Intensive management disrupts belowground multitrophic resources transfers in response to drought

31 Modification of soil food webs by historical land management may alter the response 32 of ecosystem processes to climate extremes, but empirical support for this is limited and the 33 mechanisms involved remain unclear. Here, we quantified how historical grassland 34 management modifies transfers of recent photosynthate and soil nitrogen through plants and 35 soil food web in response to drought, using in situ 13 C and 15 N pulse-labelling in paired 36 intensively and extensively managed fields. We show that intensive management decreased 37 plant carbon capture, its transfer through key components of food webs and soil respiration 38 compared to extensive management. Drought only affected carbon transfer pathways in 39 intensively managed grasslands, by increasing plant C assimilation but decreasing its transfer 40 to plant roots, bacteria and Collembola. However, drought lowered the reduction of added 41 nitrate to nitrous oxide in extensively managed grassland only. Our findings indicate that 42 intensive management disrupts fluxes of recent photosynthates belowground, which impaired 43 resistance of this process in response to drought. By contrast, extensive grassland management 44 provides a greater potential to buffer impacts to drought by promoting the transfer of recent 45 photosynthate belowground. Our work highlights that capture and rapid transfer of 46 photosynthate through multitrophic networks is a key process for maintaining grassland 47 resilience to drought.

resilience to drought. 48 49 Introduction 50 All organisms within ecosystems are interlinked by energy flows in complex 51 multitrophic networks, and changes in the network structure modify these energy flows 1 . 52 Theoretical evidence suggests that shifts in food web structure play an important role in 53 regulating the stability of soil functions following perturbations and impair their ability to 54 buffer future extreme climatic events 2-5 . Grasslands are under threat from ongoing degradation 55 caused by multiple co-occurring drivers, including management intensification and climate 56 extremes 6 . Drought events are a recurring phenomenon in many ecosystems and are predicted 57 to increase in frequency and intensity in the coming decades 7,8 . Consequently, there is a need 58 to understand the interactions between these different drivers to inform sustainability policy 59 aimed at protecting the multiple ecosystem services that grasslands provide 6,9 . 60 Intensive grassland management, characterised by the regular use of inorganic 61 fertilisers and high livestock stocking densities, is known to decrease plant diversity 10 , 62 decrease the abundance and diversity of arbuscular mycorrhizal (AM) fungi 11 and soil biota 63 12,13 , and induce shifts in the composition of soil microbial communities [12][13][14] . Such changes 64 have important consequences for biogeochemical cycles because soil food webs associated 65 with agricultural intensification, including shifts in the relative abundance of bacterial and 66 fungal energy channels, are often linked with faster nutrient mineralisation rates, which could 67 potentially contribute to greater losses of carbon (C) and nitrogen (N) from soil [15][16][17][18] . A critical 68 gap in our knowledge concerns how shifts in food web structure and the relative abundance of 69 fungi and bacteria modulate transfers of C from plants to below ground pools and fluxes, and 70 the capture of growth-limiting nutrients by plants from soil energy channels. 71 Recent empirical studies indicate that intensive management can decrease the resistance 72 to drought of plant productivity and soil respiration 19,20 of soil food web biomass 19 , and of C 73 allocation to soil microbial communities 21 . However, it is becoming apparent that the stability (20 kg N ha -1 ). Three hours later, a composite soil sample was taken consisting of 3 x 1 cm 141 diameter cores from each subplot. The following morning, vegetation in the same subplots 142 were labelled with 99 atom% 13 C-CO2 (Sigma aldrich). We used an air-tight chamber 143 constructed with a plastic bell cloche (approx. 20 L) equipped with 2 small fans to disperse the 144 gas, and a rubber septum at the top to inject the gas. Prior to the start of the 13 C labelling ca. 145 10:00 GMT, the photosynthetic rate was measured using an infrared gas analyser (EGM-4, PP 146 Systems, Hitchin, UK) to determine the timing of the CO2 injections. During approximately 2-147 3 hours, 25 mL of 13 C-CO2 were regularly injected through the septum for a total of 250 ml per 148 subplot. The twelve subplots within a site (i.e. both extensive and intensive management 149 regimes) were labelled at the same time. The three sites were labelled over 3 consecutives days 150 for logistical reasons. 151 Immediately after the 13 C-labelling, a small subsample of plant shoots (ca. 0.5 g), 3 152 small soil cores (1 cm diameter), and gas samples were taken from each plot. At 1, 2, 5, 10 and 153 20 days after the 13 C-CO2 pulse labelling, gas samples were taken and a fifth of the soil from 154 the subplots (metal ring) was harvested ( Figure S1). We refer to day 1 as being approximately 155 24h after 13 C labelling and 36h after 15 N addition. The holes created were filled back with sand 156 to minimise gas exchange from the exposed surface and disturbance. Four supplementary cores 157 outside the plots were taken per field to determine the 13 C and 15 N natural abundance signatures 158 of each C and N pool. 159 Plants were divided into shoot and root fractions, washed (for the roots fractions only), 160 oven dried at 60°C and weighed prior to analysis. All samples were analysed for total C and N 161 content and δ 13 C and δ 15 N signatures using an elemental analyser (PDZ Europa ANCA-GSL, 162 Sercon Ltd, Crewe, UK) coupled to a 20-20 isotope ratio mass spectrometer (Sercon Ltd, 163 Crewe, UK). A portion of soil was sieved and freeze dried prior to the PLFA extractions. From 164 the remaining soil, mesofauna were extracted using Tullgren funnels over 7 days and stored in 165 16:1ω5 is used widely for estimating AM fungal biomass, its use can lead to uncertainties 185 because bacteria can contribute to this pool 35 . However, in our case, the 13 C-enrichment of this 186 PLFA was very distinct from all bacterial PLFAs, which gave confidence in its use to estimate 187 AM fungal biomass and its 13 C uptake. The δ 13 C value of each PLFA molecule was corrected 188 for the C added during derivatization using the formula where CFAME , CMeOH , and CPLFA denote the number of carbon atoms in the FAME, methanol, 191 and PLFA, respectively, and δ 13 CFAME and δ 13 CMeOH are the measured 12 C/ 13 C isotope ratios of 192 the FAME and methanol, respectively (methanol δ 13 C = −29.3‰). While the fungi are Ganihar 39 for the other orders. The samples were further grouped into 7 main trophic groups 207 in order to have sufficient material to analyse 13 C and 15 N: detritivorous Collembola, 208 detritivorous mites (oribatid, astigmata and prostigmata mites), annelids, other detritivorous 209 (detritivorous coleoptera, myriapoda and diptera larvae), herbivores (hemiptera and 210 thysanoptera) predaceous mites (mesostigmata and predaceous prostigmata) and predaceous 211 fauna (arachnida, chilopoda, predatory coleoptera and symphyla). Each of these groups was 212 transferred into a tin capsule in 70 % ethanol, oven-dried and weighed prior to analysis. All the samples were analysed for total C and N content and δ 13 C and δ 15 N using a Flash EA 1112 214 Series Elemental Analyser connected via a Conflo III to a DeltaPlus XP isotope ratio mass 215 spectrometer (Thermo Finnigan, Bremen, Germany). 216 217

Gas samples 218
Gas samples were taken by placing a 1.2 L dark chamber over the gas sampling core. 219 Immediately after the closure of the chambers, and after 10, 20, and 30 minutes, 15 mL gas 220 samples were taken from the headspace of the chamber using a gas-tight syringe fitted with an 221 SGE syringe valve and transferred into pre-evacuated 12 mL gas-tight vial (Labco Ltd. UK). The isotopic concentration data was converted from δ 13 C and δ 15 N values (‰) to atom % 234 excess 13 C and 15 N by subtracting the atom % 13 C and atom % 15 N of unlabelled controls from 235 each enriched sample 40 . 13 C-and 15 N-enrichment is independent of the pool size and is 236 indicative of the replacement of C or N from the pool by newly incorporated plant-derived 13 C 237 or fertiliser-derived 15 N.
Although the plants were pulse-labelled with the same amount of 13 C-CO2, differences 239 in photosynthetic and respiration rates caused the initial amount of 13 C fixed to differ among 240 subplots. To compare the relative transfer of 13 C from the plant shoot into belowground C pool 241 (roots, microorganisms and soil fauna) in the different systems, the results were expressed as a 242 percentage of the initial plant shoot 13 C enrichment. 243 The net incorporation of the 13 C or 15 N tracer into the different carbon/nitrogen pools 244 for each subplot was calculated as: 245 where C pool is the amount of C in each pool (gC per subplot) and atom% 13 C pool is the atom 247 % excess of 13 C of each pool. The same equation is used for the net incorporation of the 15 N, 248 replacing 13 C and C by 15 N and N. In order to evaluate the recovery of the stable isotopes in 249 the system, we calculated the fraction of the isotope in the different compartments as the ratio 250 of the net incorporation amount in a compartment relative to the total initial uptake of 13 C by

1) Grassland management, soil food web composition and soil functioning 283
Land management modified plant communities and soil properties, with greater aboveground 284 plant biomass, soil pH, nitrate concentration, and bulk density, and lower water holding 285 capacity and belowground plant biomass in intensively managed grasslands compared to 286 extensively managed grasslands (Fig S2). The influence of land management on soil CO2 efflux 287 was not consistent through the sampling period (interaction management*time, F=2.86, P = 0.0162, Fig S3), with higher CO2 efflux in extensively managed compared to intensively 289 managed grassland at time 0 and the opposite at day 20 ( Fig S3). There was a trend of higher 290 soil N2O efflux in extensively managed grassland compared to intensively managed grassland 291 on the first day after the pulse labelling, but this was not significant (interaction 292 time*management, P =0.08, Fig S3)   Significance was tested by Monte Carlo permutation against 999 random datasets and 299 variables with P > 0.05 were kept (see table S2). 300 Collembola was the most abundant group (30,949 ± 3,718 ind.m -2 , 80% were within the 301 entomobryomorpha group), followed by detritivorous mites (25,114 ± 2,772 ind.m -2 , 88% were 302 oribatids) and predatory mites (20,094 ± 2,279 ind.m -2 ). Soil community composition differed 303 significantly between extensive and intensive grassland (PERMANOVA, F = 5.12, P < 0.001). 304 Despite this finding, there was substantial overlap in community composition (Fig. 2) and we 305 detected significant differences in the distribution of data, with more dispersion in soil 306 communities in extensive grassland (PERMISP, F = 9.1, P = 0.033). With the soil fauna pooled 307 by trophic groups, the biomass of detritivorous mites (mainly Oribatids) was smaller (F = 7.03, 308 P = 0.0096), while the biomass of other decomposers (mainly diptera larvae) was greater 309 (F=9.08, P = 0.0034) in intensively managed compared to extensively managed grasslands 310 ( Figure S4). Intensive management did not have a significant impact on bacterial, fungal, AM 311 fungal and actinobacteria biomass, nor the fungal/bacterial ratio across grassland sites (Fig S4). pools. There were no significant effects of grassland management: see Table S3 for statistical 331 results. Grey colour indicates no data available. 332 333 Plant leaves in the intensively managed grassland were significantly less enriched in 13 C, 334 consistently across the three sites, than in extensively managed grassland (Fig. 3, P = 0.015). 335 The maximum 13 C-enrichment in plant shoots occurred at the end of the labelling period and 336 was (on average) 0.55 and 0.39 atom% excess in extensively and intensively managed 337 grassland, respectively. The enrichment decreased substantially 1 day after labelling to an average of 0.13 and 0.11 atom% excess (Fig. 3). The uptake of 15 N in plant shoots was 339 unaffected by grassland management, but increased gradually over two days after the labelling 340 and reached a plateau of 4.5-5 atom % excess (Fig 4). Land management had no detectable 341 influence on the 13 C-or 15 N-enrichment of plant roots (Fig. 4 and 5). 342 Among microorganisms, fungi had the highest enrichment of 13 C, whereas among soil 343 fauna, Collembola had the highest enrichment of 13 C (Fig. 3). 13 C enrichment was greatest in 344 AM fungi and detritivorous mites in extensively managed compared to intensively managed 345 grassland (Fig 3, P < 0.05, see Table S3). No management effects on 15 N enrichment of soil 346 fauna were detected (Fig. 4, P > 0.05, see Table S3). Although there was a greater 13 C-347 enrichment of the soil CO2 efflux in extensively managed compared to intensively managed 348 grassland (P < 0.001, Fig 4,

3) Effect of land management on the response of C and N flow to drought 352
During the rain exclusion, soil moisture was reduced on average by 56 ± 0.4 vol. % in 353 intensively managed grassland and 74 ± 0.4 vol. % in extensively managed grassland over the 354 last 27 days recorded (Fig S1). 355 In intensively managed grassland, the drought increased the biomass of microbial 356 communities and of detritivorous mites, but reduced the biomass of plant shoots and of other 357 detritivores (Fig. S5). In extensively managed grassland, the drought increased the biomass of 358 detritivorous mites and predatory mites, but decreased the abundance of actinobacteria PLFA 359 and biomass of Collembola (Fig. S5). 360 In extensively managed grassland, drought had no detectable effect on the uptake of 13 C by 369 plants, its transfer to roots and the soil food web, or soil 13 C-CO2 efflux (Fig. 5, Table S4). 370 However, in intensively managed grassland, drought increased plant shoot 13 C enrichment (Fig.  371 5, Table S4. P = 0.004) and 13 C relative enrichment of soil CO2 efflux (Fig. 5, table S4, P = 372 0.032), but decreased its transfer to roots, bacteria and Collembola (Fig. 5, table S4

4) 13 C and 15 N allocation 378
One day after labelling, 26 % of 13 C fixed by plants remained in the shoots and 2.6 % 379 was recovered in the roots (Fig S6). A similar pattern was observed with the 15 N tracer, for 380 which 39 % was on average recovered in the plant shoots and 12% in the plant roots (Fig S6)  381 one day after the labelling. Intensive management decreased the 13 C recovery in plant roots but 382 increased the 15 N recovery in plant shoots (Fig S6, Table S5). 383 Analysis of 13 C and 15 N pools revealed that in extensively managed grassland, soil 384 fauna tended to store more photosynthesized carbon and fertiliser-derived N compared to 385 intensively managed grassland (Fig S7). This pattern becomes even clearer when considering 386 the relative allocation (Fig S6, panel C). Indeed, extensive management increased the 13 C 387 recovery in AM fungi, detritivorous mites and predatory mites, and the 15 N recovery in 388 detritivorous mites (Fig S6, Table S5, P < 0.05). Drought only decreased the recovery of the 389 15 N in the plant roots ( Fig S6, Table S5, P < 0.001) and had no significant effect on the recovery 390 of the tracers in other C and N pools (Fig S6, Table S5, P > 0.05). 391 in the intensively managed grassland, indicating impaired resistance of this process, likely 405 through a decoupling of above-below ground interactions (Fig. 6). In contrast, belowground 406 fluxes of recent photosynthate in extensive grasslands were unaffected by drought, indicating 407 greater potential to buffer impacts of climate extremes on above-below ground interactions. 408

409
The effect of drought on plant C assimilation and its allocation belowground differed 410 between the two grassland management types. Despite decreases in aboveground biomass, 411 drought increased plant shoot C uptake and decreased C transfer to roots in intensively 412 managed grassland, but had no impact on these C transfers in extensively managed grassland. 413 This finding is consistent with recent reports of greater resistance and faster recovery of plants 414 in abandoned relative to managed grasslands due to larger belowground root and fungal 415 networks in the former, which improves water access compared to intensively managed 416 grasslands 19,21,46 . In our study, the greater uptake of C by plants in intensively managed relative 417 to extensively managed grasslands could be a compensatory effect following the release of the 418 drought. Indeed, fast-growing plants, which dominate the plant community of intensively 419 managed grasslands, typically have an ability to open their stomata more quickly when drought 420 is released compared to slow-growing plant species 47 , which dominate the plant community 421 in extensively managed grasslands. Our results confirmed that intensive management reduces 422 plant diversity and promotes species with lower root-to-shoot ratio. These differences are 423 explained by the fact that intensive management promotes fast-growing plant species, which 424 store resources in roots to facilitate regrowth after cutting, while extensive management 425 promotes plants that invest in root growth, rather than storage, to access soil resources 21 . 426 However, in our study, historic management intensity had no detectable effect on root 13 C 427 enrichment, indicating a similar rate of root C allocation of newly incorporated photosynthates 428 in both management regimes, despite large differences in their biomass allocation.
Fungi and Collembola were a major conduit of recent photosynthate-derived C; on 430 average 5.8 % of fungal C and 7.6 % of Collembola C came from plant photosynthate at their 431 enrichment peak. The high 13 C-enrichment of non-mycorrhizal fungal PLFA compared to the 432 AM fungal PLFA is surprising, and implies an important role of non-mycorrhizal fungi in 433 channelling plant-derived C into the soil food web, supporting other recent findings 48-51 , and 434 reflecting that saprotrophic fungi form a significant portion (20-66 %) of microbial biomass in 435 a grassland rhizosphere 52 . Although the recovery of the 13 C tracer was greater in bacterial 436 PLFA compared to AM fungal or general fungal PLFA (see Fig. S5), the conversion of PLFA 437 to biomass is higher for fungi, and this means the absolute amount of 13 C in bacterial biomass 438 was less than for fungal biomass (11.8 nmol of the PLFA 18:2ω6,9 = 1 mg of fungal C while depending on the duration of drought, they can start to regrow within one to several days 64-66 . 467 In extensive grassland there was marginal greater emission of 15 N-N2O (which we 468 attribute primarily to the nitrate reducing processes of denitrification and nitrate 469 ammonification) immediately after the ammonium nitrate was injected into the soil and during 470 the following day ( Figure S3). This contradicts the general idea that intensive management of 471 soil enhances the production of N2O 67,68 . This is likely to reflect the higher C availability in 472 extensively managed grassland providing the reductant for more sustained nitrate reduction 69 . 473 Moreover, in extensive grassland, drought did not modify N2O efflux, but it decreased its 15 N-474 enrichment compared to the control (in intensive grassland, there was a significant effect at 475 day 1 only). Soil moisture has consistently been shown to be one of the most important 476 parameters affecting soil oxygenation and thus determining N2O production rates 70 and ME. We are very grateful to the landowners and farmers for allowing us to perform our 504 experiment and sample their fields. We also thank Lucy Frotin and Juliette Papelard for help 505 in the lab and field, L. Harrold for plant and N2O isotope analysis, and B. Thornton and G. 506 Martin for the PLFA and soil fauna isotope analysis.