Linking Soil CO2 Efflux to Individual Trees: Size-Dependent Variation and the Importance of the Birch Effect

Soil CO2 efflux (FCO2) is a major component of the terrestrial carbon (C) cycle but challenges in explaining local variability hamper efforts to link broad-scale fluxes to their biotic drivers. Trees are the dominant C source for forest soils, so linking tree properties to FCO2 could open new avenues to study plant-soil feedbacks and facilitate scaling; furthermore, FCO2 responds dynamically to meteorological conditions, complicating predictions of total FCO2 and forest C balance. We tested for proximity effects of individual Acer saccharum Marsh. trees on FCO2, comparing FCO2 within 1 m of mature stems to background fluxes before and after an intense rainfall event. Wetting significantly increased background FCO2 (6.4 ± 0.3 vs. 8.6 ± 0.6 s.e. μmol CO2 m−2s−1), with a much larger enhancement near tree stems (6.3 ± 0.3 vs. 10.8 ± 0.4 μmol CO2 m−2s−1). FCO2 varied significantly among individual trees and post-rain values increased with tree diameter (with a slope of 0.058 μmol CO2 m−2s−1cm−1). Post-wetting amplification of FCO2 (the ‘Birch effect’) in root zones often results from the improved mobility of labile carbohydrates and further metabolization of recalcitrant organic matter, which may both occur at higher densities near larger trees. Our results indicate that plant-soil feedbacks change through tree ontogeny and provide evidence for a novel link between whole-system carbon fluxes and forest structure.


Introduction
Soil carbon dioxide efflux (F CO2 ) is the predominant contributor of CO 2 to the atmosphere from terrestrial ecosystems, with the balance between net photosynthesis and F CO2 largely determining whether a given ecosystem constitutes a net carbon (C) source or sink [1]. Recent decades have seen a shift in how F CO2 is conceptualized-from a flux largely reflecting ecosystem-specific decomposition and its response to soil temperature and moisture, to a process highly influenced by active and mutual exchanges between plants and soil biota [2]. Such feedbacks between plants and soil biota play a stabilizing role in numerous ecosystem functions and allow for plant regulation of biogeochemical cycles [3,4]; however, the specific role of biota is often neglected in studies of F CO2 conducted at higher levels of aggregation, because belowground sources are difficult to disentangle, and because abiotic signals, especially temperature, may be more clearly expressed on spatially integrated data [2]. As a result, the representation of plants and associated biota in ecosystem process models remains rather simplistic and temperature generally stands as the most common driving variable considered in ecosystem models of soil C processes [5][6][7]. However, without deeper insight into the biotic drivers of soil F CO2 , it will remain extremely challenging to link localized observations to broader scale, ecosystem-level fluxes and to develop process-driven predictions of future F CO2 dynamics [2,4].
Studies quantifying F CO2 in forest ecosystems commonly find high spatial variability that is not readily explained by edaphic or other environmental variables [8][9][10]. Spatial patchiness in F CO2 partially derives from the poor mobility of belowground C inputs through the soil matrix, with further temporal variability contributed by the responses of autotrophs and heterotrophs to short timescale (hourly to daily) meteorological variability [11,12]. A prominent phenomenon exemplifying the spatiotemporal complexity of localized F CO2 involves the dramatic spikes in F CO2 that follow rainfall events, also known as the 'Birch effect' [13,14]. The Birch effect is commonly attributed to the sudden mobilization of labile carbon inputs (e.g., energy-rich carbohydrates from litter and root exudates) by rainwater, which primes microbial assemblages, allowing further decomposition of lower-quality organic matter (e.g., from macroscopic necrotic tissues and humified organic matter) [14]. Increased osmoregulatory activity and rapid turnover of lysed microbes may also contribute to enhanced C fluxes [14]. Linking biotic sources of F CO2 to the broad-scale outputs of ecosystem models may therefore require attention to high temporal variability in meteorology and could further benefit from spatially explicit consideration of C inputs [15,16]. F CO2 depends directly on primary productivity, and, surprisingly, productivity effects can be even larger than temperature effects [17,18]. Tree girdling experiments and isotopic tracing measurements suggest that as much as 50% of CO 2 emitted from forest soils is sustained by recently produced photosynthates [19,20], implying that physiological changes occurring in tree canopies likely impact F CO2 . The ability to isolate the contribution of individual trees to F CO2 has further been demonstrated in a sparsely-treed savanna, where the diurnal pattern of F CO2 in individual tree root-zones was larger in amplitude than could be explained by temperature, and lagged the diurnal up-regulation of photosynthesis, corresponding to the transport time of sugars from foliage to roots [21]. Recognizing the potential to isolate tree-level F CO2 motivates further efforts to link such observations to traits governing the quantities of C exported from trees to external C sinks [4].
A number of recent studies have documented detectable local effects of tree proximity and even tree species on F CO2 [3,[22][23][24], but the hypothetical mechanisms behind these effects are numerous and remain poorly resolved. Studies have also described effects of tree size on F CO2 in forest ecosystems, generally finding higher F CO2 in the immediate neighborhood of larger trees [25][26][27][28][29][30]. However, counterexamples exist [10,31], and several chrono sequence studies have found reduced F CO2 with stand age in even-aged plantations [32,33], or no consistent relationship [34]. Rodríguez-Calcerrada et al., [35] tested for associations between visible crown health indicators and F CO2 in an open woodland in Spain but found that increased plant recruitment near declining trees offset potential reductions in F CO2 related to tree decline. The character of C supply could theoretically shift as trees age from active extrusion of labile root exudates by more productive trees to less bioavailable necrotic tissues, but the connection between local tree effects on F CO2 and tree senescence or health status has received little attention.
Age-related trends in tree physiology are particularly large and well-documented in Acer saccharum Marsh. (hereafter sugar maple), a dominant tree in northern hardwood forests in Eastern North America. Leaf-level photosynthetic capacity peaks at intermediate sizes and declines later in ontogeny [36], and there is evidence for declines in whole-tree leaf area through ontogeny [37]. These trends are expected to constrain the amount of C available for belowground allocation, consistent with age-related declines in concentrations of non-structural carbohydrates in other sinks, such as wood [38]. As the capacity to replace fine roots declines with overall function, roots are more likely to be constructed to last longer, have increased C:N ratios, and be more resistant to decomposition [39]. If accumulation of high-C:N organic matter in tree root zones is an important chemical change that occurs through ontogeny, then the moisture-induced priming of microbial communities (Birch effect) may likewise vary through tree ontogeny. Additional processes associated with large old trees that may affect F CO2 include the potential for warmer soils beneath sparser canopies and increased maintenance respiration of older root tissue, Soil Syst. 2021, 5, 7 3 of 11 paralleling age-related trends in leaf respiration [40]. The magnitude of Birch-effect F CO2 pulses could be more pronounced near large trees, but studies to date have not examined potential local tree size influences on Birch effect peaks in F CO2 .
In the present study we collected spatially explicit F CO2 data in a large mapped forest plot to address the following questions: (1) Does F CO2 in the immediate vicinity of adult sugar maple stems differ from background soil flux? (2) Is there detectable tree-to-tree variation in root zone F CO2 , and if so, is such variation dependent on tree diameter, growth rate, or health status? (3) How does F CO2 in background vs. root-zone locations respond to soil conditions, in particular wetting following an intense rainfall event?

Study Site
The Haliburton Forest Dynamics Plot (HFDP), is located within Haliburton Forest and Wildlife Reserve Ltd., Ontario, Canada (44 • 55 N, 78 • 45 W), and belongs to the CTFS-ForestGEO global network of large-scale forest research plots [41]. HFDP encompasses several forest community types characteristic of the Great Lakes-St. Lawrence region. Total plot area is approximately 13.5 ha, of which approximately two thirds is occupied by a stand of shade-tolerant hardwoods. These interior, upland communities are dominated by sugar maple (Acer saccharum Marsh.), followed by American beech (Fagus grandifolia Ehrh.), and yellow birch (Betula alleghaniensis Britt.); the sub-canopy species striped maple (Acer pensylvanicum L.) is also abundant. Soils in the Haliburton region are shallow, predominantly sandy loams derived from till of the underlying granitic bedrock. Annual precipitation is approximately 1050 mm, with a mean annual temperature of 5 • C [42].

Data Collection
Flux collars consisted of 5-cm-long cross-sections of 10-cm diameter PVC plumbing, beveled with a power sander. The beveled edge of the collar was twisted into soil just enough to hold firmly (~1 cm); extra care was taken to ensure that collar insertion did not damage tree root tissues. If obstructions were encountered or a root was damaged, the collar was relocated within 20 cm of its initial location. Collars were always allowed to settle for at least 48 h before sampling.
Background sites were selected to ensure that locations were displaced at least two meters from any trees ≥10 cm diameter at breast height (dbh, measured using a diameter tape at 1.3 m height). Each background site consisted of a set of three collars, arrayed in a triangle with an approximate distance of one meter between collars. Background sites were distributed among 21 fixed points on a 57-m grid, with nine more randomly selected locations ( Figure 1); a combination of uniform and random points was used to achieve broad coverage of spatial variation yet ensure that background estimates were not biased by unobserved factors with spatially regular distributions. The locations of background sampling locations were adjusted slightly upon collar installation to exclude local low points with a tendency to accumulate water and locations with shallow soils or exposed bedrock that prevented insertion of flux collars. Patches of understory vegetation may also contribute to soil F CO2 [35], and so were also avoided. After study trees were selected, six soil efflux collars were inserted at a distance of approximately one meter from the base of each tree. Collars were distributed uniformly around each tree, except where obvious visible obstructions (vegetation, rocks, decayed logs) interfered. Several tree-level measurements were made to assess whether tree vitality influenced root-zone FCO2. Tree dbh was recorded at the time of FCO2 sampling using a diameter tape and compared to a former census measurement collected in the year 2011 to estimate mean annual growth increment over the two-year period (radial increment in cm/y). Crown transparency measurements from a maple health survey conducted in July 2009 with moosehorn densiometers [43] were also used as a health status indicator for sampled trees.

Pre-and Post-rain Sampling Intervals
The first round of sampling was done from 17-19 August 2013, with a subset of ten background and two root zone locations resampled on August 24th. A 42.2-mm rain event occurred on August 25th (Figure 2, Historical Climate Data, Environment Canada, Government of Canada) followed by a second round of sampling between August 27th and 29th. We were additionally able to validate the correspondence between temperature data with regional weather station and local temperatures at the soil surface with a network of 46 temperature loggers (Logtag, Auckland, New Zealand), which recorded soil temperatures just below the organic layer, spaced on a 28-m square grid throughout the study site. Mean daily soil surface temperatures and mean daily air temperatures from the regional station day were highly correlated (r = 0.94, data not shown). Root zone F CO2 observations were collected from trees that were well spaced from neighboring individuals (minimum distance 3 m) on level terrain. Sampled focal trees varied from 10 to 55 cm dbh (corresponding to a crown radius of~2 to 5 m). The majority of sampled trees were located on the interior plateau at the center of the forest plot, where composition is dominated by Acer saccharum and individual trees are well spaced (Figure 1). Canopy cover in the area is dense and spatially uniform and leaf-litter collection traps indicate that organic inputs in this section are also spatially uniform (data not shown). Measuring trees in the same neighborhood had the added advantage of minimizing underlying variation in soil drainage, texture, depth, and chemistry.

Data Processing and Quality Control
After study trees were selected, six soil efflux collars were inserted at a distance of approximately one meter from the base of each tree. Collars were distributed uniformly around each tree, except where obvious visible obstructions (vegetation, rocks, decayed logs) interfered. Several tree-level measurements were made to assess whether tree vitality influenced root-zone F CO2 . Tree dbh was recorded at the time of F CO2 sampling using a diameter tape and compared to a former census measurement collected in the year 2011 to estimate mean annual growth increment over the two-year period (radial increment in cm/y). Crown transparency measurements from a maple health survey conducted in July 2009 with moosehorn densiometers [43] were also used as a health status indicator for sampled trees.

Pre-and Post-Rain Sampling Intervals
The first round of sampling was done from 17-19 August 2013, with a subset of ten background and two root zone locations resampled on August 24th. A 42.2-mm rain event occurred on August 25th (Figure 2, Historical Climate Data, Environment Canada, Government of Canada) followed by a second round of sampling between August 27th and 29th. We were additionally able to validate the correspondence between temperature data with regional weather station and local temperatures at the soil surface with a network of 46 temperature loggers (Logtag, Auckland, New Zealand), which recorded soil temperatures just below the organic layer, spaced on a 28-m square grid throughout the study site. Mean daily soil surface temperatures and mean daily air temperatures from the regional station day were highly correlated (r = 0.94, data not shown).
FCO2. To correct for temperature effects we modeled the temperature response of FCO2 using an exponential equation (FCO2 = a*exp[b * Tsoil]), where Tsoil = mean soil temperature to 10-cm depth, a and b are fitted parameters) to the full dataset of soil collars. Q10 provides a concise indicator of the temperature sensitivity of a FCO2 temperature responses curve and was estimated to allow for comparison among similar temperature responses analyses (Q10 = exp[b * 10]). A t-test was conducted to compare soil temperatures during prerain sampling and post-rain sampling.

Data Processing and Quality Control
In total, 30 background and 18 root zone fluxes were sampled before wetting and 14 background and 22 root zones after wetting. Soil CO 2 efflux (F CO2 ) observations were made with a portable infrared gas analyzer (IRGA) with soil chamber attachment (Li6400 with Li-6400-09 attachment, LiCor, Lincoln, Nebraska). Mean temperature of the top 10 cm of soil was also recorded with a Li-6400-09TC thermocouple probe. The IRGA cuvette was sealed to the collar and approximately 30 to 60 s were allowed for CO 2 readings to stabilize before measurements. At least three F CO2 readings were logged, spaced at 30-s intervals, which were averaged prior to analysis. If collars demonstrated extreme values (greater than two standard deviations from the mean, or >50 µmol m −2 s −1 ), the collar was removed and data from this location were not integrated into the analysis (this was done for 4 out of 278 collars).

Data Analysis
A two-way analysis of variance (ANOVA) was conducted to compare F CO2 of background to root zone locations, before and after rainfall. Where duplicate observations were available, the average of the two dates was used. For root zone F CO2 values associated with tree stems, each tree's F CO2 was calculated by averaging over its associated collars prior to analysis.
Increases in soil temperature characteristically result in multiplicative increases in F CO2 . To correct for temperature effects we modeled the temperature response of F CO2 using an exponential equation (F CO2 = a*exp[b * T soil ]), where T soil = mean soil temperature to 10-cm depth, a and b are fitted parameters) to the full dataset of soil collars. Q 10 provides a concise indicator of the temperature sensitivity of a F CO2 temperature responses curve and was estimated to allow for comparison among similar temperature responses analyses (Q 10 = exp[b * 10]). A t-test was conducted to compare soil temperatures during pre-rain sampling and post-rain sampling.
The statistical residuals from the exponential model temperature response model provided temperature-corrected F CO2 indices, which were then compared to tree-level characteristics. To assess among-tree variation in root zone F CO2 , an additional one-way ANOVA with tree as a factor was performed, using temperature-corrected F CO2 values for pre-and post-precipitation root zone fluxes. Simple linear regressions were conducted to determine whether tree size, recent growth, or crown transparency were statistical predictors of root zone respiration. All analyses were performed in the 'stats' package of the R statistical programming language [44].

Pre-vs. Post-Precipitation F CO2
Soil CO 2 efflux preceding the rainfall event averaged 6.43 (±0.31 s.e.) µmol CO 2 m -2 s -1 and 6.25 (±0.29 s.e.) µmol CO 2 m −2 s −1 for background and root zones, respectively, which were not statistically distinguishable. Following the August 25th rainfall event, F CO2 significantly increased to 8.61 (±0.64 s.e.) µmol CO 2 m −2 s −1 and 10.79 (±0.41 s.e.) µmol CO 2 m −2 s −1 for background and root zone fluxes (p wet < 0.001), respectively. A significant root zone by pre-/post-rainfall interaction term was also detected (p RZ*Wet = 0.005), indicating that root zones fluxes experienced a larger enhancement following the rainfall event relative to background conditions ( Figure 3). The statistical residuals from the exponential model temperature response model provided temperature-corrected FCO2 indices, which were then compared to tree-level characteristics. To assess among-tree variation in root zone FCO2, an additional one-way ANOVA with tree as a factor was performed, using temperature-corrected FCO2 values for pre-and post-precipitation root zone fluxes. Simple linear regressions were conducted to determine whether tree size, recent growth, or crown transparency were statistical predictors of root zone respiration. All analyses were performed in the 'stats' package of the R statistical programming language [44].

Temperature Effects on FCO2
In addition to increased soil moisture, soil temperatures were also higher during post-rain sampling (

Temperature Effects on F CO2
In addition to increased soil moisture, soil temperatures were also higher during post-rain sampling (14.8 (±0.09 s.e.) • C vs. 16.8 (±0.09 s.e.) • C; Figure 4a). We detected a statistically significant influence of temperature on F CO2 and this relationship was accurately captured with an exponential model (F CO2 = 0.6189 × e 0.1651 . Tsoil ; p < 0.001) (Figure 4b), From this relationship, we further estimated a Q 10

Tree effects on FCO2
Post-rain temperature-corrected root zone FCO2 was positively related to tree dbh (Figure 5b). Following rain, root-zone FCO2 increased at a rate of 0.058 μmol m −2 s −1 for every centimeter increase in tree dbh (p = 0.040, r 2 = 0.154). We did not detect a significant effect of tree size on pre-rain FCO2 (Figure 5a). The other metrics examined, radial increment and

Tree Effects on F CO2
Post-rain temperature-corrected root zone F CO2 was positively related to tree dbh (Figure 5b). Following rain, root-zone F CO2 increased at a rate of 0.058 µmol m −2 s −1 for every centimeter increase in tree dbh (p = 0.040, r 2 = 0.154). We did not detect a significant effect of tree size on pre-rain F CO2 (Figure 5a). The other metrics examined, radial increment and crown transparency, were not able to explain any variation in temperature corrected effluxes among trees either pre-or post-rain. Although tree size appeared to be the only metric to influence F CO2 , ANOVA results showed statistically significant variation among individuals in temperature-corrected efflux, both before and after the rainfall event (p < 0.001 for both ANOVAs) ( Figure 5), indicating additional among-tree variation.
Soil Syst. 2021, 5, x FOR PEER REVIEW 8 of 12 Figure 5. Relationships between individual tree characteristics and temperature-corrected soil CO2 efflux. Error bars represent standard error (n = 6) in each individual tree's root zone flux. Significant variation was detected among individual tree root zones both before (panels (a,c,e)) and after (panels (b,d,f)) wetting (pdry and pwet < 0.001). Post-rain variation in soil root zone fluxes was significantly related to tree dbh (solid regression line) but not to other variables (dotted regression lines).

Figure 5.
Relationships between individual tree characteristics and temperature-corrected soil CO 2 efflux. Error bars represent standard error (n = 6) in each individual tree's root zone flux. Significant variation was detected among individual tree root zones both before (panels (a,c,e)) and after (panels (b,d,f)) wetting (p dry and p wet < 0.001). Post-rain variation in soil root zone fluxes was significantly related to tree dbh (solid regression line) but not to other variables (dotted regression lines).

Discussion
Our objective was to determine whether the contribution of individual trees to soil F CO2 could be detected by comparing root zone F CO2 to background variation, and to Soil Syst. 2021, 5, 7 8 of 11 further link aboveground properties of A. saccharum trees to belowground C processes in a closed-canopy forest. Under typical conditions, root-zone F CO2 did not differ from background F CO2 but was significantly higher following an intense precipitation event. We also detected significant variation among individual root-zone F CO2 estimates, both before and after rainfall. Temperature-corrected F CO2 increased significantly with tree size, consistent with prior studies [25][26][27][28][29][30]; however, this pattern was only detected for post-precipitation F CO2 . We did not find any trend with tree health indicators (either recent tree growth increment or crown transparency), suggesting that the size-dependent pattern in F CO2 is not related to tree senescence.
Transient 'Birch effect' increases in F CO2 in response to wetting is a well-documented phenomenon [13,14]. Elevated F CO2 typically persists for several days after moisture increases and is mainly attributed to the improved mobilization of labile carbon compounds, which primes microbial communities for further degradation of less bioavailable substrates [14]. Root exclusion studies have also found that the Birch effect depends on root presence [45]. Our results indicate that the Birch effect is both more pronounced in the vicinity of tree stems and is greater in magnitude near larger trees. Prior studies examining tree-size effects on F CO2 have not explicitly tested for Birch-effect patterns, but Søe and Buchmann [26] found that a best-fitting model describing variation in F CO2 included both tree size (mean dbh of trees within 4 m radius), and volumetric soil moisture content; Schwendenmann and Macinnis-Ng [30] noted relatively dry soil conditions near the base of large trees that could enhance the gas diffusion, and so increase F CO2 .
Several mechanisms might contribute to the tree-size-dependent Birch effect detected in the present study. Mobility of labile C substrates, such as root exudates, is limited by diffusion, and these labile C compounds could thus accumulate in the immediate vicinity of their source [46]. Larger trees with larger canopies are generally more productive and may export larger quantities of carbohydrates to associated soil biota, which accumulate until rain-induced mobilization. It is furthermore likely that necrotic tissues such as dead fine roots, which are an important carbon source for heterotrophic respiration [47], occur at higher densities in the root zones of larger trees. These higher-density pools of low-quality organic matter are more accessible to microbial communities following wetting events. An additional possibility consists of a more biophysical mechanism, where larger canopy areas may intercept larger quantities of moisture, leading to higher stemflow and higher moisture near the bases of large trees [48].
Among-tree variation in root zone F CO2 was not significantly correlated with tree size prior to rainfall, but we did detect significant variation among individuals. This additional variation among locations could be explained by topography and unobserved edaphic variables, but also invites deeper explorations of how tree health status could also regulate soil F CO2 . For example, Hancock et al. [49] found that the early stages of infection of Fagus grandifolia by the invasive pathogen Neonectria (beech-bark disease) corresponded with sharply increased local F CO2 . Geddes et al. [50] reported anomalous ecosystem-level C losses from a stand co-dominated by Fagus grandifolia at HFWR and considered the onset of beech-bark disease a likely explanation. We speculate that pathogen infection may more generally contribute to high tree-to-tree variation in F CO2 and component processes.
Although biotic factors have been found to dominate soil F CO2 in studies conducted at comparable scales to the present analysis [17,18], we still found that correcting for temperature effects was essential for detecting tree size effects on soil F CO2 . Soil temperatures were clearly sensitive to changes in ambient air temperature over the duration of the study, with impacts on F CO2 (Figure 3). However, the estimated sensitivity of F CO2 to temperature in the present study was substantially higher than estimates from comparable forest soils (compare our estimated Q 10 of 5.8, to a prior estimate of 3.0 at HFWR [8], and 3.5 at a comparable site in New England [51]). This high sensitivity is almost certainly due to coupled increases in soil moisture and temperature, which likely had a compound effect on soil F CO2 . This observation further highlights the need to sample over wider ranges of variation in meteorological conditions, whereas studies of plant-soil systems often aim to narrow the range of environmental variation in an effort to improve their ability to detect specific biotic interactions.
The ability to anticipate ecosystem-level responses of soil respiration to the changing climate is still limited by our understanding of relationships between soil moisture and microbial respiration [16]. The Birch effect is frequently left out of ecosystem-level models of C exchange entirely [16], even though this has been shown to result in a~25% underestimation of soil C emissions, at least in northern deciduous forests [52]. Changes in the isotopic composition of C emissions over very short periods following moisture pulses indicate that inorganic C can also contribute to the Birch effect in some systems [53]. Uncertainty in these sources of C emissions limits the applicability of otherwise analytically tractable process-based models that incorporate the kinetics of C supply and microbe consumption [54]. Our results suggest that Birch effect fluxes may be specifically enhanced in forests with large, old trees, potentially indicating an underestimation of carbon losses from old-growth forests without explicit consideration.

Conclusions
The present study compliments prior work suggesting strong autotrophic control over F CO2 . Our work replicates prior studies indicating a size-dependent increase in F CO2 , while also demonstrating a strong ontogenetic trend in carbon flux associated with the Birch effect. Detecting tree size effects on soil CO 2 efflux invites the investigation of additional biophysical and biogeochemical relations that may be impacted by the ontogeny and/or allometry of tree root functioning and organic matter production. Unifying physiological mechanisms with ecosystem modeling efforts requires replication of comparable measurements across tree species and over a wider range of environmental variation, including rainfall intensity, rainfall frequency, and seasonal variation in belowground photosynthate allocation.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.