the surface depending on overall soil depth.Samples 126 were stored in airtight plastic containers for transport back to the laboratory.127 128 2.3 Laboratory analyses 129 130 Samples were dried, homogenized, and sent to the Department of Agronomy Soil 131 Testing Lab at Kansas State University.Foliar samples were weighed into 50 ml Kimax 132 digestion tubes for elemental analysis.Boiling chips and 5 ml of nitric acid at 50% 133 strength were added to the tubes, which were covered with plastic wrap to react 134 overnight.The following day, 5 ml of perchloric acid was added to predigested plant 135 material and tubes were then placed on a cold techator digestion block.Temperatures 136 were set to 200°C and digests were heated for approximately three hours or until white 137 fumes appeared and the acid was clear (and colorless when cooled to room temperature) 138 (Gieseking et al., 1935).Tubes were diluted to 25 or 50 ml with deionized water and 139 mixed by inverting twice.Each digested sub-sample was then analyzed for P, Mg, K, Ca, 140 iron (Fe), manganese (Mn), zinc (Zn), and SO 4 using a Varian Model 720-ES ICP Optical 141 Emission Spectrometer.Analytical precision (σ) for element analysis was <0.0001%.142 Base cations (Ca, Mg, K, and Na) were extracted from soil sub-samples using 1 143 M ammonium acetate (pH 7.0).The resulting supernatant was then analyzed using a 144 Varian Model 720-ES ICP Optical Emission Spectrometer.Trace elements (e.g.Mn, Zn, 145 etc.) were not measured on soil sub-samples.Soils were not analyzed for P. 146 Total carbon and nitrogen concentrations (weight %) in both plant and soil sub-147 samples were measured using a LECO CN 2000 combustion analyzer.Analytical 148 precision (σ) was approximately 0.06% for carbon and 0.006% for nitrogen.149 150 2.4 Statistics 151 152 To compare elemental concentrations among different sample sites, we used box 153 plots, unpaired t-tests (assuming unequal variance), principal components analysis 154 (PCA), and Pearson correlation coefficients.We calculated principal components using 155 the ade4 package in R (Dray and Dufour, 2007).Correlation tables were calculated using 156 the R base stats package (R Core Team).157 158 3. Results and discussion 159 160 Biogeosciences Discuss., https://doi.org/10.5194/bg-2018-443Manuscript under review for journal Biogeosciences Discussion started: 15 November 2018 c Author(s) 2018.CC BY 4.0 License.
pattern of five elements in the soil-C, N, Ca, K, and Mg-correspond with 171 the dominant role of decomposing plant material controlling the concentration of 172 nutrients in forested soils (Aerts and Chapin, 1999).These results are consistent with a 173 global compilation of 10,000 soil profiles that indicate ranking of nutrient concentrations 174 from shallow to deep (Jobbagy and Jackson, 2001) (Fig. 2, Fig. S1).The decline in the 175 bioavailable K, Ca, and Mg in the mineral horizon suggests removal by mineralization or 176 microbial immobilization (Qualls et al., 1991;Qualls, 2000).177 190 Element concentrations did not vary with site latitude or elevation likely as a 191 result of the constricted latitude and elevation range of the study sites.Soil and foliar N, 192 K, and Mg were all positively correlated (p<0.001) with each other (r=0.42-0.75)(Figs.193 S1, S2, S3, and S4).In soils, Ca was positively correlated (p<0.01) with Mg (0.74), K 194 (0.54), N (0.47), and C (0.39), whereas in plant tissues Ca was poorly correlated (p>0.05)195 with P (-0.21),K (-0.21), and Mg (0.15), C (0.05), and N (0.02).Foliar C was negatively 196 correlated (p<0.05) with Mg (-0.53),N (-0.41),P (-0.38), and K (-0.30).In soils, Na was 197 inversely related (p<0.05) to both C (-0.52) and N (-0.42).In vegetation, Zn was 198 correlated (p<0.05) with Mn (0.54), K (0.33), and C (-0.26).Trace elements Fe and Mn 199 were weakly correlated (p>0.05) with C, N, P, K, and Mg in plant tissues.200 Biogeosciences Discuss., https://doi.org/10.5194/bg-2018-443Manuscript under review for journal Biogeosciences Discussion started: 15 November 2018 c Author(s) 2018.CC BY 4.0 License.A multivariate analysis of elemental concentrations clearly separated foliar (n=72) 201 from litter/soil samples (n= 44) (principal component analysis [PCA]; Fig. 4).Axis 1 of 202 the PCA explained 58.7% of total variation in the soil and foliar elements.The 203 concentrations of several elements were negatively loaded on Axis 1: C (-0.905), K (-204 0.809), Ca (-0.791), Mg (-0.799), and N (-0.706).C/N was correlated to Axis 1 (-0.535)205 but more weakly than C, N, and the base nutrients.Axis 2 explained 22.4% of the 206 variation and was primarily driven by N and Mg (0.577, 0.381, respectively), and was 207 strongly influenced by the high foliar N concentrations of Alnus viridis (n=2).Foliage 208 samples from different plant species were dispersed along Axis 2. 209 Leaf carbon, nitrogen, and phosphorus content are considered important 210 biochemical traits in PFT modeling, as seen in the TRY plant database (Kattge et al., 211 2011).We measured additional elemental concentrations to understand inter-and intra-212 species variation in biogeochemical traits.In live vegetation, N, P, K, Mg, and SO 4 are all 213 240 Biogeosciences Discuss., https://doi.org/10.5194/bg-2018-443Manuscript under review for journal Biogeosciences Discussion started: 15 November 2018 c Author(s) 2018.CC BY 4.0 License.centurycontained significantly less C on average than samples from sites that burned 241 prior to 1901 CE (t = -2.26,df= 9.60, p= 0.0484).In contrast, N (t = -1.46,df= 10.54, p= 242 0.1731) and other soil nutrient concentrations (Fig.5; Ca, K, Mg not shown) did not 243 significantly differ between these two populations.244The significant difference in soil C suggests that a single fire event in the early 245 20 th century at several sites reduced the pool of soil C, and that the legacies of those fires 246 are still detectable today (Fig.5).These results are consistent with ecosystem modeling 247 experiments, which find that both fire frequency and severity are the dominant drivers of 248 C dynamics in sub-alpine coniferous forests over centennial to millennial timescales 249(Hudiburg et al., 2017).Furthermore, the greater variance in C concentrations observed in 250 soils burned in the early 20 th century (σ 2 =37.31) versus soils not burned (σ 2 =12.44)(Fig.2515)may also reflect aspects of fire history, and specifically high spatial variability in fire 252 severity and post-fire recovery.253Theresults from this study are broadly consistent with ecosystem model 254 simulations suggesting that changing fire regimes under climate change scenarios have 255 the potential to alter C stored in forested ecosystems by changing the frequency and 256 intensity of wildfire events (Hudiburg et al., 2017;Kelly et al., 2016).We show that even 257 in a relatively small region of subalpine forests in the Rocky Mountains, an assumption 258 of a single value for C stocks would not hold, and are instead highly dependent on a 259 spatially heterogeneous fire history extending back for a least a century.The importance 260 of this variability in determining post-fire C dynamics implies that equilibrium scenarios 261 extrapolated from a single fire event in one location are a poor assumption when 262 simulating fire regimes in Earth System models at spatial scales larger than an individual 263 site (Hudiburg et al., 2017).264 265 3.3 Evaluating nutrient limitations 266 267 The process whereby forest vegetation typically returns to pre-fire conditions after 268 several years or decades following wildfire events is referred to as biogeochemical 269 resilience (McLauchlan et al., 2014;Smithwick, 2011).This resilience is determined, in 270 part, by the availability of growth limiting nutrients, such as N and P (Güsewell, 271 2004;Schreeg et al., 2014).In non-N 2 -fixing plant species, foliar N and P concentrations 272 reflect soil N and P availability (Reich and Oleksyn, 2004).Previous studies have shown 273 that a N/P mass ratio in foliar material of about 10-20 is optimal for plant growth (Aerts 274 and Chapin, 1999;Ingestad and Lund, 1979).N/P ratios >16 can indicate P-limited 275 biomass productions, whereas N/P ratios <14 are suggestive of N-limited plant growth 276 (Aerts and Chapin, 1999;Koerselman and Meuleman, 1996).277 The average (± 95% confidence intervals) N/P values for sampled foliage (n=72) 278 was 9.8 ± 0.6.This suggests a N-limited growing environment.Although N/P ratios of 279 leaves are often used to evaluate nutrient status in plants, several factors could complicate 280 Biogeosciences Discuss., https://doi.org/10.5194/bg-2018-443Manuscript under review for journal Biogeosciences Discussion started: 15 November 2018 c Author(s) 2018.CC BY 4.0 License.this interpretation: (1) N/P ratios of old leaves may be different than new foliage growth 281 (Schreeg et al., 2014).When plants are older, nutrients are reallocated to active 282 meristems (e.g., young leaves, shoot tips) (Güsewell, 2004).We did not account for 283 foliage or plant age during sampling.(2) All foliage was recovered in the summer, 284 therefore N/P ratios could be low because samples were taken during the period of active 285 growth (Méndez and Karlsson, 2005;Rivas-Ubach et al., 2012).(3) As a result of N 286 limitation, biomass allocation to roots can increase at the expense of foliage (Andrews et 287 al., 1999;De Groot et al., 2003).(4) P uptake can be enhanced in response to deficiencies 288 through several mechanisms, such as exudation of enzymes or acids (Dakora and Phillips, 289 2002) and association with mycorrhizal fungi (Colpaert et al., 1999).(5) Productivity 290 might also be limited by other elements (van Duren and Pegtel, 2000), solar irradiance, 291 and/or climatic factors (Spink et al., 1998).292 To assess potential nutrient limitations resulting from wildfire, we compared 293 average N and P concentrations in foliar material from plants grown in soils that were 294 burned in 1910 and 1918 CE (n= 29) with sites not burned (n= 43) since the start of the 295 Fire Atlas (Morgan et al., 2014).We found no significant difference in average N (t = -296 0.13, df= 47.76, p= 0.8948) or P (t= 0.99, df= 53.48, p= 0.3258) concentrations between 297 the two sample groups.Likewise, average C (t= -0.40, df= 65.96, p= 0.6901) 298 concentrations did not significantly differ between foliar samples taken from burned 299 versus unburned locations.This indicates that, although past wildfires reduced the pool of 300 soil C, they did not affect the concentration of growth limiting nutrients measured in 301 vegetation a century later.and soil samples from a network of sites located in the northern 307 U.S. Rocky Mountains indicate the spatial distribution of key elements in subalpine 308 forested ecosystems.The concentration of C and nutrients (N, K, Mg, Ca) in soils is 309 highest in the upper litter and organic horizons and decreases at depth in the mineral soil, 310 consistent with previous studies.Comparing the two plant functional types, needle leaf 311 plants contain higher concentrations of C, while broadleaf material is enriched in N and 312 other trace elements (Mg, SO 4 ).Sites that were burned in a regionally extensive wildfire 313 in the early 20 th century contained significantly lower C on average in the organic soil 314 horizon, compared with sites that burned prior to the 20 th century.This highlights the 315 important role of wildfires as a dominant driver of soil C dynamics in sub-alpine forests, 316 with legacies that can last for more than a century (Hudiburg et al., 2017).Furthermore, 317 the high degree of variance in soil C concentrations among burned soils is consistent with 318 the inherent spatial heterogeneity in fire severity seen in contemporary fires.This spatial 319 heterogeneity adds to additional complexity Earth Systems Modeling efforts to represent 320 fire across space and time.The low average values of foliar N/P ratios (9.8 ± 0.6) suggest 321 Biogeosciences Discuss., https://doi.org/10.5194/bg-2018-443Manuscript under review for journal Biogeosciences Discussion started: 15 November 2018 c Author(s) 2018.CC BY 4.0 License.
The samples were collected during the middle of the growing season from sites 85 with similar elevation, climate, and bedrock geology, but with varying fire histories over 86 the 20 th century.Our approach isolates local plant and soil forming processes that affect 87 the distribution of key elements, and compares sites that are reported to have burned in 88 extensive fires in 1910 and 1918 CE to those that have not burned since before 1901 CE 89 (Morgan et al., 2014).We evaluate nutrient status and characterize patterns of element 90 distribution within and among the leaves of several tree and shrub species in the context 91 of associated bioavailable nutrient data from upper soil horizons.We use statistical 92 methods to infer the transfer and distribution of key elements in soil and vegetation.

Of the 14 sites, seven experienced wildfire in 1910 or 103 1918 CE, while the other seven have no indication of fire activity back to 1901 CE 104 (Morgan et al., 2014)(Table 1). The regional climate is classified as modified maritime 105 with warm, dry summers and cool, wet winters. The study region is underlain with
The 100 vegetation at all sites is subalpine forest dominated by subalpine fir (Abies lasiocarpa), 101Engelmann spruce (Picea engelmannii), lodgepole pine (Pinus contorta), and mountain 102 hemlock (Tsuga mertensiana).