Phosphorous fertilization and soil pH affect the growth of deciduous trees in a temperate hardwood forest

Abstract

Forest ecosystems with altered nutrient limitations are a common legacy of acidic deposition in North America. Continued acidic deposition has lowered soil pH and revealed phosphorus (P) limitations in many temperate forest ecosystems. Previous studies exploring P limitations or co‐limitations are often short term, and thus may potentially show a response to limitation that is not sustained over time. To better understand how a forest's response to P limitation and acidic deposition can change over time, we added P, limestone to raise pH, and a cross‐treatment where both P and limestone were added to 3 different northeastern Ohio forest stands over a 12‐year period. We tracked diameter at breast height of the trees annually, conducted foliar nutrient analyses, and collected tree roots to assess treatment impacts on mycorrhizal colonization. We analyzed our dataset in three sections: the first 6 years after manipulation, the latter 6 years, and the entire 12‐year period. These sections allowed us to compare differences between early responses to manipulation and later responses. Here, we found that P additions increased basal area growth across multiple species and throughout the entire study, confirming that our forest trees are P‐limited. Cross‐treatments similarly increased basal area growth, but not as much as P additions alone. Some species saw waning effects of treatment in the second half of the study. This could be due to changes in weather patterns, an adjustment of the study system's equilibrium, or the emergence of beech leaf disease in 2014, which has led to the decline of Fagus grandifolia. Early successional species such as Acer rubrum began to benefit from treatments after beech leaf disease killed canopy F. grandifolia trees, perhaps first being light‐limited, but later able to take advantage of the nutrient additions and pH alteration of their soils. Our results suggest that in forests subject to acidic deposition, soil P may co‐limit tree growth, but responses are species dependent.

INTRODUCTION

Ongoing human industrial activity in North America is changing the way forest ecosystems function and the ability of organisms to grow and survive within them. Some of these ecosystem changes are a result of the long-term deposition of acidic chemicals from industrial emissions (Cogbill & Likens, 1974; U.S. EPA, 2023a; Watmough & Dillon, 2003). Although sulfur dioxide (SO₂) and nitrogen oxide (NO_x) emissions have been reduced by the Clean Air Act Amendments of 1990 (Aber et al., 1998; U.S. EPA, 2023b; Greaver et al., 2012; Li et al., 2016), previous and still ongoing emissions have left behind a legacy of acidic depositions. Acidic deposition from SO₂ and NO_x emissions has negatively affected forest ecosystems in North America by changing the soil chemistry over decades (Bailey et al., 2005), eventually leading to additional nutrient limitations (Aber et al., 2003), reductions in understory plant diversity (Zarfos et al., 2019), and overall declines in forest health and productivity (Driscoll et al., 2003; Tamm, 1976; Walter et al., 2021; Watmough & Dillon, 2003).

Biogeochemical and ecosystem theorists have posited that younger, glaciated forest soils are primarily nitrogen (N) limited, while older, unglaciated soils are primarily phosphorus (P) limited (Crews et al., 1995; Walker & Syers, 1976). Past fertilization studies have corroborated these theories (Finzi, 2009; LeBauer & Treseder, 2008; Magill et al., 2000), but several studies have reported either P limitations (Gonzales & Yanai, 2019; Goswami et al., 2018; Mohren et al., 1986; Naples & Fisk, 2010) or N and P co-limitations (Elser et al., 2007; Fisk et al., 2014; Hedwall et al., 2017; Vadeboncoeur, 2010) in light of anthropogenic activity. This shift in nutrient limitation has been demonstrated before, most notably in Douglas Fir stands in the Netherlands, where it took 27 years for the system to shift from N limitation to P limitation (Mohren et al., 1986). The authors believed that this shift was due to human industrial activity adding large amounts of N to the atmosphere, writing that any leaching losses of P would have been negligible within their study system (Mohren et al., 1986).

The shift that Mohren et al. (1986) observed in forest soil nutrient limitations was likely driven by N enrichment or saturation. Heavy N deposits on forest soils, as a result of NO_x emissions, can cause N enrichment and may eventually lead to N saturation in some forests (Aber et al., 1998; Berendse et al., 1993). Additionally, within the United States, largely unregulated ammonia (NH₃) emissions have increased since 2002 and are highest in the eastern United States and the Ohio Valley (Benish et al., 2022; Li et al., 2016). While not acidic, NH₃ emissions contribute to the overall high rate of N deposition across the United States (Benish et al., 2022; Li et al., 2016). The alleviation of N limitations via N saturation has revealed the potential for a transactional P limitation within forest ecosystems (Hong et al., 2022; Vitousek et al., 2010). In the eastern United States, local studies of foliar and litter N and P concentrations have suggested P limitation in some forests (e.g., Burke et al., 2023; Gonzales & Yanai, 2019; See et al., 2015). However, to date, a widespread shift from N to P limitation in forests of Ohio, to our knowledge, has not been documented, nor has widespread P limitation in glaciated forest soils across the eastern United States (Crowley et al., 2012). P limitation, though, can also be experienced in soils that are acidic, regardless of N status. This is because acidic soil conditions, such as those created by acidic precipitation, allow for the mobilization of toxic ions such as aluminum ions (Al⁺) and a loss of base cations (Driscoll et al., 2003; Fenn et al., 2006). Acidic conditions promote the binding of P with Al⁺, immobilizing P and rendering it unavailable for uptake by trees and other plants (Fenn et al., 2006; Lawrence et al., 1997; Liu et al., 2019; Reuss, 1983). Understanding this anthropogenically mediated nutrient limitation and its effects on plants growing within these P-limited environments is essential to predict how temperate forest ecosystems will differ in the future as human industrial activity continues.

To our knowledge, only a small number of studies have examined P limitations within temperate deciduous forests in the eastern United States, despite these forests occupying 26 states and hundreds of millions of acres (National Park Service, 2022). The few studies that have investigated responses to P fertilization in temperate forests, however, all show short-term effects. Goswami et al. (2018) explored basal area growth of overstory trees within a P × N fertilization study at Multiple Element Limitation in Northern Hardwood Ecosystems (MELNHE), but their observations only took place over four growing seasons. Fisk et al. (2014) examined the response of inorganic N and P availability to fertilization with these same nutrients over one growing season at MELNHE, and Shan et al. (2022) examined root biomass and growth 5 years after beginning fertilization with N and P, also at MELNHE. Finzi (2009) reported that Great Mountain Forest in Connecticut was still N limited and did not find strong evidence for either P limitation or N and P co-limitation, but the study only took place after 2 years of fertilization.

While the work done in these studies is informative for growth patterns and short-term ecosystem consequences over a few growing seasons, observing forest ecosystems for a relatively short period of time might not capture the full extent of a forest community's response to altered nutrient availability resulting from either fertilization or anthropogenic emissions. The current study aimed to fill this gap in literature surrounding long-term P limitation in temperate hardwood forests. Using an ongoing nutrient manipulation experiment with 12 years of data from three northeastern Ohio forests, we aimed to determine how mature trees responded to nutrient limitations caused by acidic deposition and how responses might change through time. To explore this subject, measures of leaf litter biomass and nutrient content, root biomass, mycorrhizal colonization, and tree growth were taken over 12 years. Because nutrient concentrations of senesced leaf litter provide information on resorption proficiency (a measure of how well plants minimize nutrient loss; Killingbeck, 1996), these data for our study provided an estimate of the nutrient status of our sites (as in Richardson et al., 2008; See et al., 2019). Leaf litter biomass, root biomass, and tree basal growth were measures of plant productivity in response to the treatments (Clark et al., 2001; Goswami et al., 2018). Mycorrhizal colonization of the dominant tree species was used to assess the strength of plant reliance on this mutualism for nutrient acquisition in ambient compared with manipulated soil conditions (Treseder, 2004) given that P-limited systems tend to have higher colonization (Almeida et al., 2019; Bahr et al., 2015). We asked the following questions: (1) Does leaf litter biomass or chemistry change with nutrient manipulation? (2) Does root biomass or mycorrhizal colonization change with manipulation? (3) Does tree basal area growth differ by nutrient manipulation, and does this response change with mycorrhizal association or among tree species? (4) Does tree growth response to nutrient manipulation change with tree size?

MATERIALS AND METHODS

Site description and data collection

The study sites are located in northeastern Ohio, USA, within three deciduous forests: Pierson Creek and Schoop Forest, managed by the Holden Arboretum in Kirtland, OH, and Squire Valleevue and Valley Ridge Farms, managed by Case Western Reserve University in Hunting Valley, OH. The average yearly temperature for this area is 8.1°C and the average yearly precipitation is 1200 mm (with 2870 mm average snowfall). These forests are dominated by Acer saccharum (sugar maple) and Fagus grandifolia (American beech) with a mix of other hardwood species including Acer rubrum (red maple) Quercus rubra (red oak), Quercus alba (white oak), Carya ovata (shagbark hickory), Carya cordiformis (bitternut hickory), Prunus serotina (black cherry), Liriodendron tulipifera (tulip poplar), Ostrya virginiana (American hop hornbeam), and Carpinus caroliniana (American hornbeam) (Appendix S1: Table S1). The most abundant tree species in our plots are sugar maple (44%), American beech (21%), red maple (11%), red oak (8%), black cherry (5%), American hop hornbeam (2%), American hornbeam (1%), and shagbark hickory (1%). All other species combined make up less than 10% of the trees within our plots (Appendix S1: Table S1). Though the exact age of the forests is unknown, they have all been conserved by their respective institutions since the 1930s. Each forest has 12 plots (20 × 40 m, 800 m²) that were established in 2009: three treatments paired with one control (four plots in a set) and three replications of a set. The treatments are as follows: Lime, where Hi-Ca limestone was added to raise soil pH; TSP, where triple super phosphate (TSP) was added to increase levels of available P; and X-Trt, where both Hi-Ca limestone and TSP were added. pH was monitored annually so that soil pH remained above 5.5 in limestone addition plots (Lime and X-Trt), as soils with a pH below 5 have increased solubilized aluminum (Goldberg et al., 1995; Thomas & Hargrove, 1984). Ambient pH in plots that did not receive limestone ranged from 3.9 to 4.1 at 2 years post-treatment initiation (Carrino-Kyker et al., 2016) and remained around 4 throughout the study. pH has been measured yearly since 2009 to maintain target pH in limestone addition plots. Soil pH values from within the first 7 years of this study can be found in Burke et al. (2021), Carrino-Kyker et al. (2016), and Kluber et al. (2012). Appendix S1: Table S2 lists soil pH values for two of the subsequent years from which we are reporting plant growth data in the current study. Pelletized limestone was added in the fall of 2009, 2010, 2012, 2013, and 2017 to maintain pH levels. TSP, also in the form of pellets, was added in the spring of 2010, 2011, 2013, 2014, 2015, 2016, and 2017. See DeForest et al. (2012) for a more detailed description of the experimental setup.

Trees within the plot boundaries with a diameter at breast height (DBH) of 6 cm or greater were tagged in 2010. The tags were nailed to the trees at 1.4 m (4.5 ft) from the ground and used as a permanent mark for subsequent DBH measurements. DBH measurements began in June 2010, which was 9 months after the initial limestone application and 2 months after the initial TSP application. DBH measurements continued annually and always in May–June to keep the season of measurement consistent. In 2013, trees that had grown to the minimum size requirement for this study were tagged and included in the annual DBH survey. In total, 1492 trees were tagged and measured yearly, unless the tree had died. Because the trees were largely visually assessed to be at 6 cm DBH or greater, some trees fell slightly below this but were still kept in the dataset (42 of the 1492 trees were below 6 cm DBH but were between 4.5 and 5.9 cm DBH at the start of the study).

Leaf litter

Wooden collection baskets were placed in each forest during the summer of 2014, 4 years after the initiation of soil treatments, to capture leaf litter. Three baskets were placed in each of the forest's 12 subplots for 36 baskets per forest and 108 in total. Leaf litter was collected every autumn from 2014 to 2022, resulting in a total of 9 years of data. Leaves were collected between 2 and 4 times a season (most commonly three times), depending on a visual assessment of canopy drop during each season. The leaf litter collections spanned the litterfall season and, thus, captured naturally occurring temporal changes in nutrient concentration (See et al., 2019). The biomass for each subplot was summed across all collections, yielding a total biomass per subplot for each season.

Each collection of leaves was dried for 3–5 days at 60°C; subsequently, mass was recorded. Leaf masses were then summed across subplots, giving 12 biomass values per forest and 36 total. Samples of leaves in each subplot from 2015 and 2021 were combined, crushed, and sent to the Plant Analysis Program in the Ag Analytical Services Lab at The Pennsylvania State University for chemical analysis. Although this lab tested for a wide range of elements, our study was concerned with those that could be impacted by our treatments: aluminum (Al), calcium (Ca), magnesium (Mg), N, and P. The leaves from 2015 and 2021 were analyzed because, at the time of their analysis, they represented the halfway point and most recent data collection of our study. Analysis results are only available at the plot level. Leaves were not separated or analyzed by species.

Root biomass and colonization

Root biomass

At each plot within all three forests, six trees were selected for root sample collection. Of the six trees selected per plot, three were arbuscular mycorrhizal fungi-associated species (AM trees—either sugar maple or red maple) and three were ectomycorrhizal fungi-associated species (ECM trees—either American beech or red oak). However, selecting three AM and three ECM trees at each plot was not always possible due to a lack of desired species or too young/unhealthy/deceased trees. In total, from all three forests (36 plots), root samples were used from 96 AM trees and 97 ECM trees.

At each selected tree, root samples were taken using a 10-cm-diameter metal soil core sampler. All root samples were collected between June 6, 2022, and July 11, 2022, which is just over a decade after treatment initiation. Approximately 1 m from the base of each tree, the litter layer was removed, and the soil core sampler was inserted 5 cm deep into the ground (as this is the zone of highest fine root growth and activity in our forests). The soil and roots extracted by the core sampler were then emptied into a plastic bag. This process was repeated on the opposite side of the tree (2 cores per tree) and bags were placed in a cooler for transport back to the laboratory. At the laboratory, soil containing roots was kept refrigerated at 4°C until processed.

During root processing, soil was separated from roots using a round sieve (2 mm) and forceps, and by rinsing roots with deionized water. Cleaned, wet roots were then placed in an oven for drying after subsamples were removed for mycorrhizal colonization estimates (see ECM colonization and AM colonization sections below). After drying completely, root mass was recorded. The dry mass in grams of each root sample was reported as root biomass.

ECM colonization

Because ECM fungi grow on the outside of tree roots, making them visible under low-level magnification, ECM colonization was quantified as the number of live ECM “root tips” per gram of root biomass for each root sample adjusted for tree DBH (n = 97). Before ECM root samples were placed in the oven for drying (as described above), each sample was placed in a 9-cm glass petri dish and examined under a dissecting microscope to count the total number of root tips colonized by ectomycorrhizal fungi (Figure 1). Root tips were removed from the root sample as they were counted to avoid counting errors. The total number of root tips was divided by the sample mass to calculate the ECM colonization of each sample and then divided by the DBH to adjust for the size of each tree. 1 $$$ \mathrm{ECM}\ \mathrm{fungi}\ \mathrm{colonization}=\frac{\left(\frac{\mathrm{No}.\mathrm{root}\ \mathrm{tips}}{\mathrm{Root}\ \mathrm{mass}\ \left(\mathrm{g}\right)}\right)}{\mathrm{dbh}\ \left(\mathrm{cm}\right)} $$$

[IMAGE OMITTED. SEE PDF]

AM colonization

Because AM fungi grow within tree roots, making them not easily visible under low-level magnification, AM colonization was quantified as the number of AM fungal gene copies per gram of root biomass for each root sample adjusted for tree DBH (n = 96). To obtain the number of AM fungal gene copies, quantitative polymerase chain reaction (qPCR) was used.

Before root samples were placed in the oven for drying (as described above), a small portion of the root sample was removed, placed in an empty 1.5-mL microcentrifuge tube, and mass was recorded. The wet roots that were removed for qPCR were corrected for water content and added to the dry mass for a measure of total dry root biomass. Microcentrifuge tubes containing roots were then immediately placed in a freezer at −80°C until removal for DNA extraction.

Frozen roots were ground under liquid nitrogen directly in the 1.5-mL microcentrifuge tubes with sterile plastic pestles.Four hundred microliters of 2% cetyltrimethylammonium bromide (CTAB) was added to the crushed roots, and 1000 μL pipette tips with the top 1–2 mm cut off were used to transfer the CTAB and crushed roots to bead beating tubes. Each bead beating tube contained 300 mg of 400 μM glass beads (OPS Diagnostics, Lebanon, NJ, USA) and 200 mg of 1 mm glass beads (Chemglass Life Science, Vineland, NJ, USA). An additional 350 μL of CTAB was then added to the bead beating tubes containing the crushed roots, and cells were lysed using a Precellys homogenizer (Bertin Technologies, Montigny-le-Bretonneux, France) where samples were beaten at 6500 rpm for 40 s. DNA isolation was then performed on each AM root sample using a phenol/chloroform procedure that precipitated DNA with 20% polyethylene glycol 8000 (see Burke et al., 2009). Isolated DNA was suspended in Tris EDTA and stored at −20°C in low retention tubes (Fisher Scientific, Florence, KY, USA) until use in qPCR.

The total number of AM fungal 18S gene copies per sample was assessed with a qPCR procedure adapted from the procedure described in Kluber et al. (2012) using primers AMG1F and AM1 (as in Hewins et al., 2015). In silico analysis has shown these primers likely bind to most members of the Glomeromycota, but notably may not amplify members of Claroideoglomeraceae, Archeosporales, or Paraglomerales due to mismatches on the 3′ end of the reverse primer, AM1 (Bodenhausen et al., 2021). qPCR with this primer set as a way to assess AM fungal colonization has been tested by others and shown to correlate well with microscopy in several plant species (Corona Ramírez et al., 2023). All materials and surfaces used during the preparation of samples for qPCR were treated with ultraviolet light for 15 min before being used. qPCR reactions were run in triplicate for each sample and contained 1× iTaqUniversal SYBR Green Supermix (Bio-Rad Laboratories, Inc., Hercules, CA, USA), 0.4 μM of each primer, 0.5 mg/mL bovine serum albumin (BSA), and 1 μL of extracted DNA in 20 μL reactions. qPCR runs were conducted on a CFX Connect Real Time System (Bio-Rad) with the following thermocycling conditions: 95°C initial denaturation, 35 cycles of 95°C for 30 s denaturation, 62°C for 90 s annealing, and 72°C for 90 s extension, and a plate read after each extension. Reaction specificity was determined with melt curves that increased temperature from 65°C to 95°C with a plate read at every 0.5°C increment and visual estimation on 2% agarose gels, and all no template controls were below detection. Each run included a five-point standard curve that ranged in value from 10⁶ to 10² copies, which was made with a transformed plasmid containing an AM fungal 18S rRNA gene segment (see Hewins et al., 2015). Threshold cycles (Cqs) were determined manually for each run to optimize the reaction efficiency (91.1%–96.2%) and standard curve r² (0.995–0.999). The number of 18S rRNA gene copies per sample DNA was averaged across the three qPCR replicates and reported as an estimate of AM fungal abundance. The number of AM gene copies was then adjusted for the mass of the root sample (corrected for water content) and the DBH of the corresponding tree. 2 $$$ \mathrm{AM}\ \mathrm{fungi}\kern0.34em \mathrm{colonization}=\frac{\frac{\mathrm{No}.\kern0.5em \mathrm{DNA}\ \mathrm{copies}}{\mathrm{Root}\kern0.34em \mathrm{portion}\ \mathrm{dry}\ \mathrm{mass}\kern0.1em \left(\mathrm{g}\right)}}{\mathrm{dbh}\ \left(\mathrm{cm}\right)} $$$

Basal area

Diameter at breast height measurements (in centimeters) were taken annually in early summer. The basal area of each tree in each year was calculated from the DBH of that year using the square meter formula described by Bettinger et al. (2017). Because differently sized trees grow at different rates, we standardized each tree's basal area before analysis following the formula used for Relative Basal Area Increment (hereinafter RBAI) as described by Goswami et al. (2018), where BA_P is the final basal area, BA_i is the initial basal area, and n represents the number of growing seasons. RBAI approximates percent growth per year. 3 $$$ \mathrm{RBAI}=\left({\left(1+\left(\frac{\left({\mathrm{BA}}_{\mathrm{P}}-{\mathrm{BA}}_{\mathrm{i}}\right)}{{\mathrm{BA}}_{\mathrm{i}}}\right)\right)}^{\left(\frac{1}{n}\right)}-1\right) $$$

We wanted to observe how the forest might be responding to conditions at various time points after soil treatment: the first 6 years (2011–2016), the last 6 years (2017–2022), and over the entire 12-year growing period (2011–2022). These growing periods allowed us to see if the forest was responding slowly to the soil additions or if there was an initial response that stabilized over time. A tree must have been alive in all 6 or 12 years of a growing period in order for it to qualify as being present in that growing period. If a tree died earlier than the end of the growth period, it was excluded from the analysis of that period. Each tree, if alive during the entire length of the study, had three different RBAI calculations—one for each period. In total, there were 1269 trees analyzed for the first 6-year growing period and 1064 trees analyzed for the second 6 year and entire 12-year growing periods. We analyzed all RBAI data using these three time points to see how the treatments impacted trees overall, by species, by mycorrhizal association, and by size class at different points in the experiment.

Data analysis

We tested treatment effects on leaf litter biomass, leaf litter chemistry, root biomass, root colonization, and RBAI with linear mixed-effects models (code lme) using the lme4 package, version 2.1-161 in R, version 4.2.2 (R Core Team, 2024). All models used the block of the forest site (Pierson Creek, Schoop Forest, or Squire Valleevue and Valley Ridge Farms) as the random effect to account for any variation across these sites. Shapiro and Bartlett tests were conducted on every model as a diagnostic assessment. In order to make the RBAI, root biomass, root colonization, and leaf litter biomass data distributed more normally and have more constant variances, we performed a square root transformation ($$$ \sqrt{x} $$$). Tukey post hoc tests were conducted after every model (with the multcomp package, version 1.4-20) to observe any differences between treatments. Significance for all models was determined at p < 0.05, but we discuss trends for p values ranging from 0.05 to 0.09.

Leaf litter

Individual year and treatment (Control, Lime, TSP, or X-Trt) were used as the fixed effects in the litter biomass models to observe how litter biomass might be changing over time and in response to the treatments. We regressed elements (Al, Ca, Mg, N, or P) against treatment in the litter chemistry models to see if there were any changes in chemistry between treatments. We separated these chemistry models by time of chemical analysis (2015 or 2021), creating 10 models total (5 elements × 2 times).

Root biomass and colonization

Root biomass and colonization data for both ECM and AM trees were standardized for tree size prior to statistical analysis. We divided dry root mass and root colonization data by the 2022 DBH of the tree from which the sample was collected We separated the analysis by AM and ECM species, creating 4 models in total with treatment as the fixed effect. A significant value from one of these models would indicate that the treatments were impacting the biomass or colonization of tree roots. Because root sampling in a forest setting can include roots of multiple individual trees, root biomass and colonization data were also analyzed without standardizing for tree size.

We then regressed 2021 leaf chemistry data against our 2022 root data to examine the effects of leaf litter chemistry on root biomass and colonization. We used P, Ca, Mg, N, Ca:P, and N:P as fixed effects and again ran separate models for AM and ECM trees. It is important to note that leaf litter was not sorted by species. This analysis included leaf litter from both mycorrhizal associations and therefore cannot be very specific. Regressions were also conducted on root biomass and colonization numbers without standardizing for tree size as a comparison.

Overall RBAI

We used the RBAI of trees within each time period as the response variable in the overall RBAI models. The change in RBAI analyses used treatment as the fixed effect. These models were repeated for each growing period for a total of 3 models of overall RBAI.

Mycorrhizal association RBAI

Within our forests, ECM fungi are associated with American beech, oak species, and hickory species, among others, while AM fungi are associated with maple species, black cherry, and tulip poplar, among others (Brundrett & Tedersoo, 2020) (A list of all tree species encountered in our plots, along with their mycorrhizal fungal association type, can be found in Appendix S1: Table S1). We first used treatment and mycorrhizal association as fixed effects to look for an interaction between treatment and mycorrhizal association. We repeated these models for each period. We then created separate models for the two mycorrhizal associations, ECM and AM, using treatment as the fixed effect. Again, we repeated these models for each time period, creating 6 models (2 mycorrhizal types × 3 time periods).

Species-specific RBAI

We then analyzed the dataset to see if RBAI varied on a species level, using species as the fixed effect of the model. After examining overall trends, we analyzed the RBAI of the 6 most abundant species in our forest that spanned all three sites: sugar maple, American beech, red maple, red oak, black cherry, and shagbark hickory. We created models to see how their RBAI changed by treatment within each period. Treatment was again used as the fixed effect. We repeated these models for each period, creating a total of 18 models (6 species × 3 time periods).

Tree size RBAI

Our old-growth forests are comprised of a wide variety of tree sizes. We used four different DBH ranges to classify the trees: size 1 (<10.0 cm), size 2 (10.0–29.9 cm), size 3 (30.0–59.9 cm), and size 4 (≥60.0 cm). The DBH in the first year of measurement was used to determine size class. For the vast majority of trees this was 2011, but some trees (107 of the 1492) were added to the study in 2013 at a time when the dataset was being re-evaluated. These size classes were determined using 2013 DBH values. We then analyzed the dataset to see if RBAI varied by size class or by the interaction between size class and treatment. Each model had both size class and treatment as fixed effects. We repeated these models for each time period, creating a total of 3 models. We also looked at the responses to treatment within each size class. Using treatment as the fixed effect, we created 4 models, one for each size class. We repeated these models for each time period, creating 12 total models.

RESULTS

Leaf litter

Leaf litter biomass did not significantly change across treatments or over time, and there was no interaction between the two (Treatment: F = 0.92, p = 0.434; Year: F = 1.13, p = 0.344; Treatment × Year: F = 0.31, p = 0.999).

Leaf litter chemistry data confirmed that our treatments made the intended nutrients of P, Ca, and Mg available for plant uptake at both the halfway point and the end of the study (Table 1). In plots that received TSP, the litter chemistry showed increases in percent P. Post hoc tests revealed that in 2015, TSP and X-Trt plots had significantly increased percent P over Control (p = 0.013 and p < 0.001, respectively; Figure 2a, Appendix S1: Table S3). In 2021, TSP and X-Trt still had increased percent P over Control (Figure 2b), but the X-Trt plots saw a difference right at the threshold for significance (p = 0.002 and p = 0.050, respectively; Appendix S1: Table S3).

TABLE 1 Ranges of measured values of nutrients from leaf litter in control plots and significance values of leaf litter chemistry linear mixed-effects models with treatment as the fixed effect.

Element	2015	2021
Range	dfn	dfd	F	p	Range	dfn	dfd	F	p
Al (mg/kg)	42–427	3	30	0.64	0.595	50–636	3	30	0.31	0.820
Ca (%)	0.86–1.31	3	30	3.93	0.018	0.79–1.31	3	30	4.14	0.014
Mg (%)	0.13–0.20	3	30	4.068	0.015	0.14–0.25	3	30	5.64	0.004
N (%)	0.67–0.83	3	30	0.090	0.965	0.69–1.06	3	30	0.29	0.833
P (%)	0.03–0.06	3	30	8.88	0.0002	0.03–0.05	3	30	5.55	0.004

[IMAGE OMITTED. SEE PDF]

In plots that received Lime, the litter chemistry increased in percent Ca and Mg (Figure 2c–f). Post hoc tests (Appendix S1: Table S3) show that in 2015, X-Trt plots had significantly increased percent Ca over the TSP plots (p = 0.022), but this increase was only marginally significant over control plots (p = 0.078). Interestingly, Lime plots did not show a significantly increased percent Ca over Control, but did show a marginal increase over TSP plots (p = 0.084). In 2015, Lime plots saw significantly increased percent Mg over the TSP plots (p = 0.005) while X-Trt plots showed a marginally significant increase over TSP plots (p = 0.073). In 2021, the Lime plots did significantly increase percent Ca over both the control and TSP plots (p = 0.027 and p = 0.032, respectively) as well as percent Mg (p = 0.005 and p = 0.002, respectively). The X-Trt plots did not have significantly increased Ca or Mg in leaf litter in 2021.

The leaf litter from Lime plots alone never saw increases in percent P (Figure 2a,b), contrary to our previous expectation that the elevating pH would create more available inorganic P. When elevating pH and TSP addition were combined in the X-Trt plots, we did see greater leaf litter percent P in 2015 than in the TSP plots alone (Figure 2a), indicating that the two treatments might be working together. However, the difference between the TSP and X-Trt plots was not significant (Appendix S1: Table S3), and in 2021, leaf litter from the X-Trt plots showed a reduction in percent P compared with TSP plots. Again, this difference was not significant (Figure 2b, Appendix S1: Table S3). Finally, N and Al concentrations were not changed in any treatment during either 2015 or 2021 (Table 1). The range of values is reported for each measured nutrient in the Control plots in Table 1.

Root biomass and colonization

Root biomass did not significantly differ across treatments for either AM or ECM trees (Table 2). Mycorrhizal colonization, however, did change with treatment for both mycorrhizal associations (Table 2, Figure 3). AM trees had higher colonization in Control and TSP than in Lime and X-Trt plots (Figure 3a). The only significant difference, however, was between X-Trt and TSP (p = 0.016). ECM trees saw decreased colonization in X-Trt plots (Figure 3b), but again this was only significantly lower than TSP plots (p = 0.039). The Lime plots alone did not decrease colonization in ECM trees as they did in AM trees (see Table 2 and Figure 3). These results were consistent with biomass and colonization data that were not standardized by tree dbh (Appendix S1: Table S4, Figure S1).

TABLE 2 Significance values of root biomass and colonization linear mixed-effects models (shown as the Treatment column) as well as results of regression analysis for Ca content and N content.

Root trait	Treatment	Ca content	N content
dfn	dfd	F	p	dfn	dfd	F	p	dfn	dfd	F	p
AM biomass	3	96	2.022	0.116	1	98	4.56	0.035	1	98	5.47	0.021
AM colonization	3	96	3.53	0.018	1	98	1.61	0.208	1	98	2.97	0.088
ECM biomass	3	90	0.89	0.451	1	92	1.97	0.164	1	92	1.25	0.267
ECM colonization	3	90	2.83	0.043	1	92	4.92	0.029	1	92	0.010	0.920

[IMAGE OMITTED. SEE PDF]

When regressed against leaf litter nutrients, only two nutrients were associated with changes in mycorrhizal colonization and biomass: Ca and N (Table 2). AM tree root biomass was negatively correlated with increased %Ca and %N (Figure 4). AM fungal colonization tended to decrease with increased %Ca content in the leaf litter (Figure 4), but this relationship was not significant (p = 0.208). Colonization was, however, positively correlated with %N (Figure 4), but this was only a marginally significant relationship (p = 0.088). ECM tree root biomass did not respond to changes in leaf litter chemistry (Table 2). ECM fungal colonization, however, was negatively correlated with %Ca content (Figure 5; p = 0.029). ECM colonization was not associated with %N content in the leaf litter (Figure 5; p = 0.920). The %Ca results were mostly consistent with biomass and colonization data that were not standardized for tree DBH (Appendix S1: Table S4), particularly the significantly negative correlation between %Ca and both AM root biomass and ECM fungal colonization (Appendix S1: Figure S2). The positive correlation between %N and AM fungal colonization, which was approaching significance for numbers corrected for DBH (Table 2) was significant when testing this relationship on uncorrected values (Appendix S1: Table S4, Figure S2). There were three significant correlations when using numbers not corrected for tree size that were not significant when the data were corrected for tree DBH: %Ca content and Ca:P were negatively associated with AM fungal colonization, and %Mg was negatively associated with AM root biomass (Appendix S1: Figure S2).

[IMAGE OMITTED. SEE PDF]

[IMAGE OMITTED. SEE PDF]

Overall RBAI

RBAI was significantly different between treatments across all growing periods (see Table 3 for p values). In the first half of the study, only TSP plots saw significantly increased mean RBAI above Control (Figure 6a; p = 0.004). In the second half, X-Trt plots significantly increased RBAI above Control (Figure 6b; p = 0.002). TSP plots also had an increased mean RBAI, but this was not significantly different from Control (Figure 6b; p = 0.239). Across the entire study, TSP plots had a significantly greater mean RBAI than Control (Figure 6c; p = 0.012). The increase from Control was marginally significant for X-Trt plots (Figure 6c; p = 0.058).

TABLE 3 Significance values of relative basal area increment (RBAI) linear mixed-effects models with treatment as the fixed effect.

RBAI trait	First half (2011–2016)	Second half (2017–2022)	Entire study (2011–2022)
dfn	dfd	F	p	dfn	dfd	F	p	dfn	dfd	F	p
Overall	3	1263	3.91	0.009	3	1058	4.25	0.005	3	1058	3.91	0.009
AM trees	3	772	5.42	0.001	3	648	3.96	0.008	3	637	5.79	0.0007
ECM trees	3	485	0.82	0.481	3	404	1.64	0.181	3	415	1.12	0.340
Acer rubrum	3	142	1.74	0.161	3	130	7.42	0.0001	3	129	5.18	0.002
Acer saccharum	3	536	4.11	0.007	3	458	1.91	0.128	3	453	3.64	0.013
Carya ovata	3	9	0.58	0.643	3	9	0.94	0.460	3	9	0.85	0.500
Fagus grandifolia	3	279	3.16	0.025	3	211	1.73	0.162	3	224	2.58	0.055
Prunus serotina	3	48	0.52	0.673	3	24	1.28	0.303	2	23	1.57	0.225
Quercus rubra	3	111	0.44	0.727	3	100	0.21	0.891	3	100	0.26	0.851

[IMAGE OMITTED. SEE PDF]

Mycorrhizal association RBAI

In all three growing periods, RBAI significantly varied between mycorrhizal associations (see Table 4 for p values). There was an interaction between mycorrhizal association and treatment over the entire study (p = 0.015) and the first half, although only marginally significant (p = 0.080). There was no interaction in the second half (p = 0.112).

TABLE 4 Significance values of relative basal area increment linear mixed-effects models with mycorrhizal association and treatment as fixed effects.

Study period	Mycorrhizal association	Treatment	Mycorrhizal association: Treatment
dfn	dfd	F	p	dfn	dfd	F	p	dfn	dfd	F	p
First half	1	1259	25.62	<0.0001	3	1259	5.32	0.001	3	1259	2.26	0.080
Second half	1	1054	7.34	0.007	3	1054	4.44	0.004	3	1054	2.0054	0.112
Entire study	1	1054	23.0065	<0.0001	3	1054	6.53	0.0002	3	1054	3.49	0.015

We then analyzed the dataset to see how each mycorrhizal association individually responded to treatment. AM-associated trees showed a significant response to treatment in all growing periods, but ECM trees did not. See Table 3 for significance values.

Species-specific RBAI

Across the entire length of the study, RBAI varied significantly among species (F = 2.61, p = 0.0001). We then examined species individually across each time period and found that some species were significantly impacted by treatment, while others were not. Out of our six most abundant species, only three saw treatment effects at any point in the study (Table 3). Sugar maple and American beech both responded initially to treatment (p = 0.007 and p = 0.025, respectively) but did not respond during the second half of the study (Figures 7 and 8). Red maple was the only species we examined that had a significant RBAI response to treatment during the second half of the study (p = 0.0001) but not the first (Figure 7). The RBAIs of shagbark hickory, black cherry, and red oak were not impacted by the treatments at any point during the study (Table 3).

[IMAGE OMITTED. SEE PDF]

[IMAGE OMITTED. SEE PDF]

Size class RBAI

Size class and treatment did not have interacting effects on RBAI despite having significant impacts separately, except for a marginally significant interaction in the second half of the study (F = 1.75, p = 0.075). Size class alone significantly impacted RBAI during the first half (F = 3.96, p = 0.008) and the entire study (F = 3.01, p = 0.029), but not during the second half (F = 0.78, p = 0.503). Smaller trees (size classes 1 and 2) saw significant responses to treatment, while larger trees (size classes 3 and 4) did not in any time period (Table 5). Generally, size class 1 trees grew the most in TSP plots, while size class 2 trees grew the most in X-Trt plots (Figure 9), although this difference was only significant for the second half of the study.

TABLE 5 Tree size class linear mixed-effects models using treatment as the fixed effect.

Size class	First half (2011–2016)	Second half (2017–2022)	Entire study (2011–2022)
dfn	dfd	F	p	dfn	dfd	F	p	dfn	dfd	F	p
1 (<10.0 cm)	3	431	2.66	0.048	3	371	1.79	0.149	3	339	3.21	0.023
2 (10.0–29.9 cm)	3	455	1.15	0.327	3	349	5.76	0.0007	3	374	2.50	0.059
3 (30.0–59.9 cm)	3	269	0.89	0.448	3	237	0.18	0.910	3	242	0.67	0.574
4 (≥60.0 cm)	3	90	0.96	0.414	3	83	1.047	0.376	3	85	1.17	0.328

[IMAGE OMITTED. SEE PDF]

DISCUSSION

Ongoing acidic deposition from industrial activity has led to N-saturated soils worldwide (Aber et al., 1998; Berendse et al., 1993), revealing P limitations in many ecosystems (Aerts et al., 1992; Goswami et al., 2018; Verhoeven & Schmitz, 1991; Vitousek et al., 2010). The results from our study corroborate previous findings, demonstrating a P limitation within our forests. Overall, mean RBAI increased in plots that received P additions, but this was more impactful for TSP plots than for X-Trt plots. Lime plots never had a significantly different mean RBAI from Control, indicating that low soil pH was not a main limiting factor of tree growth. We found that response to treatment varied with mycorrhizal colonization, tree species, and over growing period time, all of which will lead to important future considerations for nutrient dynamic research in temperate forests.

Does leaf litter biomass or chemistry change with nutrient manipulation?

Leaf litter biomass did not change with treatment. However, at both time points, the litter chemistry data confirmed that our additions were successful at making nutrients available to the trees. Plots that received TSP had higher concentrations of P in the leaf litter. Similarly, plots that received limestone had higher concentrations of Ca and Mg. We originally believed that limestone additions would also make inorganic P more available, as raising the soil pH would free P from binding with Al⁺ (Fenn et al., 2006; Reuss, 1983). While this might be occurring on a small scale, there were no detectable changes in P concentration within the leaf litter of the limestone addition plots, and this result is further supported in the literature (Monfort-Salvador et al., 2015; Rineau et al., 2010). Interestingly, the %N and %P values in leaf litter from Control plots (Table 1) were at levels that suggest highly proficient resorption for P, but not N (less than 0.05% P in senesced leaf litter is indicative of complete P resorption; Killingbeck, 1996). This offers some evidence that our forests naturally have low P availability, especially because TSP addition raised %P to levels above this threshold value (Figure 1; Killingbeck, 1996). Natural P limitation in these forests is also supported by our previous work where N:P ratios from fresh leaves of the herbaceous plant Podophyllum peltatum (mayapple) suggested P limitation in the absence of fertilization (Burke et al., 2023).

Does root biomass or mycorrhizal colonization change with nutrient manipulation?

We found that root biomass did not change with treatment, but that both AM and ECM colonization did. AM colonization decreased in both the Lime and X-Trt plots, while the TSP plots had similar colonization levels as Control. ECM colonization only decreased in X-Trt plots. Paired with leaf litter chemistry data, both AM and ECM colonization were negatively associated with percent Ca, and AM colonization was positively associated with percent N and negatively associated with Ca:P. These negative correlations with percent Ca and Ca:P offer some evidence for the suggestion that forests with chronic N deposition may have nutrients other than P (such as Ca) that have become limiting (Crowley et al., 2012), especially as these base cations may leach from systems experiencing deposition of acidic N-containing compounds.

Our results differ from the body of literature surrounding mycorrhizal response to nutrient manipulation. P fertilization has consistently been shown to decrease mycorrhizal colonization in both AM and ECM associated species (Treseder, 2004), and that under P-limited conditions, colonization tends to increase (Almeida et al., 2019; Bahr et al., 2015). Changes in N concentration are generally believed to be negatively associated with colonization (Treseder, 2004), but we observed a positive association. This could be due to greater growth and proliferation by the AM fungi in conditions with more N (Treseder & Allen, 2002). Calcium addition has been demonstrated to increase ECM colonization (Bakker et al., 2000; Børja & Nilsen, 2009; Monfort-Salvador et al., 2015). Although we did not see a significant increase in ECM colonization in the Lime plots, we did not observe the same decrease that we did with AM colonization. The X-Trt plots, however, did significantly decrease in ECM colonization. Our results also contrast with previous findings within our study system. Two years after nutrient additions, Carrino-Kyker et al. (2016) found that AM colonization increased in Lime and X-Trt plots. One possible reason for these contrasting results is the season of sampling. We collected samples in June, but other studies examining the impacts of fertilization on mycorrhizal fungi sampled in the early fall (Børja & Nilsen, 2009 in autumn; Carrino-Kyker et al., 2016 in August and September; Almeida et al., 2019 in October). Mycorrhizal fungi are known to vary seasonally (Hewins et al., 2015; Welsh et al., 2010), which could potentially alter their response to the abiotic environment. There is some support in the literature that sampling date should not matter for colonization level (Bahr et al., 2015; Nowotny et al., 1998), even within the forests studied here (Carrino-Kyker et al., 2019), but this has been widely disputed (Almeida et al., 2019; Bakker et al., 2000; Bashian-Victoroff et al., 2023; Bohrer et al., 2004; Voríšková et al., 2013). It is important to note, however, that previous work suggests sampling date matters more in systems that recently received nutrient additions, whereas around 5 years post-manipulation, survey date does not impact colonization results (Bakker et al., 2000). In the present study, the most recent nutrient additions occurred in 2017 and sampling was conducted in 2022, making the difference in sampling season the reason for the difference in response by the AM fungi in our previous studies compared to the current one unlikely.

Another explanation, then, might be that the present study sampled roots 12 years after continual nutrient additions, whereas earlier work conducted on this study system took place only 2 years after experiencing manipulation (Carrino-Kyker et al., 2016). Balser et al. (2002) discuss the potential for microbial communities to reach a new “equilibrium” post-disturbance. They posit that after initial alterations, microbial communities will respond in certain, predictable ways before eventually adapting to their new conditions. This change in response will take a long period of time that is not well understood or standard across ecosystems (Balser et al., 2002). Perhaps earlier studies conducted within our forest sites were observing the initial mycorrhizal response to disturbance, but we are now seeing different results as the ecosystem settles into a new equilibrium.

What is interesting to consider is that leaf litter P, which should reflect nutrient availability, was unaffected by limestone addition between Control and Lime plots, despite large decreases in AM colonization in the Lime plots relative to Control. We previously found reductions in phosphatase enzyme production in limestone addition plots, which suggests reduced dependency on organic P and, thus, greater inorganic P availability that is biologically realized (Carrino-Kyker et al., 2016). However, these data on phosphatase enzymes coincided with direct measurements of inorganic P that were reduced with limestone addition. We suggested that this was due to greater assimilation by soil organisms and plants at the more neutral pH (which is supported in the literature; e.g., Hinzinger, 2001), but the leaf litter P results of the current study do not support this unless the trees are allocating the assimilated P to a different plant organ and, thus, have not been measured in our current or previous studies. If P assimilation does increase with limestone addition, but the assimilated P is being stored in organs other than the leaves, that could explain the reduced AM colonization we see here where the trees may no longer rely on their AM fungal associations for P acquisition. However, the declines in AM colonization in limestone addition plots may also reflect a different underlying nutrient limitation not measured within our study. One effect of acidic deposition is loss of base cations, including Ca, from soil, thereby reducing soil fertility. Reductions in AM colonization with limestone addition could reflect a potential role for AM fungi in Ca acquisition in these forest systems, as it has been shown in apple trees (Fu et al., 2023). Ca increases in Lime plots would therefore reflect a reduction in the mutualism once the underlying nutrient limitation has been overcome, similar to reductions in AM colonization following P addition as described in many studies. However, the relationship between AM colonization and Ca content of forest soils will require further study.

Does basal area growth differ by nutrient manipulation? Does this response vary by mycorrhizal association or species?

In the first half of our study, plots that received only TSP had the greatest increase in mean RBAI. Both Lime and X-Trt plots did not have significantly greater mean RBAI values than Control, but they were slightly elevated. P addition alone was the clear driver of growth during this period and was not as impactful when combined with limestone. Interestingly, mycorrhizal association was significant for determining response to treatment. At this association level, it appeared as if AM trees responded to treatment, but ECM trees did not. However, further breaking it out by species revealed that both AM and ECM trees did respond, but only one from each association: sugar maple and American beech, respectively. Both of these species followed the same trends as the overall analysis: mean RBAI was only significantly increased in TSP plots. American beech trees did seem to have a slightly negative growth response to limestone additions, as their mean RBAI values decreased from Control in both Lime and X-Trt plots, although these differences were not significant. This result is not surprising because American beech trees are often distributed on soils with lower CaO content (van Breemen et al., 1997). This distribution is thought to result in part from competition with sugar maple trees, which are stronger competitors than American beech under calcareous conditions (Kobe, 1996; van Breemen et al., 1997). Adding limestone to our plots may have allowed sugar maple to out-compete American beech trees, resulting in the suppressed growth response of American beech in Lime and X-Trt plots. Sugar maple trees did see slightly elevated mean RBAI values above Control in these plots, but this result was not significant, nor was it as large a response as in the TSP plots.

In the second half of our study, plots that received a combination of limestone and TSP had a significantly greater mean RBAI value than the Control. P addition alone did not significantly raise the mean RBAI despite having a very similar mean RBAI to the X-Trt plots. Mycorrhizal association was again a significant factor, as only AM trees responded to treatment. In this period, however, red maple responded while sugar maple did not. Notably, American beech trees did not have any growth responses to treatment in the second half of the study. This may be due to an adjustment to the treatments but is most likely a result of the increasing presence of beech leaf disease, first observed in 2014 at the Holden Arboretum, where two of the three study sites are located (Burke et al., 2020; Carta et al., 2020). By the end of the study, nearly 97% of accessioned American beech trees on Holden property showed symptoms of beech leaf disease, potentially inhibiting these trees' response to nutrient additions within our experiment (Burke et al., 2020). Beech leaf disease is also documented at the third forest in our study, Squire Valleevue and Valley Ridge Farms. Since 2015, nearly 22% of American beech trees in our plots have died (B. Shepherd, personal communication).

In light of beech leaf disease, one likely explanation for red maple's late response to treatment is that the loss of American beech trees led to increased light availability. Stebbins Gulch, a nearby forest at the Holden Arboretum, had a canopy gap of just over 8% in 2015. But in 2021, after the onset of beech leaf disease, the canopy gap reached a peak of nearly 16% (M. Watson, personal communication). As a plastic species considered both early and late-successional, red maple trees are able to grow quickly in smaller forest gaps (Abrams, 1998; Hibbs, 1982; Lawrence et al., 2017). This is demonstrated within the forests studied here, as red maple has begun to fill in the canopy gaps in Stebbins Gulch (M. Watson, personal communication). Previous research has found that sugar maple trees are more sensitive to nutrient stress than red maple trees and that this stress is easily exacerbated in high light conditions (St. Clair & Lynch, 2005). In the beginning of this study, sugar maple trees were relying on the treatments to alleviate their stresses, while red maple did not respond. Later, when American beech trees began to die and light became more available, the stress of growing on acidic soils may have worsened for sugar maple trees, potentially explaining the lack of response of sugar maple in the second half of the study. Red maple trees, on the other hand, were likely not nearly as nutrient-limited as sugar maple trees because they are tolerant of nutrient-poor situations (St. Clair & Lynch, 2005). After 2015, when light limitations were alleviated, red maple trees likely took advantage of the available nutrients and increased light availability, growing more in response. Similar results have been found in ecosystems facing beech bark disease (Lawrence et al., 2017).

Throughout the entire study, only half of the tree species we examined responded to treatment. This could be due to the individual species' tolerance of nutrient-poor sites, but it could also be because our sample size for these species was much smaller than that of the other species. Shagbark hickory, black cherry, and red oak made up approximately 1%, 3%, and 9.5% of the trees within our plots that survived the entire 12 years, while red maple, sugar maple, and American beech comprised around 12%, 43%, and 21% of the trees studied here. Future studies should be conducted on these species individually to better assess their responses to pH and nutrient manipulation.

Furthermore, mycorrhizal association was not always an accurate predictor of tree response. Analyzing RBAI responses to treatment at this level meant that the changes in American beech RBAI would have been missed entirely. Additionally, red maple trees would have been characterized as responsive to treatment in the first half of the study when they were not. This emphasizes the importance of species-level analyses, much like the findings of Lance et al. (2020).

Does response to treatment change with tree size?

Throughout each time point in our study, only small trees responded to treatments (DBH < 29.9 cm). Larger trees (DBH > 30.0 cm) did not respond to treatment at any point. Similar findings occurred at MELNHE. Goswami et al. (2018) found a P-limitation within their forest stands, but later, Hong et al. (2022) found greater mean RBAI in plots that received N-additions rather than P-additions. This discrepancy, according to Hong et al. (2022), results from the size of the trees studied. Goswami et al. (2018) tracked the growth of smaller trees (mean DBH of 18 cm) while Hong et al. (2022) tracked the growth of larger trees (mean DBH of 24 cm). Our size 3 and 4 classifications were much larger (30.0–59.9 cm and ≥60.0 cm). Perhaps a similar result is occurring within our forests, where smaller trees are P-limited and larger trees are not. Additionally, large trees could have widespread root systems that do not stay within the confines of our plots. Our plots measure 20 × 40 m and are at a minimum of 10 m apart. However, mature red maple can have roots measuring 20–25 m in length (Wilson, 1964) and red oak can have a maximum of 15 m long roots (Stone & Kalisz, 1991). If any mature trees are on the edges of our plots, their root systems could spread to neighboring plots and access the treatments within those soils.

CONCLUSION

The northeastern hardwood forests studied here show significant responses to P addition, suggesting that our forests are P-limited, although this limitation may be present for only smaller tree size classes. Previous studies that did not find evidence for P limitations may not have taken place over long enough periods of time for the effects of this limitation to be observed. Mycorrhizal fungi have been shown to alleviate some of the stresses associated with nutrient limitations (Burke et al., 2021; DeForest et al., 2012); however, we saw reductions of AM colonization where limestone was added to raise soil pH. This suggests that AM colonization itself may not be responding to P or N limitations within our system, but possibly to base cation loss from forests that results from acidic deposition. Observed large decreases in American beech populations due to beech leaf disease have led to significant reductions in canopy cover and could be impacting the growth responses of other tree species, as plastic species such as red maple began to take advantage of the increased light availability while sugar maple slowed in growth. One important conclusion from this dataset is that species-specific information is needed to further understand how a forest might respond to acidic deposition and disease introduction, as well as changes in nutrient availability across the landscape. Simply grouping trees into mycorrhizal associations might lead to interesting insights at a functional group level, but our study demonstrates that these groups might not be representative of individual species responses. We suggest future research be conducted within systems affected by beech leaf disease to better understand individual species responses and untangle the nutrient dynamics of a forest affected by disease and acidic deposition.

ACKNOWLEDGMENTS

We would like to thank Sheryl Petersen and Emily Galloway for earlier efforts in data curation and management. In addition, we thank Mike Watson for his canopy assessments, Brianna Shepherd for her Fagus grandifolia mortality analysis, our Research Renegades volunteers for their long-term assistance with processing leaf litter collections, and many summer interns who collected DBH data. This work was supported by funding from the Holden Arboretum Trust, the Corning Institute for Education and Research, and the R. Henry Norweb, Jr. Fellowship for Scientific Research in Horticulture. Initial funding for this work was supported by the National Science Foundation (grant number DEB-0918167).

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

DATA AVAILABILITY STATEMENT

Primary data and R code (Jahn et al., 2024) are available from Zenodo: .

References

Aber, J. D., C. L. Goodale, S. V. Ollinger, M.‐L. Smith, A. H. Magill, M. E. Martin, R. A. Hallett, and J. L. Stoddard. 2003. “Is Nitrogen Deposition Altering the Nitrogen Status of Northeastern Forests?” BioScience 53(4): 375–389. [DOI: https://dx.doi.org/10.1641/0006-3568(2003)053%5B0375:INDATN%5D2.0.CO;2].