Understanding Forest Nitrogen Dynamics: A Methodological Overview

This document details the meticulous methodologies employed in studying nitrogen isotope ratios and their relationship with various environmental and forest parameters across Sweden. From the precise collection of tree core samples to advanced statistical modeling, each step was designed to provide robust insights into forest health and environmental change.

Sample Collection and Preparation

Tree cores were sourced from the extensive Swedish National Forest Inventory (NFI) archive, covering a significant period from 1961 to 2018. To ensure independence and broad geographical representation, a systematic sampling approach was utilized: a grid comprising 250 square cells (50 × 50 km) was overlaid across Sweden.

Cores were selectively chosen from dominant or co-dominant Norway spruce and Scots pine trees, specifically those aged between 41 and 60 years. Samples were restricted to mesic site types with slopes of 20% or less to maintain consistency. For each species, samples were randomly picked from each grid cell, spanning six decades from the 1960s to the 2010s.

Specific sampling years included 1961 and 1977, followed by three-year periods for subsequent decades (1986–1988, 1996–1998, 2006–2008, 2016–2018). This multi-year approach for later decades was necessary to accumulate sufficient sample numbers due to reduced NFI sampling intensity over time. A total of 1,609 samples were collected for analysis.

Each tree core sample, collected at breast height (1.3 m), underwent meticulous preparation: the outer bark, cambial layers, and the most recent annual ring were carefully removed.

Subsequently, a 10-year annual growth segment was precisely separated for chemical analysis. Dissections were performed using a stainless steel surgical blade under a stereo microscope to ensure accuracy. Samples were identified by their intermediate 10-year growth increment year; for example, 1961 collected samples representing the 1951–1960 period were referred to as 1955.

Measurement and Analysis of N Isotopes

Nitrogen isotope ratios (δ15N) were rigorously analyzed at the Central Appalachians Stable Isotope Facility (CASIF). The analytical setup involved a Carlo Erba NC2500 elemental analyzer coupled with a Thermo Finnigan Delta V+ isotope-ratio mass spectrometer.

For each analysis, approximately 10 mg of wood from a radial core slice was accurately weighed, placed in a tin capsule, and subsequently analyzed for its δ15N content. Samples were pre-processed to eliminate potential interferences from CO2 and water vapor.

Data normalization was achieved against the Ambient Inhalable Reservoir (AIR) scale, employing a two-point normalization curve with established internal standards.

The analytical precision for δ15N was remarkably high, recorded at 0.3‰. The δ15N value is conventionally expressed using the standard delta notation:

$${{\rm{\delta }}}^{15}{\rm{N}}=({R}{\mathrm{sample}}{/R}{\mathrm{standard}}-1)\times 1,000$$

Here, R sample and R standard denote the ratios of heavy to light isotopes in the sample and standard, respectively, with results presented in parts per thousand (‰). AIR serves as the designated standard for nitrogen.

Climate, Nr Deposition, and Forest Data

To enrich the analytical models, supplementary data were drawn from various external databases, providing crucial context for the observed δ15N values. These data were integrated into a linear mixed-effects model.

The external data encompassed three main categories:

• Climate parameters: This included mean annual temperature (expressed in °C) and atmospheric CO2 concentrations (in ppm). An additional variable, relative temperature change, was calculated using 1961 as the reference year.
• N deposition parameters: Key metrics here included NHx and NOy (measured in g m−2), their ratio (NHx:NOy), and the 10-year average total N deposition.
• Forest stand parameters: Data on total basal area, serving as a proxy for forest biomass, and stand age were also incorporated.

Monthly temperature data were extracted from CRU TS v4.07, while CO2 concentration data were compiled from historical measurements and NOAA's Global Monitoring Laboratory. Monthly wet plus dry deposition figures for NHx and NOy were obtained from the comprehensive ISIMIP3a dataset.

Swedish NFI Growth Data

The Swedish National Forest Inventory (NFI) is a vital source of forest information, operating through annual systematic sampling of Swedish forests. Its design employs a stratified systematic cluster approach, utilizing both permanent and temporary circular plots. Tree cores for this study were primarily extracted from temporary plots, each with a 7-meter radius. Annually, the NFI undertakes over 100,000 individual tree measurements.

Growth rate estimation methodologies differ between plot types. For permanent plots, growth is determined by calculating the difference in volume estimates between successive inventories. For temporary plots, radial increment measurements from tree cores are combined with regression models to estimate growth.

These individual estimates are then synthesized into a weighted mean, achieving a high level of accuracy with an uncertainty of 1% or less.

Total volume growth for mesic P. sylvestris- and P. abies-dominated forests was acquired and subsequently converted into stand-level growth data by dividing it by the respective forest area.

Statistical Methods

All data analysis was meticulously performed using R Statistical Software, employing a suite of advanced statistical techniques.

Initial Analyses

• Linear regressions were initially performed to evaluate δ15N as a function of time for each tree species. These analyses were conducted across four distinct geographical regions: north, central, southeast, and southwest Sweden.
• Following these regressions, a two-way analysis of variance (ANOVA) was executed. Tukey’s post hoc comparisons were then applied to precisely identify significant differences in regression slopes between regions for each species.

Linear Mixed-Effects Model

• A linear mixed-effects model (implemented via the lme function from the nlme package in R) was constructed. The selection of the best predictor combination for δ15N was achieved through a rigorous stepwise forwards and backwards selection process, guided by Akaike Information Criteria (AIC).
• Initial fixed effects considered included atmospheric CO2, mean annual temperature, total Nr deposition, total basal area, stand age, tree species, and various interactions. Grid cell was incorporated as a random effect to account for spatial variability.
• The final model retained all variables except stand age and the interaction between CO2 and total Nr deposition. Multi-collinearity among variables was carefully checked using the Variance Inflation Factor (VIF).
• This primary model was then re-run with variations, replacing total Nr deposition with NHx, NOy, and the NHx:NOy ratio individually. Marginal, conditional, and partial R2m values were calculated for assessment.
• Further model variations included one that incorporated latitude and longitude, and another that substituted mean annual temperature with a temperature change variable.
• Finally, δ15N was modeled as a function of scaled variables using the emmeans function, and the slopes of significant variables were estimated using the emtrends function.

Forest Volume Growth Analysis

• To thoroughly investigate the influence of forest volume growth, a new variable—'relative forest volume growth change'—was calculated for each grid cell and species.
• Two distinct linear mixed-effects models were subsequently constructed. These models included the newly calculated relative forest volume growth change, along with either absolute forest volume growth or basal area, as additional fixed factors. Grid cell continued to be included as a random factor. Marginal and conditional R2 values were calculated to evaluate the models' explanatory power.