EFFECTS OF FERTILIZATION ON POREWATER NUTRIENTS, GREENHOUSE-GAS EMISSIONS AND RICE PRODUCTIVITY IN A SUBTROPICAL PADDY FIELD

SUMMARY Suitable fertilization is crucial for the sustainability of rice production and for the potential mitigation of global warming. The effects of fertilization on porewater nutrients and greenhouse-gas fluxes in cropland, however, remain poorly known. We studied the effects of no fertilization (control), standard fertilization and double fertilization on the concentrations of porewater nutrients, greenhouse-gas fluxes and emissions, and rice yield in a subtropical paddy in southeastern China. Double fertilization increased dissolved NH4+ in porewater. Mean CO2 and CH4 emissions were 13.5% and 7.4%, and 20.4% and 39.5% higher for the standard and double fertilizations, respectively, than the control. N2O depositions in soils were 61% and 101% higher for the standard and double fertilizations, respectively, than the control. The total global warming potentials (GWPs) for all emissions were 14.1% and 10.8% higher for the standard and double fertilizations, respectively than the control, with increasing contribution of CH4 with fertilization and a CO2 contribution > 85%. The total GWPs per unit yield were significantly higher for the standard and double fertilizations than the control by 7.3% and 10.9%, respectively. The two levels of fertilization did not significantly increase rice yield. Prior long-term fertilization in the paddy (about 20 years with annual doses of 95 kg N ha−1, 70 kg P2O5 ha−1 and 70 kg K2O ha−1) might have prevented these fertilizations from increasing the yield. However, fertilizations increased greenhouse-gas emissions. This situation is common in paddy fields in subtropical China, suggesting a saturation of soil nutrients and the necessity to review current fertilization management. These areas likely suffer from unnecessary nutrient leaching and excessive greenhouse-gas emissions. These results provide a scientific basis for continued research to identify an easy and optimal fertilization management solution.

In China, 90% of paddies are in the subtropics, such as in Fujian, Jiangxi and Hunan Provinces. Developing effective strategies to increase the cost-effectiveness of rice agriculture, enhance crop yield and mitigate GHG emissions from paddies in subtropical China to minimize future food shortages and adverse climate change is thus a global objective of national importance. We pursued this objective by (1) determining the emissions of CO 2 , CH 4 and N 2 O in response to the application of different amounts of fertilizers in paddy fields; (2) exploring the effect of amendment amount on the concentrations of porewater nutrients; and (3) assessing the impacts of the fertilizer applications on crop productivity. The results obtained in this study will provide a scientific basis for the suitable management of fertilization for rice agriculture, and the evaluation of most current fertilization strategies for increasing yield and decreasing GHG emissions.

Study site
Our study was conducted at the Wufeng Agronomy Field of the Fujian Academy of Agricultural Sciences in Fujian Province, southeastern China (26.1°N, 119.3°E) (Supplementary Figure S1, available online at https://doi.org/10.1017/ S0014479718000078). A field experiment was carried out during the early paddy season (16 April to 16 July) in 2014. Air temperature and humidity during the study period are shown in Figure S2. The soil of the paddy was poorly drained, and the proportions of sand, silt and clay particles in the top 15 cm of the soil were 28%, 60% and 12%, respectively (Wang et al., 2015). Other properties of the top 15 cm of soil at the beginning of the experiment were bulk density, 1.1 g cm −3 ; pH (1:5 with H 2 O), 6.5; organic carbon (C) content, 18.1 g kg −1 ; total nitrogen (TN) content, 1.2 g kg −1 and total phosphorus (TP) content, 1.1 g kg −1 (Wang et al., 2015). The water level was maintained at 5-7 cm above the soil surface throughout the growing season by an automatic water-level controller, and the paddy was last drained 2 weeks before harvest.
We established triplicate plots (10 × 10 m) for two treatments and a control in which rice seedlings (cultivar, Hesheng 10) were inserted to a depth of 5 cm with a spacing of 14 × 28 cm using a rice transplanter. The age of the rice seedlings was 21 days when they were transplanted. The two treatments consisted of standard fertilization and fertilization with a double amount of the standard fertilization. The control plots were not fertilized. The standard fertilization treatment consisted of three applications of N-P 2 O 5 -K 2 O (16-16-16%; Keda Fertilizer Co., Ltd. Jingzhou, China) and urea (46% N) fertilizers. The first application was 1 day before transplantation at rates of 42 kg N ha −1 , 40 kg P 2 O 5 ha −1 and 40 kg K 2 O ha −1 . The second application was broadcasted during tiller initiation [7 days after transplanting (DAT)] at rates of 35 kg N ha −1 , 20 kg P 2 O 5 ha −1 and 20 kg K 2 O ha −1 . The third application was broadcasted during panicle initiation (56 DAT) at rates of 18 kg N ha −1 , 10 kg P 2 O 5 ha −1 and 10 kg K 2 O ha −1 . These doses are the most commonly used in paddy fields in subtropical China (with annual doses of 95 kg N ha −1 , 70 kg P 2 O 5 ha −1 and 70 kg K 2 O ha −1 ), which constitute about 90% of the paddy fields in China (Wang et al., 2015). The double-fertilization treatment consisted of twice the amounts of the standard fertilization but the same schedule. The plots had previously been fertilized with amounts equal to the standard fertilization during a period of about 20 years (Wang et al., 2015). All control and treatment plots received the same amount of water. The field was plowed to a depth of 15 cm with a moldboard plow and was levelled 2 days before rice transplantation.

Measurement of CO 2 , CH 4 and N 2 O emissions
Static closed chambers were used to measure CO 2 , CH 4 and N 2 O emissions. The chambers were made of polyvinyl chloride (PVC) and consisted of two parts, an upper transparent compartment (100 cm high, 30 cm wide, 30 cm long) placed on a permanently installed bottom collar (10 cm high, 30 cm wide, 30 cm long). Each chamber had two battery-operated fans to mix the air inside the chamber headspace, an internal thermometer to monitor temperature changes during gas sampling and a gas-sampling port with a neoprene rubber septum at the top of the chamber for collecting gas samples from the headspace. We deployed three replicate chambers in each plot. A wooden boardwalk was built for accessing the plots to minimize the disturbance of the soil during gas sampling. The chambers were installed on soil with plants inside. The temperature increase inside the chambers was measured and taken into account in the calculations.
Gas flux was measured weekly in all the chambers. Gas samples were collected from the chamber headspace using a 100-mL plastic syringe with a three-way stopcock 0, 15 and 30 min after the deployment of the upper compartment. The samples were immediately transferred to 100-mL air-evacuated aluminium foil bags (Delin Gas Packaging Co., Ltd., Dalian, China) sealed with butyl rubber septa and transported to the laboratory for the analysis of CO 2 , CH 4 and N 2 O. CO 2 , CH 4 and N 2 O concentrations in the headspace air samples were determined by gas chromatography (Shimadzu GC-2010 and Shimadzu GC-2014, Kyoto, Japan) using a stainless steel Porapak Q column (2 m long, 4 mm outer diameter and 80/100 mesh). A methane conversion furnace, flame ionization detector (FID) and electron capture detector (ECD) were used for the determination of the CO 2 , CH 4 and N 2 O concentrations, respectively. The operating temperatures of the column, injector and detector for the determination of CO 2 , CH 4 and N 2 O concentrations were adjusted to 45, 100 and 280°C; to 70, 200 and 200°C and to 70, 200 and 320°C, respectively. These temperatures were the optimum temperatures for the different parts of the instrument. Helium (He) (99.999% purity) was used as a carrier gas (30 mL min −1 ), and a make-up gas (95% argon and 5% CH 4 ) was used for the ECD. The gas chromatograph was calibrated before and after each set of measurements using 503, 1030 and 2980 μL CO 2 L −1 in He; 1.01, 7.99 and 50.5 μL CH 4 L −1 in He and 0.2, 0.6 and 1.0 μL N 2 O L −1 in He (CRM/RM Information Center of China) as primary standards. CO 2 , CH 4 and N 2 O fluxes were then calculated as the rate of change in the mass of CO 2 , CH 4 and N 2 O per unit of surface area and per unit of time by using a closed-chamber equation (Ali et al., 2008): where F is the corresponding gas flux (mg/mg m −2 h −1 ), M is the molecular weight, V is the height of the chamber above the water surface (m) and T is the air temperature inside the chamber (°C).

Global warming potential
GWP is typically estimated using CO 2 as the reference gas, and a change in the emission of CH 4 or N 2 O is converted into 'CO 2 -equivalents' (Hou et al., 2012). The GWP for CH 4 is 34 (based on a 100-year time horizon and a GWP for CO 2 of 1), and the GWP for N 2 O is 298 (Myhre et al., 2013). The GWP of the combined emission of CH 4 and N 2 O was calculated as (Ahmad et al., 2009)

Collection of soil porewater
Porewater was sampled in situ once a week from 16 April to 16 July 2014. Three specially designed PVC tubes (5.0 cm inner diameter) were installed to a depth of 15 cm in each plot. Porewater samples were collected using 50-mL syringes and then separated into two parts: About 10 mL was injected into pre-evacuated 20-mL vials and the remaining 40 mL was injected into 100-mL sample vials. The samples were stored in a cool insulated box in the field until transported to the laboratory where they were stored at −20°C until the analysis of nutrients and GHG concentrations.

Measurement and calculation of dissolved CO 2 , CH 4 and N 2 O concentrations
The sample vials were thawed at room temperature and then vigorously shaken for 5 min to equilibrate the CH 4 concentrations between the water and the headspace. Gas samples were collected from the headspaces of the vials and analysed for CO 2 , CH 4 and N 2 O concentrations by gas chromatography (Shimadzu GC-2010 and Shimadzu GC-2014, Kyoto, Japan; see above for more details).
The concentrations (C) of CO 2 , CH 4 and N 2 O dissolved in the water were calculated following Ding et al. (2003) where Ch is the CO 2 , CH 4 and N 2 O concentration (μL L −1 ) in the air samples from the vials, Vh is the volume of air in the vial (mL) and Vp is the volume of the water in the vial (mL).

Measurement of soil, porewater properties and rice yield
The soil temperature, pH, salinity, redox potential (Eh) and water content of the top 15 cm of soil were measured in triplicate in situ in each plot on each sampling day. Temperature, pH and Eh were measured with an Eh/pH/temperature meter (IQ Scientific Instruments, Carlsbad, USA), salinity was measured using a 2265FS EC meter (Spectrum Technologies Inc., Paxinos, USA) and water content was measured using a TDR 300 meter (Spectrum Field Scout Inc., Aurora, USA). The concentrations of NH 4 + , NO 3 − , TN and TP in the porewater were determined using a sequence flow analyser (San ++ , SKALAR Corporation production, Breda, The Netherlands). The concentration of dissolved organic carbon (DOC) was determined using a Total Organic Carbon (TOC) analyser (TOC-V CPH, Shimadzu Corporation, Kyoto, Japan) and a filter paper of pore diameter 60 μm. Rice yield was determined at the harvesting stage by manual harvest (Wang et al., 2015).

Statistical analysis
Differences in soil properties; CO 2 , CH 4 and N 2 O emissions; dissolved porewater CO 2 , CH 4 and N 2 O concentrations and porewater nutrient concentrations among the treatments and control were tested for statistical significance by repeated-measures analyses of variance (ANOVAs). The relationships between mean GHG emissions and the soil properties, dissolved porewater GHG concentrations and porewater nutrient concentrations were determined by Pearson correlation analysis. These statistical analyses were performed using SPSS Statistics 18.0 (SPSS Inc., Chicago, USA). We analysed the effects of multiple soil variables as fixed factors on the production rates of the three GHGs using general linear models with and without spatial correlation. We used linear (lm) and mixed (lme) functions with the 'nlme' and 'lme4' R packages. We chose the best model for each dependent variable using the Akaike information criterion. We used the MuMIn R package in mixed models to estimate the percentage of variance explained by the model.

R E S U LT S
GHGs dissolved in porewater and emitted from the paddy CO 2, CH 4 and N 2 O emissions varied significantly among most sampling dates (p < 0.01; Table S1), but the treatments and the interaction of sampling date and treatment were not significant (p > 0.05). CO 2 flux generally remained low (< 354 mg m −2 h −1 ) during the first 29 DAT but then increased to a seasonal peak (> 2811 mg m −2 h −1 ) between 29 and 71 DAT ( Figure 1a). The rice was nearly ripe by 71 DAT, with a corresponding decrease in CO 2 emissions until harvesting in July. The CH 4 emissions were low soon after rice transplantation and peaked by 71 DAT in all treatments (Figure 1b). The paddy was drained after the rice reached maturity, with a corresponding decrease in CH 4 emissions until harvesting in July.
Mean CO 2 emissions were 13.5% and 7.4% higher for the standard and double fertilizations, respectively, than the control. Mean CH 4 emissions were 20.4% and 39.5% higher for the standard and double fertilizations, respectively, than the control. Mean N 2 O soil depositions were 61% and 101% higher for the standard and double fertilizations, respectively, than the control, mostly due to the lower (negative) values at 36 DAT (Figure 1c), despite no overall effect of the treatments on N 2 O emission determined by the mixed linear models (Table 1). Dissolved CO 2, CH 4 and N 2 O concentrations varied significantly among sampling dates (p < 0.01; Table S1). Treatments and the interaction between date and treatment were significant for the dissolved CO 2 concentration but not for dissolved CH 4 and N 2 O concentrations (p > 0.05; Table S1; Figure 2). Mean dissolved CO 2 concentration was 11.4% lower for the standard fertilization and 95.0% higher for the double fertilization than the control. Mean dissolved CH 4 concentrations were 25.8% and 18.9% lower for the standard and double fertilizations, respectively, than the control. Mean dissolved N 2 O concentrations were 21.0% and 73.5% higher for the standard and double fertilizations, respectively than the control. The mixed linear models (with plot and time as random factors) showed that standard fertilization increased CO 2 emissions and soil Eh and salinity, whereas double fertilization increased CH 4 emission, porewater CO 2 concentration, soil salinity and porewater NH 4 + and DOC concentrations (Table 1). This mixed linear model analysis is more robust to detect the effects of treatments than the previously commented ANOVA. CO 2 emission was correlated positively with soil Eh, temperature, water content and porewater NH 4 + concentration and negatively with soil pH and porewater TN, TP and DOC concentrations (Table S2).

Soil and porewater properties of the paddy
Soil Eh, salinity and porewater NH 4 + , TP and DOC concentrations, soil pH, temperature, water content and porewater NO 3 − and TN concentrations varied significantly among sampling dates (p < 0.01; Table S3; Figures 3 and 4). The interaction between sampling date and treatment had significant effects on soil pH, temperature, water content and porewater NO 3 − and TN concentrations (p < 0.05). Mean soil pH, Eh, salinity, water content and temperature for the standard and double fertilizations differed by <10% from the control. Mean porewater NH 4 + concentrations were 114.8% and 213.7% higher for the standard and double fertilizations, respectively, than the control. Mean porewater NO 3 − concentrations were 17.8% and 30.9% higher for the standard and double fertilizations, respectively, than the control. Mean porewater TN concentrations were 19.7% and 42.0% higher for the standard and double fertilizations, respectively, than the control. Mean porewater TP concentrations were 45.7% and 213.3% higher for the standard 404 W E I Q I WA N G et al.   and double fertilizations, respectively, than the control. Finally, mean porewater DOC concentrations were 9.3% and 11.7% higher for the standard and double fertilizations, respectively, than the control.

Relationships among GHG emissions, dissolved GHG concentrations and soil and porewater properties
The mixed linear models (with plot and time as random factors) showed that CH 4 emissions were positively correlated with soil water content (Table S2). N 2 O emission was positively correlated with porewater total TN concentration and negatively with soil pH. Porewater CO 2 concentrations were positively correlated with soil salinity and water content and negatively with porewater TN, TP, NH 4 + and DOC concentrations. Porewater CH 4 concentrations were positively correlated with soil water content, and porewater N 2 O concentrations were positively correlated with soil temperature and porewater NH 4 + concentrations. Soil relationships varied between the treatments (Tables S4 and S5). Seasonal CO 2 emission was positively correlated with soil Eh (p < 0.05; Table 2) and temperature (p < 0.01; Table S4) in all plots. Seasonal CH 4 emission was positively correlated with soil salinity (p < 0.01) and water content (p < 0.05) in all plots. Dissolved CO 2 concentration was positively correlated with soil water content (p < 0.05), while dissolved N 2 O concentration was negatively correlated with soil temperature (p < 0.05; Table S4) in all plots.

Rice productivity and GWP
The average rice yield was about 5.2% higher for the standard fertilization and about 1.4% lower for the double fertilization than the control, which were not significantly different ( Table 2). The contribution of CO 2 to total GWP (> 85%) was higher than that of CH 4 and N 2 O. The total GWPs for all emissions were 14.1% and 10.8% higher for the standard and double fertilizations, respectively, than the control. The total GWPs per unit yield were significantly higher by 7.3% and 10.9% for the standard and double fertilizations, respectively, than the control.

Effects of fertilization on CO 2 emissions
Mean CO 2 emissions were higher for the standard and double fertilizations than the control, for several potential reasons. First, fertilization, such as N fertilization, promotes the deposition of photosynthetically derived C into soil organic carbon (SOC) pools. Then, soil respiration increases when inputs of active C substrates increase (Ge et al., 2015). Second, fertilizer can provide many nutrients for microbial growth (Inselsbacher et al., 2011), and the increase in microbial activity promotes soil respiration and thus CO 2 emission (Adewopo et al., 2015). Third, NH 4 + from fertilizers can be oxidized to NO 3 − when paddies are drained, increasing the soil NO 3 − concentration. This NO 3 − would be reduced when the paddies are reflooded, producing CO 2 (Wang et al., 2015). Moreover, NH 4 + amendment in our study would be associated with ferric reduction, increasing the production and release of CO 2 (Luo et al., 2016). Ferric reduction should also decrease the number of iron plaques (by the higher solubility of Fe 2+ than Fe 3+ ) on the rice roots, which would increase the transport of gases throughout the rice plants (Huang et al., 2012). Transport by rice plants is the most important pathway of gas emission to the atmosphere (Wassmann and Aulak, 2000). Decreases in the number of iron plaques will promote root ventilation, so more CO 2 is produced and transported through the internal system of interconnected gas lacunae in plants. The positive correlation between soil redox reactions and CO 2 emission is consistent with this result.
CO 2 emission varied seasonally, increasing with rice growth and temperature. Temperature controls CO 2 production and emission not only by increasing soil microbial activity (Vogel et al., 2014), but also by altering plant respiration (Slot et al., 2013), substrate availability and quality, species composition, water availability and aerobic/anaerobic conditions (Davidson and Janssens, 2006;Inglett et al., 2012;Sihi et al., 2016a, b). Higher temperatures increase CO 2 soil emissions in subtropical wetlands (Inglett et al., 2012). C quality primarily influences Soil Organic Matter (SOM) decomposition at low temperatures, while at high temperatures nutrient availability controls SOM decomposition in subtropical wetlands (Sihi et al., 2016b). In our study, we have obtained consistent results with these previous reports. Soil CO 2 concentration in porewater increased with temperature, especially in doublefertilized soils (Figures 3 and 4), showing higher organic matter decomposition with temperature, mainly in double-N-fertilized soils. Given that the major fraction of rice croplands in Southeast Asia corresponds to puddled/wetland conditions, understanding complex interactions in such environments is important for improving our capacity for future projections under a warming climate for both natural and agricultural systems.

Effects of fertilization on CH 4 emissions
Mean CH 4 emissions were 20.4% and 39.5% higher for the standard and double fertilizations, respectively, than the control. As stated above, fertilization promotes the deposition of photosynthetically derived C into SOC pools (Ge et al., 2015). Such C can contribute up to 52% of the CH 4 emissions from paddy soils by the exudation of labile organic C from roots to the rhizosphere, which will then produce methane. These results are consistent with the lack of fertilization effects on rice yield. While fertilization can enhance photosynthesis, more photosynthates are allocated to root exudates, and this reduces allocation to growth and yield. The other 48% of the CH 4 is emitted from old soil C (Minoda et al., 1996), promoting CH 4 production and emissions (Minoda et al., 1996). Fertilization, especially N fertilization, will also increase the availability of nutrients, which will promote CH 4 production and emissions from microbes (Naik et al., 2015). Since N fertilizer was provided in NH 4 + form, it could have inhibited CH 4 oxidation because of their structural resemblance and thus enzymatic substrate competition (Gulledge and Schimel, 1998). Nevertheless, most studies testing for different methods and substances for N fertilization have observed enhancement of CH 4 production and emission (Pathak, 2010), which is in agreement with our results.
CH 4 emission varied seasonally and emissions were low soon after rice transplantation when soil was not strictly anaerobic. The correlation of soil redox reactions with CH 4 emission supported this finding (Data not shown). CH 4 emissions were also low during the final ripening and drainage periods. These results agreed with those by Minamikawa et al. (2014), in which a lowering of the water table decreased the abundance of the methanogenic archaeal population and hence CH 4 production and increased the abundance of methanotrophs and thus CH 4 oxidation.

Effects of fertilization on N 2 O emissions
N 2 O emission was low throughout the growing season, with no obvious pattern of seasonal variation (Figure 2). The paddies in our study region are strongly N limited (Wang et al., 2015), so together with low levels of soil O 2 , most of the N 2 O produced is likely reduced to N 2 , and this would lead to very low emissions or even a net uptake of N 2 O (Zhang et al., 2010). Pulses in ammonium and nitrate availability after a fertilization have been related to N 2 O production (Pathak et al., 2002). Specific N 2 O fluxes and the contribution of nitrifying and denitrifying bacteria are controlled mainly by soil moisture (Davidson et al., 1993). However, the results of our study showed that the N added by fertilization had not been sufficient to raise N 2 O emissions in these paddies with low N concentrations.

Best management practices to reduce GWP
Our results suggested that the application of fertilizer had increased the impacts of rice agriculture on climate change, with higher total GWPs per unit yield compared to the controls. The fertilizations did not significantly increase the rice yield but they increased the soil porewater nutrient concentrations, which has the potential risk of nutrient loss, eutrophication and higher costs. Judicious use of fertilization should be reconsidered in a sustainable agriculture, and our results provide strong evidence that the current strategy of fertilization in most rice croplands in subtropical China over several years will saturate the soil fertility, increasing the release of nutrients to continental water and favouring CH 4 and CO 2 production and emission without increasing rice yield. Our findings suggest that alternating years of standard and low fertilizations could decrease water pollution and mitigate GHG emissions without decreasing rice yield, an issue to be further studied.