Tillage-induced short-term soil organic matter turnover and respiration

11 Tillage induces decomposition and mineralisation of soil organic matter (SOM) by the 12 disruption of macroaggregates and may increase soil CO2 efflux by respiration, but 13 these processes are not well understood at the molecular level. We sampled three 14 treatments (mineral fertiliser = MF, biogas digestate = BD, unfertilised control = CL) of 15 a stagnic luvisol a few hours before and directly after tillage, and four days later from a 16 harvested maize field in Northern Germany and investigated these samples by 17 pyrolysis-field ionization mass spectrometry (Py-FIMS) and hot-water extraction. 18 Before tillage, the Py-FIMS mass spectra revealed differences in relative ion intensities 19 of MF and CL compared to BD most likely attributable to the cattle manure used for the 20 biogas feedstock and to relative enrichments during anaerobic fermentation. After 21 tillage, the CO2 effluxes were increased in all treatments, but this increase was less 22 pronounced in BD. We explain this by a restricted availability of readily biodegradable 23 carbon compounds and, possibly an inhibitory effect of sterols from digestates. 24 Significant changes in SOM composition were observed following tillage. In particular, 25 lignin decomposition and increased proportions of N-containing compounds were 26 detected in BD. In MF, lipid proportions increased at the expense of ammonia, 27 ammonium, carbohydrates and peptides, indicating an enhanced microbial activity. 28 2 SOM composition in CL was unaffected by tillage. Our analyses provide strong 29 evidence for significant short-term SOM changes due to tillage in fertilised soils. 30

correlations between CO 2 efflux and the turnover of soil organic carbon (SOC) after 48 tillage by first order kinetics (La Scala et al., 2008). Admittedly, these correlations do 49 not causally explain which organic components are mineralised. Furthermore, SOM-50 CO 2 -efflux-relationships are influenced by the type of soil amendment (Fiedler et al., 51 2015). 52 Biogas digestate is a relatively new type of soil amendment, and its long-term stability 53 in soil is still under debate as recently reviewed by Möller (2015). Consequently, it is 54 not clear how long-term application of biogas digestates would alter the composition of 55 SOM, and tillage effects on short-term SOM turnover in biogas digestate-amended soils 56 are almost unstudied. Even short-term changes of SOM may have strong effects on 57 nutrient availability and plant productivity. A better understanding of the immediate 58 6 discarding data noise at the beginning and the end resulting from chamber deployment 146 and removal (for details see help file for function fluxx of package flux). Thus, each CO 2 147 flux was estimated at least from 98 concentration measurements. Only linear fluxes 148 with a concentration change of at least 10 ppm, a normalised root mean square error 149 (NRMSE) ≤ 0.15 and a coefficient of determination (R 2 ) of at least 0.85 were included 150 in further analyses. We assumed linearity of concentration change and did not test for 151 non-linearity since 95.1% of the obtained linear regressions had R² ≥ 0.95. 152 To obtain reference data from before tillage operations, the undisturbed site was 153 measured hourly between 7 a.m. and 1 p.m. on 19 October 2012 (i.e. between harvest 154 and tillage). The intervals between measurements before, during and after tillage 155 operations were varied to effectively capture the development of CO 2 . The 156 measurements immediately after the tillage operations were conducted within one 157 minute by inserting the collars and putting on the airtight chambers. The timeline (24  158 till 29 October) of tillage events, soil samplings and the respective CO 2 measurements, 159 together with soil temperature, is shown in Fig. 1. After this period, CO 2 measurements 160 were performed hourly before noon on 1, 5 and 9 November. 161

Soil sampling and analyses 162
Three replicates of bulk soil samples were taken between 0 -10 cm depth (depending 163 on unevenness of soil surface due to tillage) directly with three soil sample rings (h = 164 6.1 cm, V = 250 cm³) in a triangular arrangement around the three bases for gas 165 sampling (see 2.2) in each treatment at three dates: 1) right before the first tillage 166 operation, 2) in the afternoon after the second tillage operation and 3) four days after the 167 second tillage operation. The resulting 27 soil samples were fixed immediately with 168 liquid nitrogen and splitted thereafter into subsamples for freeze-drying and for oven-169 drying at 60° C. 170 For Py-FIMS, the freeze-dried samples were finally ground and homogenized by a 171 planetary ball mill. Then, about 2 g were transferred into a Petri dish with a spatula and 172 three crucibles were filled by drawing them across. These subsamples of about 5 mg 173 were thermally degraded in the ion source (emitter: 4.7 kV, counter electrode -5.5 kV) 7 of a double-focusing Finnigan MAT 95 mass spectrometer (Finnigan,Bremen,175 Gemany). The samples were heated in a vacuum of 10 -4 Pa from 50 °C to 700 °C, in 176 temperature steps of 10 °C over a time period of 15 minutes. Between magnetic scans 177 the emitter was flash heated to avoid residues of pyrolysis products. The Py-FIMS mass 178 spectra of each sample were gained by the integration of 65 single scans in a mass range 179 of 15 -900 m/z. Ion intensities were referred to 1 mg of the sample. Volatile matter was 180 calculated as mass loss in percentage of sample weight. For plotting, the three replicates 181 of each sample were then averaged to one final survey spectrum. Moreover, 182 thermograms were compiled for the total ion intensities. The assignment of marker 183 signals to chemical compounds from the survey spectra were interpreted according to 184 Leinweber et al. (2013) to obtain the relative abundance of ten SOM compound classes: 185 1) carbohydrates, 2) phenols and lignin monomers, 3) lignin dimers, 4) lipids, alkanes, 186 alkenes, bound fatty acids and alkyl monoesters, 5) alkylaromatics, 6) mainly 187 heterocyclic N-containing compounds, 7) sterols, 8) peptides, 9) suberin, and 10) free 188 fatty acids. 189 Subsamples of oven-dried and sieved soil (2 mm) were used for determination of total 190 and hot water-extracted C and N. For determination of total C and N, 1 g of ground soil 191 was analysed with a vario Max CN Element Analyzer (elementar Analysensysteme  192 GmbH, Hanau, Germany) based on high temperature combustion at up to 1200 °C with 193 subsequent gas analysis. For hot-water extraction, 20 g of soil were boiled in 40 ml 194 deionized water for 60 minutes (Leinweber et al., 1995). After filtration with pleated 195 filters (240 mm, 80 g m -2 ) by Munktell (Falun, Sweden), extracts were analysed with a 196 DIMATOC 2000 (DIMATEC Analysentechnik GmbH, Essen, Germany) for 197 determination of hot-water extractable organic C (HWC) as well as of organic and 198 inorganic bound N, often referred to as 'total nitrogen bound' (HWN). These 199 measurements of organic C and total nitrogen bound are based on the principle of 200 thermal-catalytic oxidation with subsequent NDIR detection and the principle of 201 chemiluminescence, respectively. For each sample, two replicates were analysed and 202 results were averaged for further calculations.  Partial least squares regression (PLSR) was used for discrimination (Barker and Rayens, 221 2003) to explore linkages between shifts in the m/z data by tillage and shifts in CO 2 222 efflux. PLSR models were built using function autopls of the R package 'autopls' 223 version 1. 3 (Schmidtlein et al., 2015) with stepwise backward selection combined with 224 a 10-fold cross-validation to substantially reduce the number of variables, i.e., to extract 225 the variables with the highest explanatory power. The PLSR procedure was repeated 226 10.000 times to yield coherent results since the obtained PLSR models differed widely 227 both in the number and in the choice of variables and, thus, in their predictive 228 performance. Based on the performance index suggested by Bauwe et al. (2015), the 229 500 'best' models were obtained and, finally, the mass signals which were utilised more 230 than 50 times in the latter models were extracted. One of the replicates in MF exhibited exceptionally low HWC and HWN values. 236 According to Dixon's Q-test, these values were outliers (one-third and half, 237 respectively, as high as for the other replicates in MF) and thus excluded from further 238 analysis. Before tillage, the soil of all treatments had similar C and HWC contents, but 239 differences appeared between MF and BD, where the N and HWN contents were 240 slightly higher in MF, resulting in narrower C/N and HWC/HWN ratios in MF (8.5 and 241 5.9, respectively) compared to BD (9.0 and 8.5, respectively) ( Table 1). The C, N and 242 HWC contents of all treatments changed only slightly after tillage, but the HWN 243 content of soil in BD increased significantly (p < 0.05) from 0.05 mg g -1 (5.6 % of N) 244 up to 0.07 mg g -1 (7.4 % of N), resulting in a significant (p < 0.05) narrowing of the 245 HWC/HWN ratio from 8.5 down to 6.0 (Table 1). 246

Pyrolysis-Field Ionization Mass Spectrometry 269
The thermograms of total ion intensity (TII) and the Py-FIMS mass spectra of the soil 270 samples of CL and MF taken before tillage were similar whereas the ones of BD 271 differed markedly from those two ( Fig. 4): The TII-thermograms of CL and MF had a 272 peak at 480 °C, but BD displayed a pronounced bimodal shape with a first volatilisation 273 maximum at about 390 °C which was less marked in CL and MF. Furthermore, the 274 mass spectrum of BD differed distinctly from the mass spectra of MF and CL, 275 especially the abundance of marker signals for carbohydrates and peptides (e.g., m/z 58, 276 60, 84, 69, 110, 126 and 162) was lower. Apart from this the spectra are dominated by 277 signals for lignin mono-and dimers (e.g., m/z 150, 208, 222, 244) as well as for 278 homologous series of alkenes and alkadienes from n-C 18 up (e.g., m/z 252, 264/266, 279 278/280, 294, 308, 322, 336, 364, 392, 406) (Fig. 4). 280 After discriminant function analysis with Wilk's λ, the resulting significant relative 281 mass signals (p < 0.001, n = 67) were further explored by PCA. The first two principal 282 components accounted for 78.3% and 8.3% of total variance. All treatments are well 283 separated from each other (Fig. 5), with CL mainly in the 3rd quadrant, MF mainly in 284 the 1st and BD spanning from the 2nd to the 4th quadrant. According to this analysis, 285 samples from MF and BD taken before the tillage events (pre) showed the largest 286 differences in composition. The PCA separated the samples taken at different dates (pre, 287 post and post + 4) in the treatments MF and BD, but not in CL. 288 Basic data of the Py-FI mass spectra and the proportions of compound classes are 289 compiled in Table 2. Approximately 46.9% of the TII in the mass spectra could be 290 explained by m/z signals assigned to the compound classes. Additionally, non-specific 291 low-mass signals and isotope peaks contributed 2.6% and 14.2%, respectively. Before 292 tillage, the volatised matter (VM) was highest in BD and increased from 5.2 to 7.1% 293 during the days after tillage. Such an increase over time was only observed for BD, but 294 it was not significant (p > 0.1). In the other treatments, a temporal increase in VM 295 occurred directly after the first tillage with disc harrow. 296 The relative (Table 2) and absolute (data not shown) ion intensities of the compound 297 classes varied across treatments before tillage and changed differently after tillage. In 298 the undisturbed soil, BD had the lowest proportions of carbohydrates, heterocyclic N-299 containing compounds and peptides and the highest proportions of lignin dimers, lipids, 300 sterols, suberin and free fatty acids. CL was characterized by higher proportions of 301 phenols and lignin monomers whereas MF ranged between BD and CL regarding the 302 proportions of these compound classes. In BD, the relative proportions of the samples 303 taken after tillage displayed significant (p < 0.1) increases of carbohydrates, phenols 304 and lignin monomers, alkylaromatics, heterocyclic N-containing compounds and 305 peptides while lignin dimers, lipids, sterols and free fatty acids decreased. In MF, the 306 proportion of lipids increased while carbohydrates and peptides decreased. No changes 307 were detected in the unfertilised treatment CL. The discrimination of relative mass 308 signals with PLSR to explain cumulated CO 2 efflux revealed mainly functional groups 309 from ketones and amides, peptides, carbohydrates as well as lignin building blocks and 310 fatty acids (Table 3). 311 Linear correlations were calculated to check relationships between HWC, HWN and 312 soil respiration as indicators of SOM dynamics (Kuzyakov, 2006;Leinweber et al., 313 1995) and the absolute signal counts of the compound classes (Fig. 6). The latter was 314 derived from Table 2   The increase in HWN in BD after tillage indicates an increase of easily mineralisable 331 organic N which probably originates from soil biomass and lysates (Ghani et al., 2003;332 Leinweber et al., 1995) and implies an accelerated microbial turnover of soil organic N. 333 This seems reasonable since the microbial community is able to adjust its structure and 334 activity relatively fast to utilise formerly protected organic matter after exposure due to 335 disruption of aggregates by tillage (Jackson et al., 2003;La Scala et al., 2008). Possibly, we did not detect it, because we took no soil samples after the first day. 339 Overall, a single amendment with biogas digestates very likely is insufficient to initiate 340 changes in bulk soil C-and N-levels. However, the increased HWN-levels in BD can be 341 ascribed to a tillage promoted microbial turnover of soil organic N, confirming that the 342 hot water extracts are a particularly sensitive approach to detect early SOM changes 343 (Haynes, 2005). 344

Soil CO 2 efflux 345
The immediate and sharp increase of CO 2 efflux from soils just after tillage is a well-346 documented response and seems to be mainly driven by the release of trapped CO 2 from 347 13 broken up aggregates by tillage (Reicosky et al., 1997). It is commonly suggested that a 348 few hours afterwards, waning of this physical outgassing is accompanied by an 349 increased soil respiration due to a better substrate supply for microorganisms from 350 disrupted aggregates as well as increased soil aeration (Grandy and Robertson, 2007). 351 The amounts of the observed fluxes are well in accordance with the findings of previous 352 studies (e. g., Rochette and Angers, 1999) and can be explained both by the magnitude 353 of the disturbance, i.e. soil comminution, and the fertilisation history of the soil (Fiedler 354 et al., 2015). 355 The smaller relative efflux from BD compared to MF and CL after tillage is remarkable 356 since before tillage the CO 2 fluxes in BD were of the same magnitude as those in MF 357 and exceeded those in CL (Fig. 2). This becomes particularly evident when we consider 358 the relation of cumulated CO 2 fluxes between the treatments before (19 October) and 359 after tillage (24 -29 October) (Fig. 3). The relatively lower CO 2 efflux from BD after 360 tillage may have different reasons. On the one hand, C originating from the digestates is 361 likely less available to soil microorganisms compared to undigested organic matter, i. e. 362 more 'recalcitrant', since the most labile C is generally consumed in the biogas reactor 363 (Möller, 2015). On the other hand, even a single application of organic amendment can 364 increase aggregate stability (Grandy et al., 2002). Therefore, the resilience against 365 disruption by tillage might be promoted, leading to a better physical protection of labile 366 soil C not contained within digestates. As a consequence, the effect of increased CO 2 367 efflux after tillage as observed in CL and MF may have been substantially reduced by a 368 relative shortage of labile substrate for soil respiration in BD. 369

Pyrolysis-Field Ionization Mass Spectrometry and synthesis 370
Generally, the Py-FIMS basic data and mass spectra (Fig. 4)  to residues of the just harvested maize. Indeed, Gregorich et al. (1996) found that these 375 are important components of maize leaves and roots as well as of the light fraction of 376 the soil under this crop. Overall, the Py-FIMS data indicate differences in SOM 14 composition between the fertilization treatments and a pronounced impact of tillage in 378 the treatments MF and BD (Fig. 5). 379 In the spectra of samples from BD, the additional peak at 390° C in the TII-thermogram 380 (Fig. 4) can be attributed mainly to phenols and lignin monomers which likely 381 originated from primary organic matter residues since this relatively low volatilization 382 temperature indicates labile and fairly undecomposed organic matter (Ludwig et al., 383 2015;Sleutel et al., 2011). It is reasonable to refer this organic matter to residues from 384 the application of BD. VM as well as TII, which are indicators of SOM content (Sorge 385 et al., 1993) and also of its stability (Ludwig et al., 2015), were larger in BD than in MF 386 and CL before tillage (Table 2). This suggests a tendency to elevated SOM due to 387 application of organic matter with biogas digestate. The increase in VM after tillage 388 might be explained by a general destabilization, perhaps by an enhanced SOM turnover 389 due to an improved microbial accessibility to relatively recalcitrant residues of BD after 390 tillage (Dao, 1998). The temporal increase in VM directly after the first tillage with disc 391 harrow in MF and CL may indicate a similarly increased accessibility of SOM. But 392 here, the newly available SOM has been depleted quickly by microbial respiration since 393 the microbial community is able to respond rapidly to disturbances of arable soils 394 (Jackson et al., 2003). In MF, this assumption is supported by the decreasing proprtions 395 of carbohydrates and by significantly decreasing relative signal intensities of m/z 17 and 396 18 (data not shown), which are assigned to ammonia and ammonium, pointing to a 397 microbial immobilisation (Mengel, 1996). Accordingly, these two m/z were also 398 selected by the PLSR as explanatory signals for CO 2 efflux (Table 3). 399 The compound classes of BD revealed the largest proportions of lignin dimers, lipids, 400 sterols, suberin and free fatty acids at the expense of carbohydrates, heterocyclic N-401 containing compounds and peptides before tillage (Table 2). Such a SOM composition 402 most likely reflects the cattle manure and plant residues of the biogas feedstock and 403 their relative depletions (amides and polysaccharides) or enrichments (lignins and long-404 chain aliphatic compounds) during anaerobic fermentation (Möller, 2015;van Bochove 405 et al., 1996). The pronounced tillage effect in this treatment, obvious from the increased 406 relative signal intensities of carbohydrates, phenols and lignin monomers, 407 alkylaromatics, heterocyclic N-containing compounds and peptides at the expense of 408 15 lignin dimers, lipids, sterols and free fatty acids following tillage (Table 2), suggests the 409 decomposition of lignin and the new formation of carbohydrates and peptides. This is in 410 line with reports of lignin decomposition faster than that of the total SOM (Leinweber et 411 al., 2008;Thevenot et al., 2010). Kalbitz et al. (2003) suggested that lignin-derived 412 moieties and lipids are utilised by microorganisms at low initial availability of 413 carbohydrates, accompanied by an accumulation of the resulting microbial metabolites 414 like carbohydrates and peptides. Recently, Rinkes et al. (2016) also found that 415 decomposers may break down lignin to acquire C for their metabolism in the absence of 416 available labile C. This suggestion is supported on the one hand by the effect of specific 417 lignins on soil CO 2 efflux (Table 3) since CO 2 is an indicator for microbial 418 decomposition activity (Kuzyakov, 2006). On the other hand, a relative increase of the 419 signals for m/z 125, 167, 185 and 203 was observed in the BD treatment (data not 420 shown) which are assigned to the bacterial cell wall products N-acetylmuramic acid and 421 N-acetylmuramyl-L-alanyl-D-isoglutamine (Bahr and Schulten, 1983). Furthermore, the 422 build-up of heterocyclic N-containing compounds might also imply a relative shortage 423 of available carbohydrates since a reduced C availability during the microbial 424 transformation of N is suggested to promote formation of heterocyclic N instead of N 425 immobilisation (Follett and Schimel, 1989;Gillespie et al., 2014;Schulten and 426 Hempfling, 1992). The increased proportion of lipids at the expense of carbohydrates 427 and peptides in MF likely results from increased heterotrophic respiration of labile 428 substrates driven by enhanced microbial activity after tillage (La Scala et al., 2008;429 Zakharova et al., 2014). The minor changes in SOM compounds in CL might be a 430 consequence of the wider HWC/HWN ratio compared to MF and BD since it indicates a 431 lower availability of labile N for microbial utilisation (Mengel, 1996). However, the 432 total C/N ratios were not critical for microbial activity (Table 1) (Kuzyakov et al., 433 2000). 434 A significant (p > 0.05) and positive correlation was observed between HWC and 435 carbohydrates in MF. This linkage was previously described by Leinweber et al. (1995) 436 and attributed to microbial biomass (Ghani et al., 2003) and labile soil C (Sparling et 437 al., 1998). In contrast, this correlation was not apparent in BD. This corroborates the 438 assumption that microorganisms in BD may have been short in available labile C. 439 Interestingly, HWN correlated positively with carbohydrates in BD. Since the major 440 part of carbohydrates in soils originates from microorganisms and their residues 441 (Gunina and Kuzyakov, 2015), this may suggest a metabolic coupling between 442 carbohydrates and HWN because many N-cycling processes are mediated microbially 443 (Isobe and Ohte, 2014). 444 Increased amounts of sterols are typically found in biogas digestates (Leinweber, 2016, 445 unpublished Py-FIMS data). In BD, the cumulated CO 2 efflux and the amount of sterols 446 was negatively correlated. This supports the suggestion of Heumann et al. (2011Heumann et al. ( , 2013) 447 that sterols may have an inhibitory effect on microorganisms of the N cycle and, thus, 448 may slow down soil respiration. However, since the amounts of sterols decreased 449 significantly after tillage in BD (Table 3) In MF, the depletion of HWC was linked to decreasing amounts of carbohydrates, 461 certainly due to increased microbial respiration, though no significant correlation with 462 CO 2 efflux was found. No modifications were detected in CL were the absence of 463 amendment may have led to a relative shortage of labile N as indicated by the higher 464 HWC/HWN-ratio which possibly prevented an enhanced microbial activity. 465