Introduction

Intensive management regimes, global warming, introduced species, and other anthropogenic alterations of biogeochemical cycles will all potentially affect the soil nitrogen (N) status of ecosystems, with possibly undesirable consequences to plant species composition and dynamics (Vitousek and others 1997). One measure of resilience or adaptive capacity in this regard would be the magnitude of variation in nutrient availability that an ecosystem can experience without undergoing significant changes in productivity or species composition (Carpenter and others 2001; Drever and others 2006). In the case of climate change and N, experimental studies with artificially raised temperatures have demonstrated substantial increases in N mineralization rates, averaging 46% with warming of 0.3–6.0°C, but with wide variation in responses due to factors such as moisture, vegetation type, and site quality (Rustad and others 2001). The difficulties in extrapolating short-term, site-specific results to larger landscapes and time frames suggests that further investigations, including retrospective analysis of historical trends, are needed to better understand the baseline sensitivity of a wide variety of soils to changing environmental conditions.

A potential tool in retrospective analyses of ecosystem N dynamics is the use of natural stable isotope ratios (δ15N), which are typically well correlated to soil N availability for tree species colonized by ectomycorrhizal (EM) fungi (Hobbie and Hobbie 2008; Craine and others 2009). It has been suggested that the relationship between soil N availability and foliar δ15N can be extended to wood as well, allowing for reconstruction of long-term records in soil N availability through tree ring analysis (McLauchlan and others 2007; Bukata and Kyser 2007; Hietz and others 2011). Some caveats in this methodology and interpretations have been commented on, however, especially in relation to internal translocation of N among tree rings (Drake and others 2011; Doucet and others 2011) and the lack of correlation in δ15N among neighboring trees (Hart and Classen 2003). More evidence of the fidelity in wood δ15N as an index of soil N availability is needed, and an effective test might be comparisons of wood δ15N along edaphic gradients where organic matter quantity (total N content) and quality (C:N ratio) underpin strongly contrasting soil N dynamics (Nordin and others 2001; Kranabetter and others 2007; Rothstein 2009).

The Coastal Douglas-fir (CDF) zone is a temperate, coniferous forest zone, the most northern extent of which is located along low elevations of southeastern Vancouver Island and the Gulf Islands of British Columbia (Canada) (Meidinger and Pojar 1991). The CDF zone is characterized by long, dry summers and mild, wet winters, creating Mediterranean-like climatic conditions that differ from the temperate rainforests associated with much of the British Columbia coastline. Our interest in N dynamics of the CDF arose because modelling scenarios of potential future climates suggest these forests will experience increasingly severe warming and drying, with an expansion of the contemporary climate envelope to include areas further north and at higher elevations (Hamann and Wang 2006). To explore these climatic effects on soils, we considered the recent past of the Pacific Northwest, which has undergone decadal-scale oscillations in climate driven by circulation patterns of the Pacific Ocean (Gedalof and others 2002; D’Arrigo and others 2005), including fluctuations in mean annual temperatures of up to 2°C (Davi and others 2006). These decadal fluctuations in climatic regime are potentially reflected in tree ring δ15N, allowing for a better understanding of both the extent of change in soil N availability and the relative influence of temperature and moisture drivers on soils. We might also improve the predictability of N dynamics in the CDF by comparing wood δ15N along edaphic gradients to address the important influence of organic matter quality on warming effects (Hart 2006).

In this study, we describe patterns in soil N availability and wood δ15N along 15 old-growth forest stands that encompass the full productivity extent of upland sites in the CDF zone of southern Vancouver Island. The objectives of our study were to: (1) calibrate wood δ15N of Douglas-fir (Pseudotsuga menziesii var. menziesii), an EM tree species, against N availability metrics of the contemporary landscape along natural productivity gradients; (2) document the stability of ecosystem N status over the previous century through increment cores and the trends in wood δ15N; and (3) test whether fluctuations in mean annual temperature, precipitation, and summer moisture deficits over this same period caused concurrent departures in wood δ15N from the long-term trend. Our hypothesis was that fractionation against the N isotope by EM fungi would be reduced by increased soil N availability and cause current wood δ15N to increase linearly with site quality (as per foliar δ15N; Kranabetter and MacKenzie 2010), and that periodic increases in temperature or precipitation in the recent past would have enhanced soil N mineralization rates (Rustad and others 2001) enough to similarly reduce δ15N depletion in increment cores.

Materials and Methods

Site Descriptions

The CDF lies in the rainshadow of the Vancouver Island and Olympic mountains at elevations below 150–380 m, depending on latitude, with mean annual temperature (MAT) ranging from 9.2 to 10.5°C and mean annual precipitation (MAP) from 647 to 1263 mm (Meidinger and Pojar 1991). Douglas-fir is the most common tree species of upland forests, with western redcedar (Thuja plicata), grand fir (Abies grandis), arbutus (Arbutus menziesii), western hemlock (Tsuga heterophylla), and bigleaf maple (Acer macrophyllum) frequently accompanying Douglas-fir depending on soil moisture and nutrient regime. The region was characterized by frequent, low intensity fires set by aboriginal peoples before European contact (mid-1800s), which with the high fire resistance of Douglas-fir can often result in complex stand structures (Bjorkman and Vellend 2010).

The study took place within 20 km of Victoria, British Columbia (Canada), on the Department of National Defence lands at Rocky Point (48°19′N, 123°34′W, elevation 60 m), Royal Roads (48°26′N, 123°28′W, elevation 50 m) and Heal’s Rife Range (48°32′N, 123°27′W, elevation 100 m), as well as Thetis Lake park (48°28′N, 123°28′W, elevation 80 m). Forests within these protected areas had never been logged, with an exception of two plots with some partial harvesting in the late 1800s. Codominant trees range in age from mature (130–250 years) to old (250–375 years). Nitrogen deposition from atmospheric pollution on southern Vancouver Island is considered insubstantial, likely less than 1 kg ha y−1 (National Atmospheric Deposition Program 2009), but these otherwise pristine forests have had exotic lumbricid earthworms introduced within the last century (Addison 2009).

Fifteen plots (25 × 25 m in size) were selected to provide three replicates of five site series (represented by climax plant communities corresponding to soil moisture and nutrient regime) of the CDF zone (Green and Klinka 1994): (02) xeric and very poor Douglas-fir, Lodgepole pine—Arbutus; (03) submesic and poor Douglas-fir—Oniongrass; (01) mesic and medium Douglas-fir—Salal; (04) moist and rich Douglas-fir, Grand fir—Oregon grape; and (06) subhygric and very rich Western redcedar, Grand fir—Foamflower. Site series are hereafter referred to by their relative nutrient regime (very poor, poor, medium, rich, and very rich). Site characteristics, vegetation, and soils were described for each plot in 2009 using standard provincial protocols (BC Ministry of Environment, Lands, and Parks and BC Ministry of Forests 1998). Average site index (m@50y) per plot was based on ages and heights for three co-dominant Douglas-fir using the British Columbia Ministry of Forests Site Tool (version 3.2B).

The majority of soils along the edaphic gradient had coarse (loamy sand) to medium (sandy loam to silt loam) textures, characterized by thin (<2 cm) to nonexistent forest floors with a dark brown, organic-enriched, acidic upper mineral horizon (an Ahf; Broersma and Lavkulich 1980) and a red colored, podzolized lower mineral horizon (Bf), corresponding to Sombric Humo-Ferric Podzols (Soil Classification Working Group 1998). Rooting depth averaged 50 cm, although soil profiles of the very poor and poor sites were occasionally shallow, with bedrock at 10–30 cm depths. Further details on plant communities, stand structure and soil profiles for the study plots are located online with the 2009 Environmental Science Advisory Committee annual reports (http://cfs.nrcan.gc.ca/subsite/esac/annual-report).

Climate Instrumental Record

Monthly temperature and precipitation records began in the city of Victoria, British Columbia, in 1899 near the harbor, and then the station was moved in 1914 to a nearby, permanent location at Gonzales (Lat 48.42, Long 123.32, elevation 70 m). Comparison with other stations in the region was used to check homogeneity of the Gonzales record and to aid filling in missing data. The temperature for the 1899–1914 period has not been adjusted for the new location because no other climate station records are available for cross comparison. Climatic moisture deficit (CMD) was calculated as the estimate of evaporative demand (Allen and others 1998) minus precipitation for each month. Summer heat:moisture index (SHM) was determined by dividing the mean warmest month temperature by the mean May–September precipitation (Wang and others 2006). Gonzales is drier than the study sites (Tuller 1987), but provides a reliable long-term index of the interannual variability in the regional climate.

Soil Sampling

In early spring of 2010 (March 29–31), at the onset of the active growing season, an in situ incubation of soils was established at five random locations within each plot (Hart and others 1994). Soil cores were extracted with a soil auger at 0–15 cm and 15–30 cm depths, then gently poured into polyethylene bags, sealed with a twist tie, and returned to the bore hole for 8 weeks. Forest floors were not sampled separately because the majority of microsites had mineral soil directly beneath the litter; if surface organic matter was found (usually 1–2 cm), it was included with the upper mineral soil incubation. The depth of sampling was noted whenever the soil profile was less than 30 cm because of the underlying bedrock. After the 8-week incubation, the soil samples were retrieved and immediately run through a 5-mm sieve for analysis of amino acids, NH4 +, and NO3 − concentrations at the MOF analytical laboratory in Victoria. One subsample was taken for moisture content (105°C for 24 h) and a second subsample was ground and resieved to 2 mm to correct for remaining coarse fragment content. Loss on ignition (650°C for 24 h) was then undertaken on these latter samples to derive a bulk density estimate based on organic matter content for each buried bag (Périé and Ouimet 2008).

Soil samples for total N and carbon (C) concentrations were collected at the same depths in July of 2010 at nine random locations and bulked into three subsamples per depth. These soils were air-dried, ground, and sieved (2 mm) for chemical analysis, including 15N abundance of the 0–15 cm depth. The coarse fragment content from these samples was determined gravimetrically and converted to a percent volume estimate using a specific gravity of 2.65 g cm3 and sampling volume corresponding to the auger (15 cm depths and 4 cm diameter). Concentrations of soil amino acids and inorganic N were converted to mass estimates for the soil profile (0–30 cm) using the bulk density values derived from organic matter content and the average coarse fragment content determined from soil chemistry sampling.

Wood Sampling

Increment cores from the three codominant Douglas-fir site index trees were taken in February 2010 from each plot. Cores were air-dried and lightly sanded only where necessary to differentiate tree rings under magnification. Samples from 3-year tree ring increments were separated with a chisel, going from 2009 back to 1900 (36 samples per core), and bulked together by sample period among the three trees. No attempt was made to cross-date the cores because mounting and sanding of the increment cores for scanning (for example, Zhang and Hebda 2004) would not allow sufficient material for isotope analysis. At one of the very rich sites, we were unable to sample earlier than the mid-1920s because the radius of the trees exceeded the length of the increment corer. In October 2010, the current year tree ring deposited over the growing season was sampled with an increment corer from nine codominant Douglas-fir trees and bulked into three subsamples. All the wood samples were then oven-dried at 60°C for 24 h and ground with a Wiley mill to less than 2 mm for natural stable isotope analysis.

Laboratory Analysis

Amino acid-N, NH4-N, and NO3-N were determined from a 5-g dry-soil equivalent of mineral soil (Hart and others 1994). A 2 M KCl solution was added to the soils at a 1:10 w/v ratio, and samples were shaken for 1 h at 20°C. Samples were clarified by centrifugation for 15 min at 850 g. The extract was pipetted from the clear supernatant into an auto-analyzer cup for analysis. The NH4-N and NO3-N in the extracts were measured colorimetrically using an Alpkem Flow System IV analyzer (OI Analytical, College Station, Texas). Amino acids in the extract were measured using a ninhydrin method (Swift and Bingil 2001). One milliliter of extract was mixed and shaken with 1 ml of ninhydrin color reagent. The samples were heated for 45 min at 105°C, then cooled and mixed with 5 ml of 2 M KCl. Absorbance at 570 nm was measured with a UV–Visible spectrophotometer and corrected for previously measured NH4 + levels in the determination of amino acid concentrations. Amino acid concentration was converted to amino acid N using an average ratio of 1.4 mol N mol−1 amino acids (Rothstein 2009). Lab standards were calibrated against standards AG1 and AG2 for nitrate (mean 16.0 mg kg−1, standard deviation [SD] 0.7), ammonia (6.3 mg kg−1, SD 0.6) and ninhydrin reactive N (10.8 mg kg−1, SD 0.8). Total C and N were measured using combustion elemental analysis with a Fisons/Carlo-Erba NA-1500 NCS analyzer (Thermo Fisher Scientific, Waltham, Massachusetts) (Kalra and Maynard 1991).

The natural abundance of the 15N isotope in soil was determined using AS autosampler of a Thermoquest (Carlo Erba Instruments) NC 2500 elemental analyzer. The mass ratio 29/28 of the sample relative to the mass ratio 29/28 of a N reference gas was used to determine the isotopic value of the sample. Lab standards were calibrated against the international standards IAEA-N1 (+0.4 per mil) and IAEA-N2 (+20.3 per mil). Natural abundance of 15N in wood (25 mg subsamples) was measured on a PDZ Europa 20-20 isotope ratio mass spectrometer fitter with sequential traps of MgClO4, NaOH on solid support (Carbosorb, Sydney, Australia), and a cold trap in liquid N2. Lab standards were calibrated against standards UCDN2 (mean −0.31 ‰, SD 0.25), nylon (−9.77 ‰, SD 0.16), USGS-41 enriched glutamic acid (+45.1 ‰, SD 0.13), and peach leaves (+1.95 ‰, SD 1.2). Nitrogen isotopic values are listed relative to air δ15N versus air (Hauck and others 1994). A technical difficulty arose with the incomplete combustion of a subset of wood samples, limited to four plots, and so these were rerun using an alternative methodology that utilized a vario EL Cube (Elementar Americas Inc., Mt. Laurel, New Jersey), with a molsieve absorption trap that releases CO2 after the N2 peak is analyzed. It is uncertain whether the two methodologies yield the same results in absolute 15N abundance, but we limited any possible methodology effects by utilizing residuals in wood δ15N for the correlation analysis.

Statistics

The study was organized in a randomized complete block design (one replicate of each site series at Thetis Lake park, Heal’s Rifle Range and Rocky Pt/Royal Rds). Site index of Douglas-fir was the independent variable in the statistical analysis, as plots were chosen based on productivity differences among plant associations, whereas the soil N indices and N isotope abundance were dependent variables. The null hypothesis was for no change in soil N indices or N isotope abundance with increases in site index, and no relationship between soil N availability and wood δ15N. The GLM procedure in SAS using Type 1 Sums of Squares (SAS Institute Inc. 2004) was used to test linear and curvilinear regressions of plot means for the dependent variables, with the best fit determined by r 2 values. Trends in wood δ15N over time were fitted to linear or curvilinear regressions for each plot using the GLM procedure and Type 1 Sums of Squares, with the best fit determined by r 2 values. The δ15N data were then detrended to accommodate potential differences in stand conditions and site history. Residual δ15N was determined for each 3-year sample period by subtracting the data point from the fitted regression. Detrended residuals in MAT, MAP, CMD, and SHM were also calculated for the corresponding 3-year period of tree ring sampling by fitting regressions to the climatic data recorded for Victoria. Pearson correlations between climatic and wood δ15N residuals were undertaken by the Correlation procedure in SAS, and included partial correlations using MAP or SHM with MAT (SAS Institute Inc. 2004).

Results

Climatic Records for Victoria British Columbia

Mean annual temperatures (MAT) have increased by 0.7°C in the past century in Victoria, whereas in 2010, the year of the study, the mean temperature of 10.5°C was close to the long-term mean of 10.2°C (Figure 1A). MAP averaged 666 mm over this same time period and has not changed significantly over time (P = 0.152, Figure 1B). Drought indices have also remained consistent during the past century, with an average CMD of 30.3 (P = 0.502) and summer heat:moisture index (SHM) of 164 (P = 0.859).

Figure 1
figure 1

A Mean annual temperature by year (gray line) and 3-year average (black line) from 1899 to 2010 (°C = −2.98 + 0.0067[year]; P < 0.001, r 2 = 0.14), and B mean annual precipitation over the same period.

Full size image

Nitrogen Status of the Contemporary Landscape

The average concentrations of amino acids, NH4 + and NO -3 (mg kg−1 soil) extracted after the 8-week incubation demonstrated strong patterns along the productivity gradient (as defined by Douglas-fir site index, SI) where each N type was progressively replaced in dominance (Figure 2A). Amino acids comprised the highest proportion of extractable N on poorer sites and declined linearly, whereas NH4 + concentrations peaked on the medium to rich sites, followed by the most productive sites where NO3 − dominated. An exception to this trend was a slight upturn in NO -3 concentrations found on poorer quality sites (Figure 2A). When converted to mass (kg ha−1), the quantity of extractable N (organic and inorganic N combined) as an index of soil N supply was positively associated with Douglas-fir site index (P = 0.001, r 2 = 0.56; Figure 2B). Total N content of the soil profile (0–30 cm) ranged from 2200 to almost 4000 kg ha−1, and was also positively correlated to Douglas-fir site index (Total N = 1297 + 58.5[SI]; P = 0.003, r 2 = 0.51). The quality of the soil organic matter, as defined by the carbon:nitrogen (C:N) ratio, ranged from 19 to 31, but a linear correlation was weaker (P = 0.026, r 2 = 0.33) than a negative curvilinear correlation with site productivity (C:N = −3.4 + 2.55[SI] − 0.05[SI]2; P = 0.001, r 2 = 0.68), where the highest C:N ratios were found among the medium quality sites.

Figure 2
figure 2

A Gross production of amino acids, NH4 + and NO3 − along the productivity gradient of Douglas-fir (each point an average of the three replicates for each plant association, with bars representing ±SE), and B amino acid + inorganic N content (kg ha−1) of the soil profile (0–30 cm) (±SE) for each plot in correlation with Douglas-fir site index (organic + inorganic N = −5.0 + 1.03[SI]; P = 0.001, r 2 = 0.56).

Full size image

The natural 15N abundance of the upper mineral soil (0–15 cm) averaged 1.75 (SE 0.18) and did not differ significantly (P = 0.120) along the productivity gradient (Figure 3A). The current year’s wood increment, in contrast, was more depleted in 15N than the soils, averaging −3.9 (SE 0.15). Wood δ15N was not correlated to soil δ15N (P = 0.878), and instead was positively but weakly correlated with site productivity (Figure 3A; P = 0.010, r 2 = 0.41). A stronger relationship was found by multiple linear regression, however, to account for variation in soil δ15N (Figure 3A). Alternatively, a similar result was obtained with Δδ15N (the difference between mean wood and soil δ15N of each plot), which ranged from 4.0 to 7.0 and was moderately well correlated negatively with site productivity (Δδ15N = 8.6 − 0.11[SI50]; P = 0.001, r 2 = 0.58) and soil N supply (Figure 3B).

Figure 3
figure 3

A Soil δ15N (0–15 cm) and current wood δ15N increment (bars ± SE) for each plot in relation to Douglas-fir site index; the multiple linear regression (SI50 = 59.3 + 6.46[ring δ15N] − 4.21[soil δ15N]; P = 0.004, r 2 = 0.61) was plotted with soil δ15N = 1.75, and B Δδ15N for each plot in correlation with soil N availability (0–30 cm) (Δδ15N = 7.4 − 0.077[org. + inorg. N]; P = 0.002, r 2 = 0.53).

Full size image

Historic Patterns in δ15N Revealed Through Increment Cores

Overall, the combined trend in wood δ15N values from 1900 to 2009 was negative (δ15N = 14.6 − 0.0087[year]; P < 0.001), with a net decline of 0.9 ‰, but the patterns diverged widely when compared among plots. To illustrate some examples, we found a slight, consistent increase in wood δ15N with time (Figure 4A); an initially elevated δ15N followed by a sharp decline and plateau (Figure 4B); and a plateau in δ15N followed by a more recent, slow decline (Figure 4C). In total, six plots had a negative trend in wood δ15N, three plots had a positive trend, three plots had a curvilinear trend, and three plots had no trend with time. The net change in 15N abundance from 1900 to 2009 across the 15 plots ranged from 0 to 3.5 ‰, whereas the interannual oscillation was commonly 2–3‰.

Figure 4
figure 4

Wood δ15N over time, 1900–2009, for three plots demonstrating a contrasting range of patterns. Residual δ15N was determined by subtracting data points from the fitted regressions.

Full size image

The wood δ15N data for each plot were detrended to standardize comparisons of residual wood δ15N in relation to climatic variables. An overall mean residual by year, or average by site type, was deemed inappropriate because of poor correlations in residual δ15N among plots; only 5 out of 105 pair-wise comparisons were significant (Pearson correlation P < 0.10). When tested individually (three examples provided in Figure 5A–C), the residual δ15N was significantly correlated (P < 0.10) with residual MAT or SHM in only 1 of the 15 plots for each variable (Table 1). We did not detect any correlations in residual δ15N with MAP (not shown), and found no improvement in residual δ15N–MAT correlations when MAP was included as a partial variable.

Figure 5
figure 5

Residual δ15N by sample period (3-year average) for three plots demonstrating a lack of correlation with residuals in mean annual temperature.

Full size image
Table 1 Pearson Correlation (P value) and Coefficient (where P < 0.05) by Plot for Wood δ15N Residuals with Annual Temperature (MAT) and Summer Heat:Moisture Index (SHM) Residuals, 1900–2009
Full size table

Discussion

The edaphic gradient reflected well-established relationships between soil N supply and stand productivity in coniferous forests, largely underpinned by soil moisture regime and geochemistry via slope position, soil texture, coarse fragment content, and parent material (Prescott and others 2000; Högberg and others 2006). The maximum soil N concentration of the upper horizon among the plots was 0.36%, which was well below the threshold suggested for saturation of plant N demands in temperate forests of the Pacific Northwest (Perakis and Sinkhorn 2011). The progressive replacement of N forms along the productivity gradient, from predominantly organic (amino acids) supplies to NH4 + and ultimately NO3 −, was generally consistent with temperate forest edaphic gradients elsewhere (Berthrong and Finzi 2006; Rothstein 2009). The results suggest a greater degree of N mineralization than found in boreal forest soils (Kranabetter and others 2007), in part because of the lack of forest floors in the CDF zone. We found infrequent “hotspots” of NO3 − (up to 50 mg kg−1 soil) on poorer sites, which was unexpected, but these anomalies might explain why shallow soils on ridges or bedrock can sometimes host a small coverage of nitrophilic plants. Many of the correlations between soil N properties (extractable N, total N, C:N ratio) and site index had only moderate r 2 values (~0.58), possibly because of other colimitations to tree growth such as water storage deep in the soil profile (Warren and others 2005).

The relationship between site productivity and wood δ15N of Douglas-fir was consistent with positive correlations found for foliar δ15N of EM trees on similar edaphic gradients (Kranabetter and MacKenzie 2010), and confirms the utility of this index in monitoring N dynamics as we had hypothesized. Our study was designed to test inherent differences in N availability among sites, rather than after a large, one-time input of N, and so the mobility of N within sapwood may have been less confounding to the results (Hart and Classen 2003; Drake and others 2011). Depletion of 15N abundance in EM trees is thought to be primarily caused by the degree of internal isotopic discrimination by mycorrhizal fungi during the transfer of N to the plant (Hobbie and Colpaert 2003; Hobbie and Hobbie 2008). Increased soil N availability typically results in increased plant δ15N (Craine and others 2009) because EM fungi retain a lower proportion of N uptake for their own metabolism on productive sites (Hobbie and Colpaert 2003), although isotopic differences in the form of available N are also a possible influence (Averill and Finzi 2011). Although it is difficult to isolate N retention effects from the supplies of inorganic N, both of which change along the productivity gradient, we did not find evidence of a curvilinear pattern in wood δ15N that would correspond to the contrasting supplies of NH4 + or NO −3 across site types. Presumably the linear correlation of 15N with gross N supply is due to the full utilization of all N forms by Douglas-fir (that is, no surplus N to plant demand; Perakis and Sinkhorn 2011), resulting in minimal fractionation of the isotope during N uptake (Hobbie and Hobbie 2008). We found that small variations in soil 15N abundance hampered the comparison of N status among sites, so Δδ15N (wood δ15N − soil δ15N) was useful in standardizing the extent of internal fractionation by EM fungi. Wood δ15N correlations were again moderately effective (r 2 ~ 0.57) in capturing variations in site productivity and soil N supply, but this was consistent with all the soil–stand relationships determined under these study parameters.

Although recognizing the primary influence of soil N supply on the fractionation of 15N, the increment cores also revealed considerable complexity in describing how these relationships may have changed over time. Temporal patterns in wood δ15N might differ among the plots with inconsistent forest succession effects on soil N availability (Smithwick and others 2009) and N forms (LeDuc and Rothstein 2010), or because of discrete disturbances from low severity ground fires or wind events (Couto-Vázquez and González-Prieto 2010; Hietz and others 2010; Beghin and others 2011). The overall negative trend in wood δ15N could imply soil N availability has generally declined over the last century in this unpolluted forest landscape, but we suspect, given the many positive or flat trends in wood δ15N, that nonedaphic factors involving site history and stand dynamics have influenced wood δ15N to some degree.

Examples of possible stand dynamic effects would include how N stress is alleviated during self-thinning of regenerating stands (as per foliar N concentrations and stand density relationships; Turner and others 2009), likely causing an increase in wood δ15N (Stock and others 2012). EM fungi require an allotment of photosynthate C from the host tree to scavenge soil N (Högberg and others 2003, 2010), and so stress (root or foliar disease, wind damage) or simply size may have negatively affected photosynthetic efficacy (Bond and others 2007) and reduced the C investment in fungi for N uptake. The most striking changes in wood δ15N over time were from younger (~150 years), even-aged stands at Rocky Pt (Figure 4B), and we speculate on the influence of EM fungal succession over this period, especially tuberculate ectomycorrhiza (Rhizopogon spp.) that are abundant in young stands (Twieg and others 2007) and might supplement N uptake through associated fixation (Li and others 1992; Paul and others 2007). Finally, we assume soil 15N abundance was relatively stable over the time period sampled, but there is the possibility of changes in isotopic signatures with extensive organic matter transformations (Billings and Richter 2006), which in these forests may have occurred to some degree after the introduction of lumbricid earthworms (Addison 2009; Dempsey and others 2011).

We hypothesized that periodic increases in temperature or precipitation over the past century would have stimulated soil N mineralization rates (Rustad and others 2001) and caused a concurrent shift in wood δ15N of the increment cores. We found instead a general lack of correlation among plots in residual wood δ15N, and in relation to climate variables, despite the large interannual fluctuations in MAT and MAP (up to 3°C and 800 mm). Part of the inconsistency in wood δ15N could be related to the sampling issues of stable N isotopes, especially in comparison to the precision garnered under standard dendrochronology methods (for example, Zhang and Hebda 2004). Among these is the cost and effort in sampling, which greatly limits replication and makes it more difficult to determine a robust average trend for plot correlations. Data collected for single trees by year, rather than for bulked trees by multiple years, might improve precision but the minimum sample size (25 mg) required for stable isotope analysis precludes this on older trees with thin tree rings. The mobility of N in sapwood potentially reduces synchronicity with short-term climatic variations, although for the majority of increment cores with heartwood the unextracted δ15N values should be less affected by N translocation (Bukata and Kyser 2005; Hietz and others 2010; Doucet and others 2011). Accurate measurement of the N isotope is also a consideration, as the very low N concentration of wood is technically challenging and likely produces data less precise than the measurement of ring width. There is also the issue of missing tree rings on moisture-stressed sites, which can be accounted for after cross-dating of increment cores but are more difficult to correct for in destructive sampling. These methodological challenges in adequately constructing historic 15N trends suggests that the data are better suited to infer more widespread, landscape-level changes in soil N availability and N stress of trees over time (McLauchlan and others 2007; Hietz and others 2011), rather than discrete changes in N availability.

Another implication to the lack of correlation in residual wood δ15N with climate variation over the last century is the possibility that soil N dynamics have been relatively insensitive to the interannual oscillations in MAT and MAP. There is arguably a high degree of inertia in soil N availability across (unpolluted) landscapes through the constraints attributed to the quantity (total N) and quality of soil organic matter (C:N ratio) (Booth and others 2005). Neither property is likely to change appreciably with yearly or decadal climatic fluctuations, and may undergo significant changes only after severe disturbances and vegetative shifts, especially to N-fixing alders (Perakis and Sinkhorn 2011). Although it is well established that increases in temperature can immediately stimulate N mineralization rates (Rustad and others 2001), it is also hypothesized that this boost may only last with a concurrent increase in soil N pools (Melillo and others 2011). These inherent limits in fertility set by soil organic matter would be consistent with the role attributed to species conversion in altering soil N dynamics, rather than a more primary influence of warming (Willis and others 1997; Niu and others 2010; Jeffers and others 2011). Perhaps the more likely outcome in the CDF to the short-term climate fluctuations of the past century were subtle shifts in the predominance of N forms (amino acids, NH4 +, NO3 −), which might have altered the competitive balance among mycorrhizal plant guilds (Kranabetter and MacKenzie 2010) but had less effect on total N availability and isotopic fractionation.

Conclusions

We validated the natural stable isotope ratios (δ15N) of wood (or more precisely by Δδ15N, the difference between wood and soil δ15N) as a measure of soil N availability and N deficiency in mature EM trees from contemporary landscapes. Interpreting temporal changes in stand N status via increment core δ15N was more challenging, partly because of the potential variation in factors affecting N supply (ground fires, forest succession, invasive earthworms), and in factors affecting stand response and N uptake (EM fungal succession, intraspecific competition, photosynthetic capacity). These issues in interpretation suggest some caution is needed in differentiating possible site-level tree physiological effects on wood δ15N from the broader climatic or anthropogenic influences on soil N dynamics. Patterns in wood δ15N might be more consistent, and interpretations more credible, in studies undertaken on uniform, even-aged stands with a less complex history of disturbance. The lack of correlation in residual δ15N with climate variation may be partly due to limitations in methodology, but might also reflect the relative stability in N supply over the previous century due to the overriding constraints of soil organic matter quantity and quality.