We report tree stem ARQ results based on measurement methods described in Hilman and Angert (2016). These methods overcome the difficulty of measuring small changes in O2 against the high atmospheric background by using a static stem chamber, in which the O2 changes are considerably larger than in an open flow chamber.

We used three different approaches to measuring ARQ: two are based on discrete gas samples of headspace air, and one is based on direct measurement of the headspace air using gas sensors (“continuous” sampling). Discrete gas samples are either taken within 30 min to several hours after chamber sealing (“instantaneous” sampling) or after the chamber has been sealed to the stem for more than 24 h, once steady-state conditions have been achieved (“steady state”). These methods and the time required for achieving steady state were confirmed by comparing with continuous measurements (Hilman and Angert, 2016). For each site and experiment described in Sect. 2.2, we identify the method used to estimate ARQ as instantaneous, steady state (for flask samples), or continuous (Table 1).

For addressing the variation in stem ARQ values across a range of tree species and environments, our study included trees located in tropical forests (Panama and Brazil), in temperate forests (Bartlett and Harvard, USA), and in a Mediterranean savanna (Spain) and a Mediterranean shrubland (Carmel Ridge, Israel). We also included five trees located on the Hebrew University campus in Jerusalem (Israel) and in the adjacent Jerusalem Botanical Gardens. The trees in Panama were part of a fertilization and litter manipulation project (Wright et al., 2011; Sayer and Tanner, 2010). No treatment effects were found (Fig. S2, this topic is not in the scope of this paper). Details about the sites, tree species, stem chambers, stem dimensions, and experiments conducted in each of the sites are presented in Table 1 and Fig. 3. The list below summarizes what data were available from the different sites and what questions in particular we addressed with these data (the numbering of the experiments matches Table 1): A.

Seasonal and/or phenological measurements of stem ARQ were performed in Jerusalem, US, and Brazil sites. In Jerusalem, five individual trees from five different species (first five species in Table 1) were measured. The phenological state of deciduous trees (all except Quercus calliprinos) was separated into four classes (Fig. 4). In the US sites, trees measured at the northern site (Bartlett Experimental Forest) had fall color development, while leaves at Harvard Forest (southern site) were still green. After analysis of flasks, we excluded results from three trees because of suspected air leakage from the chamber (O2>20 % after 6 days of stem incubation). In Brazil, six trees were measured. After analysis we excluded results from 4 out of 12 instantaneous measurements because of a weak signal (O2>20.7 % and SD>0.1 after 3 h of incubation).

Sap flux density was monitored in nine trees at the site in Spain using heat ratio method (HRM) sensors (SFM1 Sap Flow Meter, ICT International). A description of the installation and measurement is presented in Methods S1. The detailed procedures for sap flux corrections and calculations are described in Perez-Priego et al. (2017). We tested whether the daily maximum sap flux density (i.e., average of measurements between 10:00 and 17:00 local time during the day of the ARQ measurement), which correlated with CO2 dissolution fluxes (Bowman et al., 2005), could explain variability in ARQ (instantaneous).

Predawn shoot water potential (Ψpd) on Carmel Ridge was measured using a pressure chamber (PMS Instrument Company, Corvallis, Oregon, USA). At each sampling time, we sampled 2–3 terminal twigs containing 5–10 leaves from each tree. The samples were wrapped in plastic, placed on ice, and measured within an hour of sampling using the pressure chamber technique (Scholander et al., 1965).

For sampling gas from inside the stem, stainless-steel tubes (1.3 cm diameter) were installed 4, 8, and 12 cm deep into the stem, in various stem heights on the same tree in Brazil where the vertical ARQ transects were measured. Installation procedure was according to Muhr et al. (2013) and tubes were sealed between sampling dates. Using rubber tubing we connected the sampling flasks to the tubes for incubation of 4 days. The flasks were then analyzed for CO2 and O2 in the Hampadah. Assuming steady state, ARQ was calculated using Eq. (3) (Angert et al., 2012).

Stem cores were extracted immediately after the chamber incubation experiment in Panama, Spain, and Jerusalem using a 1.2 cm diameter cork borer. The outer bark and green tissues, as well as sapwood sieves (with paler color than the phloem tissues), were removed from the cores. The cores were cut into longitudinal halves, wrapped with moist gauze cloth to avoid desiccation, and placed in an air-tight incubation system to which glass flasks (two or three) were connected by Swagelok Ultra-Torr fittings (Swagelok, Solon, OH, USA, Fig. S4). At the end of the incubation period, the flasks were closed and analyzed in the Hampadah. Since the incubations took place in a closed system (no diffusive exchange with outside air), the changes with time in [CO2] and [O2] are assumed to be linear, and ARQ can be calculated using Eq. (1).

In Panama and Spain the incubations were started immediately upon core extraction, at ambient temperature, and lasted 8 and 3 h, respectively. In Jerusalem the cores were kept on moist gauze cloth for 2 h before being sealed in the incubation system and kept at 25 ∘C in an environmental chamber. Repeated incubations were performed in series, with the incubation systems flushed in between with ambient air. Simultaneously, from each tree, four leaves from an understory branch were cut and inserted into the same incubation systems, for the same incubation durations. The O2 uptake rate (nmol O2 g FW-1 s-1) was calculated as follows (adopted from Pruyn et al., 2002): O2uptakerate=ΔO2100×VHT×MFW×Vm×109, where ΔO2 is the decrease in [O2] during the incubation, VH is the volume of headspace (mL), T is the incubation period (s), MFW is the fresh weight (g), Vm is the molar volume, and 109 converts units to nmol. We dried the samples in an oven at 60 ∘C for 2 days for the dry weight.

In Brazil, stem cores were extracted by using a 5.15 mm diameter increment corer. After bark was removed the cores were cut to a length of 6 cm each and then allowed to equilibrate with the atmosphere for 6–8 h, while continually being kept moist. After equilibration, each core was transferred to an incubation chamber equipped with flasks. Prior to starting the incubation, a few milliliters of water were added to keep the core tissue moist. In this case, incubations were left at room temperature (∼25 ∘C) for 24 h before flasks were closed and removed.

All statistical analysis was done using JMP (JMP^®, JMP Pro 13, SAS Institute Inc., Cary, NC, USA). Repeated-measures analysis of variance (ANOVA) was used to evaluate how the interaction of tissue (stem core and leaves) with ARQ and O2 uptake varies with time in the repeated incubations of the tissues from the trees in Jerusalem. Mauchly's test indicated violation of sphericity in the ARQ response in the repeated incubations experiment (χ2=18.132, P=0.021); therefore, the Greenhouse–Geisser adjusted F test was chosen. One-way ANOVA followed by Tukey–Kramer HSD (honestly significant difference) was used to perform comparisons among time points in every tissue. A Student's t test was used for comparisons between stem cores and leaves at each time point.

ARQ (instantaneous) values of nine Q. ilex trees in Spain were invariable (mean ± SD of 0.42±0.04) in comparison with the larger variation in maximum daily sap flux density among these trees (0.15±0.05 m3 H2O m-2 h-1), and no correlation was found between the ARQ and sap flux density (r2=0, P=0.9891).

Mean ARQ ± SD values (steady state) of the trees at the Carmel Ridge site were 0.62±0.06, 0.68±0.07, and 0.69±0.08 for spring, summer, and winter, respectively. Repeated-measures analysis of variance found no significant difference between seasons (F2,2=2.52, P=0.28), while Ψpd varied significantly with seasons (F2,2=207.85, P=0.0048). During summer, Ψpd was -2.65 MPa, much lower than the spring and winter values (-0.64 and -0.86 MPa, respectively).

In the Jerusalem day–night campaigns, ARQ (instantaneous) values ranged between 0.52 and 1.05, across all trees, seasons, and sampling times (Fig. 5). Predawn ARQ values exceeding than daylight values (by amounts larger than the differences between duplicates) were observed during the summer in M. domestica and in the upper chamber on Q. calliprinos. No significant diurnal effect was found in repeated-measures analysis of variance of the breast height chambers, neither when results of all the trees were grouped by season nor when results were grouped by stem chamber. In continuous measurements of M. domestica, with ARQ values obtained every 4 h, ARQ during the night (0.70; n=12) was not significantly (P>0.76 in a Student's t test) greater than in the day (0.71; n=11; Fig. 6). The variations among the nighttime values were best explained using temperatures measured 235 min before the ARQ measurement (r2=0.84, P=0.0001, ARQ=0.01×temperature(∘C)+0.54). With the same time lag, the coefficient of determination for the daytime values is r2=0.44 (P=0.0266).

In Q. calliprinos, measured over 3 years in Jerusalem, ARQ did not differ significantly (P>0.33 in a Student's t test) between breast height and stem base (ARQ of 0.56 vs. 0.59, respectively, n=14, Fig. 4). For P. occidentalis measured for the same period the ARQ measured at breast height was significantly higher than ARQ measured at the stem base (0.74 vs. 0.64, respectively, n=12, P=0.003 in a Student's t test, Fig. 4). For a single S. micranthum tree in Brazil, ARQ values measured at heights of 6.5 and 11 m above the ground were similar to ARQ measured at breast height (Fig. 7), but also show differences with the stem base. In this tree, ARQ measured in March (0.46±0.11; wet season) was lower than in October (0.89±0.16; dry season). The in-stem ARQ values ranged between 0.25 and 0.56, with average ± SD of 0.46±0.07 in both seasons and at all stem positions and depths. The in-stem ARQ, as well as [CO2] values, had no clear vertical trend (Figs. 7; S5).

The average ARQ values of the stem core incubations were similar to the stem chamber ARQ for the four sites and trees where these comparisons were made (Fig. 8). When incubations were repeated over time for Q. ilex stem cores and leaves, significant effects of time, tissue (leaves, stem cores), and their interactions (time × tissue) on ARQ and O2 uptake rates were observed. ARQ of the stem cores increased from 0.44±0.08 (mean ± SD, n=4) after 3 h to 0.94±0.03 at the end of the experiment (32 h; Fig. 9). The ARQ of incubated leaves of the same trees showed higher initial ARQ of 0.80±0.02, with an increase over time to 0.92±0.02.

The ARQ measured in stem chambers installed on 85 individual trees of nine species including tropical, temperate, and Mediterranean forest trees was considerably and almost universally lower than 1.0. ARQ values as low as 0.7 could indicate that lipids were used exclusively as substrates for respiration, but current understanding suggests this scenario is implausible. However, this understanding relies on low and constant lipid concentrations over seasonal sampling (Hoch et al., 2003); daily changes in lipid concentrations and RQ were measured in response to shading and drought treatments, indicating this substrate might be more important than commonly thought (Fischer et al., 2015; Hanf et al., 2015). Nevertheless, many of the measured ARQ values were below 0.7, so substrate use alone cannot explain them. Additionally, as ARQ values above 1.0 are expected when lipids are produced (De Vries et al., 1974), ARQ<1.0 resulting from lipid metabolism must be mirrored with ARQ>1.0 at a different time (assuming the lipids are produced locally). However, ARQ almost never exceeded 1.0. The results demonstrate that O2 influx to the stems usually exceeded the CO2 efflux, regardless of tree species, site, season, and time of day. Assuming O2 uptake provides a measure of in situ respiration (due to the low solubility of O2) and carbohydrates are the main substrate, values of ARQ averaging 0.59 indicate that on average 41 % of the CO2 produced by respiration was not locally emitted to the atmosphere, but apparently retained in the stem.

For sites where we have time series data for the same individuals, considerable variations in ARQ values were observed over 2 years in Brazil (Fig. S3) and over 3 years in Jerusalem (Fig. 4). A decrease in ARQ values was often observed during entrance to dormancy for the deciduous trees in Jerusalem, as well as an apparent minimum in ARQ for P. atlantica and Q. calliprinos in spring (Fig. 4). The autumn decrease seems to be in agreement with the finding of significantly lower ARQ for Bartlett Experimental Forest, where leaves were beginning to senesce, compared to the more southerly Harvard Forest, where leaves were still green.

The possibility of measurement artifacts as the source for the low ARQ values seems unlikely, as Hilman and Angert (2016) previously demonstrated the validity of the measurement methods and the box-model approach. Further support comes from the small mean difference (0.02) between the instantaneous and steady-state ARQ measured for the same tree, which reflects overall agreement between the measures. The considerable scatter around perfect agreement (Fig. 2), expressed also in RMSD of 0.15, could be taken as an indication that the measurement methods differ significantly. However, since the model assumes constant ARQ with time, and temporal changes in ARQ are obviously present as shown in Figs. 5 and 6, the scatter could also be attributed to temporal differences in the time integrated by the two types of measurement: the instantaneous sampling was typically conducted a few days before the steady-state sampling on the same tree. Additionally, the precision for instantaneous ARQ was lower than for steady-state values, due to smaller changes in O2 over the shorter time periods. This may also contribute to the scatter in Fig. 2 (Hilman and Angert, 2016). We also found strong similarities between ARQ measured on intact stems using chambers and ARQ determined by incubating stem cores (Fig. 8). This provides another, indirect, confirmation that the low ARQ values obtained with the stem chamber measurement approaches are measuring something that is occurring in the stem tissues.

Given the low solubility of O2, stem flux ARQ values <1.0 (or potentially <0.7 for “fat trees”) are the result of respired CO2 either being exported from the site of respiration before it can be emitted to the atmosphere or being refixed during biosynthesis processes within the stem. As noted earlier, a second possibility is non-respiratory O2 uptake, e.g., by oxidases and hydroxylases that are O2-consuming enzymes, most notably used in lignin biosynthesis. However, stoichiometric analysis of this pathway shows that the CO2 produced from the sucrose that is the lignin's substrate usually exceeds the O2 consumption, so that the net effect of lignin biosynthesis should be a local increase in ARQ (Amthor, 2003). To the best of our knowledge, there are no other significant O2-consuming processes in tree stems that might affect the ARQ value.

We conclude that the low stem ARQ must be the result of CO2 being locally fixed or transported away from the site of respiration. If CO2 dissolution and DIC transport is the main export mechanism, we would expect ARQ to increase with temperature (i.e., according to solubility changes with temperature), be anti-correlated with sap flow (McGuire and Teskey, 2004; McGuire et al., 2007; Bowman et al., 2005), and to increase with height in the stem (Hölttä and Kolari, 2009). Three observations support the idea that this export mechanism controls some of the variability in ARQ. First, nighttime ARQ in M. domestica was indeed correlated with temperature, an expected trend given the greater temperature sensitivity of the CO2 solubility in comparison with O2 (Gevantman, 2018). Second, P. occidentalis had higher ARQ values in the upper stem position, especially during the growing season (Fig. 4). Third, relatively high ARQ values were observed at 0.2 m above the ground in the S. micranthum tree (Fig. 7), which may reflect a burst of in-stem CO2 that originated from belowground respiration (McGuire and Teskey, 2004; Levy et al., 1999). However, in most of our observations ARQ did not vary as expected if CO2 dissolution and transport were the main CO2 export mechanism.

When sap flux density was measured directly, it did not explain the variation in ARQ among Q. ilex trees in Spain. Mean ARQ values were fairly stable over spring, summer, and winter (0.62–0.69) for Q. calliprinos in the Carmel Ridge site, while the transpiration stream probably varied greatly between seasons if related to Ψpd. Additionally, during dormancy when no leaves were in place to force the transpiration stream, we found ARQ values <1.0 in four deciduous trees (black markers in Fig. 4). Sap flow rates are assumed to decline during the night, but ARQ values <1.0 during nighttime were measured in five species, and in most cases no nocturnal increase in ARQ in comparison to daytime values was observed (Figs. 5, 6). Thus, the temperature dependency observed for the M. domestica tree during the night, which explained variability in ARQ values between 0.65 and 0.75, must be a second-order control on ARQ variability and cannot explain the big deviation from unity (according to the linear fit, an ARQ of 1.0 is expected at the unreasonable temperature of 63 ∘C). Also, the vertical transects of ARQ for Q. calliprinos and S. micranthum, including in-stem ARQ for the later (Figs. 4, 7, S5), showed no consistent pattern of ARQ increasing with stem height, unlike the ARQ increase with height measured in the P. occidentalis (Fig. 4).

ARQ values measured in the stem core incubations, where tissues are isolated from the influence of transport in the xylem stream, were well below 1.0 and similar to the chambers' values (Figs. 8, 9). The in-stem ARQ measured in the S. micranthum was likewise <1.0, but although the proximity to the xylem was greater, the values were not necessarily lower than the surface ARQ (Fig. 7). It is likely that in-stem ARQ values are influenced by dissolution in the xylem water, but the question is the following: what is the contribution of in-stem CO2 to the CO2 efflux from the stem surface? There are contradicting assessments, and the influence likely is related to wood anatomy. For example, studies of ring- and diffuse-porous species observed tight covariations of in-stem CO2 and surface efflux and have interpreted this as evidence of strong influence of in-stem CO2 concentrations (Teskey and McGuire, 2002, 2007; Steppe et al., 2007), while other studies conducted on conifers with tracheid anatomy inferred only marginal influence of in-stem processes on surface efflux (Ubierna et al., 2009; Maier and Clinton, 2006). Nevertheless, observations of covariation in in-stem [CO2] and CO2 efflux do not necessarily represent cause-and-effect relationships (Maier and Clinton, 2006). Muhr et al. (2013) utilized the difference in 14C signature of in-stem CO2 (5 cm deep) and surface efflux to estimate that <20 % of total emitted CO2 originates from the inner stem in three tropical non-coniferous tree species. Small contribution of in-stem CO2 to the surface efflux can be easily explained by the slow diffusion through wood of all three anatomical groups (Sorz and Hietz, 2006). The woody diffusional barrier can explain the apparent decoupling between ARQ, sap flux density, and Ψpd presented above. A major contribution from respiratory activity concentrated in the outer stem tissues to overall stem respiration would further reduce sap flow effects on surface fluxes (Hölttä and Kolari, 2009; Maier and Clinton, 2006; Ubierna et al., 2009).

An alternative explanation for low ARQ values could be the fixation of CO2 by biosynthesis with engagement of the enzyme PEPC, which is able to fix respired CO2. Indirect evidence for PEPC activity can be found in the increase in the ARQ values with time in our repeated incubations, while cellular activity was retained as reflected in O2 uptake rates (Fig. 9). Such a pattern may reflect a biochemical process, e.g., CO2 fixation by the enzyme PEPC, which decreases with time due to self-inhibition by the accumulation of the products (Kai et al., 1999; Huber and Edwards, 1975). PEPC fixation rates can easily explain the retained CO2, according to mass balance calculation for the stem core incubations and published PEPC fixation rates in young tree stems (Table 2). Assuming refixation is important, the fact that ARQ measured from intact stems is almost always lower than unity indicates that the fixation products, organic acids like malate and citrate or amino acids, are not inhibiting the fixation or being oxidized locally and are further metabolized or allocated elsewhere in the stem. The malate can be transported in the xylem stream as indicated by an upwards concentration increase in Acer platanoides stems (Schill et al., 1996). A possible fate of the malate might be similar to C4-like photosynthesis observed in tobacco, where xylem-transported malate contributes C to photosynthesis in leaves (Hibberd and Quick, 2002). Alternatively, the fixation products might be exported via the phloem. One possible sink is excretion of organic acids to the rhizosphere as root exudates, which can account for ample fraction of overall gross primary productivity (GPP) in forests (Abramoff and Finzi, 2016; Finzi et al., 2015). Indications for the transport of organic acids from upper parts of the plant to the roots have already been reported (Hoffland et al., 1992; Shane et al., 2004).

Overall, our results suggest that CO2 dissolution and removal in the xylem stream are not the main cause of the low ARQ values that are common to the trees we measured. We speculate the observed ARQ values resulted by PEPC refixation, with possible cumulative effects of some dissolution and transport, partial lipid metabolism, and some non-respiratory O2 consumption. Corticular photosynthesis may theoretically influence ARQ, but in a complex manner; with an assumed O2/CO2 exchange ratio of 1.0 and given that all other processes yield ARQ<1.0, the photosynthesis will reduce the CO2 and O2 concentration gradients between stem and atmosphere in the same absolute numbers, which will cause a further ARQ decrease. However, in our measurements photosynthesis was prevented by shading the measured stem surface. Additionally, most evidence for significant corticular photosynthesis comes from twigs and young stems (Pfanz et al., 2002; Ávila et al., 2014), while stems in the current study were mature. Wood anatomy may further impact ARQ by modifying the contribution of internal stem processes on surface fluxes. The numerous mechanisms potentially responsible for ARQ probably varied with the broad range of species and wood anatomies we investigated. Further research to pursue the potential role of PEPC, including direct measurement of PEPC activity, would be needed to assess whether PEPC plays a role in lowering ARQ values to the levels observed. To complete the stem C balance, additional evaluation of the relations between the in-stem and the stem surface fluxes is also needed, as well as analysis of how organic and amino acids vary in the stem.

From a whole ecosystem perspective, if respired CO2 in the stem returns to the atmosphere elsewhere (e.g., in the soil, canopy), the overall ecosystem–atmosphere C fluxes will not be affected, and high ARQ associated with the release of the transported CO2 will balance the low ARQ in the stem. Such ARQ>1.0 values are expected in the rhizosphere where organic acids are decomposed. In the canopy, greater refixation of internal C is expected to increase the photosynthetically oxidative ratio (O2produced/CO2consumed), as the internally transported C replaces the atmospheric CO2 when assimilation is measured. Additionally, such internal transport can (i) cause a discrepancy between the measured aboveground and belowground CO2 effluxes and the locations where respiration is actually occurring (Aubrey and Teskey, 2009) and (ii) lead to false attribution of respiration responses to environmental conditions. Moreover, the different long-term temperature sensitivity of CO2 efflux and O2 influx is of interest and might explain part of the gap between modeled and observed Q10 values of tree respiration (Griffin and Prager, 2017). For example, decrease in ARQ with rising temperature (due to higher PEPC activity for example) might result in a slow increase in CO2 efflux, whereas the respiration rate (O2 uptake) is actually increasing sharply, together with the internal C flux. Future studies should determine how temperature and nutrients control long-term changes in ARQ and aim to identify the biochemical process that control the low ARQ reported by the current study.