Research Determinants of legacy effects in pine trees – implications from an irrigation-stop experiment Roman Zweifel1 , Sophia Etzold1, Frank Sterck1,2 , Arthur Gessler1,3 , Tommaso Anfodillo4 , Maurizio Mencuccini5,6 , Georg von Arx1 , Martina Lazzarin1,7, Matthias Haeni1, Linda Feichtinger1 , Katrin Meusburger1 , Simon Knuesel1, Lorenz Walthert1 , Yann Salmon8,9 , Arun K. Bose1,10 , Leonie Schoenbeck1 , Christian Hug1, Nicolas De Girardi1, Arnaud Giuggiola1, Marcus Schaub1 and Andreas Rigling1 1 Swiss Federal Institute for Forest, Snow and Landscape Research WSL, 8903 Birmensdorf, Switzerland; 2Forest Ecology and Management Group, Wageningen University, 6701 Wageningen, the Netherlands; 3Institute of Terrestrial Ecosystems, ETH Zurich, 8092 Zurich, Switzerland; 4Dipartimento Territorio e Sistemi Agro-Forestali, University of Padova, 35020 Legnaro, Italy; 5 ICREA, 08010 Barcelona, Spain; 6CREAF, Universidad Autonoma de Barcelona, 08193 Barcelona, Spain; 7Horticulture and Product Physiology, Wageningen University, Wageningen 6701 the Netherlands; 8Institute for Atmospheric and Earth System Research/Physics, University of Helsinki, 00100 Helsinki, Finland; 9Institute for Atmospheric and Earth System Research/Forest Sciences, University of Helsinki, 00100 Helsinki, Finland; 10Forestry and Wood Technology Discipline, Khulna University, 9208 Khulna, Bangladesh Summary Author for correspondence:  Tree responses to altered water availability range from immediate (e.g. stomatal regulation) Roman Zweifel to delayed (e.g. crown size adjustment). The interplay of the different response times and pro- Tel: +41 79 703 04 91 cesses, and their effects on long-term whole-tree performance, however, is hardly under- Email: roman.zweifel@wsl.ch stood. Received: 4 December 2019  Here we investigated legacy effects on structures and functions of mature Scots pine in a Accepted: 24 March 2020 dry inner-Alpine Swiss valley after stopping an 11-yr lasting irrigation treatment. Measured ecophysiological time series were analysed and interpreted with a system-analytic tree model. New Phytologist (2020) 227: 1081–1096  We found that the irrigation stop led to a cascade of downregulations of physiological and doi: 10.1111/nph.16582 morphological processes with different response times. Biophysical processes responded within days, whereas needle and shoot lengths, crown transparency, and radial stem growth Key words: cambial activity, drought stress, reached control levels after up to 4 yr only. Modelling suggested that organ and carbon ecological memory, irrigation experiment, reserve turnover rates play a key role for a tree’s responsiveness to environmental changes. osmoregulation, point dendrometer, radial Needle turnover rate was found to be most important to accurately model stem growth stem growth, TreeNet. dynamics.  We conclude that leaf area and its adjustment time to new conditions is the main determi- nant for radial stem growth of pine trees as the transpiring area needs to be supported by a proportional amount of sapwood, despite the growth-inhibiting environmental conditions. encompassing various carry-over or lagged effects (also called eco- Introduction logical memory effects). As a consequence, the stronger a legacy Physiological responses of plants in general, and trees in particu- effect is, the lower becomes the predictive power of current con- lar, are often explained by current environmental conditions in, ditions for a physiological response (Meinzer et al., 2013; Jump for example, ecophysiological models (Steppe et al., 2006; et al., 2017; Zweifel & Sterck, 2018). It is thus also a measure for Zweifel et al., 2007; Medlyn et al., 2011), wood formation stud- the degree of decoupling of plant responses from the concurrent ies (Drew et al., 2010; Rathgeber et al., 2016; Delpierre et al., environmental conditions (Kannenberg et al., 2019b; c). 2019) or assessments of climate-vegetation dynamics (Liu et al., There is fast-growing statistical evidence for such legacy effects 2018; Ma et al., 2019). However, a wide range of evidence docu- (Ogle et al., 2015; Jiang et al., 2019; Peltier & Ogle, 2019). ments the limitation of this common approach accounting for Annual stem growth of trees, as an example, is generally not opti- concurrent environmental drivers only and, instead, strongly sug- mally explicable with current-year conditions (R2 = 20–40%) and gests to additionally consider past conditions (Anderegg et al., including past conditions has been shown to result in greater frac- 2015; Ogle et al., 2015; Jump et al., 2017; Zweifel & Sterck, tions of explained variance (R2 increased by 30% and more (Ogle 2018; Kannenberg et al., 2019a). This effect is commonly et al., 2015). But the question remains, how do legacy effects take described with the term ‘legacy effect’ (Huang et al., 2018; Peltier place at a mechanistic, physiological level? By contrast with statis- et al., 2018) and is in this work used as a generic term tical evidence for the importance of legacy effects, little Ó 2020 The Authors New Phytologist (2020) 227: 1081–1096 1081 New Phytologist Ó 2020 New Phytologist Trust www.newphytologist.com This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. New 1082 Research Phytologist information is known about the potential mechanisms. Zweifel We propose three alternative response types of the trees grow- & Sterck (2018) recently proposed an approach considering the ing in the irrigation-stop plots (Fig. 1) in order to discuss timing turnover rates of organs and reserves as a way to link past condi- and strength of legacy effects in different organs and the interplay tions to the current plant response. They employed a model among them at a whole-tree level: showing that differences in turnover rates can theoretically (1) In the first type, the physiological and morphological vari- explain different levels of responsiveness to current conditions. ables (i.e. sap flow, tree water deficit, stem growth, needle and Trees with generally longer turnover rates of leaves, sapwood and shoot length, and crown transparency) return to the respective carbon reserves (> 5 yr) were found to be less responsive to a sud- levels of the control trees in the first year after the irrigation stop. den change in the environmental conditions than trees with This type of response is not affected by past conditions (at least shorter organ turnover rates of 1 or 2 yr. It means that the former not to the extent as compared with the control) and is called here type of trees is better able to buffer short-term negative environ- a type of response with ‘no legacy effect’. mental impacts. However, it also means that it takes longer to (2) In the second type, the different variables are maintained recover from severe impacts, in terms of re-building structures of above the level of the control for several years before returning to the predisturbance level (Zweifel et al., 2000; Weber et al., 2007; it. This response type is called a ‘positive legacy effect’ because Galiano et al., 2011; Zweifel & Sterck, 2018). past conditions induce a positive physiological or morphological The turnover rate of an organ or a reserve is defined as the time response temporarily remaining above the level of the control. in which the underlying structure (e.g. the needles of the crown, or We hypothesise that such trees may benefit from reserves accu- the carbon molecules of the carbon reserve) is on average replaced. mulated during the irrigated years. Consequently, the turnover rate of needles strongly affects the (3) Finally, in the third type, the physiological responses drop adjustment time of the total leaf area to changed conditions. For below the control level before returning to it. This response is the sapwood, as another example, the turnover rate translates into called ‘negative legacy effect’, as the past conditions alter the the average number of years a tree-ring remains part of the sap- tree’s response in a negative way by shifting the process (tem- wood before it is turned into heartwood. Therefore, this individual porarily) below the control level. This case reflects the idea that tree-ring with its specific number and distribution of water con- the ‘overbuilt irrigated trees’ are less well adjusted to the reduc- ducting elements and thus particular hydraulic properties, can tion in soil water after the irrigation stop than control trees and affect the water conductance of a tree for a longer or a shorter time therefore perform worse. period depending on its functional life-span. The value of such an organ and reserve turnover approach lies in its intrinsic capacity to Materials and Methods mechanistically link past conditions to a current plant’s response (Jump et al., 2017; Zweifel & Sterck, 2018). Site Here we test this idea of linking past and present environmental conditions to current physiological and morphological responses The long-term irrigation experiment is located in a Scots pine via organ and reserve turnover rates with field data from mature forest (Pinus sylvestris L.) at the northwest-exposed slope close to Scots pine trees which were experimentally irrigated since 2003 and cut off from the treatment after 11 yr at the end of 2013 (Dob- bertin et al., 2010; Sch€onbeck et al., 2018; Brunner et al., 2019). The setup of the Pfynwald irrigation experiment allows comparing control trees that were never irrigated with trees that received con- tinued irrigation treatment until present, with trees for which irri- gation was recently stopped (Bose et al., 2020). Scots pines in this dry inner-Alpine valley (Valais) have been reported to suffer from increased mortality during the past decades (Zweifel et al., 2009; Rigling et al., 2013; Etzold et al., 2019), a phenomenon caused by longer and more frequent dry periods (Bigler et al., 2006; Szejner et al., 2019). The irrigation treatment in the Pfynwald experiment added c. 500–600 mm water per year. This doubling of the natural precipitation increased stem radial growth, leaf area, needle and shoot lengths, whereas crown transparency decreased (Dobbertin et al., 2010; Eilmann et al., 2013; Sch€onbeck et al., 2018). Furthermore, the irrigated trees adjusted their root production to the irrigation in Fig. 1 Alternative physiological or morphological responses of irrigated terms of a significantly increased fine-root biomass (Brunner trees to a stop of the treatment in comparison with nonirrigated control trees. A positive legacy effect suggests that a response variable remains et al., 2009; 2019; Herzog et al., 2014). Over the 11 yr of irriga- above the level of the control for several years. A tree without a legacy tion, these trees thus developed new (overbuilt) structures and effect returns to the level of the control trees without delay. A negative adjusted their carbon reserves to the growth increase (Sch€onbeck legacy effect leads to a response variable below the level of the control et al., 2018). before returning to the control level. New Phytologist (2020) 227: 1081–1096 Ó 2020 The Authors www.newphytologist.com New Phytologist Ó 2020 New Phytologist Trust New Phytologist Research 1083 the bottom of the Swiss Rhone valley in the driest part of the Additionally, we applied a system-analytical tree model Valais, (46°180 N, 7°360 E, 615 m asl) and close to the dry edge of (Zweifel & Sterck, 2018) that bridges the influence of past condi- distribution of Scots pine. Mean annual temperature was 9.2°C tions to the present physiological response on an annual level and mean annual precipitation 518 mm (1971–1990; Support- (Fig. 2). The model quantifies legacy effects on the organ and ing Information Fig. S1) (Wehren et al., 2010). reserve status of a tree with the help of turnover rates of leaves, The forest is described as Erico-Pinetum sylvestris with a sapwood and carbon reserves. Those response variables, which mean tree height of 10.8 m, a stand density of 730 stems ha1, were measured and modelled (radial stem growth, needle and and a basal area of 27.3 m2 ha1 (Dobbertin et al., 2010). A shoot length), were compared in order to quantify the explana- description of trees equipped with dendrometer and sap flow tory power of the ‘turnover approach’ to catch different patterns sensors can be found in Table S1. For the irrigated plots, pre- of legacy effects (Fig. 1). cipitation has been approximately doubled since 2003 (addi- tional 500–600 mm yr1) by adding water from a nearby Environmental measurements channel during night of the growing season (April–October) using sprinklers of 1 m height (Herzog et al., 2014; Bose et al., Meteorological measurements were recorded 2 m above the 2020). End of 2013, the irrigation was stopped in the upper canopy on top of a scaffold, c. 13 m above ground. Air tempera- third of each irrigated plot. The soil is a shallow Regosol char- ture (Sensirion, St€afa, Switzerland), relative humidity (Sensirion) acterised by low water retention capacity. All plots had the and precipitation (Young tipping bucket 52203, Michigan, USA) same exposition. were continuously recorded at a 10 min interval. Data obtained in nearby meteorological stations of MeteoSwiss (www.mete oswiss.admin.ch) were employed to fill gaps and to correct obvi- General setup ous instrumental errors by applying filters and using simple We measured microclimatic variables in air and soil (data resolu- regressions. tion: 10 min), physiological variables (10 min) and crown mor- Soil water content was measured with time domain reflectome- phological variables (annual) on trees in the different subplots: tries (TDRs) (Tektronix 1502B cable tester, Beaverton, OR, control plots which never were irrigated (control), treated plots USA) at soil depths of 10, 40 and 60 cm from 2002 to 2013. In with irrigation until the end of 2013 (irrigation stop), and irri- spring 2014, the soil water measurement equipment was partially gated plots until the end of 2017 (irrigation). replaced and relocated. Since 2014, 10 HS-Sensors (Decagon Fig. 2 System-analytical tree model to quantify legacy effects according to Zweifel & Sterck (2018). The model consists of four elements that describe the tree status at the beginning of the year (crown, bud, carbon reserve and sapwood). The model further takes up the key processes of radial stem growth, crown growth, needle length growth (NL), shoot length growth (SL), and carbon reserve growth. The elements are linked to a network with positive linear functions (indicated as arrows with numbers) weighted by a weighting factor (WF). Additionally, the turnover (TO) rates for the crown (needles), the sapwood and the carbon reserves quantify the time that is needed to renew the respective organ or reserve. With this network of functions, the new status of crown, bud, carbon reserve and sapwood are calculated. The model is run by an index for environmental conditions (Supporting Information Methods S2). Functions and additional explanations for the numbered arrows are given in Methods S1. Ó 2020 The Authors New Phytologist (2020) 227: 1081–1096 New Phytologist Ó 2020 New Phytologist Trust www.newphytologist.com New 1084 Research Phytologist Devices, Pullman, WA, USA) were installed in all treatments and irrigation stop and irrigation) was used as a proxy for the devel- recorded soil water content at 10 and 80 cm soil depth. Soil water opment of the total leaf area. content data were gap-filled and homogenised for the period from 2011 to 2017 with the help of overlapping periods of differ- System-analytical tree model ent soil measurement methods and devices. The logging devices used were from DecentLab (DecentLab GmbH, D€ ubendorf, The model applied (Zweifel & Sterck, 2018) is based on physio- Switzerland) and Campbell (CR1000, Logan, UT, USA). logical key processes such as crown growth, radial stem growth and carbon reserve growth (Fig. 2). The key processes described as linear functions (with weighting factors, Methods S1) are Physiological measurements driven by environmental conditions (linearly coupled to tree Stem radius (SR) changes were measured with point dendrome- water relations), and the status of crown, buds, carbon reserves ters at breast height (ZN11-T-WP; Natkon, Oetwil am See, and sapwood of the past year. We refer to this approach as a sys- Switzerland) on nine trees (three trees per treatment) consisting tem-analytic model (Vester, 2007) in which absolute mass or of a T-shaped carbon fibre frame anchored in the stem with three energy balances are not quantified, but the individual responses stainless steel rods and a potentiometer. The dendrometers, of system components are quantified relative to each other and including cables and loggers, have a low temperature sensitivity relative to a ‘normal’ response (value = zero). This approach is of < 0.3 µm per °C and SR data were not further corrected for particularly valuable to assess a system responsiveness, respectively temperature sensitivity. Furthermore, tree water deficit-induced the legacy effects of system parts (organs), in our case a tree. reversible stem shrinkage and swelling (TWD) and growth-in- The simulation in annual steps starts with the status of buds, duced irreversible increment (GRO) were calculated with the R crown, carbon reserve and sapwood that are the result of the past package TREENETPROC (Haeni et al., 2020), according to the environmental conditions and the related processes over time. approach of Zweifel et al. (2016) assuming no cell growth during The initial status needs to be set before the first iteration. The periods of stem shrinkage. length of time considered to affect the status of organs and Stem sap flow was measured simultaneously with Granier-type reserves is defined by the turnover rates determining over how sensors (UP GmbH, Ibbenb€ uren, Germany) on the same trees many years the respective structure is built and renewed, respec- where the dendrometers were mounted. The two needles of the tively. The model quantifies the current-year environmental sensors were drilled 5 cm into the sapwood and insulated from impact on all the processes involved and calculates the new status direct sunlight. Data resolution was 10 min. Sapwood depth was of the organs and reserves at the end of the year according to the found to range between 23 and 107 mm in a sampling of 20 irri- network of functions (Fig. 2; Methods S1). gated and 20 control trees in 2013 (other trees than the ones The model input is an annual environmental index calculated equipped with sap flow sensors). The mean sap wood width for from water supply (precipitation plus irrigation), soil water con- control and irrigated trees was 47 and 51 mm, respectively tent, air temperature, and radiation (Methods S2). (Fig. S2), found to be not significantly different (t-test, P > 0.05). The model consists of linear functions with weighting factors Sap flow was calculated from individual stem diameters and an (Methods S1) for which the input-, operating- and output-ranges average sapwood depth of 5 cm (370–425 cm2). are limited to index values ranging from 1 (very poor status), Sensors were powered and logged by devices establishing a over 0 (average) to +1 (very good/improved status). This way, local mesh network around a central base station with data trans- annual changes of environmental conditions (also expressed in mission to the related online database (DecentLab). values between 1 and +1) alter the status of structures (crown, bud, sapwood) and reserves (carbon reserve) according to the lin- ear functions, their weighting factors and their turnover rates. Sta- Crown morphology measurements tus values are thus always relative to an average value of zero. Needle and shoot lengths were measured from 36 trees (10 con- Values above zero always mean a status or a response above the trol, 12 irrigated, 14 irrigation stop). Three branches from each average, a value below zero a status or response below the average. of these trees were selected from the top and the middle parts of The weighting factors (WF) for the linear functions between the crown, resulting in 116 sampled branches. Shoot lengths on the status and the processes of the model were parameterised every branch were measured for the years 2011 to 2017 and aver- (Table S2) for the control trees (never irrigated, control), for the aged for the three treatments. Needle length of each shoot was irrigated trees (irrigated), and for the trees treated with irrigation determined by measuring five randomly selected needles close to until the end of 2013 and for which irrigation was removed after- the centre of the shoot. wards (irrigation stop). Annual crown transparency was assessed by a visual rating in The model was run for two scenarios for all three treatments 5% steps of all (c. 800) trees of the experimental area using refer- (control, irrigation, irrigation stop) with the respective parameter ence photographs ranging from 0% (a fully foliated tree) to 100- sets. Scenario I mimics a tree with no ecological memory in % (a dead tree) as described by Dobbertin et al. (2005). The tree which the model parameters ‘turnover rate of needles’, ‘turnover crown foliage is judged relative to the optimum foliage of an rate of sapwood’, and ‘turnover rate of carbon reserves’ were set average tree of the same size and species. The average crown to 1 yr (abbreviation ‘NoMemo’). Scenario II (abbreviation transparency of all trees within the different treatments (control, ‘Memo’) mimics a tree with more realistic turnover rates of New Phytologist (2020) 227: 1081–1096 Ó 2020 The Authors www.newphytologist.com New Phytologist Ó 2020 New Phytologist Trust New Phytologist Research 1085 needles (5 yr), sapwood (50) and carbon reserves (10) based on (orange line) dropped to the same level as the control plot (red empirical findings. The number of Scots pine needle cohorts at line) or even slightly below already in 2014. Generally, the water Pfynwald was measured to be between three and five (data not content of the soil with its low water holding capacity responded shown). The average sapwood turnover rate was found to be quickly to changes in the treatment, which became visible during between 45 and 55 yr (Fig. S2). Most uncertain was the estima- periods of irrigation outages and during winter-time when the tion of the turnover rate of the carbon reserve (Gessler & Trey- irrigation was stopped. dte, 2016). The metabolically active carbon reserve was reported to be 1–2 yr old (Gaudinski et al., 2009), whereas the average age Tree water deficit of all stem carbon reserves was found to be 10 yr (Carbone et al., 2013). Richardson et al. (2015) explained this finding by the The irrigation-stop trees had a significantly lower tree water presence of two storage pools: a fast-turning and a slow-turning deficit (TWD) than the control trees during the irrigation period one. They assumed that the slow pool, however, was large enough until the end of 2013 (Fig. 4). After switching off irrigation, the that it cannot be ignored as a store of reserves and that it is over average annual TWD of the irrigation-stop trees generally simplistic to assume a single pool that turns over quickly. As a increased, however not linearly over the entire season. TWD consequence, we set the average turnover rate of the carbon pool remained reduced from March to July 2014–2016 (Fig. 4) before to 10 yr. returning to the TWD level of the control trees during the later summer months. This intraseasonal pattern disappeared in the fourth year (2017) after the irrigation stop. Statistical methods SR data were aligned, cleaned and gap-filled for gaps < 120 min Sap flow data with the R-based (R Core Team, 2019) package TREENETPROC (Haeni et al., 2020). Data gaps equal or longer than 120 min Before stopping the irrigation treatment, sap flow rate of the irriga- remained in the data set as NA. tion-stop trees was strongly increased during the summer months. The WF of the system-analytic model were calculated in an During this time, sap flow reached values up to the double the optimisation process with Excel’s function Solver, maximising rates of the control trees, however with a wide variation among the determination coefficients for annual stem growth (GRO), individual trees (Fig. 5). During winter-time, when the irrigation needle (NL) and shoot length (SL) between consecutive years. was stopped, both treatments behaved in a similar way. After the These three variables were measured in the field as well as explic- irrigation was stopped at the end of 2013, sap flow rates generally itly modelled and thus qualified to be used for the optimisation dropped significantly below the rates of the control trees during process. Parameterisation was run in three steps grouping WF the summer months, but remained increased in spring when soil according to their direct effect onto NL, SL and GRO. In the water availability was generally higher than in summer (Fig. 3). first step the two WF directly affecting NL were parameterised by This pattern persisted over the entire measurement period until optimising the determination coefficients between measured and 2017. Interestingly, the short response time to reduced soil water modelled NL. In the second and third step, the same was per- availability became also visible during short-term failures of the formed with SL (2 WF) and GRO (11 WF). irrigation system (e.g. in July/August 2011, Fig. 5): as soon as the The sensitivity of the model output to changes in the (set) irrigation was stopped, the sap flow of the treated trees started to turnover rates of needles, sapwood and carbon reserves was tested decrease and went markedly below the sap flow level of the control with a stepwise variation of each of the three turnover rates trees after c. 1–2 wk without irrigation. On a mean annual scale, (Fig. S3). The average change in the output for annual stem sap flow remained downregulated below the level of the control growth was used as an indication for the model sensitivity to trees for all 4 yr measured after stopping the irrigation. changes in the turnover rates. The turnover rate of carbon reserves had the lowest impact (weight 1) on model output, fol- Radial stem growth lowed by sapwood turnover (2.3), and needle turnover had by far the highest impact on model output (2180) (Table S3). Radial stem growth (GRO) – deduced from SR measurement The explanatory power of the simulations was quantified with including irreversible bark and wood growth – was markedly the determination coefficient of linear regressions between mea- increased in the irrigated compared with the control trees sured and modelled means and standard errors of means per (Fig. 6). On average, the irrigated trees grew two to three times treatment. faster and started growth earlier than the control trees, however, with considerable individual differences. GRO decreased gradu- ally after the irrigation stop but remained significantly above the Results growth rate of the control trees for three more years. In the fourth year after the irrigation was stopped (2017), the difference in Soil water content GRO disappeared between the two treatments. An interesting The irrigation treatment starting in 2003 considerably increased detail is the often occurring GRO increase towards the end of the the soil water content of the treated plots (Fig. 3). However, with year which goes in parallel with the general soil rehydration at the stop of the irrigation in 2013, the level of soil water content this time of the year (Fig. 3). Ó 2020 The Authors New Phytologist (2020) 227: 1081–1096 New Phytologist Ó 2020 New Phytologist Trust www.newphytologist.com New 1086 Research Phytologist Fig. 3 Mean daily time courses of soil volumetric water content (10–80 cm) in control plots (no irrigation, red), irrigated plots (blue) and plots where irrigation was stopped at the end of 2013 after 11 yr of treatment (orange). The irrigation was active during the nonfreezing period of the year. Periods of missing irrigation during the irrigation period indicate outages of the irrigation system (blue horizontal line). Data resolution: 1 h. New Phytologist (2020) 227: 1081–1096 Ó 2020 The Authors www.newphytologist.com New Phytologist Ó 2020 New Phytologist Trust New Phytologist Research 1087 Fig. 4 Time courses of tree water deficit (TWD) of irrigation stop (orange lines) and control trees (red lines) of Scots pine over 7 yr. At the end of 2013, the irrigation treatment (blue horizontal bars) was stopped. A TWD of zero means a fully hydrated tree. Increased TWD indicates stem shrinkage and thus an increased lack of water in the stem, meaning increased drought stress. Bold lines show the mean of three trees; the thin lines indicate the SE of the mean. Data resolution: 10 min. control trees and kept this trend in the following years Crown morphological measurements (Fig. 7b). SL remained high in the first year after stopping Needle length (NL) of the new cohort responded in the first irrigation and strongly responded in the second year after the year after the irrigation was stopped with a marked decrease irrigation stop (Fig. 7c). SL remained markedly below the (Fig. 7b). The needles grew even shorter than the ones of the control in the following years. Ó 2020 The Authors New Phytologist (2020) 227: 1081–1096 New Phytologist Ó 2020 New Phytologist Trust www.newphytologist.com New 1088 Research Phytologist Fig. 5 Running means of sap flow rates over 24 h of trees (n = 3) of the irrigation stop (orange) and the control (red) trees for the years 2011 to 2017 (Scots pine). Bold lines show the mean, the thin lines indicate the SE of the mean. Irrigation (blue horizontal line) was stopped at the end of 2013. The lower panel of each year shows the difference in sap flow rates between treated and control trees. Areas coloured in orange indicate higher sap flow rates of the treated trees, whereas areas coloured in red indicate higher sap flow rates of the control trees. Data resolution: 10 min. Crown transparency (Fig. 7d) and GRO (Figs 6, 7e) showed Simulated tree responses the most distinct delays in their response. In both cases, the irri- gation-stop trees needed 3 to 4 yr to reach the same values as the The simulations were run for all treatments (control, irrigation stop control trees. and irrigation) using the two alternative scenarios (Memo and New Phytologist (2020) 227: 1081–1096 Ó 2020 The Authors www.newphytologist.com New Phytologist Ó 2020 New Phytologist Trust New Phytologist Research 1089 NoMemo). The NoMemo scenario with a model parameterisation and stem growth than the Memo scenario (Table 1). The only not allowing for legacy effects (Table S2) was not able to accurately exception was with NL, which was insensitive to the turnover rates simulate the measurements of the irrigation-stop trees and led to a of needles, sapwood and carbon reserves and showed the same sim- consistently lower explanatory power of the response variables SL ulation output for both scenarios (Figs 7, S3; Tables 1, S3). Fig. 6 Mean radial stem increments (GRO) of wood and bark of Scots pine trees in control (no irrigation, red) and irrigation-stop (orange) plots. Bold lines show the mean of three trees, the thin lines indicate the standard error of the mean. The irrigation (blue horizontal bars) was stopped at the end of 2013. The red line in 2016 ends prematurely because of a logger failure. Ó 2020 The Authors New Phytologist (2020) 227: 1081–1096 New Phytologist Ó 2020 New Phytologist Trust www.newphytologist.com New 1090 Research Phytologist By contrast, the Memo scenario with the more realistic turnover 11 well watered years. Are they able to benefit from potentially rates of needles (5 yr), sapwood (50) and carbon reserves (10) was accumulated reserves or do they rather suffer from mal-adjusted able to predict the measured annual courses of radial stem growth structures? In the following section, we discuss why both and SLs significantly better for all treatments (control, irrigation responses were found (positive and negative legacy effects; Figs 1, stop and irrigation). The increased explanatory power of the Memo 7), depending on the organ or reserve that is considered scenario compared with the NoMemo scenario ranged between (Table 2). With the help of the system-analytical tree model, we 13.8% and 52%, depending on the variable (Table 1). further discuss why the turnover rates of organs and reserves are Contrary to the above described measurement–model relation- able to explain, at least partially, the different legacy effects ships, the relationship between the environmental index ENV and (Fig. 7). Finally, we speculate what processes could explain the the modelled radial stem growth (GRO) was stronger in the measured fact that irrigation-stop trees grow better than the con- NoMemo scenario than in the Memo one (Table 1), showing that trol trees despite the reduced soil water availability. increasing the memory effect (by increasing the turnover rates) is reducing the responsiveness of GRO to current environmental con- Tree water relations – mostly negative legacy effects ditions. The first response of a tree to the sudden reduction of water is of a biophysical nature. Drought-stressed trees close their stomata Chronology of responses and the respective legacy effects and thus save water (Hetherington & Woodward, 2003; Zweifel The irrigation-stop treatment led to a reduction of the soil water et al., 2012). The general reduction of the annual sap flow of the content to a level comparable with – or even slightly lower than – treated trees after the irrigation stop (Fig. 5) was thus most likely the control plots (Fig. 3), and a cascade of tree physiological and to be a consequence of the stomatal behaviour of an oversized morphological responses started (Figs 5–7). On an annual transpiring area, suffering from a reduced water supply. On an timescale, NL, sap flow, and TWD responded already in the first annual timescale, the response of sap flow and TWD resembles a year (Table 2). SL responded strongly in the second year, while negative legacy effect (Fig. 1; Table 2), that is the downregulation radial stem growth and crown transparency did not respond of tree water relations is even stronger than the one observed for abruptly but gradually, needing c. 4 yr to reach the level of the the control trees which grew under a continuous lack of water. trees in the control plots (Fig. 7; Table 2). However, there remained an important difference to this general There appeared positive and negative legacy effects as shown in annual pattern: during spring and early summer, sap flow Fig. 1. Radial stem growth and crown transparency showed a distinct remained higher (Fig. 5) and TWD remained lower (Fig. 4) than positive legacy effect (annual growth remaining above the control, in the control trees. We hypothesise that the enlarged root crown transparency remaining below the control; Figs 6, 7), whereas biomass in the uppermost soil layer (Brunner et al., 2019) might the immediate biophysical responses related to tree water relations, have led to an increased soil water uptake during the time when that is sap flow (Fig. 5), TWD (Fig. 4) and also NL (Fig. 7), the general soil water availability was still high in this first phase responded negatively (on a mean annual scale; Table 2) but with dis- of radial stem growth (Fig. 6). As a consequence, the soil dried tinct intra-annual deviations from the annual patterns (Figs 4, 5). out even faster than in the control plots (Fig. 3) and sap flow Shoot length responded with a positive legacy effect in the first year dropped in the second half of the summer below the rates of the after the irrigation stop (SL growth remained larger) and only nega- control trees (Fig. 5). Less likely, but not excluded, is the possibil- tively thereafter (shoots grew shorter than the control), that is, SL ity that deep roots tapped water in soil layers not covered by the responded with a negative legacy effect with a 1-yr delay. soil water sensors. However, if this was the case, we expected a persistent positive effect on the sap flow, which was not mea- sured. Discussion How to live with an overbuilt tree structure when soil water Needle and shoot lengths – negative legacy effects (with gets short? delay) The investigated pine trees grew larger and denser crowns during New built needles responded with an immediate length reduction the 11 yr of irrigation compared with the control trees (Dob- in the year after the irrigation was stopped (negative legacy effect, bertin et al., 2010) and the functional structures such as sapwood that is needles grew shorter than the control), whereas the shoots (Sch€onbeck et al., 2018) and roots (Brunner et al., 2019) became responded 1 yr later (Fig. 7; Table 2, negative legacy effect with adjusted to support this enlarged crown (Enquist & Niklas, 1 yr delay), a phenomenon that has been reported for pine trees 2002; Choat et al., 2012). This response demonstrates the effect before (Dobbertin et al., 2010; Feichtinger et al., 2015). As an of water-limited conditions on tree growth on the one hand, and explanation for the NL reduction with drought, it has been pro- the general tree response to a release from limiting conditions on posed that the general decrease in leaf water potentials also the other hand, already documented several times also in other reduces turgor pressure in the crown and thus, turgor pressure studies from this area (Feichtinger et al., 2014; 2015; Herzog becomes limiting for needle growth (Myers, 1988; Giuggiola et al., 2014; Grossiord et al., 2018). The novel aspect here is how et al., 2018; Guerin et al., 2018). This lowered water potential trees with an overbuilt structure deal with a water shortage after was obviously not affecting the shoot growth at the same time New Phytologist (2020) 227: 1081–1096 Ó 2020 The Authors www.newphytologist.com New Phytologist Ó 2020 New Phytologist Trust New Phytologist Research 1091 (a) (b) (c) (d) (e) Fig. 7 Measurements (full lines) vs model results (broken lines) of two scenarios ‘NoMemo’ (crossed squares) and ‘Memo’ (circles) of Scots pine. ‘NoMemo’ excluded any type of memory effects by setting the turnover rates of needles, sapwood and carbon reserves to 1 yr. Scenario ‘Memo’ set the turnover rates to more realistic values, that is 5 yr (needles), 50 yr (sapwood) and 10 yr (carbon reserves). (a) Environmental index (Supporting Information Methods S2) calculated for the control (red symbols), the irrigation stop (orange symbols) and the irrigated plots (blue symbols). Positive values indicate favourable growth conditions above the average, negative values indicate poor growth conditions below the average. (b) Measured and modelled needle lengths. (c) Measured and modelled shoot lengths. (d) Mean measured crown transparency and modelled crown status (proxy for leaf area). (e) Measured and modelled mean annual growth increments (GRO). Missing measurements or scenarios in some panels indicate not available data. Ó 2020 The Authors New Phytologist (2020) 227: 1081–1096 New Phytologist Ó 2020 New Phytologist Trust www.newphytologist.com New 1092 Research Phytologist Table 1 Explanatory power of model simulations for measured variables stem growth (GRO), needle length (NL), and shoot length (SL) of Scots pine. Trees in control plot Trees in irrigation-stop plot Trees in irrigated plot R2_NoMemo R2_Memo D R2_NoMemo R2_Memo D R2_NoMemo R2_Memo D Explanatory power of model for measurements GRO 19.2% 68.8% 49.6% 83.3% 97.1% 13.8% na na na NL 5.9% 5.9% 0.0% 79.2% 79.2% 0.0% 67.6% 67.6% 0.0% SL 1.4% 16.2% 14.8% 34.7% 86.6% 52.0% 52.5% 87.5% 35.0% Explanatory power of ENV for modelled variable GRO GRO 92.4% 19.8% 72.7% 88.7% 56.9% 31.8% 93.0% 63.1% 29.9% 2 Listed are the determination coefficients of a linear regression (R ) between the measured and modelled annual values for control, irrigation stop and per- manently irrigated trees. Additionally, R2 was calculated for the environmental index ENV and the modelled GRO. D is the difference between the two sim- ulation scenarios ‘Memo’ and ‘NoMemo’. A positive value indicates an improvement of the explanatory power of the scenario ‘Memo’. na, not available. Table 2 Responsiveness of annual means of Scots pine responses after stopping the irrigation. SAP TWD NL SL CT GRO First strong response (yr) 1 1 1 2 Gradual Gradual Matching control level (yr) >4 4 1 >4 4 4 Legacy type Negative (but see Negative (but see Negative Negative with Positive Positive intraseasonal responses) intraseasonal responses) 1 yr delay Listed are the number of years after the irrigation stop in 2013 when the measured variables returned to the level of the control trees. Measurements refer to sap flow (SAP), tree water deficit (TWD), needle length (NL), shoot length (SL), crown transparency (CT), and radial stem growth (GRO). CT and GRO did not show a strong response but were more of a gradual nature. The legacy type refers to the scheme in Fig. 1. (Table 2). The reason for that is not fully understood since both alone but is positively influenced by the past, more favourable, shoots and needles are predetermined in the buds built in the pre- conditions. The ‘Memo’ model balanced the status of the accu- vious year (Chen et al., 1996) and exposed to the same environ- mulated carbon reserves and particularly the enlarged total leaf mental conditions. However, the different responses of needles area – two positive effects – against the direct impact of the envi- and shoots may be related to their tissue-specific exposures to the ronment (negative effect) on radial stem growth (Table S2). The low water potentials or eventually with tissue-specific osmoregu- simulation results support this statement by demonstrating that lation processes (Lazzarin et al., 2019), as hypothesised also for the ‘Memo’ model (Table S2) is able to simulate the observed radial stem growth (Coussement et al., 2018). The model applied growth pattern always better than can the ‘NoMemo’ model with the ‘Memo’ parameterisation (Fig. 7) – inducing a strong (Fig. 7; Table 1). dependency of the shoot growth on conditions of the past year and thus reducing the dependency from current environmental Turnover rates of organs and reserves determine legacy conditions – was able to simulate the legacy effect with a 1-yr effects delay. However, when cutting the functional link to the past year (s) – by setting the turnover rates of needles, sapwood and carbon The turnover rate of needles of the pine trees was measured to be reserves to 1 yr – the model was no longer able to simulate the 3–5 yr at our site, corresponding with the legacy effect in crown measured shoot lengths accurately (Fig. 7; Table 1) and thus transparency and radial stem growth which lasted c. 4 yr. The showing the importance of accounting for the turnover rates of ‘Memo’ parameterisation of the model assumed a turnover rate organs and carbon reserve in the model to explain the trees’ for the needles of 5 yr (which indicates that only c. 20% of the lagged response to current conditions. total leaf area is replaced every year) and this model parameter (Table S2) was found to be crucial to be able to track the observed annual stem growth dynamics. Radial stem growth and total leaf area: positive legacy Interestingly, a sensitivity analysis of the ‘Memo’ model gave effects the highest correlation between model output and measured After stopping irrigation, annual stem growth and total leaf area radial stem growth data when a needle turnover rate of 3 yr was (as indicated by measured crown transparency data) clearly used (Fig. S3). This implies that the initially set needle turnover remained above the level of the control trees for 3 more years rate of 5 yr could be adjusted to 3 yr in order to further increase before matching the control level (Fig. 7). This is assigned to a the goodness-of-fit. Furthermore, the turnover rates of sapwood positive legacy effect (Fig. 1) and indicates that this high growth and carbon reserves contributed to the modelling quality, how- rate cannot be explained by current environmental conditions ever with much lower weights (Table S3). New Phytologist (2020) 227: 1081–1096 Ó 2020 The Authors www.newphytologist.com New Phytologist Ó 2020 New Phytologist Trust New Phytologist Research 1093 According to our crown transparency measurements (Fig. 7), is related to cellular osmoregulation in the bark including the the irrigation-stop pines took c. 4 yr to adjust their leaf area to cambium (e.g. O’Brien et al., 2014; Lintunen et al., 2016) imply- the new conditions. Furthermore, the simulation of the crown ing an active investment in growth during drought. Turgor pres- status (proxy for total leaf area) indicated that the adjustment sure in living tissues is not only determined by physical time might even be longer, as crown status remained still conditions, that is the dryness of air and soil, but is also affected increased after 4 yr (of modelling) (Fig. 7). Additionally, the lag by active biological processes such as the increase of the osmotic of 4 yr matched exactly the time during which measured radial potential in cells by sugar loading (Badalotti et al., 2000; De stem growth rates appeared to be increased in relation to the Schepper & Steppe, 2010, 2011; Barraclough et al., 2018, expected values from the control trees, and it was also about the 2019a,b; Lazzarin et al., 2019; Michelot-Antalik et al., 2019). duration in which the needle biomass was totally renewed (needle We speculate that the irrigation-stop trees avoided the low turgor lifetime was measured to be between 3 and 5 yr). An argumenta- pressure in the cambium with the mobilisation of osmotically tion that sets these facts into a causal relationship as carried out active compounds. Recent work at the same research site showed, in our model, intrinsically implies a partial decoupling of growth indeed, that such active osmoregulation might take place in pine from current environmental conditions as clearly supported by trees (Mencuccini et al., 2017; Lazzarin et al., 2019). However, it our results (Table 1). was not particularly measured for the irrigation-stop trees and the osmoregulation effect was mainly measured in the branches and not in the stem (Lazzarin et al., 2019). Leaf area determines radial stem growth Proposing an osmoregulation mechanism also raises the ques- ‘The resources grow the leaves and the leaves grow the tree’ is an tion whether these additional resources for keeping high growth old saying of some foresters and which gets reappraised with this rates are available and for how long they could last. There is evi- study. The rapid response of the needle growth to the current dence that the irrigated trees had more reserves available than the conditions fits the first part of the statement, whereas the delayed control trees. A recent study of pine trees at the same site showed stem growth response supports the second part. Physiologically, an overall increase in nonstructural carbohydrates (NSC) in the we argue that the enlarged leaf area of the irrigation-stop trees stems of irrigated trees (von Arx et al., 2017) and Sch€onbeck demands an adequate water supply system that can support the et al. (2018) found stem growth and NSC being positively related high potential transpiration (Anfodillo et al., 2016). In other to total leaf area, which was larger for the irrigated trees com- words, the sapwood area (particularly the annual sapwood incre- pared with the control. We hypothesise that the irrigation-stop ment) is needed to support the potential evaporative demand of trees relied on the additional carbon reserves and kept the reduc- the crown. This balanced relationship between leaves and their tion of available energy for extra growth under drought condi- amount of xylem (Martinez-Vilalta et al., 2009) requires a corre- tions in balance with the rate of reduction of the leaf area. As sponding amount of xylem at the stem level, also well known as long as a tree is able to keep its radial stem growth high with an pipe model theory, which basically assumes a distinct number of extra investment of energy to support the large crown, it may be xylem conduits per leaf area (Shinozaki et al., 1964; Mencuccini able to reduce its crown size slowly. Obviously, no reserve is & Grace, 1996; Sterck & Zweifel, 2016). As a consequence, we infinite and becomes exhausted eventually. At this time, the tree’s assume a direct physiological causality between leaf area and leaf area should be reduced to a size that is in balance with the wood growth (Zweifel et al., 2006; Fatichi et al., 2014; Zweifel new dry conditions. & Sterck, 2018), which can lead to an increased radial stem growth despite drought. However, what mechanism is able to Conclusions explain that? The sudden reduction in soil water availability after the irrigation stop did not lead to a rapid decrease in radial stem growth of pine Radial stem growth despite drought stress: towards an trees, as we expected. Instead, radial stem growth was found to be explanation in line with a slow reduction of the leaf area, taking 4 yr of time We propose two speculative explanations for the increased radial to reach the level of the never irrigated control trees. From a func- stem growth rates despite the drought stress conditions in the irri- tional point of view, we conclude that leaf area imposes radial gation-stop treatment: (1) the enlarged root biomass (Herzog stem growth in order to keep the balance between transpiring et al., 2014; Brunner et al., 2019) improved the water uptake surface and supporting sapwood. capacity, and (2) osmoregulation actively increased turgor pres- The modelling results suggest turnover rates of organs and car- sure in the cambium in order to reach the threshold for cell bon reserve as important determinants of legacy effects on trees. growth (Hsiao & Xu, 2000; Coussement et al., 2018). The first Particularly, crown size and stem growth seem to be strongly point (1) is discussed in the second paragraph of the Discussion determined by past conditions and processes due to the needle and could explain the more efficient water uptake with higher sap lifetime of c. 4 yr, affecting the direct growth response to current flow rates and lower tree water deficits allowing these trees to environmental conditions. In other words, we showed that the keep cellular turgor pressure high enough to allow for cell divi- biological predisposition of a pine tree is able to strongly decou- sion and cell enlargement (Lockhart, 1965; Petit et al., 2011; ple growth from current environmental conditions. Further we Ortega et al., 2012). (2) A second effect we speculate about here propose that an osmoregulation mechanism may help to explain Ó 2020 The Authors New Phytologist (2020) 227: 1081–1096 New Phytologist Ó 2020 New Phytologist Trust www.newphytologist.com New 1094 Research Phytologist the increased radial stem growth despite the suddenly reduced Leonie Schoenbeck https://orcid.org/0000-0001-9576-254X availability of soil water. Future work is called to particularly Frank Sterck https://orcid.org/0000-0001-7559-6572 focus on the species-specific aspects of these findings. Lorenz Walthert https://orcid.org/0000-0002-1790-8563 Roman Zweifel https://orcid.org/0000-0001-9438-0582 Acknowledgements References We thank the forestry services of the Pfynwald and the Wasserwerk Anderegg WRL, Schwalm C, Biondi F, Camarero JJ, Koch G, Litvak M, Ogle Chippis for providing the land for the research site and part of the K, Shaw JD, Shevliakova E, Williams AP et al. 2015. Pervasive drought infrastructure. We particularly thank Konrad Egger (Forest Region legacies in forest ecosystems and their implications for carbon cycle models. Leuk) for his continuous technical support. We acknowledge the Science 349: 528–532. involved networks TreeNet, Swiss Long-term Forest Ecosystem Anfodillo T, Petit G, Sterck F, Lechthaler S, Olson ME. 2016. Allometric Research Programme LWF, MeteoSwiss, the Integrated Carbon trajectories and stress: a quantitative approach. Frontiers in Plant Science 7: 1–6. Badalotti A, Anfodillo T, Grace J. 2000. Evidence of osmoregulation in Larix Observation System ICOS, and Swiss ForestLab for data and tech- decidua at Alpine treeline and comparative responses to water availability of two nical support. Further we acknowledge the financial support by the co-occurring evergreen species. Annals of Forest Science 57: 623–633. Federal Office for the Environment, FOEN (00.0365.PZ I 0427- Barraclough AD, Cusens J, Zweifel R, Leuzinger S. 2019. Environmental drivers 0562, 09.0064.PJ/R301-0223), the Swiss National Science Foun- of stem radius change and heterogeneity of stem radial water storage in the dation SNF (20FI21_148992, 20FI_173691), the Natural Envi- mangrove Avicennia marina (Forssk.) Vierh. Agricultural and Forest Meteorology 280: 1–12. ronment Research Council NERC (#NE/I011749/11), and the Barraclough AD, Zweifel R, Cusens J, Leuzinger S. 2018. Daytime stem Academy of Finland (#1312571, #323843). swelling and seasonal reversal in the peristaltic depletion of stored water along the stem of Avicennia marina (Forssk.) Vierh. Tree Physiology 38: 965–978. Barraclough AD, Zweifel R, Cusens J, Leuzinger S. 2019. Disentangling the net: Author contributions concomitant xylem and over-bark size measurements reveal the phloem- generated turgor signal behind daytime stem swelling in the mangrove RZ (idea, paper concept, data acquisition, data analyses, mod- Avicennia marina. Functional Plant Biology 46: 393–406. elling, interpretation of results, writing); SE (data analyses, inter- Bigler C, Braeker OU, Bugmann H, Dobbertin M, Rigling A. 2006. Drought as pretation of results, writing); FS (modelling, interpretation of inciting mortality factor in Scots pine stands of the Valais, Switzerland. results, writing); AG (interpretation of results, writing); TA (in- Ecosystems 9: 330–343. terpretation of results, writing); MM (interpretation of results, Bose AK, Moser B, Rigling A, Lehmann MM, Milcu A, Peter M, Rellstab C, Wohlgemuth T, Gessler A. 2020. Memory of environmental conditions across writing); GvA (data acquisition, interpretation of results, writ- generations affects the acclimation potential of scots pine. Plant, Cell & ing); ML (data acquisition, interpretation of results, writing); Environment 43: 1288–1299. MH (data acquisition, data analyses, interpretation of results, Brunner I, Herzog C, Galiano L, Gessler A. 2019. Plasticity of fine-root traits writing); LF (data acquisition, data analyses, interpretation of under long-term irrigation of a water-limited Scots pine forest. Frontiers in results, writing); KM (data acquisition, data analyses, interpreta- Plant Science 10: 1–10. Brunner I, Pannatier EG, Frey B, Rigling A, Landolt W, Zimmermann S, tion of results, writing); SK (data analyses, interpretation of Dobbertin M. 2009. Morphological and physiological responses of Scots pine results, writing); LW (interpretation of results, writing); YS (in- fine roots to water supply in a dry climatic region in Switzerland. Tree terpretation of results, writing); AKB (interpretation of results, Physiology 29: 541–550. writing); LS (interpretation of results, writing); CH (data acquisi- Carbone MS, Czimczik CI, Keenan TF, Murakami PF, Pederson N, Schaberg tion, interpretation of results, writing); NDG (data acquisition, PG, Xu XM, Richardson AD. 2013. Age, allocation and availability of nonstructural carbon in mature red maple trees. New Phytologist 200: 1145– interpretation of results, writing); AG (data acquisition, interpre- 1155. tation of results, writing); MS (fund rising infrastructure, inter- Chen HJ, Bollmark M, Eliasson L. 1996. Evidence that cytokinin controls bud pretation of results, writing); AR (idea research infrastructure, size and branch form in Norway spruce. Physiologia Plantarum 98: 612–618. research coordination, fund rising infrastructure, interpretation Choat B, Jansen S, Brodribb TJ, Cochard H, Delzon S, Bhaskar R, Bucci SJ, of results, writing). Feild TS, Gleason SM, Hacke UG et al. 2012. Global convergence in the vulnerability of forests to drought. Nature 491: 752–755. Coussement JR, De Swaef T, Lootens P, Roldan-Ruiz I, Steppe K. 2018. ORCID Introducing turgor-driven growth dynamics into functional-structural plant models. Annals of Botany 121: 849–861. Tommaso Anfodillo https://orcid.org/0000-0003-2750-9918 De Schepper V, Steppe K. 2010. Development and verification of a water and Georg von Arx https://orcid.org/0000-0002-8566-4599 sugar transport model using measured stem diameter variations. Journal of Experimental Botany 61: 2083–2099. Arun K. Bose https://orcid.org/0000-0001-8581-1651 De Schepper V, Steppe K. 2011. Tree girdling responses simulated by a water Linda Feichtinger https://orcid.org/0000-0001-9176-1056 and carbon transport model. Annals of Botany 108: 1147–1154. Arthur Gessler https://orcid.org/0000-0002-1910-9589 Delpierre N, Lireux S, Hartig F, Camarero JJ, Cheaib A, Cufar K, Cuny H, Maurizio Mencuccini https://orcid.org/0000-0003-0840- Deslauriers A, Fonti P, Gricar J et al. 2019. Chilling and forcing temperatures 1477 interact to predict the onset of wood formation in Northern Hemisphere conifers. Global Change Biology 25: 1089–1105. Katrin Meusburger https://orcid.org/0000-0003-4623-6249 Dobbertin M, Eilmann B, Bleuler P, Giuggiola A, Pannatier EG, Landolt W, Andreas Rigling https://orcid.org/0000-0003-1944-4042 Schleppi P, Rigling A. 2010. Effect of irrigation on needle morphology, shoot Yann Salmon https://orcid.org/0000-0003-4433-4021 and stem growth in a drought-exposed Pinus sylvestris forest. Tree Physiology 30: Marcus Schaub https://orcid.org/0000-0002-0158-8892 346–360. New Phytologist (2020) 227: 1081–1096 Ó 2020 The Authors www.newphytologist.com New Phytologist Ó 2020 New Phytologist Trust New Phytologist Research 1095 Dobbertin M, Hug C, Mizoue N. 2005. Using slides to test for changes in crown wood anatomy and drought timing across the eastern US. Ecology Letters 22: defoliation assessment methods part II: application of the image analysis system 119–127. CROCO. Environmental Monitoring and Assessment 102: 167–178. Kannenberg SA, Novick KA, Alexander MR, Maxwell JT, Moore DJP, Phillips Drew DM, Downes GM, Battaglia M. 2010. CAMBIUM, a process-based RP, Anderegg WRL. 2019b. Linking drought legacy effects across scales: from model of daily xylem development in Eucalyptus. Journal of Theoretical Biology leaves to tree rings to ecosystems. Global Change Biology 25: 2978–2992. 264: 395–406. Kannenberg SA, Novick KA, Phillips RP. 2019c. Anisohydric behavior linked to Eilmann B, Dobbertin M, Rigling A. 2013. Growth response of Scots pine with persistent hydraulic damage and delayed drought recovery across seven North different crown transparency status to drought release. Annals of Forest Science American tree species. New Phytologist 222: 1862–1872. 70: 685–693. Lazzarin M, Zweifel R, Anten N, Sterck FJ. 2019. Does phloem osmolality affect Enquist BJ, Niklas KJ. 2002. Global allocation rules for patterns of biomass diurnal diameter changes of twigs but not of stems in Scots pine? Tree partitioning in seed plants. Science 295: 1517–1520. Physiology 39: 275–283. Etzold S, Zieminska K, Rohner B, Bottero A, Bose AK, Ruehr NK, Zingg A, Lintunen A, Paljakka T, Jyske T, Peltoniemi M, Sterck F, von Arx G, Cochard Rigling A. 2019. One century of forest monitoring data in Switzerland reveals H, Copini P, Caldeira MC, Delzon S et al. 2016. Osmolality and non- species- and site-specific trends of climate-induced tree mortality. Frontiers in structural carbohydrate composition in the secondary phloem of trees across a Plant Science 10: 1–19. latitudinal gradient in Europe. Frontiers in Plant Science 7: 1–15. Fatichi S, Leuzinger S, Korner C. 2014. Moving beyond photosynthesis: from carbon Liu ZH, Ballantyne AP, Poulter B, Anderegg WRL, Li W, Bastos A, Ciais P. source to sink-driven vegetation modeling. New Phytologist 201: 1086–1095. 2018. Precipitation thresholds regulate net carbon exchange at the continental Feichtinger LM, Eilmann B, Buchmann N, Rigling A. 2014. Growth scale. Nature Communications 9: 1–10. adjustments of conifers to drought and to century-long irrigation. Forest Ecology Lockhart JA. 1965. An analysis of irreversible plant cell elongation. Journal of and Management 334: 96–105. Theoretical Biology 8: 264–275. Feichtinger LM, Eilmann B, Buchmann N, Rigling A. 2015. Trait-specific Ma J, Xiao XM, Miao RH, Li Y, Chen BQ, Zhang Y, Zhao B. 2019. Trends and responses of Scots pine to irrigation on a short vs long time scale. Tree controls of terrestrial gross primary productivity of China during 2000–2016. Physiology 35: 160–171. Environmental Research Letters 14: 1–14. Galiano L, Martinez-Vilalta J, Lloret F. 2011. Carbon reserves and canopy Martinez-Vilalta J, Cochard H, Mencuccini M, Sterck FJ, Herrero A, Korhonen defoliation determine the recovery of Scots pine 4 yr after a drought episode. JFJ, Llorens P, Nikinmaa E, Poyatos R, Ripullone F et al. 2009. Hydraulic New Phytologist 190: 750–759. adjustment of Scots pine across Europe. New Phytologist 184: 353–364. Gaudinski JB, Torn MS, Riley WJ, Swanston C, Trumbore SE, Joslin JD, Medlyn BE, Duursma RA, Eamus D, Ellsworth DS, Prentice IC, Barton CVM, Majdi H, Dawson TE, Hanson PJ. 2009. Use of stored carbon reserves in Crous KY, de Angelis P, Freeman M, Wingate L. 2011. Reconciling the growth of temperate tree roots and leaf buds: analyses using radiocarbon optimal and empirical approaches to modelling stomatal conductance. Global measurements and modeling. Global Change Biology 15: 992–1014. Change Biology 17: 2134–2144. Gessler A, Treydte K. 2016. The fate and age of carbon – insights into the Meinzer FC, Woodruff DR, Eissenstat DM, Lin HS, Adams TS, McCulloh KA. storage and remobilization dynamics in trees. New Phytologist 209: 1338–1340. 2013. Above- and belowground controls on water use by trees of different Giuggiola A, Zweifel R, Feichtinger LM, Vollenweider P, Bugmann H, Haeni wood types in an eastern US deciduous forest. Tree Physiology 33: 345–356. M, Rigling A. 2018. Competition for water in a xeric forest ecosystem – effects Mencuccini M, Grace J. 1996. Developmental patterns of above-ground of understory removal on soil micro-climate, growth and physiology of hydraulic conductance in a Scots pine (Pinus sylvestris) age sequence. Plant, Cell dominant Scots pine trees. Forest Ecology and Management 409: 241–249. & Environment 19: 939–948. Grossiord C, Sevanto S, Limousin JM, Meir P, Mencuccini M, Pangle RE, Mencuccini M, Salmon Y, Mitchell P, Holtta T, Choat B, Meir P, O’Grady A, Pockman WT, Salmon Y, Zweifel R, McDowell NG. 2018. Manipulative Tissue D, Zweifel R, Sevanto S et al. 2017. An empirical method that experiments demonstrate how long-term soil moisture changes alter controls of separates irreversible stem radial growth from bark water content changes in plant water use. Environmental and Experimental Botany 152: 19–27. trees: theory and case studies. Plant, Cell & Environment 40: 290–303. Guerin M, Martin-Benito D, von Arx G, Andreu-Hayles L, Griffin KL, Michelot-Antalik A, Granda E, Fresneau C, Damesin C. 2019. Evidence of a Hamdan R, McDowell NG, Muscarella R, Pockman W, Gentine P. 2018. seasonal trade-off between growth and starch storage in declining beeches: Interannual variations in needle and sapwood traits of Pinus edulis branches assessment through stem radial increment, non-structural carbohydrates and under an experimental drought. Ecology and Evolution 8: 1655–1672. intra-ring d13C. Tree Physiology 39: 831–844. Haeni M, Kn€ usel S, Wilhelm M, Peters RL, Zweifel R. 2020. treenetproc - clean, Myers BJ. 1988. Water-stress integral - a link between short-term stress and long- process and visualise dendrometer data. R package v.0.1.4. Github repository. term growth. Tree Physiology 4: 315–323. [WWW document] URL https://github.com/treenet/treenetproc [accessed 15 O’Brien MJ, Leuzinger S, Philipson CD, Tay J, Hector A. 2014. Drought December 2019]. survival of tropical tree seedlings enhanced by non-structural carbohydrate Herzog C, Steffen J, Pannatier EG, Hajdas I, Brunner I. 2014. Nine years of levels. Nature Climate Change 4: 710–714. irrigation cause vegetation and fine root shifts in a water-limited pine forest. Ogle K, Barber JJ, Barron-Gafford GA, Bentley LP, Young JM, Huxman TE, PLoS ONE 9: 1–11. Loik ME, Tissue DT. 2015. Quantifying ecological memory in plant and Hetherington AM, Woodward FI. 2003. The role of stomata in sensing and ecosystem processes. Ecology Letters 18: 221–235. driving environmental change. Nature 424: 901–908. Ortega JKE, Munoz CM, Blakley SE, Truong JT, Ortega EL. 2012. Stiff Hsiao TC, Xu LK. 2000. Sensitivity of growth of roots versus leaves to water mutant genes of Phycomyces affect turgor pressure and wall mechanical stress: biophysical analysis and relation to water transport. Journal of properties to regulate elongation growth rate. Frontiers in Plant Science 3: 1–12. Experimental Botany 51: 1595–1616. Peltier DMP, Barber JJ, Ogle K. 2018. Quantifying antecedent climatic drivers Huang MT, Wang XH, Keenan TF, Piao SL. 2018. Drought timing influences of tree growth in the Southwestern US. Journal of Ecology 106: 613–624. the legacy of tree growth recovery. Global Change Biology 24: 3546–3559. Peltier DMP, Ogle K. 2019. Legacies of more frequent drought in ponderosa Jiang P, Liu HY, Piao SL, Ciais P, Wu XC, Yin Y, Wang HY. 2019. Enhanced pine across the western United States. Global Change Biology 25: 3803–3816. growth after extreme wetness compensates for post-drought carbon loss in dry Petit G, Anfodillo T, Carraro V, Grani F, Carrer M. 2011. Hydraulic constraints forests. Nature Communications 10: 1–9. limit height growth in trees at high altitude. New Phytologist 189: 241–252. Jump AS, Ruiz-Benito P, Greenwood S, Allen CD, Kitzberger T, Fensham R, R Core Team. 2019. R: A language and environment for statistical computing. Martinez-Vilalta J, Lloret F. 2017. Structural overshoot of tree growth with Vienna, Austria: R Foundation for Statistical Computing. climate variability and the global spectrum of drought-induced forest dieback. Rathgeber CBK, Cuny HE, Fonti P. 2016. Biological basis of tree-ring Global Change Biology 23: 3742–3757. formation: a crash course. Frontiers in Plant Science 7: 1–7. Kannenberg SA, Maxwell JT, Pederson N, D’Orangeville L, Ficklin DL, Richardson AD, Carbone MS, Huggett BA, Furze ME, Czimczik CI, Walker Phillips RP. 2019a. Drought legacies are dependent on water table depth, JC, Xu XM, Schaberg PG, Murakami P. 2015. Distribution and mixing of old Ó 2020 The Authors New Phytologist (2020) 227: 1081–1096 New Phytologist Ó 2020 New Phytologist Trust www.newphytologist.com New 1096 Research Phytologist and new nonstructural carbon in two temperate trees. New Phytologist 206: Zweifel R, Sterck F. 2018. A conceptual tree model explaining legacy effects on 590–597. stem growth. Frontiers in Forests and Global Change 1: 1–9. Rigling A, Bigler C, Eilmann B, Feldmeyer-Christe E, Gimmi U, Ginzler C, Zweifel R, Zeugin F, Zimmermann L, Newbery DM. 2006. Intra-annual radial Graf U, Mayer P, Vacchiano G, Weber P et al. 2013. Driving factors of a growth and water relations of trees – implications towards a growth vegetation shift from Scots pine to pubescent oak in dry Alpine forests. Global mechanism. Journal of Experimental Botany 57: 1445–1459. Change Biology 19: 229–240. Sch€onbeck L, Gessler A, Hoch G, McDowell NG, Rigling A, Schaub M, Li MH. 2018. Homeostatic levels of nonstructural carbohydrates after 13 yr of Supporting Information drought and irrigation in Pinus sylvestris. New Phytologist 219: 1314–1324. Shinozaki K, Yoda K, Hozumi K, Kira T. 1964. A quantitative analysis of plant Additional Supporting Information may be found online in the form – the pipe model theory I. Basic analysis. Japanese Journal of Ecology 14: Supporting Information section at the end of the article. 97–105. Steppe K, Saveyn A, Vermeulen K, Lemeur R. 2006. A comprehensive model for Fig. S1 Distribution of annual precipitation over Switzerland simulating stem diameter fluctuations and radial stem growth. Acta Horticulture 718: 35–42. and the location of the research site Pfynwald. Sterck F, Zweifel R. 2016. Trees maintain a similar conductance per leaf area through integrated responses in growth, allocation, architecture and anatomy. Fig. S2 Frequency distribution of sapwood rings per tree and Tree Physiology 36: 1307–1309. sapwood widths. Szejner P, Belmecheri S, Ehleringer JR, Monson RK. 2019. Recent increases in drought frequency cause observed multi-year drought legacies in the tree rings of semi-arid forests. Oecologia 192: 241–259. Fig. S3 Sensitivity of model output to changes in the turnover Vester F. 2007. The Art of interconnected thinking: tools and concepts for a new rates of needles, sapwood and carbon reserves. approach to tackling complexity. Berlin, Germany: MC.B Verlag. von Arx G, Arzac A, Fonti P, Frank D, Zweifel R, Rigling A, Galiano L, Gessler Methods S1 Model equations. A, Olano JM. 2017. Responses of sapwood ray parenchyma and non-structural carbohydrates of Pinus sylvestris to drought and long-term irrigation. Functional Ecology 31: 1371–1382. Methods S2 Calculation of environmental index. Weber P, Bugmann H, Rigling A. 2007. Radial growth responses to drought of Pinus sylvestris and Quercus pubescens in an inner-Alpine dry valley. Journal of Table S1 Diameter at breast height and tree height for the pine Vegetation Science 18: 777–792. (Pinus sylvestris) trees equipped with dendrometer and sap flow Wehren B, Weingartner R, Schadler B, Viviroli D. 2010. General characteristics sensors and group into subplots with treatment. of alpine waters. In: Bundi U, ed. Alpine waters. Berlin/Heidelberg: Springer Nature, 17–58. Zweifel R, Bangerter S, Rigling A, Sterck FJ. 2012. Pine and mistletoes: how to Table S2 Model parameters and their values for the three differ- live with a leak in the water flow and storage system? Journal of Experimental ent treatments and the two scenarios. Botany 63: 2565–2578. Zweifel R, Haeni M, Buchmann N, Eugster W. 2016. Are trees able to grow in Table S3 Sensitivity analyses of model output to changes of the periods of stem shrinkage? New Phytologist 211: 839–849. Zweifel R, Item H, H€a sler R. 2000. Stem radius changes and their relation to turnover rates of needles, sapwood and carbon reserves. stored water in stems of young Norway spruce trees. Trees 15: 50–57. Zweifel R, Rigling A, Dobbertin M. 2009. Species-specific stomatal response of Please note: Wiley Blackwell are not responsible for the content trees to drought – a link to vegetation dynamics. Journal of Vegetation Science or functionality of any Supporting Information supplied by the 20: 442–454. authors. Any queries (other than missing material) should be Zweifel R, Steppe K, Sterck FJ. 2007. Stomatal regulation by microclimate and tree water relations: interpreting ecophysiological field data with a hydraulic directed to the New Phytologist Central Office. plant model. Journal of Experimental Botany 58: 2113–2131. New Phytologist (2020) 227: 1081–1096 Ó 2020 The Authors www.newphytologist.com New Phytologist Ó 2020 New Phytologist Trust