Skip to main content
NCBI home page
EPA Author Manuscripts logoLink to EPA Author Manuscripts
. Author manuscript; available in PMC: 2022 Oct 27.
Published in final edited form as: Environ Res Commun. 2020 Feb 27;2(2):1–17. doi: 10.1088/2515-7620/ab77f8

Positive correlation between wood δ 15N and stream nitrate concentrations in two temperate deciduous forests

Robert D Sabo 1, Andrew J Elmore 2, David M Nelson 2, Christopher M Clark 3, Thomas Fisher 2, Keith N Eshleman 2
PMCID: PMC9610404  NIHMSID: NIHMS1579159  PMID: 36313933

Abstract

A limitation to understanding drivers of long-term trends in terrestrial nitrogen (N) availability in forests and its subsequent influence on stream nitrate export is a general lack of integrated analyses using long-term data on terrestrial and aquatic N cycling at comparable spatial scales. Here we analyze relationships between stream nitrate concentrations and wood δ15N records (n = 96 trees) across five neighboring headwater catchments in the Blue Ridge physiographic province and within a single catchment in the Appalachian Plateau physiographic province in the eastern United States. Climatic, acidic deposition, and forest disturbance datasets were developed to elucidate the influence of these factors on terrestrial N availability through time. We hypothesized that spatial and temporal variation of terrestrial N availability, for which tree-ring δ15N records serve as a proxy, affects the variation of stream nitrate concentration across space and time. Across space at the Blue Ridge study sites, stream nitrate concentration increased linearly with increasing catchment mean wood δ15N. Over time, stream nitrate concentrations decreased with decreasing wood δ15N in five of the six catchments. Wood δ15N showed a significant negative relationship with disturbance and acidic deposition. Disturbance likely exacerbated N limitation by inducing nitrate leaching and ultimately enhancing vegetative uptake. As observed elsewhere, lower rates of acidic deposition and subsequent deacidification of soils may increase terrestrial N availability. Despite the ephemeral modifications of terrestrial N availability by these two drivers and climate, long-term declines in terrestrial N availability were robust and have likely driven much of the declines in stream nitrate concentration throughout the central Appalachians.

Keywords: nitrogen, nitrate, stable isotopes, forest, stream, oligotrophication

Introduction

Recent changes in climate, atmospheric chemistry, and disturbance have the potential to influence the productivity and export of nutrients from forested catchments by either constraining or enhancing nitrogen (N) availability [Elmore et al 2016, McLauchlan et al 2017, Peñuelas et al 2017]. Terrestrial N availability may increase from warming soils further stimulating mineralization, atmospheric N deposition adding N to soils, or changes in microbial communities increasing mineralization rates as acidified soils recover from decades of elevated acidic deposition [Aber et al 1989, Rustad et al 2001, Sinsabaugh et al 2004, Sinsabaugh 2010, Brookshire et al 2011, Oulehle et al 2011, Lawrence et al 2015, Oulehle et al 2017, Lawrence et al 2019]. On the other hand, greater N demand resulting from factors such as longer growing seasons or elevated atmospheric CO2 concentrations could reduce N availability [Ollinger et al 2002, Norby et al 2016]. As inferred from foliar and wood chemistry and net mineralization and nitrification rate measurements in North American forests, the latter is supported by evidence that N availability has declined in temperate forests during the past three decades [McLauchlan et al 2010, Durán et al 2016, Elmore et al 2016, McLauchlan et al 2017, Groffman et al 2018]. Stream nitrate export has also declined since the mid-1990s throughout many forests of the eastern United States [Eshleman et al 2013], which could result from declines in terrestrial N availability [Elmore et al 2016]. However, few studies have co-located records of terrestrial N availability and stream water chemistry with which to assess this hypothesis.

A general lack of paired analysis of terrestrial and aquatic N cycling at comparable spatial scales limits understanding of the drivers of long-term trends in terrestrial N availability and its subsequent influence on stream nitrate export. Stream nutrient data are often only obtained for limited periods or measured only periodically [Argerich et al 2013, Durán et al 2016], and such snapshots may not be adequate for comparison with longer-term indicators of terrestrial N cycling, such as deduced from tree-ring δ15N records [Gerhart and McLauchlan 2014]. While stream water nutrient data are generally considered integrative of an entire watershed [Likens 2017], in-stream processing could lead to an incongruity between stream and terrestrial N datasets [Scanlon et al 2010]. Tree-ring δ15N records have been suggested to be an integrated metric of soil N supply relative to plant demand [McLauchlan et al 2007, Howard and McLauchlan 2015, Sabo et al 2016a, McLauchlan et al 2017]. Wood δ15N integrates the effects of multiple fractionation pathways on N isotopes as it is transformed and lost in the ecosystem (e.g. nitrification, denitrification, etc). Overall, sites that have greater N availability tend to have higher rates of nitrification, denitrification, and leaching [Aber et al 1989, Galloway et al 2003, Pardo et al 2006, Pardo et al 2007, Templer et al 2007]. Greater N loss rates and fractionation leads to more 14N being lost from the system thus resulting in more isotopically positive δ15N signal remaining in the soil inorganic nitrogen (IN) pool available to plants [Sabo et al 2016a]. Overall, it has been hypothesized that stream nitrate and tree-ring δ15N should be positively correlated through time [McLauchlan et al 2007, Burnham et al 2016, Sabo et al 2016a], since stream nitrate concentrations are also thought to index terrestrial N availability [Aber et al 1989]. Long-term records of terrestrial N availability preserved in tree-ring δ15N data are typically normalized to the mean δ15N value of tree-ring time series to focus on temporal patterns [McLauchlan et al 2007], but doing so masks potential information about spatial variation in terrestrial N availability contained in non-normalized δ15N values and its relationship with stream nitrate [LeDuc et al 2013].

Catchments where long-term records of stream nutrient export exist and multiple tree-ring δ15N records of multiple species can be obtained represent promising locations for assessing the relationship between stream nitrate export and terrestrial N availability. For example, Sabo et al (2016b) observed higher mean (non-normalized) δ15N tree-ring values across a catchment in the Adirondack Mountains with higher mean annual flow-weighted nitrate concentrations relative to an adjacent catchment with lower nitrate concentrations (n = 20 tree-ring δ15N records per catchment), as well as declines in δ15N values of tree rings and stream nitrate concentrations for one of two subcatchments. However, a limitation of that study was that the water chemistry data spanned a period of only 11 years and that the most important tree species in one of the study catchments (Fagus grandifolia; American beech) could not be cored because it suffered from extensive disease and mortality [Sabo et al 2016a]. Thus, uncertainty remains about whether tree-ring δ15N records are representative of the relationship between terrestrial N availability and stream N export [Sabo et al 2016a].

To assess the relationship between terrestrial N availability and stream N export and to identify candidate drivers of the temporal variability of wood δ15N, we assembled dendroisotopic records for 96 trees and stream nitrate records for six small forested catchments in the central Appalachian Mountains. Rather than restrict our efforts to describing the temporal variability of δ15N for specific tree species, which ultimately may not be representative of catchment-wide changes in terrestrial N availability [Burnham et al 2016], we used a randomized sampling design to capture catchment-wide changes in wood δ15N. Furthermore, we used absolute non-normalized δ15N values to explore the spatiotemporal variation in terrestrial N availability and its relationship with stream nitrate concentration. Our objective was to test the hypothesis that spatial and temporal variation in terrestrial N availability observed in tree-ring δ15N records is associated with the spatial and temporal variation of stream nitrate concentration, respectively. To elucidate some of the potential drivers of terrestrial N availability, relationships among acidic deposition (nitrogen (N) + sulfur (S) deposition), precipitation, temperature, and forest disturbance on wood δ15N were also assessed.

Methods

Site description

The Upper Big Run catchment (UBR, 1.63 km2) is located within the Appalachian Plateau physiographic province in the western panhandle of Maryland, USA (figure 1, supplemental figure 1 is available online at stacks.iop.org/ERC/2/025003/mmedia). The underlying geology within the watershed consists of folded sedimentary rocks of Devonian through Mississippian age [Eshleman et al 1998]. Soils (Ultisols and Inceptisols) are primarily composed of stony loams of the Dekalb/Gilpin and Meckesville series [NRCS 2009]. According to the National Vegetation Classification System, the two primary ecological forest systems that occur within the catchment are Northeastern Interior Dry-Mesic Oak Forest and Appalachian Hemlock-Hardwood Forest [GAP 2011]. The watershed is 91% forested with the remainder consisting of meadows, roads, cropland, and a power-line right-of-way. Various silvicultural activities have been carried out in Upper Big Run since the 1970s and multiple insect defoliation events have been observed in the late-1980’s and mid-2000’s (supplemental figure 1; [Eshleman et al 1998, Townsend et al 2012].

Figure 1.

Figure 1.

Open in a new tab

Maps of Upper Big Run and the five headwater catchments of Paine Run subjected to tree coring. The five headwater catchments were labeled PR1000, PR2000, PR3000, PR4000, and PR5000 from west to east since they contain unnamed tributaries, and these site IDs corresponded with individual trees sampled in the respective catchments.

The Paine Run catchment (PR, 12.7 km2) is located in the Blue Ridge in Shenandoah National Park, Virginia (figure 1, supplemental figure 1). The watershed falls within a designated wilderness area. Surficial geology of Paine Run is composed of phyllite, quartzite, and metasandstone. Soils (Ultisols and Entisols) are primarily composed of channery or stony loams of the Hartleton or Drall soil series. Numerous rubble islands are scattered throughout the catchment [NRCS 2009]. Paine Run is composed of secondary growth forests that have not been logged since the early 20th century. Vegetation surveys from the 1980’s show that chestnut oak (Quercus prinus) and various species of pine (Pinus spp.) were the dominant species. Widespread oak mortality in the early 1990’s was associated with repeated gypsy moth defoliation in the upper elevations of the watershed [Scanlon et al 2010]. Today, common forest types that occur within the catchment are Southern and Central Appalachian Oak Forest and Cove Forest along with scattered North-Central Appalachian Circumneutral Cliff and Talus systems [GAP 2011]. The five headwater catchments of Paine Run subjected to study each contain unnamed tributaries to the mainstem. As such, the catchments were labeled PR1000, PR2000, PR3000, PR4000, and PR5000 from west to east. These labels corresponded to the individual trees sampled in the respective catchments (supplemental table 5).

Dendroisotopic records

Thirty trees at UBR and 66 trees at PR were cored at randomly located plots (figure 1). In the field, 706.86 m2 circular plots were established at the sampling points at UBR, and 225 m2 square plots were established within the five PR subcatchments (e.g., PR1000, PR2000, etc). Tree species composition surveys were completed during the fall and winter of 2014; all stems >5 cm diameter at breast height (dbh) within each plot were measured and identified. The bole of one individual of the species with the highest summed dbh in each plot was cored twice using a 5.15 mm incremental borer. In total, 18 different species were sampled at the two sites (supplemental table 5). The cores were returned to the lab, dried in an oven at 60 °C, sanded, and stored until ring widths were identified using CooRecorder software [Larsson 2009]. Two to three-year increments were generally cut from one bore per tree using a razor blade and stored in 96-well plates prior to δ15N analysis. Rings were cut along the grain in slices to ensure weighted sampling across years, thus representing a true weighted-average across each increment.

Approximately 10 mg of wood from the two to three-year increments for the 1980–2013 period was subsampled and used for δ15N and %N analysis. The midpoint of the aggregated range (e.g., 2003) was identified, and the preceding even year increment assigned. Results falling within 1986 to 2010 period were reported. Following the exact procedures described in Sabo et al (2016b), prepared samples (n = 1322) were analyzed for δ15N using a Carlo Erba NC2500 elemental analyzer (CE Instruments, Milano, Italy) interfaced with a Thermo Delta V+ isotope ratio mass spectrometer (Bremen, Germany). Following combustion in an elemental analyzer, a Carbosorb trap was used to remove CO2 and a magnesium perchlorate trap was used to remove water vapor. The δ15N data were normalized to the air reference standard scale using a two-point normalization curve with internal standards calibrated against USGS40 and USGS41 [Qi et al 2003, Brand et al 2014]. The analytical precision among runs (1σ) of an internal wood standard was 0.3‰. Catchment-scale tree ring δ15N values were estimated by taking the arithmetic mean of all tree ring δ15N values in a given time period for all trees within a watershed [Sabo et al 2016a].

Atmospheric deposition, water quality, disturbance, and climatic data

Stream chemistry and discharge have been monitored at the outlet of UBR and PR since 1990 and 1992, respectively [Eshleman et al 1998, Scanlon et al 2010, Eshleman et al 2013]. Synoptic spring baseflow sampling in each of the five PR subcatchments was conducted from 1992–1994 period and, after a long cessation, was restarted in 2007 and repeated annually through 2016. All water samples were analyzed for nitrate concentrations using ion chromatography. The spatial relationship between mean catchment-scale δ15N and mean spring baseflow nitrate concentrations at PR for the period of observation was explored using linear regression. At UBR, mean annual flow-weighted concentrations, estimated using a multi-parameter loading model [Eshleman et al 2013], were compared against one year lagged catchment-scale δ15N values at the UBR site using simple linear regression. The one year lagged values were used because there was an assumed lag between N available in a given growing season and subsequent flush in the following dormant season. Likewise, linear regression was used to test the relationship between spring baseflow nitrate concentrations and one year lagged catchment-scale δ15N values at the PR headwater catchments. The relationship between mean April nitrate concentrations at UBR catchment-scale δ15N was also evaluated. Linear interpolation between catchment-scale δ15N time periods (e.g., between 1992 and 1994) was used to provide an annual record of catchment-scale δ15N values that is directly comparable to stream records since tree rings were analyzed in 2–3 year increments. For the Paine Run sites, this allowed three additional spring baseflow concentrations to be compared against catchment-scale δ15N (i.e., 1993, 2007, and 2009) and 10 additional mean April and mean annual flow-weighted concentration values at Upper Big Run, thus allowing the complete variability of observed nitrate concentrations to be considered. The difference in the number of spring samples between UBR and PR was due to the hiatus in sampling at PR. The imputation procedure described above may result in the underestimation of the coefficient standard error by increasing the degrees of freedom, so we carried out a complimentary comparison of non- interpolated one year lagged catchment-scale δ15N values and stream nitrate concentrations to further test the robustness of the relationship.

Disturbance within 45 m of the center of individual plots was described using the Disturbance Index (DI, [Healey et al 2005] applied to Landsat 5, 7 and 8 (TM/ETM+/OLI) data. As such, disturbance was quantified in 8 to 12% of the area in each catchment. All available tier I surface reflectance and quality assurance quality control (QAQC) data for Landsat 5, 7, and 8 were extracted for each cored plot and downloaded using the Google Earth Engine. Using the provided QAQC information, surface reflectance observations collected under suboptimal conditions (cloudy, cloud shadow, snow and ice, etc) were removed from the data set. From the data that passed quality control, the six Landsat multispectral bands were reduced to three orthogonal indices of brightness (B), greenness (G), and wetness (W) through the tasseled cap transformation [DeVries et al 2016]. The Disturbance Index (DI) is a simple linear combination of these indices (DI = B-(G + W)), where greener and wetter pixels indicate less disturbance and brighter, dryer, and browner pixels indicate greater disturbance. A continuous stable forest period (~5 to 15 years) for each individual plot was identified by referencing local logging maps provided by the Maryland State Forest Service and the North American Forest Dynamic data product, ‘Forest Disturbance History from Landsat, 1986–2010’ (supplemental figure 1). To standardize for sun-canopy-sensor geometry, we organized all ‘stable forest’ DI observations by day of year and used locally weighted regression models (i.e., LOESS) to model the average DI phenology. From the LOESS fit, expected DI for each day of year was estimated that account for canopy phenology and sun angle effects that reoccur each growing season. The difference between the observed and modeled DI values (ΔDI) was calculated for all observations, representing disturbance above (positive values) or below (negative values) the mean DI during the stable period for any given day of year. A mean growing season (May 1st to September 30th) ΔDI value was calculated using all Landsat observations that fell within the years corresponding to the cut tree ring segments. Similar approaches using various empirical models to describe Landsat phenology in forests have been applied elsewhere (e.g., [Zhu et al 2012, Elmore et al 2016] with the overall goal being to normalize for intra-annual variation so that inter-annual variation can be quantified. It should be clarified that mean catchment ΔDI was not quantified at the catchment scale (i.e., sampling all pixels within a watershed) and related to stream nitrate concentrations. This relationship has already been demonstrated in multiple studies [Townsend et al 2004, McNeil et al 2007, Eshleman et al 2009, Townsend et al 2012]. The primary motivation for this analysis was to ascertain the effect of ΔDI on wood δ15N.

Annual wet deposition of inorganic N (IN) and sulfur (S) along with annual temperature and precipitation for UBR and the PR sub-catchments were extracted for the locations of individual trees. Data sets used included wet deposition annual gradient maps published by the National Atmospheric Deposition Program for the 1980–2015 period and climate data published by the PRISM Climate group [Latysh and Wetherbee 2012]. Annual temperature values were determined by averaging extracted monthly temperatures from PRISM climatic maps. Little to any intra-site variation at UBR and PR in the climatic and deposition variables was observed due to the coarse spatial resolution (~4 km), so only the site averages were reported. Annual values of temperature, precipitation, and wet deposition of N and S deposition corresponding with the cut tree ring segments were averaged by taking the simple arithmetic mean and used in later statistical analyses to predict the inter-annual variation in tree-ring δ15N.

Annual inorganic S and N wet deposition, ΔDI, mean annual temperature, and mean annual precipitation were used as predictor variables to explain the inter-annual variation in tree ring δ15N. To avoid multicollinearity and to gain greater confidence in causal relationships, the linear effect of time was assessed first, and then removed from both predictor and response variables associated with each individual tree. Thus, the residuals from the suite of predictor variables were leveraged to explain detrended tree-ring δ15N residuals using multiple linear regression in SigmaPlot 14.0. Multicollinearity among the predictor variables was assessed using variance inflation factor (VIF, [Graham 2003]).

Results

No long-term trends in annual precipitation or temperature were observed at our study catchments over the 25- year analysis period (figure 2(A)). LOESS curves of the annual change in disturbance index (ΔDI) of individual tree time series at UBR and PR suggested a period of disturbance from the mid-1980s to early 1990s for the population of trees (figure 2(B), supplemental figures 2–7). ΔDI generally declined after the late 1980s and early 1990s at all PR headwater catchments, which is consistent with remote sensing based disturbance classification products (using different methodologies) detecting disturbance in the late 1980s and early 1990s (figure 2(B), supplemental figure 1). ΔDI at UBR declined throughout the 1990s following logging then increased after 2000, coincident with another round of logging activities and reported incidents of gypsy moth defoliations (figure 2(B), supplemental figure 1; [Townsend et al 2012]). S and N deposition declined nearly 50% throughout the period of record at both sites (figure 2(C)). During this period, catchment-scale δ15N significantly decreased (p < 0.01) at all headwater catchments at PR and UBR throughout the period of record except for PR4000 (figure 3, supplemental table 1). Consistent with these aggregated values, 80 of the 96 individual trees showed linear declines in tree-ring δ15N of which twenty were significant (p < 0.05, supplemental tables 4, 5). Only 11 of the remaining 16 trees with positive linear increases in tree ring δ15N were significant (p < 0.05).

Figure 2.

Figure 2.

Open in a new tab

Time series of (A) annual temperature, precipitation, (B) mean disturbance index (ΔDI) for the population of sampled trees in individual catchments, and (C) annual S and N deposition at Paine Run (PR) and Upper Big Run (UBR). For (B) only, the LOESS functions fit to all ΔDI observations that corresponded to individual tree ring segments within a catchment are shown (raw ΔDI data illustrated in supplemental figures 2–7).

Figure 3.

Figure 3.

Open in a new tab

Time series of catchment-scale δ15Nfor Upper Big Run (UBR) and the five Paine Run (PR) sub-catchments and associated regression lines. Simple linear regression results indicated significant declines in catchment-scale δ15N for all catchments except for PR4000(UBR: y = −0.017 + 33.86, R2 = 0.65, p < 0.001; PR1000: y = −0.056 + 107.86, R2 = 0.68, p < 0.001; PR2000: y = −0.055 + 107.57, R2 = 0.70, p < 0.001; PR3000: y = −0.032 + 60.48, R2 = 0.47, p=0.003; PR4000: y = 0.013x−26.65, R2 = 0.06, p = 0.35; PR5000: y = −0.023 + 43.81, R2 = 0.60, p < 0.001). Raw tree-ring δ15 N time series illustrated in the supplemental (supplemental figures 2–7).

Across basins and over time at individual basins, catchment mean wood δ15N values were observed to have positive linear relationships with stream nitrate concentrations. Across 5 sub-basins at PR, mean spring baseflow nitrate concentrations (averaged across all years) increased linearly with increasing catchment mean wood δ15N (R2 = 0.86; P = 0.023, table 1). The slope of this relationship equates to a gradient in spring base flow nitrate concentration from 0.09 to 0.73 mg N l−1 being associated with a gradient in catchment mean wood δ15N from −3.6 to −1.5‰ (figure 4(A)). The portion of variance in nitrate concentration explained by the spatial gradient in mean wood δ15N observed at PR was removed by taking the difference between the mean predicted stream nitrate concentration from the observed nitrate concentration value in a given year. Temporal variation in wood δ15N was a significant model effect on nitrate concentrations (figure 4(B)), with a positive effect of catchment wood δ15N observed through time at four of the five Paine Run catchments (table 2). The results from this analysis were consistent with the coefficient estimates and standard error using the non-interpolated catchment- scale δ15N data (table 2). While the variance in stream nitrate concentrations explained was generally similar, the relationships were only significant in two of the four Paine Run catchments using the non-interpolated catchment-scale δ15N time series (table 2). Estimates of slope coefficients generally increased with increasing mean wood δ15N (p = 0.17, figure 4(C)). This positive correlation suggests that stream nitrate concentrations were least sensitive to annual variation in wood δ15N at sites with low wood δ15N (e.g., PR1000) and were most sensitive at sites with high wood δ15N at Paine Run (e.g., PR5000, figure 4(C)). Like the findings from PR, mean annual flow-weighted nitrate and spring baseflow nitrate concentrations at UBR were positively correlated with catchment wood δ15N (R2 = 0.88; P < 0.0001 and R2 = 0.43; P = 0.0017, respectively figure 4(B)); and these relationships were robust when using the non-interpolated catchment-scale δ15N time series with equivalent, significant slope estimates (table 2).

Table 1.

Results of the linear regression analysis of mean catchment-scale wood δ15N as a predictor of mean spring baseflow nitrate concentrations over the period of record among the five Paine Run (PR) headwater catchments.

PR spatial model Coefficients Standard error P-value R2
Intercept 1.29 0.21 <0.01 0.86
Slope 0.35 0.08 0.02
Open in a new tab

Figure 4.

Figure 4.

Open in a new tab

(A) Relationship between mean catchment-scale δ15N and spring baseflow nitrate concentrations across space over the period of record among the Paine Run (PR) and Upper Big Run (UBG) headwater catchments (y = 0.35x + 1.29, R2 = 0.86, p = 0.023 for data from PR). The error bars illustrate the standard error of the mean. (B) Relationships between 1-year lagged catchment scale δ15N and nitrate concentrations at individual PR headwater catchments and UBR; all regressions were significant except for PR4000 (UBR FWC: y = 0.79x + 0.62, R2 = 0.85, p < 0.01; UBR SBC (mean April):y = 0.48x + 0.47, R2 = 0.40, p < 0.01; PR1000: y = 0.14x + 0.55, R2 = 0.67, p < 0.01; PR2000: y = 0.25x + 0.53, R2 = 0.67, p < 0.01; PR3000: y = 0.40x + 1.05, R2 = 0.45, p = 0.04; PR4000: y = 0.69x + 0.94, R2 = 0.14, p < 0.32; PR5000: y = 2.11x + 4.23, R2 = 0.91, p < 0.01). (C) The linear relationship between mean catchment-scale δ15N and slope estimates based on the relationship between 1-year lagged catchment scale δ15N and nitrate concentrations for the PR catchments (y = 0.74x + 2.48, R2 = 0.19, p = 0.17). The data point from UBR was not included in the regression.

Table 2.

Results of the linear regression analyses modeling the effect of 1-year lagged catchment scale wood δ15N on (1) observed mean annual flow-weighted and mean April nitrate concentrations at Upper Big Run (UBR), (2) spring baseflow nitrate concentrations at individual Paine Run (PR) headwater catchments after factoring out the influence of the spatial gradient and using the interpolated and non-interpolated wood δ15N time series.

Coefficients Standard Error P-value R2
UBR Mean Annual Flow-Weighted Concentration Interpolated Non-Interpolated Interpolated Non-Interpolated Interpolated Non-Interpolated Interpolated Non-Interpolated
Intercept 0.62 0.613 0.03 0.0295 <0.01 <0.001 0.85 0.88
Slope 0.79 0.793 0.07 0.0843 <0.01 <0.0001
UBR Mean April Concentration
Intercept 0.47 0.48 0.05 0.0593 <0.01 <0.001 0.40 0.521
Slope 0.48 0.559 0.13 0.17 <0.01 0.008
PR1000
Intercept		 0.55 0.378 0.14 0.066 <0.01 0.011 0.67 0.843
Slope 0.14 0.0859 0.04 0.0181 <0.01 0.018
PR2000
Intercept 0.53 0.842 0.18 0.271 0.03 0.053 0.67 0.473
Slope 0.25 0.219 0.07 0.102 <0.01 0.122
PR3000
Intercept 1.05 1.507 0.46 0.515 0.06 0.061 0.45 0.548
Slope 0.40 0.446 0.17 0.184 0.04 0.094
PR4000
Intercept 0.94 1.984 1.01 1.76 0.38 0.341 0.14 0
Slope 0.69 0.855 0.64 1.17 0.32 0.518
PR5000
Intercept 4.23 5.014 0.50 0.931 <0.01 0.013 0.91 0.832
Slope 2.11 2.164 0.25 0.474 <0.01 0.02
Open in a new tab

Using the simple linear regression models (table 2), catchment-scale δ15N was used to model the temporal variability in spring baseflow and mean annual flow-weighted nitrate concentrations from 1990 to 2010 for UBR and 1992 to 2010 for PR. The 1-year lagged catchment-scale δ15N regression models were effective in capturing peak nitrate concentrations in the early 1990s followed by a decline to the 2000s in five of the six catchments (figure 5). Furthermore, the models were successful in generating the generally stable stream nitrate concentrations at UBR during the 2000s. Predicted nitrate concentration time series were generally smoother than the observed spring baseflow nitrate concentration time series (one sample per year), likely reflecting the interpolation of wood δ15N values between the measured 2 to 3-year increments, wood δ15N values being minimally affected by annual discharge, and the fact the N has the potential to translocate between adjacent tree rings.

Figure 5.

Figure 5.

Open in a new tab

Predicted and observed mean annual flow-weighted nitrate concentrations at Upper Big Run (UBR) (panel A) along with predicted and observed spring baseflow nitrate concentrations for the five Paine Run (PR) headwater catchments and UBR (panel B-G).

After removing the linear effect of time, all predictor variables except for temperature and precipitation were significant in explaining the inter-annual variation in residual tree ring δ15N (table 3). Residual tree-ring δ15N decreased with increased S and N deposition (p < 0.001) and ΔDI residuals (p = 0.016). As such, higher rates of disturbance and acidic deposition decreased tree-ring δ15N. Among the significant predictor variables and after accounting for trends over time, S and N deposition was the most influential with the highest sum of squares, whereas ΔDI had the lowest sum of squares (table 3). The multiple linear regression model, though significant (p < 0.001), explained very little of the inter-annual variation of detrended tree-ring δ15N (R2 = 0.041) which could be partly due to the analytical precision of tree-ring δ15N. If using non-detrended data in a generalized linear model with a categorical variable of a tree identification code (i.e., the factor), the suite of predictor variables plus time explained upwards of 90% of the variance with similar slope coefficient estimates for S and N deposition as predicted by the multiple linear regression model (supplemental table 3), but multicollinearity amongst the predictor variables was identified (VIF > 14 for year and S and N deposition).

Table 3.

Results of the multiple linear regression analysis assessing the effects of detrended annual S and N deposition, disturbance (ΔDI), annual precipitation, and annual temperature on detrended tree ring δ15N (n = 973).

COEFFICIENT STANDARD ERROR SUM OF SQUARES P STANDARDIZED COEFFICIENT VIF
INTERCEPT 0.000 236 0.0133 0.986
S AND N DEPOSITION RESIDUAL −0.001 33 0.000 220 5.920 <0.001 −0.19 1.027
ΔDI RESIDUAL −0.994 0.405 1.170 0.014 −0.0793 1.051
PRECIPITATION RESIDUAL 0.001 38 0.001 05 0.231 0.189 0.045 1.158
TEMPERATURE RESIDUAL 0.0245 0.0349 0.0853 0.482 0.0237 1.147
REGRESSION p < 0.001 R2 = 0.043
Open in a new tab

Discussion

Our results strongly show that spatiotemporal variability in stream nitrate concentrations is correlated with a widely-used proxy for terrestrial N availability: wood δ15N. Positive relationships between catchment-scale δ15N and stream nitrate concentrations are consistent with the understanding that wood, foliar, and soil δ15N records are indicators of terrestrial N availability [Craine et al 2009, Gerhart and McLauchlan 2014]. Across space at the PR headwater catchments, those catchments with higher mean wood δ15N exhibited higher stream nitrate concentrations. This result suggests that catchments with more positive wood δ15N have higher terrestrial N availability (i.e., N supply greater than plant demand) and greater nitrate export. A similar observation was made in a watershed study comparing mean annual flow-weighted nitrate concentrations and non-normalized catchment-scale δ15N values in the Adirondacks, but that result was not considered conclusive since one of the predominant species in that study was not sampled because it was diseased [Sabo et al 2016a]. Moving beyond and consistent with dendroisotopic records, forested catchments with corresponding lower foliar and soil C:N ratios also typically have higher stream nitrate concentrations throughout the Northeast [Aber et al 1998, Aber et al 1989]. Further expanding this spatial analysis to include other catchments with stream nitrate concentrations and dendroisotopic records is needed to see if broader positive spatial relationships between stream and terrestrial N availability emerge despite differences in the underlying geology, forest management strategies, species composition, successional status, and topography among forest (akin to global foliar [N] and δ15N gradient relationships, Craine et al 2009).

Over time, periods with higher wood δ15N were associated with periods of higher stream nitrate concentrations regardless of the stream metric used (spring baseflow, annual flow-weighted, or mean April concentration). Catchment-scale δ15N regression models also successfully described the decadal stability of mean annual flow-weighted nitrate concentrations at UBR and were generally predictive of spring baseflow nitrate concentrations in five of the six catchments. These results indicate that a variety of nitrate concentration records may be used to compare against future records of wood δ15N. Similar to our results, a recent study found that wood δ15N records from a single species, tulip poplar (Liriodendron tulipifera), could explain inter-annual variation of stream nitrate concentrations at a catchment in Fernow Experimental Forest [Burnham et al 2016]. Other observational studies have suggested similar relationships through time [McLauchlan et al 2007, Sabo et al 2016a]. In combination, our results and previous studies suggest that wood δ15N models can be used to fill in temporal gaps in, and potentially extend, water quality records back in time. Furthermore, the positive correlation between wood δ15N and stream nitrate concentrations, a widely held index of terrestrial N availability [Vitousek and Melillo 1979, Aber et al 1989], provide useful information in the ongoing debate about the usefulness and interpretability of δ15N signatures in plant tissue [Craine et al 2018, Craine et al 2019, Hiltbrunner et al 2019]. While some argue that the δ15N signatures in plant tissue cannot be easily leveraged as a proxy for terrestrial N availability relative to plant demand due to the complexities associated with accounting for a plant’s unique physiological aspects interacting the with the source, transport, and isotopic fractionation of nitrogen [Tomlinson et al 2016, Burnham et al 2019, Hiltbrunner et al 2019], the positive relationship between stream nitrate concentrations and wood δ15N provide robust evidence indeed that δ15N signatures in plant tissue record terrestrial N availability [Craine et al 2009, Elmore et al 2016, Craine et al 2018].

At our sites, a positive correlation between the sensitivity of stream nitrate concentrations to wood δ15N was observed. Lower significant slope coefficient estimates (slope < 0.4) corresponded to catchments with lower mean catchment-scale δ15N (< −2.5‰) and nitrate concentrations over the period of record, suggesting the sensitivity of stream nitrate concentrations to changes in catchment-scale δ15N through time is partly a function of the amount of N available relative to plant demand at a site. This speculation is consistent with evidence that forested catchments that retain minimal amounts of atmospherically deposited N (i.e., the most N rich sites) have the greatest absolute reduction in stream nitrate yields and flow-weighted concentrations following declines in atmospheric N deposition [Eshleman et al 2013]. As such, forested areas with higher N availability relative to plant demand will demonstrate a greater per unit decline in nitrate concentrations as wood δ15N declines, evaluated on an annual basis. Whether this linear relationship holds outside the range of δ15N values observed at these catchments (−2.5 to 0‰), reaches an inflection point (i.e., a phase shift), or is a non-linear continuous response should be pursued in future research [Aber et al 1989, Lovett and Goodale 2011]. Further investigation into the nature of the terrestrial availability and nitrate concentration relationship is important to determine the trajectory of forest recovery from historically high rates of atmospheric N and S deposition.

Throughout the period of record, catchment-scale δ15N suggests that terrestrial N availability relative to plant demand generally declined through time, but trends at the scale of an individual tree were more variable. These clear declines in terrestrial N availability at the scale of the catchment and associated variability in individual tree-ring δ15N trends are consistent with other dendroisotopic studies reporting over a similar period throughout the eastern United States [Burnham et al 2016, Elmore et al 2016, Sabo et al 2016a, Mathias and Thomas 2018]. While species specific sensitivities to atmospheric N and S deposition [Horn et al 2018], soil acidification status [Sabo et al 2016a], and disturbance [Howard and McLauchlan 2015] may modify the trajectory of N availability, it is clear that N availability for forests in general has declined. This decline in terrestrial N availability, described by some as ‘oligotrophication’ [Elmore et al 2016, McLauchlan et al 2017, Groffman et al 2018], may contribute to future declines in forest productivity, especially as atmospheric N deposition continues to decline and CO2 concentrations continue to rise [Eshleman et al 2013, Lloret and Valiela 2016, Wang et al 2017, Craine et al 2018]. Although it is uncertain what the ultimate impacts of declining N availability will be on the terrestrial compartment, it is clear that less terrestrial N is reaching the already oligotrophic streams of Upper Big Run and Paine Run [Dodds et al 1998].

Our water quality analysis demonstrated that stream nitrate concentrations are largely driven by changes in terrestrial N availability, but what are the drivers of longer-term trends and the general inter-annual variation in terrestrial N availability? Recent studies have applied a suite of statistical approaches to explain δ15N through time [Elmore et al 2016, Mathias and Thomas 2018], but collinearity issues among potential predictor variables and time make it difficult to statistically attribute the influence of various environmental drivers on wood δ15N. We applied a conservative approach in which the linear correlation with time was removed from both the response and predictor variables to (1) better identify causal relationships between predictor and response variables by comparing unique information in the time series and (2) gain greater confidence in the sign and error of coefficient estimates by eliminating multicollinearity. Our resulting regression model was significant, but only explained a small portion (4%) of the detrended residual δ15N of the 96 trees. This low explanatory power speaks to the complexity of statistically modeling individual tree-ring δ15N responses using spatially coarse environmental datasets, particularly when trends over time are evaluated separately and removed from the time series. Furthermore, our decision to sample wood in 2–3 year segments rather than annual increments reduced our ability to explain variation at higher temporal frequencies. Changes in the δ15N signature of atmospheric N deposition was also not considered in this analysis since such data are not available. Some have argued that the impacts of the δ15N signature in atmospheric N deposition have minimal influence on plant tissue δ15N [McLauchlan et al 2017, Craine et al 2018], although δ15N of deposition is potentially influential in certain locales with strong emission sources [Saurer et al 2004]. Leveraging sub-annual datasets like identifying maximum disturbance in a given year [Townsend et al 2004] or spring timing [Elmore et al 2016] or collapsing the variance of the detrended residual δ15N into a catchment-scale δ15N residual, might improve the explanatory power, but generally would require data with higher temporal frequency or spatial density.

Despite the aforementioned limitations, the significant coefficient estimates offer some insights into forest- wide drivers of terrestrial N availability and parallel observations from other studies using different methods to track changes in terrestrial N and carbon cycling After removing the long-term trend of declining S and N deposition and declining wood δ15N, S and N deposition exhibited a negative relationship with wood δ15N (table 3), suggesting that lower rates of acidic deposition are associated with enhanced N availability on inter- annual scales. It has been observed in the lab and in the field that decreased soil pH and aluminum mobilization can lead to increased retention of soil organic matter and shifts in microbial activities, but increases in pH can lead to increased loss of organic matter and nitrogen [Sinsabaugh et al 2004, Sinsabaugh 2010, Oulehle et al 2011, Oulehle et al 2017, Lawrence et al 2019]. The predominantly deciduous forests of UBR and PR lie on poorly buffered, infertile soils and are likely experiencing deacidification as the soils regain base cations and limit toxic aluminum mobilization after decades of elevated rates of acidic deposition [Lawrence et al 2015, Kline et al 2016, Groffman et al 2018]. Studies that have observed deacidification suggest that organic pools of carbon and N are now being mineralized and lost to streams at both experimentally manipulated and reference catchments [Oulehle et al 2011, Johnson et al 2014, Rosi-Marshall et al 2016, Oulehle et al 2017, Lawrence et al 2019]. As soils continue to recover from past acidic deposition, increased ammonification/nitrification and subsequent denitrification, an emission pathway that greatly enriches soil 15N [Kendall and McDonnell 2012], may be a positive offset to the decline in wood δ15N and terrestrial N availability. Ultimately, the magnitude and duration of this release of legacy N and its impact on terrestrial N availability and stream nitrate loss is likely conditional on the proportion of unprocessed atmospheric nitrate reaching streams [Sabo et al 2016b] and forest growth responses to increased base cation availability [Battles et al 2013], longer growing seasons [Elmore et al 2016], increased CO2 [Norby et al 2016], and improved air quality [Mathias and Thomas 2018].

Wood δ15N also declined in response to disturbance, suggesting that ephemeral disturbances [McNeil et al 2007], consisting mainly of defoliation events and selective silvicultural activities, act to reduce N availability in these systems over the long-term. While it is apparent that short-lived pulses of nitrate loss to streams are often observed immediately after a disturbance [Eshleman 2000], the loss of this nitrate via increased denitrification may be insufficient to increase the isotopic signature of plant available N pools in the soils and it is unclear if this pulse of N is even available to defoliated plants. Furthermore, the enhancement of a long-term growth sink ultimately reduces N availability in the ecosystem. This finding is consistent with a study carried out in UBR and the surrounding Savage River State Forest which reported that the N content of leaves was lower at sites that had experienced disturbance relative to undisturbed sites [McNeil et al 2007]. Similar conclusions have also been reported based on soil δ15N data from forests of the Pacific Northwest [Perakis et al 2015]. An alternative, though less likely, explanation to the effects of disturbance of wood δ15N and terrestrial N availability described above is that increased nitrification rates spurred by decreased vegetative uptake may temporarily reduce δ15N signature of nitrate in the soil. This decline in the isotopic signature of plant available nitrogen in the soil is a transient situation where the system can have a simultaneous increase in N availability yet have a reduction in the δ15N signature of plant available nitrate. The interplay of short and long-term impacts of disturbance on δ15N in soils and plants and more importantly on terrestrial N availability in ecosystems should be further explored.

Declines in terrestrial N availability over the past 30 years drove reductions in stream nitrate concentration and loss at our study sites, consistent with reported declines in nitrate export from forested catchments throughout the mid-Atlantic and the Northeast over a similar time period [Kothawala et al 2011, Eshleman et al 2013]. The impacts of decreased nitrate loss to streams may help diminish the occurrence of episodic acidification in headwater systems and chronic eutrophication in downstream rivers, lakes, and estuaries—welcomed improvements to these chronically impaired aquatic ecosystems. Overall, it is likely that forest N availability has and will continue to impact downstream N availability, and that these trends can be monitored through dendroecological evaluation of wood δ15N.

Supplementary Material

Sup. Material

Acknowledgments

We thank Robin Paulman, Joshua Tabora, Susan Snow, Ian Cheek, Vanessa Cunningham, Stephanie Siemek, Ian Smith, and Michael Furlong for assistance in the field and with stable isotope analyses. I also thank Jim Garlitz and Katie Kline along with other personnel of the Appalachian Laboratory water chemistry lab for carrying out water quality analyses. Many thanks also to Dr Eric Davidson who provided valuable guidance in this research effort of my dissertation [Sabo 2018]. We thank Drs J Renee Brooks and Stephen LeDuc for helpful reviews on previous manuscripts. This research was made possible through the support of an Environmental Protection Agency, Science to Achieve Results Graduate Fellowship (#FP-91749901-0, to RDS), National Geographic Young Explorer’s Grant (to RDS), and a teaching assistantship through the Marine-Estuarine- Environmental Sciences (MEES) Graduate Program (to RDS). This research was also supported in part by an appointment to the Research Participation Program for the US Environmental Protection Agency, Office of Research and Development, administered by the Oak Ridge Institute for Science and Education through an inter-agency agreement between the US Department of Energy and EPA. The views presented here are those of the authors and do not represent official views or policy of the US Environmental Protection Agency (EPA) or any other US federal agency. Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the US Government.

Footnotes

Disclaimer: The views expressed in this article are those of the authors and do not necessarily represent the views or policies of the US Environmental Protection Agency.

References

  1. Aber J, McDowell W, Nadelhoffer K, Magill A, Berntson G, Kamakea M, McNulty S, Currie W, Rustad L and Fernandez I 1998. Nitrogen saturation in temperate forest ecosystems BioScience 48 921–34 [Google Scholar]
  2. Aber JD, Nadelhoffer KJ, Steudler P and Melillo JM 1989. Nitrogen saturation in northern forest ecosystems BioScience 39 378–286 [Google Scholar]
  3. Argerich A, Johnson S, Sebestyen S, Rhoades C, Greathouse E, Knoepp J, Adams M, Likens G, Campbell J and McDowell W 2013. Trends in stream nitrogen concentrations for forested reference catchments across the USA Environ. Res. Lett 8 014039 [Google Scholar]
  4. Battles JJ, Fahey TJ, Driscoll CT Jr, Blum J D and Johnson C E 2013. Restoring soil calcium reverses forest decline Environmental Science & Technology Letters 1 15–9 [Google Scholar]
  5. Brand WA, Coplen TB, Vogl J, Rosner M and Prohaska T 2014. Assessment of international reference materials for isotope-ratio analysis (IUPAC Technical Report) Pure Appl. Chem 86 425–67 [Google Scholar]
  6. Brookshire E, Gerber S, Webster JR, Vose JM and Swank WT 2011. Direct effects of temperature on forest nitrogen cycling revealed through analysis of long-term watershed records Global Change Biol. 17 297–308 [Google Scholar]
  7. Burnham MB, Adams MB and Peterjohn WT 2019. Assessing tree ring δ15N of four temperate deciduous species as an indicator of N availability using independent long-term records at the Fernow experimental forest, WV Oecologia 191 971–81 [DOI] [PubMed] [Google Scholar]
  8. Burnham MB, McNeil BE, Adams MB and Peterjohn WT 2016. The response of tree ring δ15N to whole-watershed urea fertilization at the Fernow experimental forest, WV Biogeochemistry 130 133–45 [Google Scholar]
  9. Craine JM, Elmore AJ, Aidar MP, Bustamante M, Dawson TE, Hobbie EA, Kahmen A, Mack MC, McLauchlan KK and Michelsen A 2009. Global patterns of foliar nitrogen isotopes and their relationships with climate, mycorrhizal fungi, foliar nutrient concentrations, and nitrogen availability New Phytol. 183 980–92 [DOI] [PubMed] [Google Scholar]
  10. Craine JM, Elmore AJ, Wang L, Aranibar J, Bauters M, Boeckx P, Crowley BE, Dawes MA, Delzon S and Fajardo A 2018. Isotopic evidence for oligotrophication of terrestrial ecosystems Nature Ecology & Evolution 2 1735. [DOI] [PubMed] [Google Scholar]
  11. Craine JM, Elmore AJ, Wang L, Boeckx P, Delzon S, Fang Y, Gray A, Guerrieri R, Gundale MJ and Hietz P 2019. Reply to: data do not support large-scale oligotrophication of terrestrial ecosystems Nature Ecology & Evolution 3 1287–8 [DOI] [PubMed] [Google Scholar]
  12. DeVries B, Pratihast AK, Verbesselt J, Kooistra L and Herold M 2016. Characterizing forest change using community-based monitoring data and Landsat time series PLoS One 11 e0147121 [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Dodds WK, Jones JR and Welch EB 1998. Suggested classification of stream trophic state: distributions of temperate stream types by chlorophyll, total nitrogen, and phosphorus Water Res. 32 1455–62 [Google Scholar]
  14. Durán J, Morse JL, Groffman PM, Campbell JL, Christenson LM, Driscoll CT, Fahey TJ, Fisk MC, Likens GE and Melillo JM 2016. Climate change decreases nitrogen pools and mineralization rates in northern hardwood forests Ecosphere 7 e01251 [Google Scholar]
  15. Elmore AJ, Nelson DM and Craine JM 2016. Earlier springs are causing reduced nitrogen availability in North American eastern deciduous forests Nature Plants 2 16133 [DOI] [PubMed] [Google Scholar]
  16. Eshleman K N 2000. A linear model of the effects of disturbance on dissolved nitrogen leakage from forested watersheds Water Resour. Res 36 3325–35 [Google Scholar]
  17. Eshleman KN, McNeil BE and Townsend PA 2009. Validation of a remote sensing based index of forest disturbance using streamwater nitrogen data Ecol. Indic 9 476–84 [Google Scholar]
  18. Eshleman KN, Morgan RP, Webb JR, Deviney FA and Galloway JN 1998. Temporal patterns of nitrogen leakage from mid-Appalachian forested watersheds: role of insect defoliation Water Resour. Res 34 2005–16 [Google Scholar]
  19. Eshleman KN, Sabo RD and Kline KM 2013. Surface water quality is improving due to declining atmospheric N deposition Environmental Science & Technology 47 12193–200 [DOI] [PubMed] [Google Scholar]
  20. Galloway JN, Aber JD, Erisman JW, Seitzinger SP, Howarth RW, Cowling EB and Cosby BJ 2003. The nitrogen cascade AIBS Bulletin 53 341–56 [Google Scholar]
  21. Gap U 2011. U.S. Geological Survey Gap Analysis Program, 20160513, GAP/LANDFIRE National Terrestrial Ecosystems 2011: U.S. Geological Survey; 10.5066/F7ZS2TM0 [DOI] [Google Scholar]
  22. Gerhart LM and McLauchlan KK 2014. Reconstructing terrestrial nutrient cycling using stable nitrogen isotopes in wood Biogeochemistry 120 1–21 [Google Scholar]
  23. Graham MH 2003. Confronting multicollinearity in ecological multiple regression Ecology 84 2809–15 [Google Scholar]
  24. Groffman PM, Driscoll CT, Durán J, Campbell JL, Christenson LM, Fahey TJ, Fisk MC, Fuss C, Likens GE and Lovett G 2018. Nitrogen oligotrophication in northern hardwood forests Biogeochemistry 141 523–39 [Google Scholar]
  25. Healey SP, Cohen WB, Zhiqiang Y and Krankina ON 2005. Comparison of tasseled cap-based landsat data structures for use in forest disturbance detection Remote Sens. Environ 97 301–10 [Google Scholar]
  26. Hiltbrunner E, Körner C, Meier R, Braun S and Kahmen A 2019. Data do not support large-scale oligotrophication of terrestrial ecosystems Nature Ecology & Evolution 3 1285–6 [DOI] [PubMed] [Google Scholar]
  27. Horn KJ, Thomas RQ, Clark CM and Pardo L 2018. Growth and survival relationships of 71 tree species with nitrogen and sulfur deposition across the conterminous US PLoS One 14 e0212984 [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Howard I and McLauchlan KK 2015. Spatiotemporal analysis of nitrogen cycling in a mixed coniferous forest of the northern United States Biogeosciences 12 3941–52 [Google Scholar]
  29. Johnson CE, Driscoll CT, Blum JD, Fahey TJ and Battles JJ 2014. Soil chemical dynamics after calcium silicate addition to a northern hardwood forest Soil Sci. Soc. Am. J 78 1458–68 [Google Scholar]
  30. Kendall C and McDonnell JJ 2012. Isotope Tracers in Catchment Hydrology (Amsterdam: Elsevier) https://www.elsevier.com/books/isotope-tracers-in-catchment-hydrology/kendall/978-0-444-81546-0
  31. Kline KM, Eshleman KN, Garlitz JE and U’Ren SH 2016. Long-term response of surface water acid neutralizing capacity in a central Appalachian (USA) river basin to declining acid deposition Atmos. Environ 146 195–205 [Google Scholar]
  32. Kothawala DN, Watmough SA, Futter MN, Zhang L and Dillon PJ 2011. Stream nitrate responds rapidly to decreasing nitrate deposition Ecosystems 14 274–86 [Google Scholar]
  33. Larsson LA 2009. CDENDRO: CooRecorder, edited by CDendro, http://cybis.se/forfun/dendro/
  34. Latysh NE and Wetherbee GA 2012. Improved mapping of national atmospheric deposition program wet-deposition in complex terrain using PRISM-gridded data sets Environ. Monit. Assess 184 913–28 [DOI] [PubMed] [Google Scholar]
  35. Lawrence GB, Hazlett PW, Fernandez IJ, Ouimet R, Bailey SW, Shortle WC, Smith KT and Antidormi MR 2015. Declining acidic deposition begins reversal of forest-soil acidification in the northeastern US and eastern Canada Environmental Science & Technology 49 13103–11 [DOI] [PubMed] [Google Scholar]
  36. Lawrence GB, Scanga SE and Sabo RD 2020. Recovery of soils from acidic deposition may exacerbate nitrogen export from forested watersheds Journal of Geophysical Research: Biogeosciences 125 e2019JG005036 [Google Scholar]
  37. LeDuc SD, Rothstein DE, Yermakov Z and Spaulding SE 2013. Jack pine foliar δ15N indicates shifts in plant nitrogen acquisition after severe wildfire and through forest stand development Plant and Soil 373 955–65 [Google Scholar]
  38. Likens GE 2017 Fifty years of continuous precipitation and stream chemistry data from the Hubbard Brook ecosystem study (1963–2013) Ecology 98 2224–2224 [DOI] [PubMed] [Google Scholar]
  39. Lloret J and Valiela I 2016. Unprecedented decrease in deposition of nitrogen oxides over North America: the relative effects of emission controls and prevailing air-mass trajectories Biogeochemistry 129 165–80 [Google Scholar]
  40. Lovett GM and Goodale CL 2011. A new conceptual model of nitrogen saturation based on experimental nitrogen addition to an oak forest Ecosystems 14 615–31 [Google Scholar]
  41. Mathias JM and Thomas RB 2018. Disentangling the effects of acidic air pollution, atmospheric CO2, and climate change on recent growth of red spruce trees in the Central Appalachian Mountains Global Change Biol. ( 10.1111/gcb.14273) [DOI] [PubMed] [Google Scholar]
  42. McLauchlan KK, Craine JM, Oswald WW, Leavitt PR and Likens GE 2007. Changes in nitrogen cycling during the past century in a northern hardwood forest Proc. Natl Acad. Sci 104 7466–70 [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. McLauchlan KK, Ferguson CJ, Wilson IE, Ocheltree TW and Craine JM 2010. Thirteen decades of foliar isotopes indicate declining nitrogen availability in central North American grasslands New Phytol. 187 1135–45 [DOI] [PubMed] [Google Scholar]
  44. McLauchlan KK, Gerhart LM, Battles JJ, Craine JM, Elmore AJ, Higuera PE, Mack M, McNeil BE, Nelson DM and Pederson N 2017. Centennial-scale reductions in nitrogen availability in temperate forests of the United States Sci. Rep 7 7856. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. McNeil BE, de Beurs K M, Eshleman KN, Foster JR and Townsend PA 2007. Maintenance of ecosystem nitrogen limitation by ephemeral forest disturbance: an assessment using MODIS, Hyperion, and Landsat ETM+ Geophys. Res. Lett 34 [Google Scholar]
  46. Norby RJ, De Kauwe M G, Domingues TF, Duursma RA, Ellsworth DS, Goll DS, Lapola DM, Luus KA, MacKenzie AR and Medlyn BE 2016. Model–data synthesis for the next generation of forest free-air CO2 enrichment (FACE) experiments New Phytol. 209 17–28 [DOI] [PubMed] [Google Scholar]
  47. NRCS U 2009. Web soil survey, URL http://websoilsurvey.ncsc.usda.gov/app/ [verified October 29, 2009]
  48. Ollinger SV, Aber JD, Reich PB and Freuder RJ 2002. Interactive effects of nitrogen deposition, tropospheric ozone, elevated CO2 and land use history on the carbon dynamics of northern hardwood forests Global Change Biol. 8 545–62 [Google Scholar]
  49. Oulehle F, Chuman T, Hruška J, Krám P, McDowell WH, Myška O, Navrátil T and Tesař M 2017. Recovery from acidification alters concentrations and fluxes of solutes from Czech catchments Biogeochemistry 132 251–72 [Google Scholar]
  50. Oulehle F, Evans CD, Hofmeister J, Krejci R, Tahovska K, Persson T, Cudlin P and Hruska J 2011. Major changes in forest carbon and nitrogen cycling caused by declining sulphur deposition Global Change Biol. 17 3115–29 [Google Scholar]
  51. Pardo L, Hemond H, Montoya J and Pett-Ridge J 2007. Natural abundance 15 N in soil and litter across a nitrate-output gradient in New Hampshire Forest Ecology and Management 251 217–30 [Google Scholar]
  52. Pardo L, Templer P, Goodale C, Duke S, Groffman P, Adams M, Boeckx P, Boggs J, Campbell J and Colman B 2006. Regional assessment of N Saturation using Foliar and Root\varvec {\delta}^{\bf 15}{\bf N} Biogeochemistry 80 143–71 [Google Scholar]
  53. Perakis SS, Tepley AJ and Compton JE 2015. Disturbance and topography shape nitrogen availability and δ15N over long-term forest succession Ecosystems 18 573–88 [Google Scholar]
  54. Peñuelas J, Ciais P, Canadell JG, Janssens IA, Fernández-Martínez M, Carnicer J, Obersteiner M, Piao S, Vautard R and Sardans J 2017. Shifting from a fertilization-dominated to a warming-dominated period Nature Ecology & Evolution 1 1438. [DOI] [PubMed] [Google Scholar]
  55. Qi H, Coplen TB, Geilmann H, Brand WA and Böhlke JK 2003. Two new organic reference materials for δ13C and δ15N measurements and a new value for the δ13C of NBS 22 oil Rapid Commun. Mass Spectrom 17 2483–7 [DOI] [PubMed] [Google Scholar]
  56. Rosi-Marshall EJ, Bernhardt ES, Buso DC, Driscoll CT and Likens GE 2016. Acid rain mitigation experiment shifts a forested watershed from a net sink to a net source of nitrogen Proc. Natl Acad. Sci 113 7580–3 [DOI] [PMC free article] [PubMed] [Google Scholar]
  57. Rustad L, Campbell J, Marion G, Norby R, Mitchell M, Hartley A, Cornelissen J and Gurevitch J 2001. A meta-analysis of the response of soil respiration, net nitrogen mineralization, and aboveground plant growth to experimental ecosystem warming Oecologia 126 543–62 [DOI] [PubMed] [Google Scholar]
  58. Sabo R, Scanga SE, Lawrence GB, Nelson DM, Eshleman KN, Zabala GA, Alinea AA and Schirmer CD 2016a. Watershed-scale changes in terrestrial nitrogen cycling during a period of decreased atmospheric nitrate and sulfur deposition Atmos. Environ 146 271–9 [Google Scholar]
  59. Sabo RD 2018. Shifting Inputs and Transformations of Nitrogen in Forested and Mixed Land Use Basins: Implications for Hydrologic Nitrogen Loss [Google Scholar]
  60. Sabo RD, Nelson DM and Eshleman KN 2016b. Episodic, seasonal, and annual export of atmospheric and microbial nitrate from a temperate forest Geophys. Res. Lett 43 683–91 [Google Scholar]
  61. Saurer M, Cherubini P, Ammann M, De Cinti B and Siegwolf R 2004. First detection of nitrogen from NOx in tree rings: a 15 N/14 N study near a motorway Atmos. Environ 38 2779–87 [Google Scholar]
  62. Scanlon TM, Ingram SM and Riscassi AL 2010. Terrestrial and in-stream influences on the spatial variability of nitrate in a forested headwater catchment Journal of Geophysical Research: Biogeosciences 115 [Google Scholar]
  63. Sinsabaugh R, Zak D, Gallo M, Lauber C and Amonette R 2004. Nitrogen deposition and dissolved organic carbon production in northern temperate forests Soil Biol. Biochem 36 1509–15 [Google Scholar]
  64. Sinsabaugh RL 2010. Phenol oxidase, peroxidase and organic matter dynamics of soil Soil Biol. Biochem 42 391–404 [Google Scholar]
  65. Templer PH, Arthur MA, Lovett GM and Weathers KC 2007. Plant and soil natural abundance δ15N: indicators of relative rates of nitrogen cycling in temperate forest ecosystems Oecologia 153 399–406 [DOI] [PubMed] [Google Scholar]
  66. Tomlinson G, Buchmann N, Siegwolf R, Weber P, Thimonier A, Pannatier EG, Schmitt M, Schaub M and Waldner P 2016. Can tree-ring δ15N be used as a proxy for foliar δ15N in European beech and Norway spruce? Trees 30 627–38 [Google Scholar]
  67. Townsend PA, Eshleman KN and Welcker C 2004. Remote sensing of gypsy moth defoliation to assess variations in stream nitrogen concentrations Ecological Applications 14 504–16 [Google Scholar]
  68. Townsend PA, Singh A, Foster JR, Rehberg NJ, Kingdon CC, Eshleman KN and Seagle SW 2012. A general Landsat model to predict canopy defoliation in broadleaf deciduous forests Remote Sens. Environ 119 255–65 [Google Scholar]
  69. Vitousek PM, Gosz JR, Grier CC, Melillo JM, Reiners WA and Todd RL 1979. Nitrate losses from disturbed ecosystems Science 204 469–74 [DOI] [PubMed] [Google Scholar]
  70. Wang R, Goll D, Balkanski Y, Hauglustaine D, Boucher O, Ciais P, Janssens I, Penuelas J, Guenet B and Sardans J 2017. Global forest carbon uptake due to nitrogen and phosphorus deposition from 1850 to 2100 Global Change Biol. ( 10.1111/gcb.13766) [DOI] [PubMed] [Google Scholar]
  71. Zhu Z, Woodcock CE and Olofsson P 2012. Continuous monitoring of forest disturbance using all available Landsat imagery Remote Sens. Environ 122 75–91 [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Sup. Material
NIHMS1579159-supplement-Sup__Material.pdf (657.1KB, pdf)

RESOURCES