Modeling the biogeochemical effects of rotation pattern 1 and field management practices in a multi-crop ( cotton , 2 wheat , maize ) rotation system : a case study in northern 3 China

The cropping system with rotations between cotton and winter wheat−summer maize (W-M) 12 is widely adopted in northern China. Optimizing the rotation pattern and related field management 13 practices of this system is crucial for reducing its negative impacts on climate and environmental 14 quality. In this study, the approach applied to identify the optimal rotation pattern with the best 15 management practice (BMP) relied on biogeochemical model simulations to determine the negative 16 impact potential (NIP) of individual management options/scenarios and a set of constraints. The 17 optimal rotation pattern and related BMP are referred to as the scenario with the lowest NIP that 18 satisfies the given constraints. All the variables of interest were generated by simulation of the 19 DeNitrification-DeComposition 95 version (DNDC95) model. The DNDC95 model validations 20 performed previously for a land cultivated with the W-M and presently for an adjacent area cultivated 21 with cotton showed satisfactory performance in simulating the variables of interest with available 22 observations. The simulations of rotation patterns indicated that proper rotation of cotton and the W-M 23 can simultaneously benefit crop yields, soil carbon sequestration and greenhouse gas mitigation. The 24 three-crop rotation pattern in a 6-year cycle could be optimized with 3 consecutive years of cotton and 25 3 continuous years of W-M cultivation. The experiments with 108 management scenarios showed that 26 the BMP for the optimized rotation pattern involved using 15% less nitrogen fertilizer (i.e., 94 and 366 27 kg N ha −1 yr −1 for cotton and the W-M, respectively) and 20% less irrigation water (i.e., 60−180 and 28 230−410 mm yr −1 for cotton and the W-M, respectively) by sprinkling than the conventional practices, 29 Biogeosciences Discuss., https://doi.org/10.5194/bg-2018-401 Manuscript under review for journal Biogeosciences Discussion started: 25 September 2018 c © Author(s) 2018. CC BY 4.0 License.

The level-II scenarios addressed five field management factors, which were the (i) fertilizer dose, 165 (ii) water amount and (iii) method of irrigation, (iv) incorporated/retained fraction of crop residues, and 166 (v) depth of tillage. The values of the five factors used for the baseline were known as the observations 167 for the conventional management practices in the experimental region (Tables S1, S3 and S4). The 168 nitrogen application rates of the baseline were 110 and 430 kg N ha −1 yr −1 for cotton and W-M, 169 respectively. Over the last few decades, the fields in this region have mostly been flood-irrigated (Liu et 170 al., 2010), which was chosen as the baseline condition. The baseline timings and water amounts were 171 established by referring to the 10-to 30-d cumulative precipitation prior to the individual irrigations 172 and the recorded timings and water amounts of conventional management practices in both plots. Thus, 173 the irrigation frequencies and annual cumulative water amounts of the baseline during the 18-year 174 period (Table S4) vary from 1 to 3 times and 75 to 230 mm yr −1 for cotton and 4 to 6 times and 290 to 175 510 mm yr −1 for W-M. In addition, 100% residue incorporation and conventional tillage to a depth of 176 20−30 cm were applied for the baseline conditions. To screen the BMP of six rotation patterns in the 177 interaction with all the considered management practices, the variation in the fertilizer amount, 178 irrigation amount and residue incorporation rate was set as 40% of the baseline to the baseline 179 (N44/172 to N110/430), 40% of the baseline to the baseline (I40 to I100) and 0 to 100% (RI0 to RI100). 180 The irrigated method and tillage factors consisted of flood (IF) and sprinkle (IS) irrigation and 181 no-tillage (T0) and reduced tillage (5 cm and 10 cm, T5 and T10) for W-M and conventional tillage 182 (20−30 cm, T20). We assumed that the frequency distributions of all the factors were uniform. Monte 183 Carlo simulations, at 1000 combined field management scenarios, were used to screen the BMP for 184 each rotation pattern, and the final BMP was selected from the BMPs of six rotation patterns in light of 185 the 6000 combined scenarios. 186 An 18-year simulation was performed to assess the biogeochemical effects of the rotation patterns 187 and management practices. These simulations were driven by the meteorological data observed at the 188 An objective method jointly relying on three constraints and NIPs was adopted in this study to 197 identify the BMP. These constraints included (i) stable or increased crop yields, (ii) stable or increased 198 SOC, and (iii) reduced NEGE. In the present study, the NEGE was the residual of the annual sum for 199 the CH 4 and N 2 O emissions minus the △SOC and was quantified as a CO 2 equivalent (CO 2 eq) quantity 200 for the 100-year time horizon, at 1 for CO 2 , 34 for CH 4 and 298 for N 2 O (IPCC, 2013). The NIP was 201 used to evaluate the potential for a climatically and environmentally integrative impact, which was a 202 price-based proxy quantity in USD ha −1 yr −1 . The NIP was determined by using the quantities of 203 individual decision variables and their coefficients as mass-scaled price-based proxies (Eq. (2)). 204 In Eq.
(2), the NEGE, NH 3 , NO, N 2 O ODM , and NL represent the multi-goal decision variables of 205 the net ecosystem GHG emission (Mg CO 2 eq ha −1 yr −1 ), NH 3 volatilization, NO emission, release of 206 N 2 O as depleted ozone layer matter and hydrological nitrogen loss (mainly by NO 3 − leaching), 207 respectively (kg N ha −1 yr −1 for all the nitrogen compound variables). The coefficients k 1 , k 2 , k 3 , k 4 and 208 k 5 are mass-scaled price-based proxies for the NEGE, NH 3 , NO, N 2 O ODM , and NL, respectively. Their 209 values as presented in Cui et al. (2014) were directly used in this study, and they were 7.00 USD Mg −1 210 CO 2 eq and 5.02, 25.78, 1.33 and 1.92 USD kg −1 N for k 1 , k 2 , k 3 , k 4 and k 5 , respectively. 211 In Eq. (2), a lower NIP indicates a better set of management practices that can exert smaller 212 negative impacts on the climate and environment. Accordingly, the BMP was identified as the scenario 213 with the lowest NIP among the scenarios that could satisfy all three constraints. According to the BMP 214 screening method used in this study, a solid validation of the cumulative emissions of N 2 O, NO, NEE 215 9

Statistics and analysis 217
The statistical criteria of the (i) index of agreement (IA) (Eq. (3)), (ii) Nash-Sutcliffe efficiency 218 index (NSI) (Eq. (4)) (e.g., Moriasi et al., 2007;Nash and Sutcliffe, 1970), (iii) determination 219 coefficient (R 2 ) (Eq. (5)) and slope of a zero-intercept univariate linear regression (ZIR) of observations 220 In this study, the ZIR analysis, variance analysis, and graphical comparison were performed with 231 In addition, the simulated annual N 2 O and NO emissions during the two consecutive experimental 253 years were comparable with the observations, with MRBs of 2% and 11%, respectively, both of which 254 were less than two times the spatial CVs (23−50%) for the measurements (Liu et al., 2014). 255 The simulated NEE flux is one component of the △SOC, which is a key factor that is considered 256 during BMP identification. The simulation suggested that the model captured the seasonal fluctuations, 257 which were negative during the cotton-growing season, but positive or neutral during the remaining 258 periods (Fig. 1e). The IA, NSI, and ZIR slope and  (Table S5). Figure 4 shows the relationship between the 281 annual average of each decision variable and the number of consecutive years of cotton monoculture 282 within the rotation pattern options. 283 The average grain yields for the cotton, wheat and maize were not significantly different among 284 the various rotation pattern options, with averages of 3.5, 4.8 and 6.7 kg dry matter ha −1 for cotton, 285 wheat and maize, respectively (Figs. 3a−c). 286 For the dynamic changes in the annual SOC stocks, the values were generally positive for W-M 287 but negative for the cotton, except for the first year after the transition to this fiber crop. As indicated 288 by Fig. 3d, the simulated SOC contents over the 18-year period increased for R 0 , R 1 , R 2 and R 3 but 289 decreased for R 4 and R 5 . The annual average −△SOC increased significantly (P < 0.001) with an 290 increase in the consecutive years of cotton monoculture from 0 to 5 within the 6-year rotation cycle 291 (Fig. 4a). The rotation pattern options with baseline management showed small variations in the CH 4 292 uptake (Fig. 3e), with the annual uptakes ranging from 1.6 to 2.1 kg C ha −1 . However, the annual 293 averages for the CH 4 uptake increased significantly (P < 0.001) with the increased consecutive years of 294 cotton monoculture (Fig. 4b). For N 2 O, the annual emissions showed large inter-annual variations (Fig.  295 3f), with a CV of 26−48%. In addition, the annual average emissions of this gas decreased significantly 296 from 4.6 to 2.6 kg N ha −1 (Fig. 4c) after increasing consecutive years of cotton monoculture (P < 0.001). 297 As a result, the NEGE was significantly promoted (P < 0.01) (Figs. 3g and 4d). leaching changed insignificantly in response to the consecutive years of cotton monoculture (Fig. 4g). 304 The NIP varied significantly among the various rotation pattern options (P < 0.001), declining 305 from 610 to 324 USD ha −1 yr −1 with increased consecutive years of cotton monoculture (Fig. 4h). For 306 the three constraints, the crop yields showed no obvious differences among the various rotation patterns. 307 Both R 0 and R 5 represent the typical rotation patterns in the region. The simulations for the former 308 indicate the greatest increase in SOC and the lowest NEGE but the highest NIP, while those for the 309 latter show the greatest SOC loss and the largest NEGE but the lowest NIP (Figs. 4a, d and h). These 310 patterns indicate that neither typical rotation pattern is sustainable. 311

Identification of best management practices 312
The Monte Carlo simulation showed that only 51, 22, 16, 9 and 16 of the 1000 combined 313 scenarios from R 0 to R 4 simultaneously satisfied the three constraints, while no scenario met all the 314 13 requirements for R 5 . The screened BMPs for the five rotation patterns indicated that a reduction in the 315 amount of fertilizer and irrigation water can be applied for all the rotation patterns, with the declines of 316 15−21% and 18−27% for the BMPs (87−94 and 334−367 kg N ha −1 yr −1 , and 55−189 and 211−418 mm 317 yr -1 for the cotton and W-M, respectively). In addition, when compared with flood irrigation, sprinkle 318 irrigation was adopted for all the BMPs except that for R 4 . The rate of residue incorporation for the 319 BMPs ranged from 55−90%, which increased with the increased consecutive cotton monoculture years. 320 Regarding the depth of tillage, except for the 10 cm and no-tillage treatment for the BMPs of R 3 and R 4 , 321 the depths of other rotation patterns were all 5 cm. Compared with the baseline of each rotation 322 excluding the R 5 , the NEGE, NH 3 volatilization, NO 3 − leaching and NIP of the BMPs decreased by 323 more than 4%, 20% 41% and 27%, respectively. When ranking the NIPs of each rotation BMP, the final 324 BMP was identified as the BMP of R 3 (N90/353_I82_IS_RI90_T10), with an NIP of 327 USD ha −1 yr −1 . 325 Although the NIP of the BMP for R 4 was slightly lower than that for R 3 , the NEGE and △SOC were 20% 326 higher and 71% lower than that of the final BMP, respectively, and the technology required for 327 no-tillage, such as planting, was not widely available. 328 The identified BMP for the cotton and W-M rotation system showed the following management 329 features: (i) both cotton and W-M are cultivated for three consecutive years within a 6-year rotation 330 cycle; (ii) the present crops and the current schedules of planting, harvesting, fertilization (date, depth, 331 and splits) and irrigation (date and times) are adopted; (iii) urea is applied at a 18% lower rate, namely, 332 90 and 353 kg N ha −1 yr −1 for cotton and W-M, respectively; (iv) 18% less water is used for irrigation 333 by sprinkling than the conventional level; (v) the rate of crop residue incorporation is 90% at harvest; 334 and (vi) conventional tillage (20−30 cm depth) for cotton but reduced tillage (10 cm depth) for W-M 335 are applied. In comparison to the R 3 baseline, i.e., the currently applied field management practices, the 336 identified BMP could produce stable crop yields and enlarge the △SOC (by 4% on average) while 337 decreasing the NEGE (by −4% on average), NH 3 volatilization (by −23% on average), NO emissions 338 (by −9% on average) and NO 3 − leaching (by −44% on average) (Table 1). 339

The uncertainty of the best management practice
the modified model for the cumulative N 2 O, NO, NEE and CH 4 increased by 0−8%, and thus reduced 342 the model uncertainty for validation at an annual scale. In addition, the relative uncertainty resulting 343 from the model validation was calculated based on the MRB and error transfer formula (Eqs. (S1-4)). 344 The MRBs of the cumulative N 2 O, NO, NEE and CH 4 for cotton and W-M were 2% and 8%, 11% and 345 11%, 10% and 4%, and 2% and 2%, respectively, and the MRB of the cumulative NH 3 for W-M was 346 6%. These percentages were used to calculate the relative uncertainty of the NIP for all 6000 scenarios. 347 For the BMP of each rotation pattern, the scenarios, for which the uncertainty ranges had some overlap 348 with that of the BMP, showed no significant differences from one another. Thus, 6, 7, 4, 3 and 0 349 alternative scenarios were selected for the BMPs of R 0 , R 1 , R 2 , R 3 and R 4 , respectively, with an average 350 relative uncertainty of 3.7%. For the final identified BMP of N90/353_I82_IS_RI90_T10 involved in 351 the R 3 rotation pattern, the relative uncertainty of the NIP was 3.1%, ranging from 317 to 338 USD ha −1 352 There were three other alternative scenarios (N94/366_I94_IS_RI75_T20, 353 N94/366_I91_IS_RI95_T10 and N97/378_I88_IS_RI70_T5) in R 3 , which indicated the trade-off 354 effects of different field managements, such as the opposite effect of reduced residue incorporation 355 (decrease △SOC) and tillage depth (increase △SOC) on the △SOC. These scenarios were also regarded 356 as alternative BMPs for the system (Table 1)