Drought intensity alters productivity, carbon allocation and plant nitrogen uptake in fast versus slow grassland communities

Abstract

1. Grasslands face more frequent and extreme droughts; yet, their responses to increasing drought intensity are poorly understood. Increasing drought intensity likely triggers abrupt shifts (thresholds) in grassland ecosystem functioning which can implicate recovery trajectories.
2. Here, we determined how drought intensity affects plant productivity, and plant–soil carbon (C) and nitrogen (N) cycling. We exposed model grassland plant communities with contrasting resource acquisition strategies (a fast- vs a slow-strategy plant community), to a gradient of drought intensity. The drought gradient ranged from well-watered to severely water-limited conditions. We identified thresholds of plant community productivity (above-ground biomass) at peak drought and 2 months after re-wetting, and measured net ecosystem exchange and ecosystem respiration of C throughout the drought and recovery phases. At peak drought and 1 week after re-wetting, we traced recently acquired C from plants to the soil and into microbial biomass and fatty acids using ¹³C pulse labelling, and measured plant and soil N.
3. At peak drought, slow-strategy plant communities were more drought resistant than fast-strategy communities, as the threshold in plant productivity occurred at a higher drought intensity for the slow- than the fast-strategy community. Shortly after re-wetting, microbial uptake of recent plant-assimilated C increased with increasing past drought intensity, coinciding with an increase in soil N availability and leaf N. Threshold responses to drought intensity at peak drought translated into non-linear recovery responses, with greater compensatory growth in the fast-strategy community. At peak drought, increasing drought intensity reduced C uptake and increased relative C partitioning to leaves and microbial biomass. Upon re-wetting, plant community strategy mediated drought intensity effects on plant and soil C and N dynamics and plant recovery trajectories. The fast-strategy community recovered quickly, with higher leaf N than the slow community, while the slow community increased C allocation to microbial biomass.
4. Synthesis. Our findings highlight that C and N dynamics in the plant–soil system display non-linear responses to increasing drought intensity both during and after drought, which has implications for plant community recovery trajectories.

1 INTRODUCTION

The increasing frequency and intensity of extreme weather events, such as drought, heat waves, and floods, threatens ecosystems (IPCC, 2021). The consequences for ecosystem functioning are difficult to predict, as ecological responses to extreme stress are often non-linear (Reichstein et al., 2013; Sippel et al., 2018) and can involve thresholds, defined as an abrupt shift in the state of an ecosystem (Groffman et al., 2006; Hillebrand et al., 2020; Turner et al., 2020). Although thresholds are likely key to understanding the ecological consequences of drought, few studies have explicitly addressed them. This could be due in part to the significant challenges involved in empirically detecting threshold responses (Hillebrand et al., 2020). Recently, thresholds were detected in the response trajectories of individual grassland plant species to drought: abrupt decreases in drought resistance were related with non-linear post-drought overcompensation responses (Ingrisch et al., 2023). Increasing drought intensity could shift plant–soil carbon (C) and nitrogen (N) dynamics, which could alter the plant community's ability to recover after drought (Karlowsky, Augusti, Ingrisch, Hasibeder, et al., 2018). Currently, productivity response to increasing drought intensity in mixed species plant communities is not well understood, and data on shifts in plant–soil C and N dynamics with increasing drought intensity are scarce.

A plant's position on the fast–slow resource economic spectrum has been suggested to inform plant drought response (Grime & Mackey, 2002; Volaire, 2018) and likely shapes how grassland communities respond to increasing drought intensity. The fast–slow resource economic spectrum is a well-defined framework that connects suites of functional traits to define a plant's growth and survival strategy (Grime, 1977; Reich, 2014). Slow-growing/conservative plants that invest in durable tissues better resist drought (Blumenthal et al., 2020; Ingrisch et al., 2018; Karlowsky, Augusti, Ingrisch, Hasibeder, et al., 2018; Pérez-Ramos et al., 2013), while fast-growing/acquisitive plants with N-rich tissues can regrow faster after drought stress has passed (Grime et al., 2000; Karlowsky, Augusti, Ingrisch, Hasibeder, et al., 2018; Lepš et al., 1982; Wilcox et al., 2021). Lesser known is how a plant community's resource acquisition strategy affects productivity thresholds with increasing drought intensity during drought, and productivity recovery afterwards. Abrupt decreases in plant productivity during drought are likely to already occur under mild drought intensities for fast-strategy plant communities as they prioritize growth over durability (Oliveira et al., 2021). Fast-strategy plants are also generally established in environments that are less water and nutrient limited compared to their slow-strategy counterparts (Padullés Cubino et al., 2022; Pérez-Ramos et al., 2012).

A plant community's resource acquisition strategy has been shown to affect plant–soil C cycling (Freschet et al., 2012) and N cycling (Abalos et al., 2019; Grassein et al., 2015; Legay et al., 2014). Drought can modify plant–soil C and N cycling (Chomel et al., 2019; Fuchslueger et al., 2014; Karlowsky, Augusti, Ingrisch, Hasibeder, et al., 2018) by decoupling plant–microorganism interactions (Karlowsky, Augusti, Ingrisch, Akanda, et al., 2018; Rudgers et al., 2020). This disconnect has been attributed to shifts in rhizodeposition, which are known to differ between fast- and slow-strategy plants (Williams & de Vries, 2020). A key outstanding question, however, is whether decoupling of plant–microorganism interactions occurs at different drought intensities for fast- versus slow-strategy plant communities, and whether this has implications for plant community drought resistance and recovery. Under ambient conditions, fast-strategy plants take up more C (CO₂) and have higher rhizodeposition than slow-strategy plants (Henneron, Fontaine, et al., 2020; Henneron, Kardol, et al., 2020), leading to accelerated N cycling in the rhizosphere and higher plant N uptake compared (Kaštovská et al., 2015). Moreover, during drought, plant communities dominated by slow-strategy plant species have been found to allocate more C to root storage and less to soil respiration than communities dominated by fast-strategy plants (Ingrisch et al., 2020; Karlowsky, Augusti, Ingrisch, Hasibeder, et al., 2018). Post-drought, plant communities dominated by fast-strategy plants recover quicker due to a higher capacity for N uptake (Ingrisch et al., 2018; Karlowsky, Augusti, Ingrisch, Hasibeder, et al., 2018). The reported shifts in C allocation during and after drought (Hartmann et al., 2020) determine which functions can be prioritized, for example, storing C in roots for re-growth (Ingrisch et al., 2020; Karlowsky, Augusti, Ingrisch, Hasibeder, et al., 2018), exuding it in the rhizosphere to maintain resource acquisition (Henneron, Fontaine, et al., 2020; Henneron, Kardol, et al., 2020), or allocating to mycorrhizal partners (Sanaullah et al., 2012). Although drought intensity likely influences plant–soil C and N dynamics during drought and recovery, these relationships are not well known.

Here, we determined how two experimental grassland plant communities with contrasting resource acquisition strategies (i.e. fast/acquisitive vs. slow/conservative strategy), respond to increasing drought intensity. In an outdoor mesocosm experiment, we used a ¹³C pulse-labelling approach to track freshly assimilated C from plants to soil and the microbial community at peak drought and during recovery. Threshold regression modelling was used to identify drought intensity thresholds in plant productivity at peak drought and during recovery. We explored shifts in plant–soil C and N dynamics below and above these productivity thresholds and tested how these dynamics differed between plant communities with contrasting strategies at peak drought and during recovery. We hypothesized that (1) there are thresholds in plant community productivity in response to increasing drought intensity, with the fast-strategy community being less resistant than the slow strategy community; (2) thresholds in post-drought plant productivity overcompensation coincide with thresholds in peak-drought plant productivity, occurring at a lower drought intensity for the fast- than the slow-strategy plant community; (3) increasing drought intensity at peak drought causes an abrupt shift in plant and soil C allocation, with the fast-strategy community increasing C allocation to soil microorganisms and the slow community retaining more C in roots; and (4) increasing drought intensity shifts plant–soil N dynamics after re-wetting, resulting in the fast-strategy community acquiring more N soon after re-wetting than the slow-strategy community.

2 MATERIALS AND METHODS

2.1 Experimental setup

An outdoor mesocosm experiment was established at the Botanical Garden, University of Innsbruck, Austria (47°16′04.1″N 11°22′46.3″E), which included five sets of mesocosms (7 L plastic pots, 21 cm Ø, 25 cm height, n = 210). Two sets were destructively harvested throughout the experiment at peak drought (set 1), early recovery (set 2), and the final three sets were used for non-destructive measurements throughout the experiment and to monitor recovery (sets 3–5), Figure S1. The drought treatment followed a gradient design, which optimizes the use of experimental units to determine thresholds and non-linear responses (Kreyling et al., 2018). We created the drought intensity gradient by determining the soil water content (SWC, g_water g_{fresh soil}⁻¹) at field capacity and then established a gradient of increasing drought intensity (increasing soil water deficit compared to field capacity, SWD) ranging from well-watered controls (20% SWD) to no water added in the drought period (98% SWD). This corresponded with a realized SWD of 18% to 95% (Figure S2B) and a realized SWC of 0.266–0.020 (g_water g_{fresh soil}⁻¹). The drought took place 21 July to 12–15 August 2020, when a rainout-shelter was installed over all mesocosms (2.5 m aluminium frame covered with light and UV-B permeable plastic, Lumisol clear AF; Folitec). The drought intensity treatment (SWD) was maintained by weighing and watering the mesocosms 5–7 times per week (Figure S2A). The drought ended when mesocosms were re-wetted to their control weight (20% SWD) over 3 days.

2.2 Plants and soil

Perennial plant species common to European mesotrophic grasslands were selected based on their traits related to the resource economic spectrum and assembled to create two model grassland plant communities with contrasting strategies: a fast-strategy community (high specific leaf area and leaf N content and low leaf dry matter content) and slow-strategy community (opposite trait values) (Reich, 2014). The fast-strategy community included the grasses Dactylis glomerata, Lolium perenne and Phleum pratense, and the forbs Leontodon hispidus, Plantago lanceolata and Rumex acetosa. The slow-strategy plant community included the grasses Anthoxanthum odoratum, Briza media and Festuca rubra and the forbs Campanula rotundifolia, Leucanthemum vulgare and Prunella vulgaris. Species were selected based on a priori trait values reported in Baxendale et al. (2014), De Long et al. (2019), and de Vries and Bardgett (2016). For confirmation, we measured species-specific leaf dry matter content, specific leaf area and leaf N content of the control communities (n = 4) at the peak drought sampling campaign according to Pérez-Harguindeguy et al. (2016). The trait values are reported in Table S1.

Seeds were sourced commercially (D. glomerata, F. rubra, L. perenne and P. pratense from Barenbrug BV, the Netherlands; B. media and L. hispidus from Cruydt Hoek, the Netherlands, and A. odoratum, C. rotondofolia, L. vulgare, P. lanceolata, P. vulgaris and R. acetosa from Jelitto, Germany). Seeds were surface sterilized (1:1 household bleach: tap water for 20 min, then rinsed with tap water), germinated in the same soil used in the experiment, grown for 2 weeks, and then transplanted.

Soil for the experiment was provided from the Botanical Garden, University of Innsbruck. Background soil physiochemical factors of the bulk soil were determined (see Supplementary Methods). The soil was a sandy loam: 53.5% sand (50–2000 μm), 35.6% silt (2–50 μm) and 10.8% clay (<2.0 μm). Initial chemical properties were as follows: 0.29% total N, 1.1 g kg⁻¹ plant available P, 3.575 plant available g kg⁻¹ K, 6.12% total C (i.e. organic and inorganic C), 7.57% organic matter (loss on ignition method) and a pH_CaCl2 of 7.67. The soil was sieved to 1 cm, mesocosms were filled with fresh soil (5.76 kg dw equivalent), moistened to a 20% SWD (0.248 g_water g_{fresh soil}⁻¹) and the weight was recorded. Two seedlings per plant species (12 individuals pot⁻¹, 346 individuals m⁻²) were transplanted 27 May–1 June 2020. Dead seedlings were replaced over 10 days. Mesocosms were maintained at 20% SWD until the drought and fertilized (20 kg N ha⁻¹ as urea) on 22 June 2020. Climatic conditions are reported from a nearby climate station (TAWES UIBK), provided by the Austrian Weather Service ZAMG (https://www.zamg.ac.at/) and the Department of Atmospheric and Cryospheric Sciences, University of Innsbruck (Figure S3A).

2.3 Plant biomass

Species-specific above-ground biomass was harvested 1 week before the drought started, at peak drought, and c. 2 months after re-wetting. Above-ground biomass was cut to 3 cm above the soil surface, dried at 60°C for 72 h and weighed. Below-ground biomass was determined at peak drought and early recovery (12 days after re-wetting) by washing soil from roots over a 0.5 mm sieve. Clean roots were dried at 60°C for 72 h and weighed.

2.4 Non-destructive monitoring

We measured net ecosystem exchange (NEE), ecosystem respiration (ER), normalized difference vegetative index (NDVI) and canopy height 18 times throughout the experiment. Measurements began on 20 July 2020, and continued through to 9 October 2020. Gross primary productivity (GPP_sat) was estimated with the paired measurements of NEE and ER (see Supplementary Methods).

2.5 Leaf and root N

Community leaf N concentration was determined by harvesting the youngest, fully expanded leaf of each plant species at peak drought and 12 days after re-wetting, followed by drying at 60°C for 72 h. At peak drought, leaves were pooled per mesocosm based on the species-specific relative abundance of their above-ground biomass. Twelve days after re-wetting, leaf N concentration was determined at the species level and then the community-weighted mean was calculated. Samples were ground to a fine powder using a Tissue Lyser II (Qiagen), weighed into tin cups (art.no. 176.9811.26; IVA Analysetechnik GmbH & Co.KG) and analysed with a Flash EA1112 elemental analyser (Thermo Electron Corporation). Community root N concentration was determined in the same way. Peak drought whole plant N pool (g N m⁻² ground area) was calculated as N concentration in above-ground biomass (leaves, given the very minor fraction of stems present in the mesocosms) and below-ground biomass (roots) multiplied by above-ground/below-ground biomass harvested at peak drought. Whole plant N uptake is the N acquired in the 12 days following re-wetting (i.e. the sum of above-ground and below-ground N pools at early recovery minus the belowground N pool at peak drought).

2.6 Soil N and pH

Ammonium (NH₄⁺-N) and nitrate (NO₃⁻-N) were determined according to Schinner et al. (1996) by shaking 7.5 g fresh soil in 30 mL 0.0125 M CaCl₂ for 1 h at 170 rpm, filtering (Macherey & Nagel 615¼, 150 mm filter paper) and analysing on a spectrophotometer (Hitachi U-2001) at 660 nm and 210 nm, respectively. Dissolved N was quantified on a TOC-L/TNM-L analyser (Shimadzu Co., Japan) after extracting 7.5 g fresh soil in 30 mL distilled water by shaking for 30 min at 120 rpm.

2.7 ¹³C pulse labelling and sampling campaigns

2.7.1 Pulse labelling

We performed a ¹³C-pulse-labelling campaign at peak drought (after 21 days of drought) and at early recovery (7 days after re-wetting) in line with Karlowsky, Augusti, Ingrisch, Hasibeder, et al. (2018) and Ingrisch et al. (2020). At each campaign, mesocosms were fit into a custom-built, air-tight frame and chamber. At peak drought fast- and slow-strategy communities were pulse labelled on 11 and 12 August 2020, respectively. At early recovery, all plant communities were pulse labelled on 20 August 2020. Microclimatic conditions were similar on all 3 days (Figure S3B). Inside the chamber, air was ventilated with fans and temperature regulated by circulating ice-cold water through tubes. During labelling, we continuously monitored CO₂ concentration and ¹³C isotope ratio (G2101i Analyzer; Picarro Inc., USA), air temperature and humidity (HMP 75; Vaisala, Finland), and photosynthetically active radiation, PAR (PQS, Kipp & Zonen, Germany). Labelling took place for 90 min between 10 am and 1 pm, beginning when PAR reached 1500 μmol m⁻² s⁻¹. When CO₂ dropped to 250 ppm, pulses of highly enriched ¹³CO₂ (99.00 atom-% ¹³C, Sigma-Aldrich) were added with a syringe to maintain approximately 50 atom-% ¹³C and 500–600 ppm CO₂ (Table S2).

2.7.2 Sampling

Natural abundance $\mathrm{\delta }$¹³C in leaves, roots, microbial biomass C, extractable organic C (EOC), and phospholipid fatty acid (PLFA) and neutral lipid fatty acid (NLFA) were determined the day prior to labelling on a separate set of mesocosms. Leaves were harvested at the species level, immediately treated by microwave to stop metabolic activity, and dried at 60°C for 72 h. Two soil cores per pot (2 cm Ø, 25 cm depth) were taken with a core, pooled and sieved to 2 mm. Soil was divided and stored at 4°C for microbial biomass C or at −80°C for PLFA and NLFA analyses. Roots that remained on the sieve were washed, treated by microwave and dried at 60°C for 72 h.

After pulse labelling, we sampled shortly after the chamber was opened (c. 15 min for leaves, 20–120 min for soil and 25–140 min for roots), after 48 h, and after 120 h. Each sampling campaign was carried out in the same way as the natural abundance δ¹³C sampling.

2.8 Plant leaf and root isotopic C composition

The δ¹³C of leaves and roots was determined by grinding dried biomass into a fine powder, weighing into tin cups (art.no. 176.9811.26; IVA Analysetechnik GmbH & Co.KG) and measuring total C and δ¹³C with elemental analysis (EA)—isotope ratio mass spectrometry (IRMS; EA 1100, CE Elantech; coupled to a Delta+ IRMS; Finnigan MAT). Leaves were pooled per community based on the relative abundance of species-specific leaf biomass and capsulized the tin cups.

2.9 NLFA and PLFA content and C isotopic composition

We characterized the uptake of recent plant-derived C into broad microbial groups by determining ¹³C incorporation into PLFAs and NLFAs, which were extracted from frozen soil according to Bligh and Dyer (1959) and Karlowsky, Augusti, Ingrisch, Akanda, et al. (2018) (see Supplementary Methods). Briefly, fatty acid methyl esters (FAMEs) were quantified by gas chromatography–flame ionization detection. Compounds were identified and biomass was derived from peak area using a standard curve with increasing concentrations of a mixture of known FAMEs (Supelco 37 Component FAME Mix; Sigma-Aldrich Chemie GmbH; BR2 and BR4 mixture, Larodan Fine Chemicals AB). FAME ¹³C isotope content was corrected for the methyl group introduced during derivatisation, and for offset and drift. We used the sum of PLFA markers i14:0, i15:0, a15:0, i16:0, a17:0, i17:0 and br18:0 for Gram-positive bacteria (Zelles, 1999), 10Me16:0 and 10Me18:0 for Gram-positive actinobacteria (Lechevalier et al., 1977), 16:1ω7 and 18:1ω7 for Gram-negative bacteria (Zelles, 1997, 1999), 18:2ω6,9c for saprotrophic fungi (Frostegård & Bååth, 1996), and the NLFA 16:1ω5 as a marker for arbuscular mycorrhizal fungi (AMF) (Mellado-Vázquez et al., 2016; Olsson, 1999). Total fungal abundance was calculated using the NLFA 16:1ω5 and PLFAs 16:1ω5, 18:2ω6,9c. Total bacteria included all markers representing bacteria (above) plus 14:0 and 15:0 (Canarini et al., 2021).

2.10 C isotopic composition of soil-EOC and microbial biomass carbon

EOC and microbial biomass C were determined using chloroform fumigation method (Voroney et al., 2007), see Supplementary Methods. Microbial biomass C was calculated as:

$$\text{Microbial biomass}\mathrm{C}=\left(\mathrm{EOC}\text{fumigated}-\mathrm{EOC}\text{nonfumigated}\right)/0.45,$$

where 0.45 is the extraction efficiency of microbial biomass C after chloroform fumigation (Vance et al., 1987). The δ¹³C signatures of EOC were determined after Scheibe et al. (2012): 900 μL extract was acidified with 35 μL 10% HCl, flushed with N₂ for 15 min and analysed in triplicate with HPLC-IRMS (Dionex UltiMate 3000 UHPLC coupled via a LC-IsoLink system to a Delta V Advantage IRMS, Thermo Fisher Scientific).

2.11 C isotopic composition of soil respiration

Soil respiration and its isotopic composition were continuously measured for 5 days after labelling using a custom-made measurement setup (see Ingrisch et al., 2020), modified to consecutively measure 23 soil respiration chambers with a C isotope analyser (Picarro G2131-i; Picarro Inc.), see Supplementary Methods.

2.12 Calculation of incorporated ¹³C

Incorporated ¹³C was calculated as the total amount of ¹³C in a plant, soil or microbial C pool in line with Hafner et al. (2012):

$$R_{\text{sample}}=\left(\frac{\mathrm{\delta }{}^{13}\mathrm{C}}{1000}\times R_{\text{VPDB}}\right)+R_{\text{VPDB}},$$(1)

$$\chi {\left({}^{13},\mathrm{C}\right)}_{\text{sample}}=\left(\frac{R\text{sample}}{R\text{sample}+1}\right),$$(2)

$$\mathit{\chi E}\left({}^{13},\mathrm{C}\right)=\chi {\left({}^{13},\mathrm{C}\right)}_{\text{sample}}-\chi {\left({}^{13},\mathrm{C}\right)}_{\text{natural abundance}},$$(3)

$$\text{Incorporated}13{\mathrm{C}}_t=\mathit{\chi E}\left({}^{13},\mathrm{C}\right)\times \text{Cpool},$$(4)

$$\chi {\left({}^{13},\mathrm{C}\right)}_{\text{microbial biomass}\mathrm{C}}=\frac{\chi {\left({}^{13},\mathrm{C}\right)}_{\mathrm{TOC}}\times \mathrm{TOC}-\chi {\left({}^{13},\mathrm{C}\right)}_{\mathrm{EOC}}\times \mathrm{EOC}}{\left(\mathrm{TOC}-\mathrm{EOC}\right)},$$(5)

where (1) R_VPDB = 0.011180, the ¹³C/¹²C ratio in Vienna Pee Dee Belemnite, (2) $\chi {\left({}^{13},\mathrm{C}\right)}_{\text{sample}}$ is the ¹³C atom fraction, (3) $\chi {\left({}^{13},\mathrm{C}\right)}_{\text{microbial biomass}\mathrm{C}}$ is calculated according to the isotopic mass balance (4) $\mathit{\chi E}\left({}^{13},\mathrm{C}\right)$ is the ¹³C enrichment in a C pool, derived from subtracting the atom fraction of ¹³C in the natural abundance sample from fraction of ¹³C in the enriched sample, and (5) incorporated ¹³C_t is the excess atom fraction of ¹³C multiplied by the respective C pool at a given time (t).

Soil respiration rate (μmol m⁻² s⁻¹) was calculated as:

$$\mathrm{SR}=\frac{f\times \left({\mathrm{CO}}_{2\text{out}}-{\mathrm{CO}}_{2\text{in}}\right)}{A},$$(6)

where $f$ is the flow rate through the chamber, CO_2out and CO_2in are the mean concentrations at the chamber outlet and in the buffer volume, respectively, and $A$ the area of the chamber (m²).

The isotopic composition of soil respiration χ(¹³C)_SR:

$$\chi {\left({}^{13},\mathrm{C}\right)}_{\mathrm{SR}}=\frac{\chi \left({}^{13}\mathrm{C}\right)\text{out}\times {\mathrm{CO}}_{2\text{out}}-\chi \left({}^{13}\mathrm{C}\right)\text{in}\times {\mathrm{CO}}_{2\text{in}}}{{\mathrm{CO}}_{2\text{out}}-{\mathrm{CO}}_{2\text{in}}},$$(7)

where χ(¹³C)_in and χ(¹³C)_out denote the atom fraction of ¹³CO₂ in the buffer volume and the chamber outlet, respectively, which were calculated from the mean concentrations of the two isotopologues ¹²CO₂ and ¹³CO₂.

The absolute rate of ¹³C label efflux in soil respiration (mg ¹³C m⁻² h⁻¹) is calculated as:

$$\text{incorporated}{}^{13}{\mathrm{C}}_{\mathrm{SR}}=\mathit{\chi E}{\left({}^{13},\mathrm{C}\right)}_{\mathrm{SR}}\times \mathrm{SR}.$$(8)

The cumulative amount of respired ¹³C label was calculated for each mesocosm according to the trapezoid rule (linear interpolation between consecutive measurements).

Relative partitioning of ¹³C was calculated by dividing the incorporated ¹³C in one compartment and time point by the total ¹³C measured in all compartments at T0.

2.13 Data analysis

Statistical analyses were performed in R version 4.1.0 (R Core Team, 2020). Figures were made using ggplot (Wickham, 2016), cowplot (Wilke, 2019) and patchwork (Pedersen, 2020). Productivity thresholds were identified using threshold regression models using the r-package chngpt (Fong et al., 2017). We considered two types of threshold responses: (i) continuous thresholds that change the relationship of productivity with SWD and (ii) discontinuous thresholds that cause abrupt changes in the value of productivity (Berdugo et al., 2020; Groffman et al., 2006). Continuous thresholds were modelled by segmented regressions (change in the model slope at a threshold) and discontinuous thresholds by step regressions (change in model intercept at a threshold). Threshold models were only considered if Akaike information criterion (AIC) was smaller than corresponding linear and quadratic regressions. Thus, the presented threshold models have superior data fit than do linear or quadratic regressions. Threshold estimations for the best model type were repeated with 1000 bootstrapped samples to estimate the distribution of each threshold.

To determine the effect of drought intensity and plant community on all dependent variables, we tested the effect of SWD (continuous), plant community (ordered factor) and their interaction using generalized additive models (GAMs) with the function gam from the r-package mcgv (Wood, 2011). GAMs allow for linear and non-linear relationships, reported in the edf statistic, where 1.0 is a linear relationship. We modelled the following relation:

$$Y={\beta }_0+{\beta }_1\left(\text{plant community}\right)+{\beta }_1\left(\mathrm{SWD}\right)+f_1\left(\mathrm{SWD},\text{plant community}\right),$$

where ${\beta }_0$ represents the intercept, ${\beta }_{1\left(\text{plant community}\right)}$ the effect of the plant community (fast- or slow-strategy), ${\beta }_{1\left(\mathrm{SWD}\right)}$ the overall effect of SWD and $f_1\left(\mathrm{SWD},\text{plant community}\right)$ the interactive effect between SWD and plant community. When a significant interaction was detected, we determined the effect of SWD within each plant community by modelling the relation (where plant community is considered a factor: fast or slow strategy):

$$Y={\beta }_0+{\beta }_1\left(\text{plant community}\right)+f_1\left(\mathrm{SWD},\text{plant community}\right).$$

Model fit was checked using gam.check from the r-package mcgv (Wood, 2011). Models with different smoothers and families were compared using AIC and the best fit model (lowest AIC) was retained. We tested how SWD affected overall ¹³C incorporation into PLFAs and NLFA of interest using principle component analysis (PCA) with the function pca from the r-package FactoMineR (Le et al., 2008) on scaled (mean 0, sd ± 1) data.

3 RESULTS

3.1 Drought intensity effects on plant productivity

Increasing drought intensity decreased above-ground biomass at peak drought and increased above-ground biomass 2 months post-drought in both plant communities with distinct thresholds (Figure 1a,b). The threshold responses were best described by step-regression models, which reflect an abrupt shift in the relationship between above-ground biomass and drought intensity. At peak drought, thresholds were detected at 65% SWD for the fast-strategy community and 75% SWD for the slow-strategy community, where above-ground biomass sharply decreased by 37.3% and 31.4%, respectively. During the 3-week drought, there were no significant interaction effects between drought intensity and plant community strategy on GPP, NDVI or canopy height (Figures S4–S6; Tables S6–S8). Thus, the slow community was more drought resistant than the fast-strategy community, in terms of above-ground biomass, but this did not translate into other measures of productivity.

FIGURE 1

Drought intensity (expressed as soil water deficit, %) effects on plant community above-ground biomass (living and senesced plant tissues) at (a) peak drought (after a 3-week drought) and (b) recovery (c. 2 months after re-wetting) for the fast- and slow-strategy community. Points indicate individual mesocosms and lines denote the best-fit thresholds regression models and their 95% confidence intervals. Colour density of the shaded boxes indicates the bootstrapped distribution of threshold estimates with the best-fit threshold indicated by the point.

Two months after re-wetting, a threshold at 75% SWD increased above-ground biomass by 38.6% and 39.2% for the fast- and slow-strategy community, respectively (Figure 1b). Considering all communities that experienced a drought intensity above the threshold detected at 75% SWD (Figure 1b), the fast-strategy community produced significantly more above-ground biomass than the slow-strategy community (F_1,58 = 4.95, p < 0.05). After re-wetting, the recovery responses of the fast- and the slow-strategy community diverged, and overshoot in GPP and canopy height began sooner for the fast- than the slow-strategy community (Figures S4 and S6), while NDVI recovery responses were similar (Figure S5).

3.2 Plant/soil carbon dynamics

At peak drought, both plant communities responded similarly to increasing drought intensity, which decreased C uptake and incorporation into roots, EOC and microbial biomass C in soil (Figure 2; Table S3). Increasing drought intensity increased the relative amount of ¹³C partitioned to leaves and microbial biomass, while ¹³C partitioned to roots decreased (Figure S7; Table S9). Overall, the fast-strategy community partitioned more ¹³C to roots than the slow community (Figures S7D). In both plant communities, increasing drought intensity decreased ¹³C incorporation into PLFA and NLFAs, with the exception of PLFAs representing actinobacteria which increased (Figure 3; Table S5). ¹³C incorporation into PLFAs/NLFA in communities that experienced a drought above the productivity threshold (identified in Figure 1) were separated from those that experienced a milder drought or control (Figure S11A).

FIGURE 2

Incorporated ¹³C in (a) leaves directly after labelling at peak drought and (b) at early recovery, and 5 days after pulse labelling in (c, d) roots, (e, f) microbial biomass C, (g, h) extractable organic C and (i, j) cumulative soil-respired ¹³CO₂. Lines indicate the best fit generalized additive model (GAM), and shaded area around the line depicts the model-predicted 95% confidence interval. Significance is noted between the plant communities (Community), over the drought gradient (SWD) and the interaction (SWD: community). Within community significance is shown with a significant interaction (fast, slow). p < 0.001***, p < 0.001**, p < 0.05. R² adjusted indicates the explained variation of the GAMs (Table S4).

FIGURE 3

Drought intensity (soil water deficit) effects on incorporated ¹³C at peak drought in phospholipid fatty acids (PLFAs) representing (a) actinobacteria, (c) Gram-negative bacteria, (e) Gram-positive bacteria, and (i) saprotrophic fungi, and (g) the neutral lipid fatty acid (NLFA) representing arbuscular mycorrhizal fungi (AMF), and at early recovery (b, d, f, h, j). For list of PLFA and NLFAs considered, see Section 2. p values show the effect of drought intensity (soil water deficit, SWD), plant community (Community), and their interaction on the response variable based on a generalized additive model (GAM). If a significant interaction was found, a second GAM model was used to test the relationship between drought intensity and the response variable within each plant communities. p < 0.001***, p < 0.01**, p < 0.05*. R² adjusted indicates the explained variation of the GAM model, for full statistical output, see Table S5.

Recovery trajectories differed between fast- and slow-strategy communities. Seven days after re-wetting, ¹³C incorporation in leaves and roots of the slow-strategy community significantly decreased with increasing past drought intensity, while the fast-strategy community was no longer affected (Figure 2b,d). In both plant communities, past drought intensity increased ¹³C incorporated into extractable organic and microbial biomass C. Soil respired ¹³C (cumulative over 5 days) decreased with increasing past drought intensity in both communities (Figure 2f,h,j). In the slow community only, increasing past drought intensity increased the relative partitioning of ¹³C to microbial biomass and decreased ¹³C partitioned to leaves (Figure S8). After re-wetting, the soil water deficit during the drought significantly affected ¹³C incorporation into fatty acids (Figure 3). Past drought intensity increased ¹³C incorporation into PLFAs representing actinobacteria in both plant communities (Figure 3b) and increased ¹³C incorporated into PLFAs representing Gram-positive bacteria in the fast community only (Figure 3f). ¹³C incorporation into the NLFA representing AMF decreased with past drought intensity (Figure 3h), while ¹³C incorporation into PLFA and NLFAs representing Gram-negative bacteria or saprotrophic fungi were no longer affected (Figure 3d,j). The ¹³C incorporated into the PLFAs/NLFA in communities that experienced a drought above the productivity threshold (identified in Figure 1) were grouped and separated from those that experienced a milder drought or control (Figure S11B).

3.3 Plant/soil N dynamics

At peak drought, increasing drought intensity affected soil N dynamics similarly for both plant communities. Soil NO₃⁻-N was non-linearly related to increasing drought intensity, first decreasing and then increasing at the highest drought intensities (Figure S9A). Soil NH₄⁺-N and dissolved N increased with increasing drought intensity in both communities (Figure S9B,C). At peak drought, leaf N concentration was not affected by drought intensity (Figure 4a; Table S4), while fast community root N concentration increased with increasing drought intensity (Figure 4c). Whole plant N pool (g N m⁻² ground area) decreased for both plant communities with increasing drought intensity (Figure S10A; Table S4).

FIGURE 4

Nitrogen (N) concentration in leaves at (a) peak drought and (b) early recovery (the soil water deficit refers to the soil water deficit at peak drought), and in roots at (c) peak drought and (d) early recovery.

Twelve days after re-wetting, we found that past drought intensity increased NO₃⁻-N and dissolved N in both communities, while soil NH₄⁺-N was not affected (Figure S9D–F). Leaf N concentration increased with past drought intensity in both communities (Figure 4b), and drought intensity no longer affected root N concentration (Figure 4d). Whole plant N uptake (g N m⁻² ground area) increased with increasing past drought intensity for the fast-strategy community only. The slow-strategy plant community's whole plant N uptake was not affected by increasing past drought intensity (Figure S10C).

4 DISCUSSION

4.1 Drought intensity thresholds of productivity

Thresholds in grassland response to increasing drought intensity are poorly understood, despite playing a key role in informing ecosystem functioning during and after drought (Grünzweig et al., 2022; Turner et al., 2020). Using model grassland communities established in outdoor mesocosms, we identified distinct thresholds in plant community productivity in response to increasing drought intensity during drought and recovery. We found that abrupt decreases in plant community productivity during drought were coupled to productivity overcompensation during recovery, as previously demonstrated for monocultures of two grassland plant species (Ingrisch et al., 2023). Thus, drought intensity plays an important role in determining plant community drought response both during drought and after re-wetting (see Figure 5 for an overview of our findings).

FIGURE 5

Summary of key findings of our major response variables: above-ground (AG) biomass, plant and microbial C uptake and leaf/root N concentration. At peak drought, increasing drought intensity reduced AG biomass, plant C uptake and microbial C uptake in both (a) fast- and (b) slow-strategy plant communities. Root N concentration (%) increased with increasing drought intensity in the fast-strategy community only. The fast-strategy community was less resistant, reaching the productivity threshold at a lower drought intensity (65% soil water deficit, SWD) than (b) the slow-strategy community (75% SWD), indicated by the black triangle. Post-drought, microbial C uptake increased with past increasing drought intensity in both plant communities soon after re-wetting (c, d), signalling a fast microbial recovery. (c) The fast-strategy community recovered quicker, and began overshooting in terms of plant C uptake sooner (i.e. within 10 days of re-wetting) than (d) the slow-strategy community (the ‘carbon uptake’ lines indicate the C taken up soon after re-wetting). Leaf N concentration (%) in both communities increased with increasing past drought intensity. Two months post-drought, we found thresholds in overshoot in both plant communities (indicated by the black triangle). Lines indicate a significant relationship was found, absence of the line indicate no relationship. Figure made with BioRender, drawings by NJO.

Confirming our first hypothesis, the threshold in plant community productivity at peak drought occurred at a lower drought intensity for the fast- than the slow-strategy community (Figure 1). This indicates that the fast-strategy community was less drought resistant than the slow-strategy community, broadly consistent with earlier studies which considered one level of drought intensity (Ingrisch et al., 2018; Karlowsky, Augusti, Ingrisch, Hasibeder, et al., 2018; Pérez-Ramos et al., 2013). Fast-growing plant species are more vulnerable to drought stress because of the trade-off between traits that enable fast resource acquisition and those that facilitate hydraulic safety, for example, leaf density and osmotic potential (Díaz et al., 2016; Reich, 2014; Wilcox et al., 2021). In contrast, slow-growing plant species have a stress-tolerating strategy (Grime & Mackey, 2002) that allows them to reduce water use, avoid turgor loss and maintain lower levels of growth throughout the drought, conferring higher resistance (Pérez-Ramos et al., 2013).

After the drought ended, the fast-strategy community had greater capacity for recovery, confirming our second hypothesis. Post-drought, the fast-strategy community recovered quicker than the slow-strategy community in terms of GPP, and 2 months after re-wetting had significantly more above-ground biomass (i.e. compensatory growth) above the productivity threshold. The quicker recovery of the fast-strategy community supports our second hypothesis and previous studies that fast-strategy communities have greater capacity to recover from drought, despite having lower resistance than slow-strategy communities (Ingrisch et al., 2018; Karlowsky, Augusti, Ingrisch, Hasibeder, et al., 2018). The overshoot in GPP was maintained until the end of the experiment and is consistent with the overshoot in above-ground biomass found 2 months after re-wetting (Figure 1b). Our results demonstrate that increasing drought intensity leads to abrupt shifts in productivity (both above-ground biomass and GPP) at peak drought, which differ between fast- and slow-strategy communities and likely have implications for plant community recovery trajectories.

4.2 Plant/soil C dynamics

4.2.1 Peak drought

Previous studies have found that drought reduces C uptake, below-ground C allocation and soil respiration (Chomel et al., 2019, 2022; Fuchslueger et al., 2014; Ingrisch et al., 2020; Karlowsky, Augusti, Ingrisch, Hasibeder, et al., 2018). Our study suggests that the extent to which such reductions occur depends on drought intensity (Figure 2). Considering the relative partitioning of recent plant-assimilated C, we found that drought intensity increased the relative proportion of C partitioned to leaves and transferred to the soil microbial biomass, while decreasing C partitioned to the roots (Figure S7). In previous research, drought effects on the relative partitioning of recent C vary, with studies reporting increases in relative C retention above-ground (Chomel et al., 2019; Sanaullah et al., 2012), increases in below-ground C allocation (Karlowsky, Augusti, Ingrisch, Hasibeder, et al., 2018) or no clear drought effect (Hasibeder et al., 2015). Such discrepancies could be caused by differences in drought intensity, as we show that C partitioning changes with increasing drought intensity (Figure S7).

In contrast to our third hypothesis, shifts in plant–soil C dynamics with increasing drought intensity at peak drought were similar between the fast- and the slow-strategy communities (i.e. there were no interactive effects between plant community and drought intensity, Figure 2). Over the drought gradient, C uptake by the fast-strategy community was higher than for the slow-strategy community, which is in line with previous research (Henneron, Fontaine, et al., 2020; Henneron, Kardol, et al., 2020). However, relative ¹³C partitioning to the compartments we measured was similar between fast- and slow-strategy plant communities (Figure S7). Previous studies have reported larger drought effects on C-cycling in grasslands dominated by fast- compared to slow-strategy plant species (Ingrisch et al., 2020; Karlowsky, Augusti, Ingrisch, Hasibeder, et al., 2018). However, in these field studies, the effect of species composition, management and edaphic soil characteristics between the grasslands could not be separated. In our controlled outdoor mesocosm experiment, we are able to isolate plant strategy effects on drought responses. Still, considering that plant resource acquisition strategy influences both plant–soil C-cycling (Henneron, Fontaine, et al., 2020; Henneron, Kardol, et al., 2020) and drought resistance (Pérez-Ramos et al., 2013), it is surprising that we did not find differences in the response of C incorporation to drought intensity for the fast- and slow-strategy communities.

In our study, small differences in C incorporation between fast- and slow-strategy communities could be a consequence of studying these perennial communities in their first year of growth. As plant communities with different resource acquisition strategies develop, differences in their plant–soil C dynamics likely increase due to contrasting leaf and root lifespan (Lind et al., 2013) and feedbacks to decomposition processes (Freschet et al., 2012). Differences in the associated microbial communities could also become stronger in subsequent growing seasons because of differences in plant–soil feedbacks between fast- and slow-strategy plants (Baxendale et al., 2014; Spitzer et al., 2021; Xi et al., 2021). While our findings show that increasing drought intensity has an overriding effect on plant–soil C allocation at peak drought, we expect that in following growing seasons, plant resource strategies could play a larger role in mediating C dynamics during drought.

Shifts in plant–soil C dynamics could disconnect plants and microorganisms during drought via changes in rhizodeposition (Karlowsky, Augusti, Ingrisch, Akanda, et al., 2018; Williams & de Vries, 2020). In field and glasshouse studies, drought has been reported to weaken the link between plants and the microbial community by reducing C flow from recent photosynthates to below-ground pools (Chomel et al., 2019; Fuchslueger et al., 2014), decreasing C incorporation into microbial biomass (Karlowsky, Augusti, Ingrisch, Akanda, et al., 2018; Karlowsky, Augusti, Ingrisch, Hasibeder, et al., 2018) and mesofauna (Chomel et al., 2019, 2022; Seeber et al., 2012). Our findings support this, as ¹³C incorporation into PLFAs representing Gram-positive bacteria, Gram-negative bacteria and saprotrophic fungi, as well as the NLFA representing AMF decreased with increasing drought intensity (Figure 3). AMF can play an important role in resource uptake (Pantigoso et al., 2022), reducing drought stress (Wu, 2017) and mitigating decreases in plant productivity (Jia et al., 2020). However, the sharp decrease in ¹³C incorporation into the NLFA representing AMF could signal that either drought disrupts the plant–AMF connection or that ¹³C is prioritized elsewhere. Previous studies considering drought effects on ¹³C incorporation in this NLFA are inconsistent, showing that ¹³C transfer during drought is maintained (Chomel et al., 2019, 2022; Fuchslueger et al., 2014) or reduced (Karlowsky, Augusti, Ingrisch, Hasibeder, et al., 2018). However, our results suggest that drought intensity could be an overlooked driver in plant C transfer to this NLFA, and that directly considering drought intensity may reconcile previously reported discrepancies. We found that at drought intensity between 42% and 65% SWD the amount of ¹³C allocated to the NLFA representing AMF was similar to the control (20% SWD), while increasing drought intensity to over 75% SWD caused a sharp decrease (Figure 3).

In contrast, the ¹³C incorporated into actinobacteria PLFAs increased with increasing drought intensity (Figure 3). Actinobacteria are spore forming and able to persist in stressful environments (Taketani et al., 2017). Their ability to tolerate drought (Naylor et al., 2017) may have enabled them to continue functioning as drought intensity increased. An increase in their relative abundance during extreme drought, which has been previously reported (Naylor et al., 2017; Pérez Castro et al., 2019), may have also contributed our finding that ¹³C incorporation into actinobacteria PLFAs increased with drought intensity.

4.2.2 Recovery

One week after re-wetting, C uptake by the fast-strategy community was no longer affected by drought intensity (Figure 2), signalling its quick recovery. This aligns with our finding that soon after re-wetting drought intensity no longer affected GPP (Figure S4). Only 10 days after re-wetting, there was evidence of GPP over-compensatory effects in fast-strategy communities that had experienced the highest drought intensities, and this was maintained until the end of the experiment. We show that the GPP overcompensation responses to drought depend on drought intensity in the early recovery phase (Figure S4). This is in line with previous studies showing quick post-drought recovery in grasslands dominated by fast-growing plants (Ingrisch et al., 2018). In the slow-strategy community, past drought intensity maintained its effect on C uptake and allocation to roots (Figure 2). In both communities, there was a shift to allocating more recently acquired C to EOC and microbial biomass C (Figure 2), and this was clearest in the slow community where increasing drought intensity reduced the relative amount of C partitioned to leaves in favour of partitioning it to microbial biomass and EOC (Figure S8). This shift in plant/soil C dynamics with increasing drought intensity may have contributed to plant community recovery. Increasing C allocation to EOC and microbial biomass reflects increased rhizodeposition and uptake by the microbial community, which could increase N availability in the rhizosphere (Henneron, Fontaine, et al., 2020; Henneron, Kardol, et al., 2020), as discussed below. Nine days after re-wetting, the microbial community had largely recovered in terms of ¹³C incorporation into NLFA/PLFAs. However, ¹³C incorporation into the NLFA representing AMF decreased with past drought intensity in both plant communities. Previous studies have found that while C allocation to the NLFA representing AMF quickly recovered after drought in a grassland dominated by fast-strategy plants, it remained lower plant communities dominated by slow-strategy plants (Karlowsky, Augusti, Ingrisch, Hasibeder, et al., 2018).

4.2.3 Plant/soil N dynamics

We found that drought altered plant/soil N dynamics (Figure 4; Figures S9 and S10), which could play a key role in post-drought productivity overcompensation (Figure 1). Twelve days after re-wetting, leaf N concentration increased with past drought intensity in both plant communities, and the relationship was stronger in the fast community (Figure 4). Whole plant N uptake (g N m⁻²) increased with increasing past drought intensity in the fast-strategy community only, while the slow-strategy community was not affected (Figure S10C). Thus, while leaf N concentration increased in both the fast- and slow-strategy plant communities with increasing past drought intensity, only the fast-strategy community increased its plant N uptake, signalling fast drought recovery. This supports our hypothesis that the fast-strategy community is more efficient in increasing N uptake soon after re-wetting. The increase in leaf N concentration and plant N uptake with past drought intensity could result from two sources of N: higher plant-available N in the soil (i.e. NO₃⁻-N, NH₄⁺-N, Figure S9), and/or remobilization of root N stored during drought in the fast community (Figure 4).

An increase in soil NO₃⁻-N upon re-wetting after drought is frequently reported (e.g. Mackie et al., 2018; Roy et al., 2016) but has rarely been studied in response to increasing drought intensity. An increase in organic N sources via dead plant material and lysed microbial cells above a drought intensity of 75% SWD likely caused the increase in NH₄⁺-N at peak drought, fuelling nitrification and the increase in NO₃⁻-N after the drought (Figure S9). Apart from microbial N transformation processes, this increase in N availability post-drought could be due to an increase in rhizodeposition with increasing drought intensity (discussed above). The more pronounced increase in leaf N concentration and whole plant N uptake with increasing drought intensity in the fast-strategy community only is consistent with earlier studies showing fast-strategy plants have higher rates of N uptake compared to slow-strategy plants (de Vries & Bardgett, 2016; Grassein et al., 2015). This facilitates a tighter coupling of C and N dynamics in the rhizosphere by accelerating N cycling via the input of labile C (Henneron, Fontaine, et al., 2020; Henneron, Kardol, et al., 2020). In the context of disturbance, fast-strategy plants may be specifically suited to recover quickly because they are able to take advantage of pulses of nutrients (Grime, 1979) caused by the Birch effect (Birch, 1958) upon re-wetting dry soil. Our findings are in line with this, as we found that although plant-available soil N increased with increasing drought intensity in both plant communities, the whole-plant N uptake increased in the fast- but not the slow-strategy plant community (Figure S10C). The contribution of a second N source in the fast community (re-mobilization of N stored in roots at peak drought) may have also contributed to its higher productivity overshoot and increased above-ground N pool, as storage of N in roots during stress and re-mobilization to leaves is an important factor for plant recovery (Masclaux-Daubresse et al., 2010).

5 CONCLUSIONS

By implementing a gradient design in outdoor mesocosms, we show that grassland responses to drought intensity are non-linear at peak drought and during recovery. Advancing on previous work based on monocultures, we also discovered that productivity thresholds differed with plant community resource strategy: slow-strategy communities were more drought resistant than fast-strategy communities, but fast-strategy communities recovered faster and had a higher degree of overshoot after re-wetting. We also found that drought intensity governed below-ground C allocation and N dynamics in fast and slow communities at peak drought, while plant community strategy modulated recovery dynamics after rewetting. Collectively, our results highlight the role of drought intensity in understanding grassland ecosystem functioning during and after drought, and suggest that increasing drought intensity can trigger major functional shifts in grasslands that affect plant community recovery trajectories and have potential long-lasting effects (Müller & Bahn, 2022).

AUTHOR CONTRIBUTIONS

Natalie J. Oram and Johannes Ingrisch contributed equally and share first authorship. Natalie J. Oram, Michael Bahn, Johannes Ingrisch, Richard D. Bardgett and Fiona Brennan designed the study. Data collection was carried out by Natalie J. Oram, Johannes Ingrisch, Georg Dittmann, Gerd Gleixner, Nadine Praeg, Paul Illmer, statistical analysis by Natalie J. Oram and Johannes Ingrisch, and the manuscript was written by Natalie J. Oram and Johannes Ingrisch with extensive input on data interpretation and writing from Michael Bahn, as well as Richard D. Bardgett, Fiona Brennan, Georg Dittmann, Gerd Gleixner, Nadine Praeg and Paul Illmer.

ACKNOWLEDGEMENTS

Authors would like to thank Laura Dobrowz, Thea Schwingshackl, Moritz Mairl, Greta Ploner, Dina in ‘t Zandt, Lena Müller, Akira Yoshikawa, Fabrizzio Protti, Deniz Scheerer, Suzanne Guyard, Andrew Giunta, Kathiravan Mohamed Meeran and Beverley Anderson who made this experiment possible and also a lot of fun. Thanks to Heike Geilmann for ¹³C analysis of the leaves and roots, Anna Reichstein for ¹³C analysis of TOC and EOC, and Stefan Karlowsky for helpful advice. Thank you to Barenbrug BV, the Netherlands for providing seeds for the experiment. NJO has received funding from the Research Leaders 2025 programme co-funded by Teagasc and the European Union's Horizon 2020 research and innovation programme under the Marie Sklowdowska-Curie grant agreement number 754380. RDB acknowledges support from a European Research Council (ERC) Advanced Grant (883621, SoilResist). The experiment was also supported by a Tiroler Wissenschaftsfonds grant to JI (grant ID F.16568/5-2019).

CONFLICT OF INTEREST STATEMENT

The authors declare no conflict of interest. Richard Bardgett is the Executive Editor of Journal of Ecology, but took no part in the peer review or decision-making process for this manuscript.

Open Research

DATA AVAILABILITY STATEMENT

Data are available on Dryad Digital Repository https://doi.org/10.5061/dryad.5qfttdzbb (Oram, 2023).