Increased leaf area index and efficiency drive enhanced production under elevated atmospheric [CO2] in a pine‐dominated stand showing no progressive nitrogen limitation

Enhancement of net primary production (NPP) in forests as atmospheric [CO2] increases is likely limited by the availability of other growth resources. The Duke Free Air CO2 Enrichment (FACE) experiment was located on a moderate‐fertility site in the southeastern US, in a loblolly pine (Pinus taeda L.) plantation with broadleaved species growing mostly in mid‐canopy and understory. Duke FACE ran from 1994 to 2010 and combined elevated [CO2] (eCO2) with nitrogen (N) additions. We assessed the spatial and temporal variation of NPP response using a dataset that includes previously unpublished data from 6 years of the replicated CO2 × N experiment and extends to 2 years beyond the termination of enrichment. Averaged over time (1997–2010), NPP of pine and broadleaved species were 38% and 52% higher under eCO2 compared to ambient conditions. Furthermore, there was no evidence of a decline in enhancement over time in any plot regardless of its native site quality. The relation between spatial variation in the response and native site quality was suggested but inconclusive. Nitrogen amendments under eCO2, in turn, resulted in an additional 11% increase in pine NPP. For pine, the eCO2‐induced increase in NPP was similar above‐ and belowground and was driven by both increased leaf area index (L) and production efficiency (PE = NPP/L). For broadleaved species, coarse‐root biomass production was more than 200% higher under eCO2 and accounted for the entire production response, driven by increased PE. Notably, the fraction of annual NPP retained in total living biomass was higher under eCO2, reflecting a slight shift in allocation fraction to woody mass and a lower mortality rate. Our findings also imply that tree growth may not have been only N‐limited, but perhaps constrained by the availability of other nutrients. The observed sustained NPP enhancement, even without N‐additions, demonstrates no progressive N limitation.


| INTRODUC TI ON
Increased atmospheric [CO 2 ] may result in higher net primary production (NPP) of forest ecosystems.Understanding when the extra carbon (C) fixed under elevated [CO 2 ] (eCO 2 ) is used for foliage production, thus increasing the amount of photosynthetic machinery, or to wood, thus supporting long-term C sequestration, is key to predicting how forests will function, and how much and how fast C will move between biosphere and atmosphere in future climates (De Kauwe et al., 2014;Walker et al., 2019).Both the magnitude of NPP response and its partitioning may vary with C limitation to growth relative to that imposed by other resources (Ellsworth et al., 2017;Franklin et al., 2012;Körner, 2006;Maier et al., 2022;McCarthy et al., 2010;Norby et al., 2010;Oren et al., 2001;Sigurdsson et al., 2013).For example, based on a global meta-analysis on nitrogen (N) mediated growth responses, eCO 2 increased tree biomass production by 32% overall; but responses of trees fertilized with N was higher (40%) than of those not provided with N (24%; Wang & Wang, 2021).Earlier findings at the Duke Free Air CO 2 Enrichment (FACE) experiment also indicated that the initial production enhancement under eCO 2 was maintained only under balanced nutrient additions (Oren et al., 2001) and varied spatially with native site quality (McCarthy et al., 2010).However, excluding the recent study on coarse roots (Maier et al., 2022), the effects of long-term N-additions on NPP and its partitioning among organs at this site have not yet been quantified.
Dynamics of accumulation live biomass of trees over stand development (Bormann & Likens, 1979) is the outcome of production of all organs and turnover of some (leaves, branches, fine roots, and non-lignified reproductive organs) and the recruitment and mortality of individual trees (curve 1 in Figure 1).Increased resource supply or improved climate may elevate the steady-state biomass is approaching (M Pot ; curve 2), the rate at which biomass increases toward M Pot (curve 3), or both (curve 4; see Groninger et al., 1999).Leaf biomass (or leaf area) shows similar dynamics but reaches a steady state much earlier than total biomass (Vose et al., 1994).Nutrient additions increase biomass accumulation rate (Albaugh et al., 2004;Trichet et al., 2008;Vose & Allen, 1988) and forests growing on better quality sites accumulate more biomass faster (curve 4; Samuelson et al., 2008;Vicca et al., 2012).In older stands, higher M Pot (curve 2) may be possible where increased resources prolong positive net biomass increment.In accruing stands, a directional climate change (increase in temperature), management (fertilization) or experiments (CO 2 enrichment), may result in both greater biomass accumulation rates and projected M Pot .It has been argued that CO 2 fertilization effects on tree/stand growth are most likely realized in young stands or on forest edges, where solar energy and soil resources are not fully utilized (Körner, 2006(Körner, , 2022) ) and the primary growth response to eCO 2 is accelerated stand development (curve 3).However, if eCO 2 enhances carbohydrate availability and facilitates acquisition or utilization of additional resources, both biomass accumulation rate and M Pot may increase (curve 4).
F I G U R E 1 Typical accumulation of live biomass in a forest stand, in a given environment, reflecting the difference between biomass growth and mortality (curve 1).Increased resource supply may increase the theoretical steady-state biomass (curve 2), the rate at which biomass increases toward steady state (curve 3), or both (curve 4).Note that the dynamics of biomass accumulation are shown for total biomass; foliage biomass shows similar trajectories but peaks earlier in time than total biomass.Letters a, b, and c in the main figure refer to the insets that show potential growth responses to elevated atmospheric [CO 2 ] (eCO 2 ) and nitrogen-additions (+N) across a range of leaf area index (L) values reflecting native site quality.Initial growth enhancement under eCO 2 (grey arrows, red symbols) relative to stands growing under ambient conditions (aCO 2 ; blue symbols) is mediated by both an increase in L and in growth per unit L with overall enhancement increasing with site quality (a).Progressive nitrogen limitation (PNL) where the initial enhancement is diminished over time (dashed black arrows; b).Nitrogen-additions may restore or increase the eCO 2 -induced growth enhancement (dashed black arrows, yellow symbols; c).
Response to increasing resource availability may include greater carbohydrate production per unit leaf area index (L) and/or a change in allocation of C between below-and aboveground.Fertilization on low fertility sites results in increased L and increased stem growth rate at comparable L (Albaugh et al., 2004).The primary reason for higher carbohydrate availability is an increase in L and canopy photosynthesis (Cannell & Dewar, 1994;Linder, 1987).However, as L increases in closed canopies, light interception increases proportionally less, and mean photosynthetic rate and production per unit L decrease (Norby et al., 2005;Vose & Allen, 1988;Waring, 1983).
Because fertilization-induced enhancement of leaf photosynthetic rate can be small and/or transient (Maier et al., 2008), increased stem wood production per unit L may mostly reflect increased fraction of carbohydrates used aboveground (Axelsson & Axelsson, 1986;Haynes & Gower, 1995).Similarly, from purely eco-physiological standpoint, the response of biomass production to eCO 2 reflects the net effects of a potential increase in photosynthetic rate at a given L, its change with increasing L, and changes in C allocation, all of which can be affected by nutrient supply (inset a, Figure 1; blue symbols represent different pre-treatment "base L" and native site quality and red symbols hypothetical responses to eCO 2 ).
When soil nutrient availability limits biomass production, the extra carbohydrates produced under eCO 2 may not be partitioned to biomass but to fast-turnover C pools (e.g., root exudates), in effect speeding up the 'atmosphere-to-atmosphere' cycling of CO 2 and decomposition of soil organic matter (Palmroth et al., 2006;Phillips et al., 2011;Prescott et al., 2020).Put differently, if tree growth is mostly N-limited, then following an initial, eCO 2 -induced, accelerated growth, reflecting increased N-uptake and/or -use efficiencies, soil N-availability becomes progressively limiting as N accumulates in plant biomass and soil organic pools (inset b; downward arrows indicate 'Progressive Nitrogen Limitation', PNL; Luo et al., 2004).
By extension, under these circumstances, the steady-state L is not affected by eCO 2 , but N fertilization (inset c; yellow symbols) may restore, and potentially increase, the eCO 2 -induced growth enhancement.However, growth enhancement may be sustained even in the absence of nutrient additions as increased C flux belowground under eCO 2 may increase root production, rooting depth, and rhizosphere C deposition, priming decomposition of older organic matter (Drake et al., 2011;Iversen, 2010;Maier et al., 2022;Norby et al., 2004).These processes may improve access to previously untapped nutrient resources, leading to 'progressive release from N limitation' (Walker et al., 2015).
Free air CO 2 enrichment (FACE) experiments enable evaluation of eCO 2 effects, and their interaction with other climatic factors, on ecosystem structure and function, thus providing insights into the mechanisms underlying the response.Norby et al. (2005) demonstrated that, while the increase in NPP under eCO 2 was similar among FACE sites in young forests, the role of structural changes (increases in L) in driving the response decreased with increasing degree of canopy closure.Some FACE experiments in closed canopy and/or mature forests showed the expected strong coupling between eCO 2 -induced enhancement of NPP and soil nutrient availability.For example, in a closed-canopy sweetgum (Liquidambar styraciflua) plantation, the initial growth enhancement under eCO 2 was manifested as higher light-use efficiency (NPP/amount of absorbed solar energy; Norby et al., 2005), but there was no long-term eCO 2 effect on NPP, and growth was deemed N-limited (Norby et al., 2010(Norby et al., , 2022;;Norby & Zak, 2011).
Duke FACE in a loblolly pine (Pinus taeda L.) plantation with broadleaved species found in the mid-to-lower canopy, was the longest-running FACE experiment to date (1994)(1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010) and the first to incorporate CO 2 -by-nutrient interaction into the experimental design (Oren et al., 2001).Nutrient additions at this site were superimposed onto variation in native site quality that resulted from minor topographical variations that generated subsurface flows (Schäfer et al., 2002) and were associated with differences in soil depth and rock content (McCarthy et al., 2010).Compared to the slightly elevated northern and southern portions of the site, the shallow valley in the middle supported higher pre-treatment growth rate and pine L (L P ; Oren et al., 2006).Although the initial L P was related to Nmineralization rate, likely reflecting the availability of other nutrients as well, some of the plots with higher N-mineralization rate were also located at lower elevations with deeper soils and greater water supply (McCarthy et al., 2010;Schäfer et al., 2002).
Elevated CO 2 increased total C flux belowground while N-additions under eCO 2 decreased it (Butnor et al., 2003;Drake et al., 2011;Jackson et al., 2009;Kim et al., 2017;Oishi et al., 2014;Palmroth et al., 2006).Fine-root NPP was 25% and biomass 24% higher under eCO 2 , and N-additions under eCO 2 resulted in a 13% decrease in fine-root biomass (Jackson et al., 2009;Pritchard et al., 2008).By the end of the CO 2 -enrichment period, aboveground pine biomass was 27% higher under eCO 2 , with no apparent interaction with N-supply, while that of broadleaved species was similar across treatments (Kim et al., 2020).Coarse-root biomass of pine and broadleaved species were 32% and 155% higher under eCO 2 , respectively, while the effect of N-additions on this quantity was small (Maier et al., 2022).
That the effect of N-additions was obvious on fast-turnover C pools but not on woody biomass is not surprising given that any addition to woody biomass is made to an already large and continuously accumulating pool.
Taken together, the native nutrient supply at Duke FACE was sufficient to sustain a substantial eCO 2 effect on NPP and standing biomass perhaps by allowing access to deeper and/or previously inaccessible nutrient pools in the soil (Drake et al., 2011;Maier et al., 2022;McCarthy et al., 2010).However, the recovery of eCO 2 -induced growth enhancement in the FACE prototype upon balanced nutrient fertilization (Oren et al., 2001) and the limited NPP response to eCO 2 in the low N-availability plots earlier in the study (McCarthy et al., 2010), put in the context of the prevailing N-limitation (Hungate et al., 2003;Luo et al., 2004), suggest Nlimitation to growth.Yet, the effects of N-additions on biomass production have not been assessed so far.
We assessed the eCO 2 × N effect on NPP (biomass production), by species group and organ, using a dataset extending 2 years beyond the termination of enrichment.We tested the following hypotheses: As projected based on the first 10 years of the experiment (McCarthy et al., 2010), eCO 2 will increase NPP more, and the enhancement ratio will be higher, in plots of higher native site quality (initial L) (H1).The effect of eCO 2 on NPP will be ephemeral in plots of low L, suggesting PNL (H2).Nitrogen-additions will counter PNL (H3).Higher NPP under eCO 2 , with or without N-additions, is driven by increases in both L and production efficiency (PE = NPP/L); while PE is expected to decrease as L and within-canopy shading increases (H4).Nitrogen additions under eCO 2 will change allocation of NPP (from below-to aboveground organs) rather than increase total production (H5).In addition, we expected the NPP response of broadleaved species to reflect a direct response to eCO 2 and N-additions and, perhaps even more, an increased light limitation imposed by the response of pine L (Kim et al., 2016).Post-eCO 2 enrichment, we expected pine NPP to decline in proportion to that in L (Kim et al., 2017).

| MATERIAL S AND ME THODS
The Duke FACE experiment was in the Blackwood Division of the Duke Forest in Orange County, NC (35°97′ N, 79°09′ W).The tract was homogeneous, relatively flat with moderately fertile soils (Oren et al., 2006).The site characteristics are described earlier, for example, in Maier et al. (2022) and an abbreviated description is provided here.The experiment was set up in a 90-ha loblolly pine (Pinus taeda, L.) stand planted in 1983 at 2.0 × 2.4 m spacing following a chop-andburn site preparation treatment.Individuals of 40 woody species, most commonly sweetgum (L.styraciflua L.), winged elm (Ulmus alata Michx.), dogwood (Cornus florida L.), and red maple (Acer rubrum L.) were found in the mid-to-lower canopy.Broadleaved species comprised 11% of basal area in 2010.The climate is warm and humid in the summer and moderate in the winter with a mean annual temperature of 15.5°C.Mean annual precipitation is 1145 mm.The soils are predominantly Enon silt-loam, characterized as moderately low fertility acidic clay-loam.Soil pH is ~6.0, and foliar concentrations of nitrogen (N; ~1.1%) and phosphorus (~0.3%) are average for midrotation loblolly pine stands in the region (Albaugh et al., 2010).
The FACE experiment consisted of eight circular experimental plots (30 m in diameter).A prototype (FACE P ) plot with an elevated [CO 2 ] (eCO 2 ) and a reference plot with ambient [CO 2 ] (aCO 2 ) were established in 1993, and six plots (three eCO 2 and three aCO 2 ; replicated FACE) were added in 1996.The eCO 2 plots received additional CO 2 to maintain atmospheric [CO 2 ] at ambient + 200 μL L −1 , while the aCO 2 plots received only ambient air (Hendrey et al., 1999).
Enrichment commenced in 1994 at FACE P and 1996 (at the end of the growing season) at FACE.
In 1998-2004, the FACE P plots received nutrient additions.Each plot was split in half by an impermeable barrier down to 70 cm and one-half of each plot was fertilized annually.Based on foliar nutrient analysis, the N application rate was set to either 5.6 or 11.2 g N m −2 (as NH 4 NO 3 or urea) balanced with other nutrients (Albaugh et al., 1998).In 2005, the rest of the FACE plots were halved, and annual fertilization commenced in all plots (including FACE P ) with only NH 4 NO 3 at the rate of 11.2 g N m −2 on one-half of each plot.
In the analyses on N-addition effects, we included data collected from the replicated FACE (2005)(2006)(2007)(2008)(2009)(2010).We were unable to detect a legacy effect of the earlier balanced nutrient additions in the data from the original FACE P plots.The CO 2 and N-additions created 4 × 4 half-plots representing four treatment combinations: aCO 2 (AR), aCO 2 + N (AN), eCO 2 (ER), and eCO 2 + N (EN).The experimental unit was a 'whole plot ' through 2004, and a 'half-plot' from 2005, both referred to as 'plot'.The treatments continued through October 2010.Trees in one half of each plot, consisting of a quarter plot of each N treatment, were harvested in early 2011 and trees in the other halves were left intact for two more years (2011-2012) for continuous measurements (Kim et al., 2020).All the data used in this paper are publicly available (Oren et al., 2023).

| Environmental variables
Air temperature and relative humidity were measured in the upper third of the canopy in each plot (Vaisala HMP35C and HMP45C;Helsinki,Finland).On the central tower above the canopy of plot 4, a sensor (Q190; LiCor, Lincoln, NB, USA) for measuring photosynthetically active radiation and an automated system (tipping bucket TR-525USW; Texas Electronics, Dallas, TX, USA) for measuring precipitation were mounted.All sensors were sampled every 30 s, and 30-min averages were logged (CR21X or CR23X; Campbell Scientific, Logan, UT, USA).Beginning in 1997, volumetric soil water (θ; m 3 water m −3 soil) of the upper 30 cm soil layer was measured continuously at four locations in each of the replicated FACE plots, and from 2001 in the FACE P plots using various sensors (Campbell Scientific, Logan, UT, USA, and Delta-T Devices, Cambridge, UK).
Soil moisture estimates for each sensor were rescaled to measured porosity (representing saturated water content) and hygroscopic minimum (Oishi et al., 2008).

| Estimates of leaf area index and NPP
As described in McCarthy et al. (2007) leaf production (all biomass production estimates are given as dry mass, g m −2 year −1 ) and projected leaf area index (L, m L 2 m G 2 ) were estimated using leaf litterfall samples collected bi-weekly/monthly from 12 litter baskets (0.16 or 0.22 m 2 ) in each plot.For our analysis, pine leaf litterfall data prior to 2004 were available only from six of the eight plots, and pine L estimates for those plots/years were taken from McCarthy et al. (2007McCarthy et al. ( , 2010)).For broadleaved species, peak L was estimated as total annual leaf litterfall mass (leaf production) multiplied by specific leaf area (m 2 g −1 ).For pine, to account for leaf longevity of ~19 months, L estimates were based on 2 years of litterfall (McCarthy et al., 2007(McCarthy et al., , 2010)).Reproductive biomass and production were also estimated based on litterfall collections (McCarthy et al., 2010).
Annual tree diameter and height measurements, combined with allometric relationships developed for trees and root systems harvested in 2011 (Kim et al., 2020;Maier et al., 2022), were used to estimate woody biomass and biomass production of each tree (stem, branches, and coarse roots [>2 mm diameter]).Woody biomass production represents our longest time series (covering all plots).
Branch production estimates included branch biomass turnover, calculated based on annual measurement of the height of the base of live crown and vertical branch biomass distribution.For pine, biomass from individual branch sampling and whole-tree branch diameter measurements of harvested trees were used to calculate vertical branch biomass distribution.For broadleaved species, the branch biomass was vertically distributed into three parts (top, middle, and bottom crowns) as whole-tree branch diameter measurements were not available.We used previously published estimates of fine-root biomass and production (Jackson et al., 2009;Pritchard et al., 2008) and partitioned stand-level estimates between species groups based on coarse-root estimates.Allocation ratios were calculated as each component NPP divided by the sum of all components.Mortality rate (% year −1 ) was calculated as: Mortality = (N ref1 −N ref2 )/N ref1 × 1 00/interval, where N ref1 is the number of reference trees in year 1 and N ref2 is the number of the reference trees survived until year 2.
Two estimates of annual biomass production (Clark et al., 2001) were calculated: (1) annual change in the biomass pool (ΔM = M 2 -M 1 ), where M 2 and M 1 are live biomass in year 2 and year 1, respectively, and (2) NPP calculated for trees that are alive during two consecutive, annual inventories plus ingrowth (recruitment of individuals reaching the minimum measurable size) during the period.That is, ΔM represents net biomass increment (accounting for mortality), while NPP denotes biomass production of living trees.Our analyses focused mainly on the NPP estimates and the role of L versus PE (PE = NPP/L, where NPP is total NPP for each species group) in driving NPP.To explore the potential effects of stand developmental stage on biomass production, we also performed a classical growth analysis (Evans, 1972).We calculated relative growth rate as RGR = ln(M 2 )-ln(M 1 ), its determinants, leaf area ratio (LAR = L/M) and biomass increment per unit leaf area (unit leaf rate, ULR = ΔM/L), and compared those at a common live biomass (a proxy for tree size) across treatments.
Enhancement ratios for each component, woody NPP (NPP W ; stem, branches, and coarse roots), ANPP, NPP, and L were calculated as 'X' at eCO 2 /'X' at aCO 2 for each plot pair (based on the pairing of plots, i.e., each eCO 2 plot to an aCO 2 plot in the experimental design).
To estimate annual enhancement, we averaged the plot-specific ratios (n = 4) and calculated standard error (SE).In addition, we calculated an enhancement ratio (at a standard L) for NPP W based on the relationships between NPP W and site quality under aCO 2 and eCO 2 .The ratio was computed as actual NPP W under eCO 2 divided by its expected value at the same L but under aCO 2 .If no relationship under aCO 2 was found, mean NPP under aCO 2 was used as the expected value.We used three proxies for site quality of each plot: N-availability index as reported in McCarthy et al. (2010), pine basal area (B P96 ), and pine leaf area index (L P96 ); the latter two estimated in 1996, the year the CO 2 enrichment began in the replicated FACE.

| Statistical analyses
We tested CO 2 effects on NPP and L using data from 1997 to 2010 productivity.In addition, we analyzed treatment differences in ANPP-to-L relationships annually using analysis of covariance.We used linear regressions to evaluate trends in average enhancement ratios.We used a t-test to evaluate among-treatment differences in temporally averaged allocation ratios.
We used Shapiro-Wilk tests and Quantile-Quantile plots (R Core Team, 2021) to test for normality of each measured response variable, and data were transformed before model fitting when appropriate and back transformed as needed (Jørgensen & Pedersen, 1998).
To evaluate temporal autocorrelation in time series of response variables (repeated measurements), we first fitted models with an autocorrelation function (acf; nlme package).We also fitted nlme models with and without a first-order autoregressive covariance structure (AR(1)).Based on the acf analyses and AIC fit statistics (comparing models with and without AR(1)), we found that temporal autocorrelation in all of the response variables was too small to justify using a more complex model and would not change the overall conclusions.Therefore, we fitted linear mixed models using the lme4 package and reported the test results in analysis of variance tables (ANOVA, type III, lmerTest with Satterthwaite approximation for degrees of freedom).Significant treatment or interaction effects were further analyzed using multiple comparisons with Tukey's test using emmeans package (R Core Team, 2021).Adjusted SEs of the models were calculated with the interactionMeans function in the package phia (De Rosario-Marinez, 2015).

| RE SULTS
Aboveground net primary production (ANPP), NPP, and leaf area index (L) increased for the first 5 years of the study period as canopy volume was filling up and, again, after two major disturbances in 2002, a severe drought and an ice storm (Figure 2; NPP estimates by organ are shown in Figure S1).After the plots were  S2 and S3).
After the treatments ended and half of the biomass was harvested in early 2011, estimates of pine ANPP, NPP and L decreased for 1 year and then recovered in the second year, likely in response to an increase in growing space and light availability for the remaining trees.

| Spatial variation in native site quality and NPP enhancement under elevated [CO 2 ]
Of our hypotheses (tests summarized in Figure 3), the first (H1) stated that the variation in the response of NPP to eCO 2 will reflect spatial variation in native site quality.We found that the absolute response of mean annual woody NPP (NPP W ) to eCO 2 increased with all three measures of site quality; N-availability index, initial pine basal area (B P96 ), and pine leaf area index (L P96 ; p ≤ .016for CO 2 × L P96 ; Figure S4).While this interaction produced an increasing enhancement ratio (i.e., the ratio of the lines shown in Figure 3a) with site quality, the relationship was driven by one plot.Indeed, the mean plot-specific annual enhancement ratio (data points in Figure 3a; averaged over annual estimates of actual NPP W under eCO 2 divided by the expected value at the same L but under aCO 2 ) was unrelated to L P96 (p = .469)and the other site quality proxies (not shown; p ≥ .173).
Progressive N limitation (PNL) would result in a decline in the plot-specific enhancement ratio of NPP over time, which we expected to be most noticeable in plots with low L P96 (H2).While the enhancement ratio of NPP W in each plot varied from year to year, contrary to our expectation, none showed a decreasing ratio (Figure 3b, p ≥ .387).Note that, in these analyses, we used NPP W , the production estimate with the longest record available from all plots.The fraction of NPP W of total NPP (average ± SE over 1997-2010) was 78 ± 1.0% under eCO 2 and 76 ± 0.9% under aCO 2 (p = .072).

| Overall treatment effects on NPP
Over 1997-2010, pine ANPP was, on average, 35% (p = .005;ranging from 26% to 54% across years) and stand ANPP 33% higher (p = .007;25%-48%) under eCO 2 compared to aCO 2 (% changes are based on unadjusted treatment means; p-values are taken from Table S1).Broadleaved ANPP was similar under both CO 2 treatments (p = .446).Enhancement of NPP for pine was of similar magnitude to that in ANPP (38%; p = .004;28%-54%) but larger for broadleaved species (52%; p = .051;37%-68%), reflecting a large coarse-root response.In both species groups, there was also a several-fold enhancement in reproductive NPP but, because the production of reproductive organs was at least an order of magnitude smaller than NPP of any other component (Figure S1 and Figure S3), the effect on total NPP was small.Because of the dominance of pine, stand NPP was 38% higher under eCO 2 (p = .003;30%-53%).
The effect of N-additions (in 2005-2010) on pine ANPP and NPP were similar under both CO 2 treatments (Figure 4a-d; Table S3; based on models without L) demonstrating no CO 2 xN interaction.
While N-additions increased pine L only under aCO 2 (Figure 4e,f; Table 1), the treatment induced an additional increase in PE of both species groups under eCO 2 (Figure 4g,h, Table 2; based on models with L).In this shorter timeseries, L P appeared to be lower in years with growing-season θ < 0.2 m 3 m −3 (Figure S5; Table 1).With L P included in the model, θ appeared to have no direct effect on pine NPP (p = .115;Table 2).The effect of θ on both L B and broadleaved NPP was statistically significant (p ≤ .029;Tables 1 and 2), however, the trend in L B with θ was slightly negative and that in NPP slightly positive (not shown), making the soil moisture effect difficult to interpret.
In sum, the eCO 2 -induced enhancement of pine ANPP and NPP and broadleaved NPP were sustained throughout the experiment, reflecting increases in pine L and in PE of both species groups, consistent with our expectations (H4).Moreover, PE was further enhanced by N-additions under eCO 2 .While the enhancement of L, ANPP, and NPP varied widely across plots (Figure 5), based on treatment averages (2005)(2006)(2007)(2008)(2009)(2010), the relative contribution to NPP of higher PE under eCO 2 and the two N treatments, was 38%-52% for pine and 67%-95% for broadleaved species (Figure 5e).
We also expected PE to decrease with increasing tree size, L, and within-canopy shading.We found that that PE declined with increasing L in broadleaved species but not in pine (Figure 3c,d).size in both species groups.For pine, and at comparable M, LAR was higher under eCO 2 (p < .001).Except for broadleaved species under aCO 2 , unit leaf rate (ULR = ΔM/L), did not change with M and was, on average, higher under eCO 2 (p < .001).

| ANPP-to-L relationships over time
To demonstrate the dynamics in the relative roles of native site quality, eCO 2 , N-additions and competition for light in driving production, we explored temporal changes in L P relative to B P96 , variation in L B relative to L P , and in ANPP relative L. We focused on ANPP because broadleaved coarse-root NPP was not correlated with L B (Table 2) and variation in NPP with L B was less well-defined than that of ANPP.
At the beginning of the experiment, plots that showed larger B P96 also carried a higher L P (Figure 7a-c and Figure S6).From 1999 onwards, at any given B P96 , eCO 2 plots showed a higher L P than those under aCO 2 , by a similar amount and regardless of N-additions.The differences among treatments disappeared after the CO 2 enrichment ended (Figure 7d).Broadleaved L, in turn, decreased with increasing L P across plots (Figure 8a-c and Figure S7), indicating that the variation in L B was driven by light availability.Within the common range in L P , eCO 2 plots had a higher L B than those under aCO 2 .
This difference appeared to diminish at higher L P and disappeared altogether during the post-treatment years (Figure 8d).Earlier in the study, the response of pine ANPP to eCO 2 appeared to be greater in plots supporting high L P but pine ANPP was similar under aCO 2 and eCO 2 in plots of low L P (Figure 7e,f and Figure S8).This is consistent with the mean response over the entire period evaluated at fixed, initial L P (L P96 ; Figure S4).Towards the end of the experiment, and with both ANPP and L P changing over time, plots across all treatments shared a common ANPP-to-L P relationship (Figure 7g).For broadleaved species, spatial variation in L B explained much of the variation in ANPP (Figure 8e-g and Figure S9).After the treatments ended, the ANPP-to-L B relationship did not change much (Figure 8h).For pine, ANPP remained higher in the previously CO 2fumigated plots compared to the ambient ones, but only at the upper end of the post-enrichment L P (Figure 7h).

| Variation in allocation ratios, biomass retention
Across all treatments, the largest proportion of pine NPP was allocated to stem (~50%) while the fraction allocated to coarse roots declined with increasing M over time (Figure S10).At the begin-  et al., 2007, 2010).Accordingly, we hypothesized that eCO 2 will increase production more in plots with higher nutrient availability, as reflected in higher initial L (H1).We found that the absolute woody NPP (NPP W ) response to eCO 2 depended on all three indices of site  S3, and for e-h, in Tables 1 and 2 S4], consistent with earlier results (Finzi et al., 2002;McCarthy et al., 2010).The enhancement ratio computed based on the CO 2 × L P96 interaction effect on NPP W increased slightly with L P96 (line in Figure 3a).However, based on an assessment of mean plot-specific enhancement ratios, an approach that reduces the effect of the fastest growing plot (with the greatest eCO 2 enhancement) on the relationships, the enhancement of NPP W was unrelated to L P96 (data points in Figure 3a).Notably, two eCO 2 plots, especially the most responsive one, sustained their response following termination of enrichment even after their L decreased (Figure 7).This could be a legacy effect of eCO 2 (that was not found in L nor C allocation belowground; Kim et al., 2017); a structural adjustment over the long-term CO 2 enrichment that had a carryover effect on stand function, perhaps related to hydraulic adjustments (Domec et al., 2009).
We further hypothesized that the effect of eCO 2 on production will be ephemeral in plots of low L P96 , suggesting PNL (H2).We found little evidence of PNL, confirming the findings from earlier studies (Drake et al., 2011;Finzi et al., 2006Finzi et al., , 2007;;Kim et al., 2020;Maier et al., 2022;McCarthy et al., 2010).We evaluated the dynamics of NPP W enhancement across the eCO 2 plots and over the entire study period and found no temporal trends (Figure 3b), thus leaving no basis for the hypothesis that N-addition will reverse PNL (H3).In other words, a sustained increase in production (shifting from line 1 to line 3 or 4 in Figure 1) was observed over the duration of the study.This contradicts an earlier observation at the FACE Prototype (FACE P ; one of two eCO 2 plots on poorer soil) that eCO 2 response of woody biomass production is ephemeral without complete nutrient addition (Oren et al., 2001); indeed, the response to eCO 2 recovered later and was sustained until the end of the study.The early attenuation of the growth response may have been a part of interannual variation triggered by that in climatic variables, yet no clear link was found.Alternatively, enhanced carbon allocation belowground (Drake et al., 2011;Maier et al., 2022;Palmroth et al., 2006), may have led to a release from nutrient limitation (Finzi et al., 2007;Walker et al., 2015) allowing a recovery of the response.
We found that increased pine NPP under eCO 2 was related to both increased L P and PE (H4).Pine PE and unit leaf rate (ULR) did not decrease with increasing L P and the associated within-canopy shading (Figures 3d and 6h), however, the expected inverse relationships were observed for the broadleaved component (only under aCO 2 for URL; Figures 3c and 6g).The results also suggested that eCO 2 may have sustained the PE of the lower canopy strata under increasing shading by pine.Because PE includes all biomass production, a sustained efficiency suggests a higher canopy photosynthesis or carbon use-efficiency with increasing L.Although similar to full fertilization studies, where increased L was not accompanied by decreasing (growth) efficiency, in these experiments, growth efficiency is often based on stem production, which is sustained in part by reallocating less carbon (C) to belowground (Albaugh et al., 1998;Waring, 1983).
In this study, both the absolute NPP of, and relative allocation to, broadleaved coarse roots increased under eCO 2 , while pine aboveand belowground NPP were similarly affected.Nitrogen-additions, TA B L E 2 ANOVA type III p-values related to linear mixed model fits (using the lmerTest package in R) for atmospheric [CO 2 ] (CO 2 ) and nitrogen-addition (N) effects on aboveground and total net primary production (ANPP, NPP) by biomass compartment and species group over 2005-2010; PE stands for production efficiency (NPP/leaf area index).Time (year), leaf area index for each species group (L B , L P and L T ), and growing season volumetric water content (θ) were included in the model as continuous variables.in turn, reduced fine-root production by 13% (Jackson et al., 2009), yet the fraction of NPP allocated to root production (coarse + fine) did not change with N-additions in either species group or CO 2 treatment, refuting H5 (Figure 3e,f).

| Sustained NPP enhancement under eCO 2 and its structural and functional drivers
The experimentally generated atmospheric [CO 2 ] (+200 μL L −1 under eCO 2 ) was 55% higher than ambient [CO 2 ] (aCO 2 ) in 1996, decreasing to 52% in 2005, and 51% at the end of the study.Accounting for site quality (as in Figure 3a) produced an estimate of eCO 2 -induced enhancement of NPP W of 40%, while the corresponding estimate based on the original block design was 38%.Averaged over the entire study period, the NPP of pine was enhanced by 38% and that of broadleaved species by 52% (Figure 2).Over the last 6 years of the experiment, these enhancements were estimated as 40% and 47% (Figure 5).Stand NPP estimates were comparable to those of pine as broadleaved NPP contributed only 10%-13% of the total production (Figure 10).Previous research demonstrated a limited NPP response to eCO 2 in the low N availability plots (Finzi et al., 2002;McCarthy et al., 2010), thus setting expectations that N-additions would induce a greater eCO 2 response.The increase in NPP under eCO 2 with N-additions was, however, relatively small and statistically significant only for pine (11% and 8% for pine and broadleaved species, respectively; Figure 5e).For pine, the increase in NPP under eCO 2 was similar above-and belowground, whereas the more than 200% increase in coarse-root NPP (Maier et al., 2022) made up the entire NPP response of broadleaved species.
We analyzed the production response using two approaches, one focusing on the components related to annual change in live biomass (ΔM; reflecting the balance between growth and tree mortality), the other on the L and PE components of NPP estimated for trees alive at the beginning and the end of a growing season.At comparable M, the RGR of both species groups was higher under eCO 2 than aCO 2 owing to higher unit leaf rate (ULR = ΔM/L) and, for pine, also higher leaf area ratio (LAR = L/M; Figure 6).Similarly, we demonstrated that NPP was higher under eCO 2 due to both greater PE (PE = NPP/L) and L (Figure 5).difference in the steady-state L between the two CO 2 treatments would eventually disappear (Körner, 2006;Norby et al., 2022).
Instead, we found a 19% higher L P in eCO 2 plots and a decrease of the post-treatment L P to that under aCO 2 (Figures 1 and 7, Figure S6; Kim et al., 2017).These results point to a direct CO 2 effect on maximum L, perhaps, as discussed above, due to the high CO 2 supply allowing leaves to maintain positive carbon balance under lower light availability (Long, 1991;Sprugel et al., 1991).Indeed, we observed higher RGRs under eCO 2 at a given size (biomass) and found no evidence for the rates of biomass increment slowing down faster under eCO 2 compared to aCO 2 (Figure 6; Kim et al., 2020).
Our results also show that the ratio between the annual change in live biomass (ΔM) and NPP was approximately 10% higher under eCO 2 , following a relationship between ΔM:NPP and NPP that did not change by treatment (Figure 9), that is, higher NPP lead to higher ΔM:NPP.This increase in 'biomass retention' in live biomass could indicate a reduction in mortality or, as suggested in a recent analysis (Walker et al., 2019), an increase in C allocation to woody biomass.
We found some indication for both changes.Consistent with earlier studies on understory L and aboveground biomass (Kim et al., 2016(Kim et al., , 2020;;McCarthy et al., 2010), we demonstrated that the lack of eCO 2 response in broadleaved L (L B ) and ANPP was likely due to light limitation.We showed that L B decreased with increasing L P across plots (Figure 8 and Figure S7) and, although L B was somewhat higher under eCO 2 where L P was low, the difference diminished with increasing L P .Higher broadleaved NPP under eCO 2 was mostly explained by higher PE (Figures 4 and 5).The increase in PE implies not only higher carbon uptake per unit leaf area, perhaps due to sustained lower photorespiration and decreased photosynthetic light compensation point (Drake et al., 1997;Long, 1991;Springer & Thomas, 2007), but also increased sink strength belowground (Körner, 2006).The largest increase in the broadleaved coarse- The sustained eCO 2 effect on NPP and its limited interaction response to N-additions, can be explained by a higher C partitioning belowground both in terms of root production allowing access to unexploited soil volume and access to nutrients in previously unavailable pools (Drake et al., 2011;Jackson et al., 2009;Maier et al., 2022;Pritchard et al., 2008).Drake et al. (2011) argued that higher carbon flux belowground (to mycorrhiza and as root exudates) under eCO 2 stimulated both soil organic matter decomposition and tree N uptake, where the role of mycorrhiza in boosting N uptake may be less important than the faster turnover of soil organic pools (Phillips et al., 2011;Pritchard et al., 2014).Although these processes are discussed in terms of N, it may not be the sole nutrient the limitation of which have been alleviated under eCO 2 by better exploitation of the soil and increased decomposition rate.
Southeastern US piedmont soils are often phosphorous (P) and potassium (K) deficient, and by the end of the study, the foliar P:N and K:N ratios estimated for these pines were low (Knier, 2022), far below optimum for loblolly pine (Albaugh et al., 2010).Results from Duke FACE P suggested co-limitation of several nutrients as growth enhancement under eCO 2 was greatly increased with balanced nutrient fertilization (Oren et al., 2001).Similarly, two studies on extremely poor fertility sites, one on P. taeda (sandy soil) and the other on Picea abies (sandy, post-glacial till), showed growth response to eCO 2 in large-tree chambers only under balanced fertilization (Oren et al., 2001;Sigurdsson et al., 2013).In other FACE experiments, absence of eCO 2 -induced growth enhancement in a mature eucalyptus (Eucalyptus tereticornis) forest was attributed to low phosphorus availability (Ellsworth et al., 2017;Jiang et al., 2020), and low availability of P or other nutrients may have contributed to similar findings in a mature beech-oak (Fagus sylvatica, Qercus petraea; Bader et al., 2013) and Norway spruce (Picea abies) forests (Klein et al., 2016).

| New site-specific allometric functions, new NPP estimates for Duke FACE
Our estimate of 38% for the average enhancement of stand NPP under eCO 2 is higher than the median of 23% presented for young forests (Norby et al., 2005) and the 27% previously reported for Duke FACE (McCarthy et al., 2010).This was expected.By the end of the study, pines grown under eCO 2 were taller for a given diameter, and based on these site-specific allometric functions pine biomass was 27% higher than under aCO 2 compared to 21% estimated using generic equations (Kim et al., 2020).Consequently, NPP estimates were higher and the difference between the CO 2 treatments larger.
Our estimates also include, for the first time, branch turnover as the live crowns ascend over time (Valentine et al., 2013).This the end of the study (Maier et al., 2022), were also larger than reported by McCarthy et al. (2010).In fact, the current broadleaved coarse-root NPP estimates under eCO 2 were similar in magnitude to estimates of leaf and stem NPP and resulted in a mean enhancement ratio of 52% for broadleaved NPP, similar to that reported for another FACE experiment on Populus tremuloides (Norby et al., 2005).
Like for P. tremuloides, in absolute terms, our broadleaved NPP estimates were low, making the enhancement ratio sensitive to errors in any of the components.
Our results, in context with earlier findings, indicate that NPP of this pine-dominated stand increased under eCO 2 .Not only was higher PE sustained over the years of CO 2 enrichment, but also the amount of pine leaf area per biomass and L P increased.This suggests that the shade-intolerant P. taeda behaved as if its self-shade tolerance increased (Zeide, 1987), which may indicate a higher steadystate biomass at the site.However, the stand was too young for us to evaluate whether the higher growth rate under eCO 2 would affect maximum tree age and C residence time (Brienen et al., 2020;Büntgen et al., 2019).We can only certainly say that the stand follows either curve 3 or 4 (Figure 1), based on processes shown in inset c, with some indication that it may not only sequester carbon faster under eCO 2 , but end up with a greater carbon storage in live biomass.Such trajectories can be simulated by models that have been evaluated against observations, including treatment-specific self-thinning curves.

| Toward improved ecosystem models
The results presented here offer opportunities to evaluate and improve ecosystem models.Ecosystem models with reasonable reproduction of observed results can be used to extrapolate results beyond the timeframe of experiments.However, ecosystem models that simulate the responses of forests to eCO 2 can be complex, and it is not always clear why a given model simulates a particular value for forest biomass or NPP.To gain robust insight into model performance, processes subordinate to total biomass or NPP need to be evaluated (Medlyn et al., 2015).These subordinate processes may be component processes of the main process excluding 2002-2003  affected by a combination of severe drought and an ice storm) collected in plots without N-additions.We used data from 2005 to 2010 to test the CO 2 × N-additions effects on the response variables.The experimental design was a randomized complete block with a split-plot where eCO 2 and N-additions were the main and split-plot effects, respectively.Treatment effects on NPP (by species group and organ) were tested using mixed effects models in nlme, lme4, and lmerTest packages in R (R Core Team, 2021).Block and block-by-CO 2 were treated as random variables (groups) with plots blocked according to the pairing of plots established at beginning of the experiment (n = 4).We evaluated the role of L in mediating the NPP responses by testing treatment effects on L, on NPP without L in the model, and on NPP with L included as an explanatory variable (covariate).In the most complete models for the NPP response, we included time (year), L, and volumetric soil moisture as continuous variables.When testing treatment effects on L we included B P96 as a covariate to account for variation in pre-treatment

| 7 of 20 PALMROTH
et al.While L fluctuated, tree size and live biomass (M) increased monotonically over time.By the end of the experiment, trees in both species groups had grown larger under eCO 2 than those under aCO 2 (Figure 6).Relative growth rate (RGR) decreased with increasing M and, within the overlapping range of M, RGR was higher under eCO 2 (p < .001).Leaf area ratio (LAR) also decreased with increasing tree F I G U R E 2 Temporal variation in treatment averages (n = 3-4) of aboveground net primary production (ANPP; a-c), NPP (d-f), leaf area index (L; g-i), and production efficiency (PE = NPP/L; j-l) for broadleaved species (left panels), pine (middle panels), and the stand (right panels).Error bars indicate SE (among plots).Ambient [CO 2 ] = AR, ambient [CO 2 ] with nitrogen-additions = AN, elevated [CO 2 ] = ER, and elevated [CO 2 ] with nitrogen-additions = EN.Vertical lines indicate the time of termination of treatments.
ning of the experiment, the largest fraction of broadleaved NPP was allocated to foliage and, by the end, allocation to foliage and stem was similar under aCO 2 , and to foliage, stem, and coarse roots under eCO 2 .Allocation ratios averaged over the last 6 years of the experiment demonstrate that, except for the increase under eCO 2 in broadleaved coarse-root NPP (p < .001) at the expense of leaves and stem (p ≤ .038), the rest of the ratios were within a few percentage points from each other across all treatments (p ≥ .091;Figure3e,f).Thus, in contrast to our expectations (H5), there was no reduction in the fraction of NPP allocated to roots (fine + coarse) due to Nadditions in either species group (p ≥ .379).Finally, we compared the two production estimates, NPP (annual growth of living trees only) and ΔM (annual change in standing biomass; Figure6).The overall slope of the relationship between ΔM and NPP was 0.78, whereas the mean ΔM-to-NPP ratio (averaged over the eight points in Figure9b) was 0.57 under aCO 2 and 0.63 under eCO 2 (p = .026),indicating that a higher proportion of NPP was accumulating in living biomass under eCO 2 .The hypotheses At this site, the initial leaf area index (L) was a good indicator of site quality, explaining most of the variation in biomass accumulation during the first decade of stand development(McCarthy

F
Atmospheric [CO 2 ] (ambient = aCO 2 and elevated = eCO 2 ) and nitrogen-addition (N) effects (2005-2010) on aboveground net primary production (ANPP; a, b), NPP (c, d), leaf are index (L; e, f) and production efficiency (PE = NPP/L; g, h) for broadleaved species (left panels) and pine (right panels).Each point represents an adjusted mean (n = 4) and error bars indicate adjusted SE.For a-d, the ANOVA results can be found in Table ; *** stand for p < .001,* for p < .05,and letter only for p < .1.Ambient [CO 2 ] = AR, ambient [CO 2 ] with nitrogen-additions = AN, elevated [CO 2 ] = ER, and elevated [CO 2 ] with nitrogen-additions = EN.TA B L E 1 ANOVA type III p-values related to linear mixed model fits (using the lmerTest package in R) for atmospheric [CO 2 ] (CO 2 ) and nitrogen-addition (N) effects on leaf area index of broadleaved species (L B ), pine (L P ), and the stand (L T ) over 2005-2010.Time (year), pine basal area in 1996 (B P96 ), and growing season volumetric water content (θ) were included in the model as continuous variables.
Note that our attempt to quantify the relative contributions of L and PE on NPP assumes that light interception increases linearly with L and, therefore, likely underestimates the role of PE.The assumption is, however, justified when the NPP-to-L relationship is close to linear, as in the current dataset F I G U R E 5 Enhancement ratio of aboveground net primary production (ANPP; a, b) and NPP (c, d) as a function of that of leaf area index (L) averaged for each treatment (a, c) and for each plot (four plots per treatment; b, d).Each point represents a mean ratio (over 2005-2010) and error bars indicate SE (among years or plots).Overall mean enhancement (2005-2010; e), where the solid section of bar represents leaf area-mediated enhancement (ΔL), and the open section shows the portion explained by production efficiency (ΔPE).Ambient [CO 2 ] = AR, ambient [CO 2 ] with nitrogen-additions = AN, elevated [CO 2 ] = ER, and elevated [CO 2 ] with nitrogen-additions = EN.(Figures7 and 8).Higher pine L P under eCO 2 explained most of the increase in pine NPP, yet there was also a substantial functional response, particularly under eCO 2 with N-additions, where higher L P and PE contributed nearly equally (Figure5).While L P responded to N-additions under aCO 2 , the lack of response under eCO 2 suggests that L P reached the maximum in each eCO 2 plot; lower maximum may reflect plot-scale limitation imposed by other resources.If the CO 2 -induced enhancement of growth under eCO 2 was primarily due to accelerated stand development, with no effect on maximum L, it would decline sometime after canopy closure, and the F I G U R E 6 Total live biomass over time (M; a, b), and relative growth rate (RGR; c, d), leaf area ratio (LAR; e, f), and unit leaf rate (ULR = ΔM/L, where ΔM is annual change in M and L is leaf area index) as a funtion of M for broadleaved species (left panels) and pine (right panels).Each point represents a treatment mean (n = 3-4) and error bars indicate SE (among plots).Grey symbols represent data from 2002-2003 when L and production estimates reflect carryover effects from disturbances (see McCarthy et al., 2006, 2010).Vertical lines indicate the time of termination of treatments.Post-treatment years (2011-2012) are excluded from (c-h).Ambient [CO 2 ] = AR, ambient [CO 2 ] with nitrogen-additions = AN, elevated [CO 2 ] = ER, and elevated [CO 2 ] with nitrogen-additions = EN.

F
Annual change in live biomass (ΔM) by treatment, averaged over 2005-2010 (a) and the ratio between ΔM and net primary production (NPP; b) as a function of NPP.The black line in (b) is calculated using the slope and intercept in (a) and the grey line in (a) is 1:1 line.Ambient [CO 2 ] = AR, ambient [CO 2 ] with nitrogenadditions = AN, elevated [CO 2 ] = ER, and elevated [CO 2 ] with nitrogen-additions = EN.F I G U R E 1 0 Time-averaged (2005-2010) aboveground net primary production (ANPP) of pine and broadleaved species, NPP of course roots (NPP R ) for each species group, and annual litterfall for the stand [all fluxes are in g m −2 year −1 ± SE (among years)] under ambient [CO 2 ] = aCO 2 , elevated [CO 2 ] = eCO 2 , and nitrogenadditions = +N.Image: mariaflaya/Depos itpho tos.com, BioRender.oft-ignored production component represented ~18% of the current aboveground woody NPP estimates at Duke FACE.Moreover, our coarse root biomass and NPP estimates, based on root harvest at (e.g., light capture and PE are components of NPP), or specific to the level of plant organs and/or plant species.The componentprocess analysis and organ-level and species-level results in this study (e.g., Figures 2 and 10, and FigureS1) will enable the validation of models at that level.Our results also will enable validation of whether models simulate realistic functional responses and responses to varying initial conditions (Figures7 and 8).Understanding model performance and biases at these scales enables the pinpointing of areas for model improvement and better understanding a model's projections of ecosystem responses to future CO 2 concentrations.Earth system models (ESM) rely on estimated responses of terrestrial carbon sinks to increasing [CO 2 ] for predictions of future atmospheric [CO 2 ].Recent analyses showed that terrestrial ecosystem sinks are greatly underestimated when ESM predictions are compared to sink estimates from inversion of atmospheric CO 2 measurements, especially in North America and Europe(Byrne et al., 2023).Ample information is available for evaluating model estimates against current ecosystem sink strength.However, there is little information for judging how fit models are to predict future sink strength as atmospheric [CO 2 ] increases.Our study, one of the very few obtained at the climate-change relevant spatiotemporal scale (ecosystem and 1.5 decades), demonstrates sustained sink enhancement due to increased leaf area, production efficiency, and carbon retention.These unique results will help further constrain model predictions.
and nitrogen (N) additions commenced in 2005, and until the treatments ended in 2010 (marked by a vertical grey line), pine ANPP under ambient CO 2 (aCO 2 ) increased by 15% and pine L (L P ) by 11% (p ≤ .060).Expressed as an enhancement ratio over the same period, L P enhancement under elevated CO 2 (eCO 2 ) declined by 8%-9% regardless of N-additions (p ≤ .070;Figures