Nitrous oxide (N 2 O) is a potent greenhouse gas due to its long residence time in the atmosphere and its strong propensity to deplete ozone. The concentration of atmospheric N 2 O was 330.7 ppb in 2018 and is increasing at a rate of 0.81 ppb per year (NOAA 2018;Tian et al. 2020). Although the atmospheric N 2 O concentration is much lower than carbon dioxide (CO 2 ), its global warming potential is 296 times that of CO 2 (Dalal et al. 2003). Thus, a small difference in estimated N 2 O emissions can considerably influence the warming potential of the atmosphere. Controlling factors, particularly precipitation and nitrogen (N) fertilization, are the two key drivers for soil N 2 O emissions (Dobbie and Smith 2003;Huang et al. 2014;Deng et al. 2016). As a result, projected climate change and anthropogenic activities will influence soil N 2 O emission and very likely positively alter global N cycling.
Soil moisture strongly influences N turnover, transference, and emission of soil N 2 O (Bollmann and Conrad 1998). There is growing evidence that the global warming has significantly altered the global hydrologic cycle at local, regional, and global scales (Borken and Matzner 2009;Reichstein et al. 2013). The intensity and frequency of precipitation have changed compared to historic climatic norms and include more extreme events with higher precipitation rates and longer duration droughts-trends that are expected to continue into the future (Knutson and Tuleya 2004;IPCC 2014;Knapp et al. 2015). For example, a model based on CMIP5 ensemble showed that the Northern Hemisphere will have more wet extremes and drought will also intensify in the north and central America (Zhan et al. 2020). Previous studies showed that N 2 O emission increased with increased precipitation, but most studies focused on the quantity of precipitation instead of precipitation patterns (Wu et al. 2011;Li et al. 2019a). How precipitation patterns involving both intensity and frequency affect soil N 2 O emission is still not clear.
Croplands are considered hot spots of soil N 2 O emission (Wolf and Russow 2000;Reay et al. 2012;Tian et al. 2016;Deng et al. 2019). A modeling synthesis reported that terrestrial CO 2 uptake is largely offset by CH 4 and N 2 O emission (Tian et al. 2016). However, the study notes that the estimation of soil N 2 O emission in croplands remains uncertain. The United States is the world's largest corn producer, with 38% of its corn crop exported to other counties in 2018-2019(NCGA 2019)). In addition to its large US area, corn has among the highest mean fertilizer rates thus pointing to its potential for high N 2 O emissions and opportunity for mitigation. Although N fertilization increases the availability of soil N and crop yield, recovery of N in crop plants is usually less than 50% worldwide (Fageria and Baligar 2005). A metaanalysis showed soil N 2 O emission exhibited a significant positive response to N enrichment in global croplands (? 105.6%) (Deng et al. 2019). Up to 20-30% of global greenhouse gas emissions may be due to N fertilization (Reay et al. 2012;Tian et al. 2016).
The effects of climate change and N application on soil N 2 O emission have been investigated in many field experiments (Zou et al. 2005;Ni et al. 2019). For example, soil N 2 O emissions in a Chinese maize field were reduced under deficit irrigation combined with reduced N-fertilizer rate treatment (Ning et al. 2019;Tian et al. 2017). For the wheat-maize rotation system, an improved management practice of water and fertilizer that via changed the fertilization rate, times and timing has the potential to reduce N 2 O and emissions (Liu et al. 2011). A study in California grassland showed greater soil moisture corresponded to a much higher emission rate in N addition plots (Aronson et al. 2019). These results indicate that responses of soil N 2 O emission could depend on the combinations of soil water availability and N fertilization. However, it is difficult or even impossible to manipulate multiple experimental factors or set many treatment levels for each experimental factor in field experiments (Beier et al. 2012;Li et al. 2019b;Rillig et al. 2019). Most field precipitation experiments only set a few levels of precipitation intensity or frequency. Biogeochemical or ecosystem models have been previously applied to overcome these shortcomings and to assess the effects of climate change or management practices on soil N 2 O emissions in agriculture (Deng et al. 2016;Chen et al. 2019;Tian et al. 2020). Based on field experimental results, models can be parameterized and validated to better 123 simulate the direction and magnitude of soil N 2 O emission response. With the help of models, a better understanding of soil N 2 O emissions resulting from precipitation or irrigation and N fertilization could help reduce N 2 O emissions.
The DeNitrification-DeComposition (DNDC) model was designed to simulate soil greenhouse gases emissions in terrestrial ecosystems (Li et al. 1992a, b). Many previous studies have demonstrated that the DNDC model can simulate the dynamics of soil N 2 O emission very well (Li et al. 1992b;Giltrap et al. 2010;Zhang et al. 2016;Ingraham and Salas 2019). For example, Giltrap et al. (2010) showed that DNDC model is a useful tool for simulating soil N 2 O, CH 4 , and CO 2 emissions on a wide range of land-use and agricultural management. Deng et al. (2016) parameterized the DNDC model based on a three-year field experiment in a cornfield in Nashville, TN, and simulated the effects of different agricultural practices on soil N 2 O emissions (Deng et al. 2016). Some studies also reported limitations of DNDC model when it was used to simulate daily N 2 O emission or to a specific time of emission (Chen et al. 2019;Foltz et al. 2019;Yue et al. 2019). Yue et al. (2019) conducted model comparison of seasonal cumulative N 2 O emission across China's cropland and showed that the DNDC underestimated N 2 O emissions compared to other three models (DAYCENT, a linear regression model, and IPCC Tier 1 emission factors method). Although there are some limitations, a calibrated and validated DNDC model can reliably simulate soil N 2 O emission across a broad range environment conditions (Taft et al. 2019).
In this study, we extended these previous model simulations with different levels of precipitation pattern changes and N fertilization using the previously calibrated DNDC model (Deng et al. 2016). This study builds on these previous studies (Deng et al. 2015(Deng et al. , 2016) ) and further tests how precipitation pattern changes (including 6 levels of intensity and 12 levels of frequency) and N fertilization (3 N levels) would interactively influence soil N 2 O emission in the cornfield. The major objectives of this study were:
(1) to assess the effects of precipitation intensity and frequency on soil N 2 O emission; (2) to simulate the effects of N fertilization on soil N 2 O emission; and (3) to quantify the interactive effects of precipitation intensity and frequency, and N fertilization on soil N 2 O emission.
We used the DNDC model (version 95; http://www. dndc.sr.unh.edu) to simulate and evaluate soil N 2 O emissions from the cornfield. The DNDC model was originally developed to simulate soil greenhouse gases including soil N 2 O emissions from croplands (Li et al. 1992a, b). This model consists a suite of biogeochemical processes, including decomposition, fermentation, ammonia volatilization, nitrification, and denitrification, and allows computation of the complex transfer and transformations of N in agriculture lands (Li et al. 1992a;Deng et al. 2018). The model includes two major components. The first component, consisting of the soil climate, crop growth and decomposition sub-models, predicts soil temperature, moisture, pH, redox potential (Eh) and substrate concentration profiles driven by ecological drivers (e.g., climate, soil, vegetation, and anthropogenic activity). The second component, consisting of the nitrification, denitrification and fermentation sub-models, predicts emission of N 2 O, CO 2 , CH 4 , ammonia, nitric oxide, and dinitrogen from the plant-soil systems. Simulated soil N 2 O emissions are primarily regulated by soil environment variables, e.g., soil temperature and water-filled pore space (WFSP), and substrate availability (e.g., dissolved organic carbon and inorganic N) (Deng et al. 2016).
Comparing simulated results with experimental measurements was used to verify and validate the model. In brief, the model validation was based on a threeyear cornfield experiment conducted at Tennessee State University Agricultural Research and Education Center (latitude 36.12°N, longitude 86.89°W, elevation 127.6 m) in Nashville, TN, USA. The model inputs included meteorological data, soil properties, crop parameters, and farming management practices (Deng et al. 2015(Deng et al. , 2016)). More detailed description of this no tillage and regular N fertilization study site, field sampling method, and model validation and results were given in Deng et al. (2015Deng et al. ( , 2016)). The local meteorological data during the experiment period were acquired from the weather station (Davis Instruments, Vernon Hills, IL, USA) at the experiment site. The cropland management practices, including planting and harvest, tillage, fertilization, and irrigation, and all of the date and quantity were determined based on the field measurements and operation. The DNDC model was firstly parameterized based on the measured initial soil physical and chemical parameters to simulate daily N 2 O emission in the ambient precipitation and typical N fertilization (control_TN) treatment (Deng et al. 2016). Then this parameterized DNDC model was used to simulate daily N 2 O emissions for the other 53 treatments.
To simulate the effects of precipitation pattern (i.e., intensity and frequency) change and N fertilization on soil N 2 O emission in the cornfield, we set 18 different precipitation patterns including the control treatment and three N fertilization rates. We used the three-year (2012-2014) ambient precipitation (including intensity and frequency) and typical N fertilization by farms as base case scenarios or controls (Deng et al. 2016). The ambient total annual precipitation was 1164, 1394, and 1367 mm in 2012, 2013, and 2014, respectively. Here we can conclude that 2012 was a relatively dry year because the annual precipitation of 2012 was reduced by 15.7% compared to the average value of the other two years. Each of these three years was simulated by altering the precipitation intensity without the change of frequency or by changing of the precipitation frequency without the change of annual total quantity. Overall, the 17 precipitation pattern changes included 5 precipitation intensity changes (2 decreased and 3 increased precipitations) and 12 precipitation frequency changes (7 less frequent precipitations and 5 more frequent precipitations). For precipitation intensity change, we set -50% (50% decrease in ambient precipitation), -30% (30% decrease in ambient precipitation), ? 30% (30% increase in ambient precipitation), ? 50% (50% increase in ambient precipitation), and ? 100% (double than the ambient precipitation) without changing the precipitation frequency. For precipitation frequency changes, we maintained the same total amount of precipitation in the years but changed the precipitation frequency by merging or splitting the precipitation events. We considered continuous precipitation days as one event. Less frequent precipitation treatments included M2, M3, …, M8 by merging 2, 3, …, or 8 precipitation events into one large event, respectively. For example, M3 merged precipitation events 1, 2, 3 as event 1, and events 4, 5, 6 as event 2. In the merged treatments, precipitation became less frequent but high intensity (LFHI). S2, S3, …, S6 were more frequent but less intensity (MFLI) precipitation treatments, implemented by splitting one precipitation event to 2, 3, …, or 6 small precipitation events, respectively. For example, we had one precipitation event with four consecutive rainy days, and split it into two precipitation events (S2). We divided four days total precipitation into two even portions. One portion was proportionally allocated to these four days (reduced by half), and another portion was allocated to the middle of the day between the last day of this precipitation event and the first day of the next event.
After we split one precipitation into 6 events, the precipitation events almost occurred daily the whole year.
As for N fertilization application treatment, we set three N fertilization rates: TN (typical N fertilization at 118 kg N ha -1 yr -1 ), LN (low N fertilization treatment, half the typical N fertilization at 59 kg N ha -1 yr -1 ), and HN (high N fertilization treatment, ? 50% of the typical N fertilization at 177 kg N ha -1 yr -1 ). The LN and HN treatments were within the range of the N fertilizer applied by local farmers to cornfields. Chicken manure (99 kg N ha -1 yr -1 ) was applied to all treatments before the seeds were planted (Deng et al. 2015), and aqueous urea ammonium nitrate was applied twice once on vegetative stage and another one on reproductive stage at the set N application levels accordingly. The detailed fertilization schedule can be found in Table 1.
To assess the interactive effects of precipitation pattern and N fertilization on soil N 2 O emission, we ran the DNDC model under the combinations of 18 precipitation pattern changes and three fertilization rates, for a total of 54 simulations. Daily dynamics of soil N 2 O emission and emissions through nitrification and denitrification were simulated. Effects of different precipitation patterns and N fertilization and their interactions on soil N 2 O emission were detected using the analysis of variation (ANOVA). Multiple comparisons were conducted using Least Square Difference (LSD) method. Data analysis was conducted using 123 IBM SPSS Statistics 21.0 for Windows (IBM Corp., Armonk, NY, USA).
Results of ANOVA showed that soil mean annual N 2 O emission was significantly affected by precipitation intensity and the interaction between intensity and N fertilization (p \ 0.01; Table 2). Compared to the soil N 2 O emission in the ambient precipitation intensity (control), the reduced precipitation intensity treatments (-50% and -30%) decreased soil N 2 O emission by -50%, -33%, respectively, and the three enhanced precipitation intensity treatments (? 30%, ? 50%, ? 100%) increased soil N 2 O emission by 38%, 67%, and 132%, respectively (Fig. 1a). Simulated soil N 2 O emission was more responsive to precipitation change under the high intensity precipitation treatments than the low intensity precipitation (Fig. 1b). There was no significant difference in soil N 2 O emission among years (p [ 0.05, Table 2). The lowest annual mean N 2 O emission was in the -50% precipitation treatment (1.1 kg N ha -1 yr -1 ) and the highest value was in the ? 100% precipitation treatment (4.9 kg N ha -1 yr -1 ) (Fig. 1a). The changes of precipitation intensity and N 2 O emission were described with polynomial functions (Fig. S1A, R 2 = 0.99). Daily dynamics of soil N 2 O emissions simulated using the DNDC model showed pulse responses, particularly during the growing seasons (April to September), and were dramatically enhanced as precipitation intensity increased from -50% to the control and to ? 100% treatments (Fig. S2). The high soil N 2 O emission peaks mostly occurred in the growing season for all three years.
We changed the frequency of precipitation events by either merging the 2-8 succeeding precipitation events together (LFHI, denoted as M2 to M8) or splitting one precipitation event into 2-6 precipitation events (MFLI, denoted as S2 to S6) without changing the total amount of precipitation (Fig. 2). The number of precipitation events increased from precipitation frequency treatment M8 to M2, to control, and from S2 to S6 (Fig. 3a). The minimum number of precipitation events of M8 and M7 was an average of 17 times per year across the three years of 2012-2014 (Figs. 3a,S3). The maximum number of precipitation events was S6, averaging to 320 times per year across the three years (Figs. 3a,S3).
Change in precipitation frequency significantly influenced soil N 2 O emissions (p \ 0.01, Table 2). Soil N 2 O emission did not vary significantly among the LFHI treatments (M3 to M8) and decreased as precipitation frequency increased from M2 to control, and to S2. Moreover, N 2 O emissions did not vary significantly among the S3 to S6 treatments (p [ 0.05, Fig. S1B). Compared to mean soil N 2 O emission over three years under the ambient precipitation (control), simulated soil N 2 O emission was increased by ? 51%, ? 54%, ? 51%, 48%, ? 40%, ? 51%, and ? 24% in LFHI treatments (M8, M7, M6, M5, M4, M3, and M2), respectively (Fig. 2b). Soil N 2 O emission in the reduced precipitation frequency treatment S2, S3, S4, S5, and S6 was decreased by -14%, -17%, -30%, -27%, and -31%, respectively (Fig. 2b). The three years of annual mean soil N 2 O emission ranged from 2.6 kg N ha -1 yr -1 to 3.3 kg N ha -1yr -1 under the LFHI treatments (M8-M2) and 1.5 kg N ha -1 yr -1 to 1.8 kg N ha -1 yr -1 under the MFLI treatments (Fig. 2a). Soil N 2 O emission was higher in the LFHI treatments, and lower in the MFLI 123 treatments (Fig. 2). MFLI precipitation treatments (S2 to S6) reduced the occurrence of soil N 2 O emission peaks (Fig. S3).
Nitrogen fertilization had a significant effect on soil N 2 O emission (p \ 0.01, Table2). Linear regression analysis showed that the soil N 2 O emission was significantly positively correlated to the N fertilization rate (Fig. 4). Soil N 2 O emission was higher in the HN treatment and lower in the LN treatment compared to the TN. Annual mean soil N 2 O emissions over the three years was 1.2, 2.1, and 3.2 kg N ha -1 yr -1 in the LN, control, and HN treatments, respectively (Fig. 4).
Soil N 2 O emission in the HN treatment was 2.7 times than that in the LN treatment. Soil N 2 O emission varied more between years at low N fertilization relative to high N fertilization (Fig. 4a).
A significant interaction on soil N 2 O emission was only found between precipitation intensity and N application (p = 0.03, Table 2). The daily soil N 2 O emission over the three years showed that soil N 2 O emission increased abruptly and reached high peaks at rainy days with N fertilization (Fig. S2). Compared to the mean soil N 2 O emission under the control_TN treatment (2.1 kg N ha -1 yr -1 ), the highest soil N 2 O emission (increased by 203%) occurred in the ? 100% precipitation intensity and the high N fertilization (? 100%_HN) treatment (Fig. 1). The lowest soil N 2 O emission was simulated in the -50% precipitation and the low N fertilization (-50%_LN) treatment (0.5 kg N ha -1 yr -1 ), where it was decreased by 75% compared to the control_TN treatment (Fig. 1).
The N fertilization with higher intensity precipitations increased soil N 2 O emission compared to the lower intensity precipitations. Soil N 2 O emission increased nonlinearly with precipitation intensity for all three N fertilization treatments (Fig. 1a). Overall, high intensity precipitation treatments (? 100%, ? 50%, and ? 30%) usually generated more soil N 2 O emission (higher than 2.5 kg N ha -1 yr -1 except the ? 30%_ LN), while the low intensity precipitation treatments (control, -30% and -50%) produced less soil N 2 O emission (lower than 2.5 kg N ha -1 yr -1 except the control_HN treatment) (Fig. 1b). We selected this threshold value of 2.5 kg N ha -1 yr -1 as it could separate all treatments into two groups at the control level mostly (Fig. 1a).
There was no significant interaction between precipitation frequency and N fertilization (p [ 0.05, Table 2). Soil N 2 O emission was not influenced by less frequent treatments (M3-M8) or by more frequent precipitation treatments but reduced from M3 to M2, the control, and S2 (Fig. S1B). Compared to the control_TN treatment, we observed that under the LN and TN treatments, LFHI precipitation increased soil N 2 O emission, but MFLI precipitation treatments reduced soil N 2 O emission (Fig. 2). However, under the HN treatment, both LFHI and MFLI increased soil N 2 O emission compared to the control and TN treatment (Fig. 2b).
The model predicted that soil N 2 O emission increased with increases in precipitation intensity. This phenomenon is prevalent in most terrestrial ecosystems, particularly in forests and grasslands, as revealed in a meta-analysis (Yan et al. 2018). Changes in precipitation intensity directly influence soil moisture conditions that determine soil N 2 O emission processes (Attard et al. 2011). Indeed, soil moisture increased with increases in precipitation intensity (Fig. 3b), and annual soil N 2 O emissions was increased with an increase in soil moisture in this study (Fig. 6a). Soil moisture is an important factor that strongly influences the processes of nitrification and denitrification -50%: a 50% decrease in ambient precipitation; -30%: a 30% decrease in ambient precipitation; control: the ambient precipitation; ? 30%: a 30% increase in ambient precipitation; ? 50%: a 50% increase in ambient precipitation; ? 100%: double the ambient precipitation. Relationships between soil N 2 O emission and N application are embedded as inlet and the shaded areas represent 95% confidence intervals (A). Asterisks indicate the level of significance (***p \ 0.001).
The horizontal line indicates that annual N 2 O emission is 2.5 kg N ha -1 yr -1 (a). The simulated treatment is denoted as ''precipitation pattern_nitrogen level'' (b) 123 (Bremner 1997;Wrage et al. 2001;Rochette et al. 2010). Both annual soil nitrification and denitrification rates were enhanced with the increases in precipitation intensity (Fig. 5a,b). Soil nitrification dominates the total soil N 2 O emission, while denitrification has more potential than nitrification in producing more soil N 2 O emission under anaerobic and saturated conditions (Wolf and Russow 2000;Mathieu et al. 2006). A previous study also showed that under an anaerobic condition with more precipitation, the fraction of N 2 O loss from enhanced denitrification can be 10 times larger than that in the nitrification process (Xu et al. 2019). Compared to the control, in our study, soil N 2 O emission, nitrification, and denitrification under the ? 100% precipitation treatment increased by 132%, 7%, and 221%, respectively (Figs. 1b,5a,b). Under the high precipitation intensity condition, the increased soil moisture and an anaerobic environment would stimulate the decomposition of organic matter and contribute to the process of denitrification (Chen et al. 2013), allowing simultaneous nitrification and denitrification (Butterbach-Bahl et al. 2013). The nonlinear and asymmetric response of soil N 2 O emissions to soil moisture and precipitation intensity treatments (Fig. 6c) showed the soil N 2 O emission was more responsive to increased precipitation than 123
Fig. 3 The precipitation events under different simulated precipitation frequency treatments (a), the soil average WFPS (%) under different simulated precipitation treatments (b) among different years. -50%: a 50% decrease in ambient precipitation; -30%: a 30% decrease in ambient precipitation; control: the ambient precipitation; ? 30%: a 30% increase in ambient precipitation; ? 50%: a 50% increase in ambient precipitation; ? 100%: double the ambient precipitation. Mn: merged n precipitation events into one event; Sn: spilt one precipitation event into n events; control: the ambient precipitation; WFPS: water-filled porosity that to decreased precipitation. Similar patterns were shown for plant productivity and ecosystem carbon fluxes (Wu et al. 2011). These were due to simultaneous increases in both nitrification and denitrification with high intensity precipitation.
In this study, the LFHI treatments reduced precipitation events, increased precipitation intensity and soil moisture for each event, as a result, soil N 2 O emission was higher than the control (Figs. 2,S3). The MFLI treatments showed opposite effects on these variables. Furthermore, soil moisture increased nonlinearly with precipitation intensity and number of precipitation events (Figs. 3,6a). As the total amount of precipitation was the same for all precipitation frequency treatments, MFLI precipitation treatments increased soil moisture, but reduced soil N 2 O emission. In a previous study, Liang et al. (2016) showed that soil N 2 O emission declined as drying-rewetting cycles increased, similar to our result. However, the relationship of soil N 2 O emission and soil moisture under the precipitation frequency treatments was quite different from that under the precipitation intensity treatments. Decreasing soil N 2 O emissions with increases in soil moisture in this study (Fig. 6d) were contradictory to the finding of Fentabil et al. (2016) who found that more frequent irrigation increased soil N 2 O emission in an orchard. The inconsistencies could be caused by the differences in the amount of water added in precipitation or irrigation treatments or the difference in ecosystems. As the same amount of precipitation was added, precipitation frequency treatments had less effects on soil annual N leaching, compared to precipitation intensity treatments (Fig. 5d). The soil N loss through surface runoff was also similar under all precipitation frequency treatments (Fig. 5c). Thus, the decreased soil N 2 O emission in the MFLI treatments could be due to the obvious decrease in soil denitrification (Fig. 5b). Precipitation interval or drought duration could influence substrate supply and microbial activities and affect soil N 2 O emission (Xiang et al. 2008;Borken and Matzner 2009). Rapid soil microbial response to frequent rewetting events and concomitant greater moisture availability often resulted in instantaneous C and N mineralization, followed by shifts in C/N of microbially available substrate, and enhanced potentially available N (Austin et al. 2004;Butterly et al. 2011;Lopez-Sangil et al. 2018). During the drought periods, the hydration and lysis of dead microbial cells would increase substrate availability (Fierer and Schimel 2002;Borken and Matzner 2009). The additional mineralization increased the release of N from soils. Meanwhile, frequent rewetting events might cause the breakage of soil macroaggregates and expose physically protected organic matter that was difficult for microorganisms to reach, and more available substrate will be supplied to microorganisms (Denef et al. 2001). In this study, we found that the LFHI treatments enhanced the denitrification process -30%: a 30% decrease in ambient precipitation; control: the ambient precipitation; ? 30%: a 30% increase in ambient precipitation; ? 50%: a 50% increase in ambient precipitation; ? 100%: double the ambient precipitation. Mn: merged n precipitation events into one event; Sn: spilt one precipitation event into n events; control: the ambient precipitation (Fig. 5b). As a result, an immediate pulse in soil N 2 O emission could be produced (Fig. S3).
Soil N 2 O emission increased with N application (Bouwman et al. 2002;Charles et al. 2017;Wang et al. 2018;Volpi et al. 2019). This was mainly because the N is the substrate for microorganisms related to soil nitrification and denitrification processes. Indeed, we found that nitrification and denitrification were enhanced by N application, particularly soil nitrification (Fig. 5b). Nitrogen leaching and N runoff were also increased, especially when exceeded N fertilization was applied.
Regarding the relationship between soil N 2 O emission and N application, both linear (e.g., Albanito et al. 2017) and nonlinear (Kim et al. 2013;Shcherbak et al. 2014;Ning et al. 2019) relationships have been reported. In this study, there were good linear relationships between N 2 O emission and N fertilization rate (p \ 0.05, Fig. 4). We also found that annual variation was larger under the LN treatment than the TN and HN treatments (Fig. 4a). Soil N 2 O emission was lower in the dry year (2012), probably due to lower nitrification as soil nitrifiers could be more limited by lower soil moisture (Brown et al. 2012). -50%: a 50% decrease in ambient precipitation; -30%: a 30% decrease in ambient precipitation; control: the ambient precipitation; ? 30%: a 30% increase in ambient precipitation; ? 50%: a 50% increase in ambient precipitation; ? 100%: double the ambient precipitation. Mn: merged n precipitation events into one event; Sn: spilt one precipitation event into n events; control: the ambient precipitation; WFPS: water-filled porosity. Asterisks indicate the level of significance (*p \ 0.05; **p \ 0.01; ***p \ 0.001)
The changes of soil N 2 O emission between the HN and LN fertilization varied from 104.2% in 2013 to 327.7% in 2012 (Fig. 4a). It indicated that soil N 2 O emission was more sensitive to N availability in a dry year. However, these results should be interpreted with caution, since we only tested three N application levels, as one of the main objectives in this study was to test the responses of the soil N 2 O emission to precipitation pattern changes under different N application levels. Further studies need to be conducted with multiple N applications to develop their Interactive effects of precipitation pattern change and N fertilization on soil N 2 O emission In this study, increased precipitation intensity with N fertilization generally increased soil N 2 O emission (Fig. 1). Soil N 2 O emission increased with precipitation intensity treatments from dry to wet years under all N fertilization treatments, but soil N 2 O emission was stimulated more under the higher N fertilization than the typical and the low N fertilization treatments (Fig. 1a). The lowest soil N 2 O emission under the -50% and LN treatment was caused by low nitrification and denitrification (Fig. 5a,b). The higher sensitivity of soil N 2 O emission to precipitation intensity was mostly caused by the stronger increases in soil denitrification under the high precipitation intensity (Fig. 5b). Under the control, -30% and -50% treatments, soil N 2 O emission was lower than 2.5 kg N ha -1 yr -1 for all N fertilization treatments except the HN with the control, and under the other precipitation intensity treatments, soil N 2 O emission was higher than 2.5 kg N ha -1 yr -1 except the LN with ? 30% and ? 50% treatments. These results indicated that water availability was a more important factor in reducing soil N 2 O emission in the field. The highest soil N 2 O emission occurred in the HN with ? 100% precipitation treatment, due to both high soil nitrification and denitrification under this condition (Fig. 1b).
In this study, we found no significant interaction of precipitation frequency and N fertilization on soil N 2 O emission. We observed different soil N 2 O emission patterns with precipitation frequency treatments, but the responses were similar under different N fertilization treatments (Fig. 2a). As precipitation frequency decreased and intensity increased, soil nitrification was enhanced, but denitrification was mostly reduced, resulting in no significant differences among the MFLI treatments or among the LFHI treatments, for all N fertilization levels.
The timings of N fertilization and precipitation/ irrigation could also influence responses of soil N 2 O emission. Applying N fertilizer on or after a rainy day produced the highest peak of soil N 2 O emission (Fig. S3), as more substrate and water availability simultaneously stimulated soil microbial activities. We also found that the annual soil N 2 O emission was unexpectedly increased in the S3 and S5 precipitation treatments under the HN treatment (Fig. 2a). The higher annual N 2 O emission might be due to the soil N 2 O emission peaks in November (Fig. S3). Higher soil N 2 O emission corresponded to higher precipitation events, but the emissions occurred always during or soon after the precipitation. When drought was followed by large precipitation, increased soil moisture would likely enhance soil N 2 O emission in periodic pulses (Aronson et al. 2019). The effects of N fertilization and precipitation on soil N 2 O emission were additive especially in the dry years (Zhang et al. 2017). In this study, the HN treatment greatly increased soil N 2 O emission and showed the greatest changes in 2012, a relatively dry year (Fig. 4a).
Based on 54 simulations of soil N 2 O emission in a cornfield over three years using the DNDC model, we investigated the main and interactive effects of precipitation pattern (intensity and frequency) change and N fertilization. Less frequent but high intensity precipitation stimulated soil N 2 O emission, particularly under the high nitrogen application. Similar results could be found in croplands and grasslands with similar climate conditions and agricultural practices. However, soil N 2 O emission in a changing environment is unlikely affected by a single factor. Understanding the interactive effects of precipitation patterns, N fertilization, and the timing of management practices provides insights regarding timing and intensity of irrigation and fertilization, and improves water and N use efficiencies in agriculture. Although we have qualitatively estimated the response of N 2 O emission to many scenarios, the changing of precipitation events to global warming may be more 123 complex than we expected. It implies that we should study the interaction between precipitation intensity and frequency rather than these two factors individually. More field experiments with combined precipitation pattern changes and N fertilization are needed to confirm the interactive effects of precipitation pattern and N application in croplands observed here.
Nutr Cycl Agroecosyst