Trade-offs and spatial dependency of rice production and environmental consequences at community level in Southeastern China

Zhejiang Province in the Yangtze River Delta is one of the most densely populated and rice-dominant agricultural regions in China (figure 1(a)). With limited arable land and increased competition for land from urban development, maintaining agricultural food production and ensuring environmental quality is an increasing challenge. Nonpoint source pollution from agricultural fields is blamed for frequent algal bloom and water pollutions in the region but solutions are lacking and policies are ineffective.

Two case study areas—Anji County (30.6391°N, 119.4842°E) and Dasong River watershed (30.6391°N, 119.4842°E) in the province (figures 1(b) and (c)) were selected, based on two previous studies (Xu 2011, Wang et al 2017). Anji and Dasong cover 1886 km² and 108.76 km², respectively. Anji County is within the Tiaoxi River watershed that drains into Taihu Lake, one of the most algal bloom prone inland lakes in China. It was estimated that about 30%–40% of riverine N sources were discharged from paddy fields (Tian et al 2012, Ye et al 2017), Similarly, the Dasong River watershed (108.76 km²) was blamed for frequent eutrophication of the Xiangshan Bay of China's east coast. Land use in this region is diverse, but rice farming may have contributed more than 25% of N discharged to rivers in the area (Nobre et al 2010).

2.2.1. Anji county data

2.2.2. Dasong river watershed data

The process-based DeNitrification-DeComposition (DNDC) model (Li et al 1992, 1994) was initially developed to simulate the dynamic interactions among different C and N pools in the soil and plant (Li et al 2001, Zhang et al 2002) and subsequently has been calibrated and validated over the past twenty years including those by the model developer (e.g. Li et al 2001, Zhang et al 2010, Deng et al 2011). The model has four primary components: climate, soil, vegetation, and management practices to govern the biogeochemical processes that determine crop yield, GHG emissions, and nitrogen leaching. With different management options, soil condition and plant information as inputs, the DNDC model was used to simulate environmental consequences and rice yield potentials.

The DNDC model was first calibrated by comparing simulated yields to reported yields from the case study areas and comparing simulated fluxes against published values (figure 3). Once calibrated, the model was used to simulate paddy rice yield, GHG emissions and N leaching for the subsequent trade-offs analyses among different management options. The model was run based on single rice rotation without considering wheat or rapeseed as the second cropping. Information on direct-seeded versus seeding-transplanting methods were not explicitly simulated.

2.4.1. Village level simulation

The calibrated DNDC model was run to simulate nitrogen loss ratio (NLR) pattern at the village level for Anji County, where NLR was defined as the ratio of nitrogen leached, including N runoff, to total N input:

$$\begin{equation} {\rm{NLR}} = \,{\rm{N}}\, {\rm{leached}}/{\rm{Total}}\, {\rm{N}}\, {\rm{fertilizer}}\, {\rm{input}} \end{equation}$$

Two simulations were made to discern the effects of climate variation and farm management practices:

• 1.  
  Historical climate scenario: simulation with historical (1991 to 2009) climate data and surveyed farm management information for five times from 1991 to 2009. Given that continuous management data was not available, the simulation period was divided into five groups, each representing the closest date when management data were surveyed (see table 5).
• 2.  
  2005 climate scenario: simulation with the historical management data, but climate data was fixed to 2005 condition.

Zoom In Zoom Out Reset image size

Table 5. Experimental design for paddy rice farming simulations in Anji County.

Group ID	Year of survey data	Group of simulation years	Climate for historical and scenarios	Climate for 2005 scenario
1	1991	1991 to 1993	1991 to 1993
2	1995	1994 to 1997	1994 to 1997
3	2000	1998 to 2002	1998 to 2002	2005
4	2005	2003 to 2007	2003 to 2007
5	2009	2008 to 2009	2008 to 2009

2.4.2. Field level simulation

The differences between simulated and observed yields were less than 5% at the county level for Anji and less than 2% at the field level for Dasong river watershed (figure 3(a)). GHG emission and N leaching data published in the literature were compiled and subsequently used to validate the reliability of the DNDC model (figures 3(b)–(d)). At the county level, simulated mean CH₄ and N₂O emission, and N leaching ranged from 175 to 300 kg CH₄-C ha yr⁻¹, 2 to 5.4 kg N₂O-N ha yr⁻¹, and 32–54 kg N ha yr⁻¹. For reference purpose, the observed values in northern Zhejiang and Taihu Lake areas are 14–571 CH₄-C ha yr⁻¹, 0.1–15.3 N₂O-N ha yr⁻¹ and 8.5–61.1 kg N ha yr⁻¹, respectively (figures 3(b)–(d)), suggesting that the simulated values in this study are well within the range of observed values in the region. At the field level, simulated CH₄ ranged from 7–320 kg CH₄-C ha yr⁻¹ and N₂O from 0.4 to 1.4 kg N₂O-N ha yr⁻¹ in 2014 in Dasong River watershed. The results indicate that model simulated fluxes were similar to observed values in the region (figures 3(b)–(c)), suggesting the validity of the model for this region.

Zoom In Zoom Out Reset image size

Fertilizer and yields trade-offs: The total N input for surveyed villages in Anji County remained around 250 kg N ha⁻¹, but reduced to less 200 kg N ha⁻¹ after 2000. However, chemical N fertilizer input (urea plus compound fertilizer) increased quickly since 1995 (figure 4(a)). While manure was the main source for N input in 1990s, chemical N fertilizer has replaced manure as the main fertilizer input since the early 2000s (figure 4(a)) as the number of villages using manure as fertilizers reduced from thirteen (13) in 1990 to four (4) in 2009. This is because chemical fertilizer, when compared to manure, is readily available for applications and far more convenient to handle, thus less labor intensive. Labor saving has been overlooked in the past as a driver for increased chemical fertilizer uses; however, as China's economy continues to grow, migration of rural young population to urban cities, leaving seniors behind in the farm fields (Chang et al 2011). The choice between manure and chemical fertilizer is thus an easy one, when labor is considered.

From the economic perspective, fertilizer prices have been kept quite low for the past two decades (1990s–2010s) due to regulation and state subsidies to the fertilizer production industry (Li et al 2014). Subsidized fertilizer price, in turn, incentivized the widespread of excessive N fertilizer uses across China (Li et al 2013). Replacing manure with chemical N fertilizer increased rice yields (figure 4(b)), but also reduced the amount of organic matter returned to soils (figure 4(c)), which led to a decrease in soil organic carbon (SOC) after 2000 (figure 4(d)).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Environmental trade-offs: Temporal changes in temperature and precipitation (figure 2) seem to have stronger impacts on methane emission and nitrogen loss from paddy rice fields. The average CH₄ emissions increased significantly from 1991 (~200 Kg C ha yr⁻¹) to 2009 (~300 kg C ha yr⁻¹) (figure 5(a)). One might speculate that increased N fertilizer applications promoted CH₄ emission. However, previous studies found N application may promote or inhibit CH₄ emission, depending on local soil texture and/or climate conditions (Cheng-Fang et al 2012, Xie et al 2010). In fact, simulation under the 2005 climate scenario (figure 5(b)) suggests that CH₄ emission would only increase marginally (11%) if climate variation were not considered. In Anji county, the elevated temperature during the 1991–2009 period (figure 2(a)) may explain the increase in CH₄ emission because higher soil temperature could potentially enhance methanogenic activity and therefore higher bio-methane production (Allen et al 2003, Sun et al 2016, Tian et al 2015). Recently, a field experiment in the region also confirms the relationship (Sun et al 2016).

Comparison of simulated N loss pattern under historical and 2005 climate scenario indicates that changes in N leaching and N₂O emission over past two decades were more strongly influenced by precipitation variation than the N rates. For instance, the amount of average N leached from paddy fields decreased from 52.5 to 35.2 kg N/ha yr⁻¹, or a 33% reduction, between 1991 and 2009 (figure 6(a)). However, when precipitation pattern holds constant at 2005 level, N leaching increased by 37.8% (figure 6(b)). This was likely caused by increased N fertilizer applications and reduced manure input because N fertilizers are much easier to be leached (Fan et al 2017). Similarly, N₂O emission fluctuates between 1991 and 2009 without a clear increasing or decreasing trend (figure 6(c)). However, a clear decreasing trend in N₂O emission can be observed when 2005 precipitation is used throughout the simulation period (figure 6(d)). The reduction in N₂O emission under 2005 climate condition is primarily caused by the elimination of manure fertilizer: several field studies found that application of manure could result in higher N₂O emission compared to chemical fertilizer (Akiyama and Tsuruta 2003, Zou et al 2006, Jin et al 2010). The contrasting results under historical and 2005 climate data scenarios clearly demonstrated the importance of taking climate variation into GHG mitigation plans.

Zoom In Zoom Out Reset image size

Fertilizer use efficiency: Between 1991 and 2009, the fertilizer use efficiency indicator, NLR values, have decreased for most villages, but substantial spatial variation existed within the county. Village level NLR between 1991 and 2009 ranged from 0.01 to 0.48 (mean = 0.19, SD = 0.15) and 0.03 to 0.37 (mean = 0.15, SD = 0.11) respectively. Improvement in NLR is most evident between 1991 and 2000, with mean NLR values decreased from 0.19 to 0.12, which is consistent with the decreasing trend in N leaching. After that, NLR rose again by 0.01 to 0.2 (mean = 0.06) for many villages, although still lower than those in the 1990s. The rebound of NLR signaled that further increasing N rates without better nutrient management would lead to lower N use efficiency. In fact, there is about a 20% gap between crop N demand and crop N uptake in 2009 (results not presented), even though N fertilizer input has increased more than 50% since 2005. This suggests there is plenty of room to improve fertilizer use efficiency and farm management practices.

Simulated results suggest that current rice management practices are far from being optimal and there is plenty of room to reduce GHG emissions without jeopardizing rice yields. However, the most effective management options for N₂O and CH₄ are quite different, depending on field soil conditions. There is a large spatial variation in CH₄ and N₂O fluxes (figure 7) among paddy fields, suggesting that both management practices and soil conditions are affecting CH₄ and N₂O emissions.

The N₂O emissions were particularly sensitive to the soil pH and SOC conditions. Soils with a higher SOC content tend to release more N₂O (figure 7) because soil organic matter provides source materials to soil microbial activities. Soils with a pH value between 6–8 tend to release more N₂O because denitrification bacteria are more active within this range (Saleh-Lakha et al 2009). Low pH (4.65) at site #1 and low SOC (11.0 g kg⁻¹) at site #3 explain why N₂O emissions are lowest among all sites (figure 7), even though N application rates in these two fields are relatively high.

Simulations with varying N input confirmed that N was overused in these fields. Annual N fertilizer input (urea plus compound fertilizer) for the 12 fields in Dasong river watershed averaged at 212 ± 121.5 kg N/ha. This input level generally exceeds crop N demand in this watershed. There seems to be a tipping point in nitrogen fertilizer applications at 37.5 kg N ha⁻¹, at which rate both yields can be maintained while N₂O emission was kept minimal (figure 8).

Field flooding is another important factor in GHG emissions. Simulated results showed that CH₄ emission is positively correlated to flooding period (figure 9) and flood depth (results not presented) because the anaerobic environment under flooding condition is more suitable for the methanogenic activity. Given that extended flooding duration does not help with increasing rice yields, it is reasonable to keep flooding days to about 70–80 d and low drainage rice strains so that the CH₄ emission from paddy fields can be substantially reduced (figure 9). In addition, results suggest that flooding depth would greatly affect CH₄ emissions. Among the 12 paddy fields, half of them presented a CH₄ rate lower than 15 kg ha yr⁻¹. This is largely because flooding depths in these fields (2–4 cm) are much lower than in other fields (8–15 cm) (table 4). Previous studies also found water-saving irrigations could reduce CH₄ emission by more than 80%, when compared with traditional irrigation practices (Xu et al 2015, Hou et al 2016). Still, field based measurements is needed for future validation to understand the unusual low CH₄ rate (7 kg C ha yr⁻¹).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Survey results indicate that farmers in Anji County tend to increase paddy rice yield (thus thus income) by growing a second crop after rice harvest, while those in Dasong River watershed simply applied more fertilizers to maintain yields. If the increasing trend of N fertilizer input continues, most farmers in the region would apply more fertilizers than needed, more than 200 kg N ha yr⁻¹. There have been some studies to suggest policy interventions, including eliminating fertilizer subsidies (Kahrl et al 2010) and imposing a fertilizer tax (Liu and Luan 2012). However, these suggestions are unlikely to be politically acceptable because increasing farmer's income (yields) is also a priority for the central government. 'Optimal N rate' is another popular policy suggestion, but the two case studies clearly demonstrated that substantial spatial-temporal variation exist at both the village and field levels, suggesting 'one-size fits all' policies will not achieve desired GHG emission reduction goal without sacrificing farmer's benefits.

At the field level, more 'smart' or adaptive fertilizer management strategies, such as the '4R' technique (Bruulsema et al 2009), which refers to applying right formulation at the right time and right rates in the right locations, are needed to synergize rice production and environment conservation. However, application of '4R' technique for small farms can be more challenging than adopting 'optimal N rate', as it requires more efforts for training and capacity building. Currently, farmers typically over-apply fertilizers as an 'insurance' or receive blanket fertilizer recommendations from local extension services, but studies in Zhejiang Province suggested that farmers are quite interested in techniques like '4R' because they enable increase in yields at lower fertilizer inputs (Wang et al 2003, Fang et al 2005). While profits are likely the main motivation for farmers to adopt new technologies, saving the environment has started to motivate farmers to take actions to reduce N inputs. Widespread of these technologies, however, would be constrained by accessibility of technical support. One possible solution is to provide real-time simulation results to farmers via smartphones or computers to allow direct access to the information needed for optimal N input. This is possible by coupling detailed database (e.g. local soil) with online information delivery, but more government support would be needed to build the capacity.