Reconciling discrepancies in measurements of vulnerability to xylem embolism with the pneumatic method

Abstract

Chen et al. (2021) reported discrepancies between pneumatic and other methods for measuring embolism vulnerability in plant xylem tissue, leading them to caution against using the pneumatic method. We show that this discrepancy arises from faulty implementation: failing to measure air discharge to sufficiently negative water potentials, they under-estimated maximum air discharge (ADmax), and hence embolism resistance. Embolism vulnerability methods identify how dehydrated a stem must become (in water potential, P50) to cause a 50% loss in xylem conductivity. Pneumatic methods require measurements to water potentials roughly twice as dehydrated as P50. Yet Chen et al. often stopped at magnitudes even less than what other methods showed P50 to be, making it mathematically impossible for their pneumatic implementation to measure true P50. Our simulations confirmed that premature stopping causes large errors in derived P50, accounting for reported discrepancies. Further, literature studies show consistent agreement between pneumatic and other methods. Thus, Chen et al.’s pneumatic results are invalid, and should not be taken to undermine confidence in pneumatic measurements. We recommend best practices, including the ‘Pneumatron’ for automated measurements, that prevent the errors of Chen et al. and make pneumatic measurements of embolism vulnerability that are fast, replicable, and accurate. Drought-induced xylem embolism is a potentially important mechanism of plant mortality during a severe drought in natural and agricultural systems (McDowell et al., 2008; Brodersen & McElrone, 2013; Choat et al., 2018), with debate on whether embolisms are a primary causal mechanism, or an associated side effect, of tree mortality (Körner, 2019). This has stimulated research efforts to measure and understand the vulnerability of plants to xylem embolism, which depend on methods for consistent and accurate measurement of xylem traits related to hydraulic failure of the transport system. The parameters measured are often quantified as P50 (or P88), the water potential at which plants lose 50% (or 88%) of xylem hydraulic conductivity during a dehydration process (Sperry et al., 2002; Meinzer et al., 2009). In a recent paper entitled ‘Quantifying vulnerability to embolism in tropical trees and lianas using five methods: can discrepancies be explained by xylem structural traits?’, Chen et al. (2021) (hereinafter simply Chen et al.) compared five methods to estimate xylem vulnerability to embolism in terms of P50. Chen et al. reported large discrepancies in vulnerability curves obtained with the recently-developed rapid pneumatic method (Pereira et al., 2016) in comparison with the bench-top dehydration method (Sperry & Tyree, 1988), the optical method (Brodribb et al., 2017) and X-ray-computed microtomography (microCT; Cochard et al., 2015). They argue that the pneumatic method, which quantifies the relative increasing volume of air discharged from embolized regions of a branch's xylem (relative to a maximum amount of air discharged, ADmax) as the branch is dehydrated in the laboratory, underestimates P50 and therefore xylem embolism resistance. Comparisons among methods are important for advancing science, but must be done carefully as there can be important differences in methodological assumptions and practices that create apparent differences in results, which are actually artefacts (Pereira et al., 2016, 2020a,b, 2021; Zhang et al., 2018; Jansen et al., 2020; Sergent et al., 2020). Here, we show that Chen et al.'s reported discrepancy in P50 results obtained from the pneumatic vs other methods is an outlier among all studies that have used it; we offer evidence that this anomalous result is likely an artefact caused by inadequate implementation of the pneumatic method by Chen et al. in comparison with others who applied this method. Since the pneumatic method is otherwise generally consistent with other methods, we conclude that it in fact offers an effective and reliable way to estimate P50 across angiosperm species. The key problem with the Chen et al. implementation of the pneumatic method is evident from their own data, which suggests that a cause of their reported discrepancy is incorrect determination of the maximum air discharge (ADmax), a key step in the pneumatic method. Chen et al's fig. 1(d) (reproduced here as Fig. 1) shows that they did not measure air discharge (blue points and line in Fig. 1) to sufficiently negative water potentials to achieve an accurate estimate of the maximum air discharge. ADmax, the maximum amount of air discharged from a branch at the end of a sequence of pneumatic method measurements, is approached when water potential values decline to levels corresponding to 100% loss of branch xylem conductivity. Correct estimation of a true maximum for ADmax is necessary to accurately normalize percent air discharge (PAD), the vertical axis of the graph in Fig. 1. Good practice is thus normally to continue measurements until evidence supports a conclusion that ADmax has in fact been reached (Pereira et al., 2016; Trabi et al., 2021). If ADmax estimates are too low, the derived percent loss of air discharge (PAD = AD/ADmax) will be consistently too high, the 50% loss point will be reached too soon, at water potential magnitudes that are too small (i.e. insufficiently negative), and P50 will be consistently underestimated. Chen et al.'s data (Fig. 1) show precisely this; indeed, their estimates of ADmax were in some cases likely to be extreme underestimates, as they were taken at water potentials that did not even reach values as large as the P50 value that they were seeking to quantify. Since the P50 derived by the pneumatic method should be less negative than the water potential at ADmax, the pneumatic P50 estimated in the example represented by the blue dashed line in Fig. 1 is forced to be substantially less negative than the dehydration method P50 represented by the dashed red line. Chen et al.'s own analysis thus demonstrates that it would not be mathematically possible for their implementation to give an accurate pneumatically-derived P50 estimate, regardless of the branch water status in the branches being measured and regardless of other details or issues that may be associated with the method. We will offer two independent lines of evidence that support the inference (from Fig. 1) that Chen et al. underestimate pneumatic P50. First, we conduct a simulation based on a new dataset that demonstrates the P50 artefact caused by a variation in ADmax and quantifies its magnitude. We use this simulation to show that the discrepancies reported by Chen et al. could arise from stopping the dehydration experiment too soon (as with the blue curve in Fig. 1). Second, we present an analysis, compiled from other recent studies and from our own measurements, that compares the deviation of the P50 derived from the pneumatic method to values derived from other methods, showing that Chen et al.'s data are anomalous relative to others in the literature. We compiled P50 and P88 data from seven published or in-review studies (Supporting Information Table S1). We added to these datasets new vulnerability curves that we measured on eight tree species from seasonal Amazon forests in Brazil, using both the pneumatic and hydraulic methods as described in Table S2 and Fig. S1. We conducted two kinds of analyses: first, we conducted simulations based on our Amazon tree species dataset to test the sensitivity of pneumatically derived measurements (P50, P88) to variation in ADmax; and second, we compared P50 and P88 measurements derived from the pneumatic method to values derived from other methods across multiple different studies. All data analyses were performed in R software (v.3.5.1; R Core Team, 2020). We used data from eight tree species from a seasonal Amazon Forest (Table S2; Fig. S1) and started with ADmax.ref, ΨPADmax.ref and P50.ref as reference values, corresponding to our ‘best’ estimates for the maximum air discharge, and associated maximum water potential and P50, for each species (according to the methods presented in the supplement). We then performed a sensitivity analysis to quantify how sensitive estimates of P50 were to artificially reduced values of ADmax and ΨPADmax that were at lower magnitudes than ADmax.ref and ΨPADmax.ref (taken for the purposes of this simulation as the ‘true’ values). We simulated several curves for each species, repeatedly recomputing PAD and vulnerability curve parameters for subsets of the curve with consecutively lower ADmax values (and correspondingly less negative xylem water potentials). We retained at least the first three most hydrated points (least negative water potentials) in all simulations to have enough points to fit an S-shaped curve. We included all simulated subsets for which the model fit converged (in some cases, a particular subset did not generate a simulated P50, because the fit did not converge due to a low number of points). We used the derived P50 values corresponding to each ADmax for each species and then quantified the error of this simulated value relative to the ‘best’ estimate as the ratio P50.simulated : a ratio less than for simulations that underestimate Thus, values of lower than measurements of air discharge stopped too at water potentials less negative than that of the Since P50 estimated from a vulnerability curve must by be in than the water potential air discharge is 50% of it is mathematically impossible for 1 measurements the Chen et al. measurement in Fig. 1) to an estimate of the true P50. values than but to are likely for a estimate using vulnerability P50 may be 50% of the Fig. to or or measurements become in the that they become less likely to errors due to the of We used the in our sensitivity analysis, using P50.ref as the reference P50 value and a for each simulated vulnerability curve as the ADmax consecutively We used low values of to potential of air discharge in studies in the literature Chen et using the P50 by a comparison method as P50.ref P50 We used reported ADmax when the data were not we estimated from in each Table S1). measurements were on multiple each an estimate of ADmax and a corresponding water potential In this we a the most negative value from among for for that branch the to quantify how negative the measurement of water potential, relative to the P50 and the most negative of the P50 values were compared across the compiled studies (Table based on the bench-top dehydration air and methods. studies show that pneumatic measurements on can be to achieve with the pneumatic (Pereira et al., 2016; Zhang et al., 2018; Sergent et al., we species from our analyses and included angiosperm we did not in our analyses the pneumatic method from the new automated pneumatic using a to consistent results and to in the estimated volume of air discharged (Pereira et al., Second, we the that the line between pneumatic P50 and P50 derived by method from the 1 : 1 line (i.e. that the of that line from and the from across all studies. We used the axis test using the et al., with the used to the different datasets from the different studies. We for differences in and in between the pneumatic and other methods in each a (i.e. in both pneumatic and other methods (as in least that error estimates to the et al., & In we were to whether the between methods is consistent across or whether a Chen et al.) from other studies. we used the model to estimate the confidence of the and the of the estimated and with and the P50 P88 estimates were among methods et al., In we in to identify data points that have the or in the least and have an extreme on the estimates by the pneumatic method and other methods The simulations of the sensitivity of pneumatic vulnerability curves their parameters P50 and P88) to using different values of ADmax and ΨPADmax are presented in Fig. we the embolism vulnerability curves by ADmax values that were less than the best estimate ADmax.ref, the correspondingly estimated P50 values were lower in on than the best estimate as for each of the species The between the P50 error P50 as a of the reference that = and a value an and the of premature stopping = across all species (Fig. lower showed that simulated P50 consistently lower than P50.ref when ADmax too low low and P50.ref line at P50.ref = as This is by a fit to the curve = with = and = vs = The of the P50 discrepancies associated with the of ADmax among species (as by the in Fig. lower in of Fig. We the values of the Chen et al. measurements their hydraulic P50 as P50.ref for the relative error of their pneumatic P50 and the on the graph as our simulations points in Fig. lower We that most of of the Chen et al. vulnerability curves were based on The Chen et al. values showed on than our simulated errors of P50.ref lower than the curve in Fig. some of their reported between pneumatic and other methods may have been caused by other (or by differences among most Chen et al. points the of our simulated that Chen et al.'s pneumatic errors are consistent with errors due to premature the datasets of all studies (Table we agreement between pneumatic estimates of P50 as compared to estimates derived from other methods (Table for the Chen et al. which as anomalous (Fig. Chen et al. the that did not show between pneumatic P50 estimates and derived from other methods, with a different from 1 (Fig. the between methods = and a in the which the error of the studies (Fig. Table The other lines with the 1 : 1 line confidence et al. its from an Table including Sergent et al. which reported apparent discrepancies for species et al., and presented the among all studies Fig. points in the data point from Chen et al. and three from Sergent et al. were as to (Fig. by the in the Fig. the pneumatic measurements reported by Chen et al. were taken during and of the dehydration with of the primary measurements the of dehydration necessary to vulnerability curves and accurate estimates of ADmax and hence of P50 we that species in Chen et al. fig. the In all cases, the xylem water potentials corresponding to ADmax for the pneumatic method were not even of the water potentials the maximum of loss of hydraulic conductivity with the other methods (Fig. with of five measurements stopping at water potential magnitudes of or less (and a of The of stopping measurements the ADmax, and the the pneumatic method P50, with lower than Fig. when Chen et al. used methods, they estimated the embolism using the maximum amount of embolized taken at negative xylem water potentials negative than Our primary is that the large discrepancy between the pneumatic and other methods reported by Chen et al. is with studies pneumatically derived embolism to other methods in angiosperm species, including published studies and a new dataset from that of et al. The discrepancies reported by Chen et al. are due to incorrect implementation of the pneumatic method, of the maximum amount of air discharged from xylem due to stopping the measurements sufficiently negative water This conclusion is here by simulations (Fig. and by the fact that between pneumatic and hydraulic methods precisely for measurements with ΨPADmax ADmax taken branch which pneumatic data points from the 1 : 1 (Fig. In cases, Chen et al. their pneumatic measurements at potentials low in that they of the xylem P50, making it mathematically impossible for their pneumatic estimates to that value obtained by other methods that they in a way that did achieve sufficiently negative water potentials). We conclude that Chen et al.'s reported results for the pneumatic method are and should not be used for analysis of embolism vulnerability in should not be taken to undermine confidence in pneumatic measurements of embolism resistance. We here that a recent in the of the pneumatic method automated Jansen et al., 2020; et al., the pneumatic method to use and it impossible for to create the of implementation error in Chen et al. The data of and thus less compared with measurements (Fig. and the (or of a of air discharge is evident from the data The may the pneumatic method to be applied including to which results is for this we on angiosperm species the et al., 2020; et al., 2021). on the and use of the software for and by and on best practices for pneumatic measurements we to Trabi et al. Chen et al. offer a of potential causes of their reported underestimates of pneumatically derived P50, which they against the pneumatic method the analysis presented other the discrepancy reported by Chen et al. and that this discrepancy likely arises from inadequate implementation of the methods that the results the evidence suggests caution should be applied to the Chen et al. measurements. as it is generally important to all potential of error in a and Chen et al. issues that we are important for of embolism vulnerability estimates to be we them on (Pereira et al., 2016; et al., 2021; et al., including to for the pneumatic method. The by Chen et al. that the pneumatic method may be measuring the vulnerability of at the end of a measured branch and these are to be by the that there are in the ADmax, which Chen et al. with a analysis on of their species embolism vulnerability curve two during the dehydration a as to how can when in air discharge has been reached fig. in Chen et that there are the of air discharged in the pneumatic that branch or may which undermine pneumatically-derived to the potential of the end of the branch due to the it is important to a potential here and from that have been for the purposes of making the pneumatic measurements become and should be an of the discharge They are not a of the being measured and should not the the of the air discharged, the of the for air in the The of in may embolism in the of measurement possible for values based on pneumatic measurements can be than embolism in et al., 2016; et al., 2021). the evidence shows that the of xylem vulnerability in the branch should not be to the or by the the because embolism the branch from embolized et al., 2016; et al., to but this as the water potential is low enough to large is derived from an embolized to an is what the measurement by with increasing In the the in confidence in the that is generally between the pneumatic and other methods in the literature Chen et al. (Fig. which suggests there is not a or vulnerability at the end of branches measured with the pneumatic method. in these are possible when the pneumatic method is (as by the in the Chen et al. but with to as by the et al. studies here have all to pneumatic method estimates consistent with other and by the automated which it to underestimate ADmax (Fig. Chen et al. conduct a measurement test to dehydration on of their species, that the embolism vulnerability curve two in air discharge what ADmax should when dehydration is to extreme negative water potentials, which they reached Fig. which from Chen et al. fig. the of the of water potential measurements to extreme values Chen et al's fig. (Fig. to our when they stopped at an water potential value (blue points and curve in Fig. Chen et al. a P50 of all the points to and points and they a P50 of which is in the of the of values from other methods they reported in the (from for hydraulic to for optical to for is water potentials negative than are the of the used by Chen et al. and and these stem water potentials were not but from a between and water potential and model were not in their paper and how of a have been water potentials between and are in the plant hydraulic would branches to a of which the negative in plants to values plant et al., 2020). This point would likely make branches to and air from the pneumatic the reported (Fig. is to data et al. make or other test of the of two and we would at Fig. to whether are evident with to the of water potential, in or in the dataset of In the and of water potentials, the first water potentials are the of whether the it be an artefact of a between and We are that the is or a problem to fit the pneumatic or that would be a problem for other method. In we that with the of new Fig. that continue the measurement until a is the of in ADmax will be for measurements the pneumatic method. is important to in that the of a reliable for datasets is not to the determination of ADmax for the pneumatic method, but to other methods, including the optical method, which the maximum embolized based on an that is at the end (Brodribb et al., et al., 2021). This with hydraulic measurements of embolism in which the maximum hydraulic conductivity is obtained in the = ADmax can be obtained at the end of a pneumatic experiment when the of in to a in xylem water potential is to methods that on hydraulic conductivity have a in sequence in these two reference points for of the pneumatic method must be to have a point in the hydrated water potential as as a (ADmax), which is when a of ADmax is to the of discharged both and on across xylem evidence that the pneumatic method that is from due to the of across et al., 2020; et al., et al., 2021; et al., 2021; Trabi et al., 2021; et al., 2021; et al., et al., across is & et al., 2015). the of is than the of a the of the of across et al., 2021). is evidence that the is and from during the first et al., the for the pneumatic method. there are some cases may be a of in the pneumatic method the to due to or we that or can be in or when the water potential is negative than (as in the of Chen et al., Fig. but we have not evidence in our own with plants and water potentials our to the of and are not of evidence from that possible are substantially to the and to the pneumatic we that or be to the of normally in even applied and that these as the method the cause potential small air et al., 2021). In we that there are to xylem hydraulic that are not and that these likely to the of in this and all methods that on xylem based on pneumatic with species, we not that of these from the large errors in ADmax in Chen et al.) that undermine the value of pneumatic measurements than errors in the other methods the of methods. in all methods, of best practices, in this including (or use of the Fig. to that an ADmax has been and of during and measurement to of of should give confidence in the results of pneumatic measurements et al., 2021). these an important should we these methods of embolism and to converge on a estimate of embolism this of our of the and of xylem in and what is being is to understand in the xylem and which the of pneumatic and how to embolism from an embolized to a et al., the a of embolism measurements to other methods, as the method et al., 2019). the of measuring embolism is based on hydraulic conductivity measurements, potential that may hydraulic measurements the effect, or of embolism, of decline of hydraulic conductivity and embolism that is not et al., et al., 2021; et al., 2021). embolism as on as on xylem with the been in of the of pneumatic is that in embolized the mechanism of xylem conductivity is In this the pneumatic method differences in the plants due to a (Pereira et al., Trabi et al., 2021; Jansen et al., et al., In we multiple lines of evidence from simulations of of error and from among datasets of other the inference that Chen et al.'s pneumatic method implementation and results for pneumatically derived embolism vulnerability Thus, being to the Chen et al. results to the we for or plant to the value and of embolism measurements obtained conducted pneumatic we would to that in on methods to and we that the pneumatic method for embolism in the xylem of plants as by automated measurements that can be with the is of these that offers in measurement and of implementation in We caution that incorrect of methods could make even faulty and can the method. and of methods should be carefully the and literature of supports the use of pneumatic as a for advancing our of the in plant xylem and embolism resistance. to and and are of the for and and from the and and were by and to the data in the Amazon in by the to the data in the Amazon the for its and the first of this and derived the and the to the analysis presented in the All to the Fig. curves based on the of air discharge and loss of hydraulic for eight angiosperm species from a seasonal Amazon Fig. S2 of how we the of stopping the branch process using the pneumatic method in to method. Fig. of the P50.ref by the corresponding simulated underestimates of ADmax as quantified by the ΨPADmax across the eight species and across all and the P50.simulated as a of the simulated ADmax, showing how they in each Fig. for each that most the model fit between P50 and P50 by a different method, with the corresponding of P50 vs P50 Fig. of fig. of Chen et al. showing the of Chen et of air volume against a of water potentials than used in their (Fig. 1). Table that P50 or P88 values from the pneumatic method with other methods. Table S2 of eight tree species at the Table using axis is not for the or of Information by the than should be to the The is not for the or of by the than should be to the corresponding for the

Topics & Concepts