The field experiment was conducted in the trial bases of Sanhu farm, Jiangling county, Jingzhou city, Hubei province, China (30°12′N, 112°31′E), between March 2017 and May 2019. This area experiences a northern, humid, subtropical monsoon climate with a mean annual temperature and average annual precipitation of 16.0°–16.4° and 900–1,100 mm, respectively. The soil in this region was classified as inceptisol, while the soil properties (0–20 cm depth) were as follows: total carbon 26.44 g/kg, total nitrogen 2.44 g/kg, organic matter 28.59 g/kg, alkali hydrolyzed nitrogen 170.88 mg/kg, total phosphorus 0.38 g/kg, available phosphorus 12.67 mg/kg, total potassium 17.76 g/kg, available potassium 159 mg/kg, and pH 6.92.

Three rice cropping systems, namely, DR, RR, and RW, were subjected with a randomized block design with three replications. The size of each experimental plot was 98 m² (14 m × 7 m), and a ridge with plastic film attached around each plot was established to prevent the flow of water and the fertilizer. The previous cultivation pattern was rice monoculture and residue return fields. Field management measures were according to local agronomic practices. Detailed information of agricultural inputs, including crop species, seed rate, agrochemical rate, diesel, and electricity usage are shown in Table 1; the date of sowing, transplanting, and harvest for the three cropping systems are shown in Table 2; local daily average temperature and rainfall during the test period are shown in Figure 1. Field management practices are provided as follows:

According to PAS 2050:2011 (British Standards Institution [BSI], and Carbon Trust, 2011), we set the system boundary from agricultural inputs production to the crops harvest (from cradle to gate) for CF and NF. CF calculation did not involve changes in soil carbon content based on the PAS 2050 principle. The functional unit of CF, NF, and WF was 1 ton grain yield.

In accordance with the definition of IPCC 2006, the CF refers to all indirect and direct greenhouse gas (GHG) emissions within the whole life cycle process of crop production quantified as CO₂ equivalents (CO₂-eq). The indirect GHG emissions include manufacturing, transportation, storage, and application of agricultural inputs, while the direct emissions refer to CO₂, N₂O, and CH₄ emissions from soils; direct CO₂ emission was not considered in this study because of the higher CO₂ fixation by crops than their emissions (IPCC, 2013). Therefore, the CF was calculated according to the following equations (Pandey and Agrawal, 2014; Wang et al., 2016):

where, AI_i is the amount of agricultural inputs including fertilizers, pesticides, electricity, diesel, seeds, and labor. EF_i is the carbon emission factors of agricultural inputs as shown in Table 3, which is quoted from Ecoinvent version 2.2 (Swiss Centre for Life Cycle Inventories, Switzerland) and Chinese Life Cycle Database (CLCD version 0.8). CE (CH₄) and E(N₂O) are the amount of methane and nitrous oxide measured directly from soils. The global warming potential (GWP) coefficients of N₂O and CH₄ at a 100-year time horizon (IPCC, 2013) are 298 and 25, expressed in kg CO₂-eq kg^–1, respectively. Y is grain yield per unit area, t ha^–1.

A static chamber (100 cm × 50 cm × 50 cm) method and gas chromatography were employed to measure direct GHG emissions (i.e., N₂O and CH₄) from soils. The material of the chamber was Plexiglas, and a tinfoil was wrapped around the chamber to keep the internal temperature from changing drastically. N₂O and CH₄ fluxes were collected between 9:00 and 11:00 one time per week; gas samples were collected at 0, 5, and 10 min after the chamber was closed. Concentrations of N₂O and CH₄ were analyzed by a gas chromatograph (Agilent 7890A), in which, N₂O was detected with an electron capture detector (ECD), and CH₄ was detected with a flame ionization detector (FCD). The N₂O and CH₄ fluxes were calculated using a linear increase of gas concentration over time. The calculation formula is as follows:

where F is N₂O and CH₄ emission rates, the units of N₂O and CH₄ are μg (m² h)^–1 and mg (m² h)^–1, respectively; dc/dt is the rate of change of gas concentration in the chamber with time during the sampling process; the units of N₂O and CH₄ are μl (m³ h)^–1 and ml (m³ h)^–1, respectively; h is the height of the chamber, 1.0 m; ρ is the gas density at standard atmospheric pressure; the unit of N₂O is 1.964 kg m^–3, and the unit of CH₄ is 0.714 kg m^–3; 273 is the absolute temperature (K); T is the temperature inside the chamber during sampling (°).

Nitrogen footprint refers to the environmental impacts of Nr loss on water, air, and soil based on ISO 14044 (ISO, 2006) and CML2002 (Guinée et al., 2002) methodology, which was characterized as eutrophication potential (EP) in this research. Similar to the CF, the calculation of NF includes indirect emissions of agricultural inputs and direct emissions from soils. On-field Nr loss mainly includes NH₃ volatilization, N₂O emission, and NO₃^– and NH₄⁺ leaching, respectively. The NF was calculated as follows (Chen et al., 2020):

where AI_j is the amount of agricultural inputs, EF_j is the nitrogen emission factors of agricultural inputs (Table 3). NV(NH₃), NE(N₂O), NL(NO₃^–), and NL(NH₄⁺) are the amount of NH₃ volatilization, N₂O emissions, and NO₃^– and NH₄⁺ leaching from the field, expressed in kg N-eq kg^–1, respectively.

where N is the nitrogen fertilizer rate (kg ha^–1). α, β, and γ are the coefficients of NH₃ volatilization loss, NO₃^–, and NH₄⁺ leaching, respectively. The α value is 0.142 (Xia et al., 2021). In accordance with the manual for China fertilizer leaching coefficients, the β value is 0.066 for DR and RR, 0.060 for RW; the γ values are 0.339, 0.165, and 0.19% for DR, RR, and RW, respectively. 17/14, 62/14, 18/14, and 44/28 are the molecular weight ratios of NH₃ to NH₃-N, NO₃^– to NO₃^–-N, NH₄⁺ to NH₄⁺-N, and N₂O to N₂O-N, respectively. 0.833, 0.238, 0.786, and 0.476 are the EP coefficients of NH₃, NO₃^–, NH₄⁺, and N₂O at a 100-year time horizon, which are sourced from the CML2002 (Guinée et al., 2002).

In this article, the utilization of blue water and green water resources is based on the measured results of the local 2-year experiment. The pollution caused by N fertilizer application was mainly considered when calculating gray WF, and the critical dilution volume method was adopted to calculate gray water demand. The calculation formulas are as follows:

where, WF_blue, WF_green, and WF_gray refer to blue, green, and gray WF, respectively, expressed in m³ t^–1, respectively. CWU_blue, CWU_green, and CWU_gray are the water consumption of blue water, green water, and gray water during crop growth period, respectively, expressed in m³ ha^–1, respectively. L_p is the amount of pollutants entering water, and 10% of the total nitrogen application amount is usually selected (Hoekstra et al., 2011). C_max is the maximum acceptable pollutant concentration in water, no more than 10 mg of nitrogen per liter of drinking water (EPA, 2005); C_nat is the concentration of the pollutant in the water in the natural state, which is usually 0.

All data were presented as the means ± SE (standard error). Statistical analysis was conducted using SPSS 26.0 (SPSS Inc., IL, Chicago, United States) and Origin Pro 9.0 (OriginLab Corporation, Northampton, MA, United States). Two-way ANOVA was employed to analyze effects of years and cropping systems on the grain yield, CF, NF, and WF, followed by Duncan multiple comparison with a significance level of 5%.

As shown in Figure 2, cultivation area of the traditional DR in central China decreased from 4.72 million ha in 2005 to 3.81 million ha in 2018. On the contrary, the planting area of RW increased by more than 1 million ha, from 2.74 to 3.80 million ha in the same period and area. What is more amazing was that the RR increased nearly four times, from 0.12 to 0.57 million ha.

The 2-year experiment showed that there were some significant differences in grain yield among the three rice cropping systems (Table 4). In 2017–2018, the grain yield in RR and RW was significantly higher than that in DR (P < 0.05); however, no significant difference was detected between RR and RW. When it came to 2018–2019, the grain yield of RR was 16.83 t ha^–1, which was significantly higher than RW (15.55 t ha^–1) and DR (15.06 t ha^–1); no significant difference was found between RW and DR. Additionally, the order of grain yield was RR > RW > DR in both years.

The average contribution of different sources to the carbon emissions is illustrated in Figure 3. It is clear that methane on-field emission was the largest cause of the whole carbon emissions with a percentage of 39.38–66.48%, closely followed by the on-field nitrous oxide, the second important source, accounting for 12.61–33.21%. Indirect GHG emissions from agricultural inputs together accounted for 20.91–34.30%, in which, synthetic fertilizer, diesel for mechanical cultivation, and electricity for irrigation were the most important sources to the carbon emissions (Table 5 and Figure 3). Furthermore, compound fertilizer in RW system was the primary single agricultural input source accounting for 11.82–11.93% of the total GHG emissions.

Table 4 shows the CF performance of the three cropping systems. The total CF were from 483.98 to 1,094.13 kg CO₂-eq t^–1 among the three modes during the 2 years. In 2017–2018, DR system had the highest CF at 1,094.13 kg CO₂-eq t^–1, significantly higher than those of 604.99 kg CO₂-eq t^–1 for RW and 567.22 kg CO₂-eq t^–1 for RR (P < 0.05), but no significant difference between RW and RR. In 2018–2019, DR maintained the highest CF of 718.97 kg CO₂-eq t^–1, 24.27 and 48.55% higher than RW and RR (P < 0.05), respectively. In addition, RW was 19.54% higher than RR. The interaction between cropping systems and year had a highly significant effect on CF (P < 0.01). Specifically, the CF of the three systems in 2 years presented the following order: DR > RW > RR.

Table 5 and Figure 4 show that NH₃ volatilization was the principal source to the Nr emissions with the proportion of 58.47% to DR, 60.76% to RR, and 64.20% to RW, respectively. The second significant contributor to the Nr emissions was NO₃^– leaching, which accounted for 28.27–29.43% among the three rice cropping systems. Additionally, N₂O had a percentage of 5.16–10.22% to the N emissions as the third important source. In contrast, all of the agricultural inputs and NH₄⁺ had little contribution to the total nitrogen emissions, accounting for only 1.93–3.00% taken together.

Ratoon rice system showed the lowest NF of 5.41 and 5.03 kg N-eq t^–1 in 2017–2018 and 2018–2019, respectively (Table 4), significantly lower than those of DR and RW in both years (P < 0.05), while the highest NF value was observed in DR, which were 6.59 and 6.07 kg N-eq t^–1 in the 2 years, respectively. Obviously, the NF value of RW was somewhere between DR and RR. Furthermore, no significant differences existed between RW and DR.

The WF value for DR system was found to be the highest at 1,338.33 and 1,103.79 m³ t^–1 in the two study years, respectively (Table 4), 24.52% higher than RR and 13.06% higher than RW in the first year (P < 0.05), while the percentages were 21.40 and 9.56% in the second year. WF values for the RW system were 10.13 and 10.81% higher than those of the RR within 2 years (P < 0.05).

Figure 5 shows that green, blue, and gray water contributed 35.64–38.37, 40.90–42.71, and 20.74–21.66%, respectively, to the whole WF for DR, while the proportions of the three categories of water for the RR mode were respectively 38.55–41.47, 36.73–38.57, and 21.79–22.88%. In addition, green water (48.85–55.72%) was the largest contributor to WF for RW. Noticeably, the blue WF among the three rice cropping systems demonstrated the order of DR > RR > RW in both years, which showed the difference in irrigation water use. Additionally, the RR mode had the smallest WF value, significantly lower than the other two modes.

In this study, the CF values among the three rice cropping systems over the 2 years were 483.98–1,094.13 kg CO₂-eq t^–1 (Table 4), which were consistent with many previous studies in central or southern China (Liu et al., 2016; Xue et al., 2016; Chen et al., 2020; Xu et al., 2020a). However, Jiang et al. (2020) reported a slightly higher CF value of DR pattern since they adopted higher direct and indirect emission factors. Liu et al. (2020d) reported a higher CF in southern China mainly because of more intensive machinery operation and extensive use of plastic films. A significantly higher CF in a DR system reported by Sun et al. (2019) possibly attributed to the adoption of higher emission factors of agricultural inputs. Results from this study are higher than the research by Cheng et al. (2015), since their calculation of CH₄ and N₂O were based on lower statistical data and emission coefficients.

In this research, direct GHG emissions, especially methane on-field emission, constituted the largest fraction to CF in all of the three rice cropping systems (Figure 3), which was consistent with many previous studies (Xue et al., 2016; Jiang et al., 2020; Xu et al., 2022). Therefore, reduction of field methane emissions would be an effective strategy for reducing CF; there are a number of appropriate farming practices that can serve this purpose. For example, Fertitta-Roberts et al. (2019) reported that noncontinuous flooding can reduce methane emissions by 44%; Liu et al. (2020d) found that N fertilizer deep placement can reduce methane on-field emission by 36–39%; Jiang et al. (2019) reported that optimizing the amount of N fertilizer can reduce GHG emissions. Additionally, synthetic fertilizer was the principal component of GHG emission from agricultural inputs (Table 5 and Figure 3), similar with the results reported by Chen et al. (2020) and Xue et al. (2016); in addition, indirect GHG emission from electricity accounts for up to 35.37% in DR system mainly because DR system needs a lot of electric power for pumping irrigation. Yu et al. (2021) suggested that GWP of the second season was significantly lower than the first season in RR, which is similar to our result. Furthermore, Figure 3 shows that CH₄ emissions of the three cropping systems in 2017–2018 were all higher than that of 2018–2019, when it came to N₂O; the results were opposite. This phenomenon could be explained by the following reasons. Methane is only produced rely on the decomposition of soil microorganisms under anaerobic conditions (Fagodiya et al., 2017), while nitrous oxide is primarily produced through nitrification and denitrification under aerobic conditions (Bouwman, 1998). However, rainfall during the rice growing season in 2018 was 7 days less than usual, and the temperature was 1.4° higher on average (Hubei Meteorological Service [HMS], 2018). Under the condition of sufficient oxygen, high temperature, and dry soils in 2018, N₂O emissions would be enhanced while the production of CH₄ was severely inhibited.

In this study, cropping systems have significant effects on yield and CF (P < 0.01) and the general trend of CF in the two study years being DR > RW > RR (Table 4). Similarly, Chen et al. (2020) showed that the CF for DR was 12.17% higher than RW, mainly attributed to higher CH₄ field emission of DR. Xu et al. (2022) indicated that the RR system reduced average annual CF by 27.37% compared with the DR system, which may be due to the annual CF of the second season in RR was significantly lower than that of DR. Therefore, it demonstrated the possibility of rational selection of cultivation mode to balance yield and footprint impact, which is consistent with many previous studies (Ali et al., 2012; Cha-un et al., 2017; Sun et al., 2019).

In this study, the NF value for the three systems was 5.03–6.59 kg N-eq t^–1 across the two test years (Table 4), which were higher than the NF of a rice monoculture system in central China (Xu et al., 2020a) mainly because of the lower application of the N fertilizer rate. Chen et al. (2020) and Xue et al. (2016) reported higher average NF for DR and RW in southern China since they employed higher NH₃ and N₂O emission factors; Pierer et al. (2014) reported a significantly higher NF of cereals in Austria, possibly because they calculated the NF of not only the production process but also the consumption process. In addition, NH₃ volatilization was the primary source of active N emission among the compared cropping patterns (Table 5 and Figure 4), which agrees with previous studies (Leip et al., 2014; Chen et al., 2020; Xu et al., 2020b). Therefore, the selection of NH₃ volatilization coefficient was very important for the calculation of NF; thus, we chose a NH₃ volatilization coefficient obtained from the long-term location tests in the study site, which was close to other test results in central China (Qi et al., 2012; Li et al., 2018). Moreover, Chen et al. (2020) and Xu et al. (2020b) suggested NO₃^– was the second major source to the NF, which was consistent with our result. A large number of previous studies showed that Nr was closely related to the N fertilizer applications (Qi et al., 2012; Pierer et al., 2014; Xu et al., 2020b). So proper nitrogen fertilizer management is crucial to the mitigation of Nr without compromising the grain yield, such as reasonably adjust the amount of nitrogen application (Qi et al., 2012) or selecting appropriate fertilizer categories (Lian et al., 2018). It is difficult to compare the NF value with others owing to the limited research on NF of the three rice cropping systems. Similar to the CF, cropping systems have a significant effect on NF (P < 0.01), and the NF value follows the tendency of DR > RW > RR (Table 4). On the contrary, Chen et al. (2020) found that NF of DR was lower than RW by 13.43%, primarily because RW had a lower grain yield.

The WF of the three rice cropping systems in this study ranged from 909.21 to 1,338.33 m³ t^–1 (Table 4), which were similar to some studies (Xu et al., 2020b; Zhai et al., 2021). Chapagain and Hoekstra (2011) also reported the green, blue, gray, and total WF of rice production in China, respectively, were 367, 487, 117, and 971 m³ t^–1, while the values of global average WF of rice planting were 346, 374, 65, and 784 m³ t^–1. The difference mainly came from gray WF due to the higher N fertilizer rates. Kashyap and Agarwal (2021) indicated higher WF value of rice and wheat in India, primarily due to the higher blue WF; local farmers used a lot of irrigation water under drought conditions. In addition, the average blue WF followed the order: DR > RR > RW (Figure 5) possibly because DR still needs a lot of irrigation water in the second season, which is significantly higher than that of RR, while RW only needs natural precipitation instead of irrigation water in the second season, which is also the reason why RW has the highest proportion of green WF. Moreover, due to the different nitrogen application rates and yield factors, both RW and DR had significantly higher gray WF than RR (Figure 5). Therefore, the key to reducing WF lies in water-saving irrigation and nitrogen reduction (Zhang et al., 2014; Livsey et al., 2019).

In this study, year and cropping system had profound impacts on grain yield and WF (Table 4). In fact, the total amount of virtual water in 2017–2018 was only 4.43–11.17% higher than that in 2018–2019, but the WF of the former was 17.50–21.25% higher than that of the latter, which was mainly because grain yield of all treatments in 2018–2019 were higher than that in 2017–2018 due to a better meteorological conditions (Figure 1).

Grain yield and environmental impacts need to be considered during rice production decision-making (Zhu et al., 2016). Compared with DR and RW systems, RR was a low-cost and high-output cultivation mode not only because of the low labor intensity but also because of the feature of harvesting twice in a single year while only sowing once (Tables 2, 3). Our results also illustrated that the RR system had a lower environmental impact due to its performances on CF, NF, and WF, but with the highest grain yield (Table 4), which was consistent with a research (Xu et al., 2022). The RR system has become a significant cropping system promoted over a wide region of southern China since the 1980s. Meanwhile, there are 3.3 million ha rice fields suitable for growing ratoon rice accounting for 30.8% of the total paddy fields in southern China (Xu et al., 2022).

Furthermore, it is important to take into account the current social environment and rural situation when appropriate agricultural policies are established (Pandey and Agrawal, 2014; Zhou et al., 2022). First, the population issue cannot be ignored because the rural population accounts for only 36.11% of the total population with a sharp decrease by 164.36 million during 2010–2020. Second, the population older than 60 years accounted for 18.70% compared with 13.26% in 2010; therefore, the problem of an aging population will deepen (NBS, 2022). Finally, the interaction of rural land transfer and urbanization also has an impact on the rural population (National Bureau of Statistics [NBS], 2021). All of these factors have resulted in a large reduction in the rural laborers; therefore, the low labor intensity rice planting mode (i.e., RR) has become an important choice.