Effects of Different Types of Water and Nitrogen Fertilizer Management on Greenhouse Gas Emissions, Yield, and Water Consumption of Paddy Fields in Cold Region of China

Water management and nitrogen (N) fertilizers are the two main driving factors of greenhouse gas emissions. In this paper, two irrigation modes, controlled irrigation (CI) and flood irrigation (FI), and four nitrogen fertilizer levels (N0: 0, N1: 85, N2: 110, and N3: 135 kg·hm−2) were set to study the effect of different irrigation modes and N fertilizer amount on greenhouse-gas emissions of paddy fields in cold region by using the static chamber-gas chromatograph method; yield and water consumption were also analyzed. The results showed that, compared with FI, CI significantly reduced CH4 emissions by 19.42~46.94%, but increased N2O emissions by 5.66~11.85%. Under the two irrigation modes, N fertilizers could significantly increase N2O emissions, but the CH4 emissions of each N treatment showed few differences. Compared with FI, appropriate N application under CI could significantly increase grain number per spike, seed-setting rate, and 1000-grain weight, thus increasing yield. Under the two irrigation modes, water consumption increased with the increase of N application rate, and the total water consumption of CI was significantly lower than that of FI. The global warming potential (GWP) of CI was significantly smaller than that of FI. The trend of GWP in each treatment was similar to that of CH4. Through comprehensive comparison and analysis of water productivity (WP), gas emission intensity (GHGI), and the yield of each treatment, we found that CI+N2 treatment had the highest WP (2.05 kg·m−3) and lowest GHGI (0.37 kg CO2-eq·kg−1), while maintaining high yield (10,224.4 kg·hm−2). The results of this study provide an important basis for guiding high yield, water-savings, and emission reduction of paddy fields in cold regions.


Introduction
Methane (CH 4 ) and nitrous oxide (N 2 O), as the major greenhouse gases emitted from farmland soils, have an important impact on global climate change. The fifth assessment report of the Intergovernmental Panel on Climate Change (IPCC) pointed out that the global warming potential (GWP) of CH 4 and N 2 O was 21 times and 310 times that of CO 2 , respectively, on the 100-year scale [1]. In 2010, agricultural non-CO 2 greenhouse gas emissions were about 5.2-5.8 Gt CO 2 -eq. Rice fields were the main source of agricultural CH 4 , and N 2 O was also emitted from paddy fields. Even though N 2 O emissions were

Experimental Design
The experiment was conducted at the National Rice Irrigation Experiment Center (127 • 40 E, 46 • 57 N) in Heilongjiang Province in 2017, and rice was planted in the experimental area for more than 20 years. Average annual temperature is 2~3 • C, while accumulated temperature is 2300~2500 • C. Average annual rainfall is 500~600 mm. The frost-free period is 128 days. The experimental area has a temperate continental monsoon climate. The soil is albic rice soil, the soil-bulk density is 1.01 g/cm 3 , and soil porosity is 61.8%. Basic physicochemical properties of soil are as follows: organic matter content, 41.8 g/kg; pH, 6.4; total N, 15.06 g/kg; total P, 15.23 g/kg; total K, 20.11 g/kg; alkali hydrolysable N, 198.29 mg/kg; available P, 36.22 mg/kg; and available K, 112.06 mg/kg. The meteorological data of the rice-growth period were recorded by automatic weather station DZZ2X (Tianjin Meteorological Instrument Factory), and are shown in Figure 1. CI and FI were used in the experiment, and the corresponding water management is shown in Table 1. CI had no water layer for the rest of the growth period except for the regreening period and each fertilizer application. The surface-water layer of the field was irrigated to a depth of 1 cm before each fertilizer application. Four kinds of N application levels were selected: no N treatment (N0: 0 kg·hm −2 ), low N application level (N1: 85 kg·hm −2 ), normal N application level (N2: 105 kg·hm −2 ), and high N level (N3: 135 kg·hm −2 ). A total of eight treatments were performed, and each treatment was repeated three times. The length and width of each plot were 10 m, respectively, and the area was 100 m 2 . Each plot had a water meter, and the irrigation and drainage of each plot were separated. The cement ridge and plastic clapboard were used for seepage between the plots to prevent moisture and fertilizer exchange.
The fertilizers tested were urea (N, 46%), superphosphate (P 2 O 5 , 12%), and potassium chloride (K 2 O, 60%). For the N fertilizer, the ratio of basal fertilizer:tillering fertilizer:spikelet-promoting fertilizer:spikelet-preserving fertilizer was 4.5:2:1.5:2. P 2 O 5 45 kg·hm −2 and K 2 O 80 kg·hm −2 were used for each treatment, K 2 O fertilizer was applied twice as basal fertilizer and 8.5 leaf age, respectively; the ratio before and after was 1:1. P 2 O 5 fertilizer was applied to the basal fertilizer once. The tested rice was Longqing 3. According to the technical requirement of rice seedlings in the cold region, sowing began when air temperature was stable at 5~6 • C, and the seedbed-soil temperature was above 12 • C. Pregerminated seeds were cultivated into the seedlings in a soil-filled seedbed. Sowing started on 20 April and seedlings were transplanted on 17 May. Planting density was 30 × 10 cm. Technical conditions such as seedling raising, transplanting, density, fertilization, and pesticide use were the same for each plot. Rice was harvested on 20 September. Note: Data before "~" are the lower limits of water, data after "~" are the upper limits of water. CI, controlled irrigation. FI, flood irrigation, θs, saturated moisture content of root layer soil, same as below.

Gas Sampling and Analysis
Field sampling of CH 4 and N 2 O was carried out from the beginning of seedling transplanting to harvest for about once a week. On cloudy or rainy days, sampling was postponed until the next sunny day. The number of sampling times was increased during the rapid rice-growth period. There were a total of 17 sampling events in this experiment. The static-chamber method was used for sampling. The length and width of the chamber and the stainless-steel base were 25 × 25 cm. The chamber was made of 5-mm-thick plexiglass, which was covered with an insulation board and tin foil to prevent solar radiation from causing temperature changes inside the chamber. A chamber with 60-cm height was used in the early growth stage of rice, and a chamber with 110-cm height was used in the late growth stage. A 12-V fan was installed inside each chamber to ensure the gas inside was evenly mixed. An electronic thermometer probe (0.1 • C) was equipped inside the chamber to correct emission errors caused by the rise of temperature. The sampling port was set on the side of the chamber, and the sampling tube was inserted into the chamber for 20 cm. The gas was sampled by a 50-mL syringe, and was then injected into the gas-collection bag (E-Switch). The upper part of the base had a 5-cm-wide and 5-cm-high sink, and the sink was filled with water before gas sampling to ensure that there was no gas exchange between the chamber and the external environment.
Samplings time were from at 10:00 to 12:00 h, with sampling once at 0, 10, 20, and 30 min, respectively, during the time of chamber closure. Before each sampling, 30 mL gas was first extracted from the sampling tube with a syringe, and then injected back into the chamber to reduce the error caused by the nonuniform mixture of gas in the sampling tube. After sampling, the air bag was taken directly to the laboratory for analysis. Gas concentration was manually analyzed by gas chromatograph (SHIMADZU GC-2010plus, Japan).
The emission flux of CH 4 and N 2 O was calculated with the following formula [7]: where F is the CH 4 flux (mg·m −2 ·h −1 ) or N 2 O flux (ug·m −2 ·h −1 ); dC dt is the slope of curve of gas concentration versus time; h is the effective height of the chamber (m); ρ is gas density at the standard state (kg·m −3 ); and T is the average temperature inside the chamber ( • C).
GWP and GHGI were used to assess the greenhouse-gas effects. Taking 100a as the time scale, the GWP of CH 4 and N 2 O gas per unit mass was 21 and 310 times that of CO 2 , respectively (IPCC, 2014); its unit is kg CO 2 -eq·hm −2 . GHGI represents the GWP per unit rice yield; its unit is kg CO 2 -eq·kg −1 .

Yield and Components
In the maturity period, a 1-m 2 block of well-grown and evenly grown rice plants from each plot was collected for yield measurement. Rice yield per unit area was calculated according to 14.5% moisture content. Spikes per unit area, grain number per spike, seed-setting rate, and 1000-grain weight were also counted at the same time.

Water Consumption and Water Productivity
Soil-moisture content was measured with a TPME-PICO64/32 soil-moisture analyzer when there was no water layer in the field. When there was a water layer, the water depth of three points in each plot was recorded at 08:00 h every day, and the average value was then calculated. When the height of the water layer was greater than the upper limit of water, drainage would be carried out and the depths of the water before and after drainage were recorded. Water consumption was calculated by the water-balance equation: where ET is water consumption (mm); P is precipitation (mm); I is the irrigation water amount (mm); K is groundwater recharge (mm); W1 and W2 are soil water storage before and after the growth period (mm), respectively; R is the drainage amount (mm); and D is the percolation amount (mm). Because groundwater in the experiment area is relatively deep, groundwater recharge does not exist, so K = 0. The soil percolation amount in the paddy fields was taken as an average value of experiment area based on the thesis of Guo [22].
The WP was obtained from rice-grain yield (Y) per unit of water consumption. The calculation formula was as follows: where WP is water productivity (kg·m −3 ).

Data Analysis
To test the differences between treatments, the data were analyzed using analysis of variance with SPSS 17.0 (IBM Corp., New York, NY, USA). The treatment means were compared with a least-significant-difference (LSD) test.

CH 4 Emissions
As shown in Figure 2, there were three emission peaks during the growth period: in the later tillering drainage, jointing and booting stage, and milk maturity stage. The average flux of each peak decreased as days after transplanting (DAT) increased. Under the same N application condition, the CH 4 emission flux of each treatment under CI was lower than that under FI except for the late growth stage of rice, which showed that CI had an overall inhibition effect on the CH 4 emission flux. The peak value of FI+N3 emission flux was 41.45 mg·m −2 ·h −1 , which was 135.24% higher than the peak of CI. The second peaks of FI and CI occurred at 51 and 57 DAT, respectively. This may be due to better soil aeration at the tillering stage under CI inhibiting the amount of methanogen [19], and drainage practice in later tillering further decreasing the amount of methanogen. After rewatering at the beginning of the jointing and booting stage, the amount of methanogen was difficult to rapidly increase in a short time. This resulted in the second peak of CI treatments later than FI for six days, and the emission peak value was less than FI. The third emission peak of CH 4 emission was at 94 DAT. In this peak, the CH 4 flux of CI was slightly higher than that of FI. This may be because in the late growth of rice, the root system of control irrigation was more developed, root exudates provided more substrates for the production of CH 4 , and the developed roots were beneficial for the release of CH 4 through the plant pathway.
CI+N2 and FI+N2 were selected to analyze the relationships between CH 4 flux and field-soil water conditions ( Figure 3). There was a significant positive correlation between CH 4 flux and soil water content (p < 0.05) in CI+N2 except in the regreening stage, while there was a significant positive correlation between methane-emission flux and water depth during the whole growth period of rice under FI (p < 0.05). Under both irrigation modes, the increase of water content and water depth promoted the increase of CH 4 emission flux.  By comparing the CH 4 emissions of each treatment in Table 2, under the same N application, CH 4 emissions of CI+N3, CI+N2, CI+N1, and CI+N0 decreased by 37.98%, 46.94%, 19.42%, and 28.23%, compared with FI+N3, FI+N2, FI+N1, and FI+N0, respectively, indicating that CI under the same N level could significantly reduce CH 4 emissions of paddy fields. For CH 4 emissions of CI treatments, CI+N3, CI+N2, and CI+N1 were significantly higher than those of CI+N0, but differences between CI+N3, CI+N2, and CI+N1 were not significant. CH 4 emissions of CI+N1 were the largest (189.79 kg·hm −2 ), which were 4.95%, 8.56%, and 20.30% higher than CI+N3, CI+N2, and CI+N0, respectively. For CH 4 emissions of FI treatments, CH 4 emissions first increased and then decreased with the increase of N application. CH 4 emissions of FI+N2 were the largest (329.48 kg·hm −2 ), which were 13.01%, 39.88%, and 49.89% higher than those of FI+N3, FI+N1, and FI+N0, respectively. CH 4 emissions of FI+N3 and FI+N2 were significantly higher than those of FI+N1 and FI+N0. However, the difference between FI+N3 and FI+N2 was not significant. Results showed that N application promoted CH 4 emissions under FI, and there existed a CH 4 emission threshold with the increase of N application. In summary, under two irrigation modes, N application promoted CH 4 emissions in paddy fields, and CI could significantly reduce CH 4 emissions. Table 2. Greenhouse-gas intensity and water productivity of each treatment. Note: Lowercase letters represent the differences between each treatment (p < 0.05), N0, N1, N2 and N3 represent 0 kg·hm −2 , 85 kg·hm −2 , 110 kg·hm −2 and 135 kg·hm −2 ,respectively, same as below. GWP: global warming potential; GHGI: gas emission intensity; WP: water productivity.

N 2 O Emissions
From Figure 4, we can see there were two emission peaks during the rice-growth period, which were in the later tillering drainage and the heading-flowering stage, respectively. In other periods, the N 2 O emission flux fluctuated at a smaller value. At each N 2 O flux peak, the N 2 O flux of CI was generally higher than that of FI. The first smaller N 2 O emission peak occurred on 45 DAT, which was possibly because of the reduction of soil-moisture content during late tillering drainage, which increased the available oxygen in the soil and finally promoted the emission of N 2 O. In addition to CIN0 and FIN0, the second emission peaks were found at 71 and 79 DAT, respectively. Fluctuation ranges were 6.63~101.68 ug·m −2 ·h −1 and 10.11~81.07 ug·m −2 ·h −1 , respectively. FI peak lagged behind that of CI; this may because the thicker water layer of FI was not conducive to N 2 O production and it blocked the diffusion of N 2 O from the soil to the atmosphere. The peak of N 2 O emissions did not appear after the application of tillering fertilizer and spikelet-promoting fertilizer, but there was an emission peak after applying spikelet-preserving fertilizer. This may be due to the low soil temperature at the tillering stage, and the low nitrification and denitrification reaction in the soil leading to a smaller N 2 O flux. Continuous cloudy and rainy days occurred after spikelet-promoting fertilizer application ( Figure 1). As a result, the effect of N application on N 2 O emissions was not captured. After applying the spikelet-developing fertilizer, the weather was sunny, effective heat radiation received by the field was larger, and suitable soil and climate conditions promoted N 2 O emissions. The relationships between N 2 O flux and field-soil water conditions are shown in Figure 5. There was a trade-off relationship between N 2 O and soil-water content in CI+N2 treatment except in the regreening stage and yellow-maturing stage, but correlation analysis showed that the relationship between them was not significant (p = 0.095). Except for 79 DAT N fertilizer application, there was a significant negative correlation between N 2 O and water depth (p < 0.05) in FIN2. In both irrigation modes, reducing soil-moisture content and field-water depth promoted N 2 O emission. However, the promotion effect was only significant under FI.
As shown in Table 2, under the same N application, N 2 O emissions of CI+N3, CI+N2, CI+N1, and CI+N0 increased by 11.85%, 10.55%, 9.88%, and 5.66% more than those of FI+N3, FI+N2, FI+N1, and FI+N0, respectively, indicating that CI treatments could significantly increase N 2 O emissions. Among CI treatments, N 2 O emissions of CIN3 were the largest (0.44 kg·hm −2 ), at 22.91%, 62.62%, and 207.73% higher than those of CI+N2, CI+N1, and CI+N0, respectively. Among FI treatments, the N 2 O emissions of FI+N3 were the largest (0.39 kg·hm −2 ), at 21.48%, 59.75%, and 190.69% higher than those of FI+N2, FI+N1, and FI+N0, respectively. There were significant differences between the levels of N application among both CI and FI treatments, N 2 O emissions increased with the increase of N application, and the increase of N 2 O emissions under CI was larger than that under FI. In summary, under the two irrigation modes, increasing the amount of N application would significantly promote N 2 O emissions. Compared with FI, CI significantly increases N 2 O emissions.

Yield and Its Components
As shown in Table 3, under the same N application, the yield of CI treatments was 5.53~27.01% higher than that of FI treatments. Among them, the increase of CI+N2 is the largest, which indicates that CI has a good effect on increasing yield. Under FI, yield increased with the increase of N application, and maximum yield was 8049.78 kg·hm −2 when N application was 135 kg·hm −2 . Under CI, yield initially increased and then decreased with the increase of N application. When N application was 110 kg·hm −2 , maximum yield was 10,224.4 kg·hm −2 . Therefore, from the perspective of irrigation mode and N application, the appropriate N application rate under CI was more conducive to a higher yield.
The yield components of two irrigation modes under different N applications are also analyzed in Table 3. For CI treatments, except for the seed-setting rate, the other three yield components had a threshold value with the increase of N application, which showed a trend to first increase and then decrease, resulting in the same trend of yield change. For FI treatments, the spikes per unit area and 1000-grain weight increased with the increase of N application, while the grain number per spike and seed-setting rate showed a tendency to rapidly increase and then slightly decrease, resulting in a continuous yield increase with the increase of N application. Under the same amount of N application, compared with FI treatments, the spikes per unit area of CI treatments decreased by 5.80~13.98%, while grain number per spike, seed-setting rate, and 1000-grain weight increased by 1.06~14.09%, 1.57~5.34%, and 1.52~2.49%, respectively. Compared with FI treatments, the increase of grain number per spike, seed-setting rate, and 1000-grain weight of CI treatments made up for the loss of panicle number per unit area, and eventually led to an increase in yield of CI treatments. Note: Lowercase letters represent the differences between each treatment (p < 0.05).

Water Consumption
As shown in Figure 6, the water-consumption trends of rice at different growth stages under different treatments were the same. Under two irrigation modes, the descending order of water consumption in different growth periods was: tillering stage > jointing and booting stage > heading and flowering stage > grain-filling stage > maturity stage > regreening stage. Due to the same water-management method in the regreening stage, and seedlings being smaller and the transpiration of plants being weaker, the water consumption of each treatment had no significant difference. In addition to the regreening stage, the water consumption of CI treatments during each growth stage of rice was significantly lower than that of FI treatments. Comparing the water consumption at different growth stages under the same N application rate, we can see that the difference of water consumption between CI and FI treatments was the largest during tillering stage. The water consumption of CI+N0, CI+N1, CI+N2, and CI+N3 decreased by 55.52%, 49.74%, 46.27%, and 45.08%, respectively, compared with FI+N0, FI+N1, FI+N2, and FI+N3. The water-saving effect of CI was the most obvious in this period. At the grain-filling, jointing and booting, heading and flowering, mature, and tillering stages, the differences between water consumption of CI and FI treatments decreased, and the water-saving effect of CI gradually became smaller. By comparing the relationship between water consumption and N application at different stages, it was found that the water consumption of each growth period increased with the increase of N application under both CI and FI. For CI, water consumption increased by 23.97~41.18%, 21.13~29.59%, 10.44~16.97%, 10.31~16.35%, and 6.19~15.77% in the tillering stage, the jointing and booting stage, the heading and flowering stage, the grain-filling stage, and the mature period under different N application rates compared with the CI+N0 treatment. However, for FI, water consumption increased by 9.69~14.34%, 9.49~13.62%, 6.63~10.2%, 4.97~11.85%, and 3.83~12.06%, respectively, compared with FI+N0 in the corresponding growth stages. By comparing the effects of N application on water consumption in different growth stages of rice under two irrigation modes, it could be seen that the effect of N application on water consumption under CI was more significant than that under FI.  Table 2, under the same N application, the total water consumption of CI+N0, CI+N1, CI+N2, and CI+N3 decreased by 40.47%, 36.68%, 35.91%, and 34.89%, respectively, compared with FI+N0, FI+N1, FI+N2, and FI+N3; CI total water consumption was significantly lower than that of FI. Compared with FI, a water layer was not established in the paddy fields in each growth stage except the regreening stage for CI. CI reduced the evaporation of field water while meeting the rice-growth requirement, and achieved the goal of saving water. Among CI treatments, the maximum water-consumption treatment was CIN3 (5197 kg·hm −2 ), which was 3.98%, 7.84%, and 23.04% higher than that of CI+N2, CI+N1, and CI+N0, respectively. Among FI treatments, the maximum water-consumption treatment was FI+N3 (7982 kg·hm −2 ), which was 2.35%, 4.89%, and 12.50% higher than that of FI+N2, FI+N1, and FI+N0, respectively. Under the two irrigation modes, increasing the amount of N application would lead to an increase of water consumption. In summary, increasing N application significantly increased water consumption, but the water-consumption amount in each growth period of rice with different N applications could still be significantly reduced under CI with different N applications compared with FI; thus, total water consumption was reduced, and the effect of CI on water-saving was significant.

Comprehensive Assessment of GWP, GHGI, and WP
From the above analysis, we know that CI increased N 2 O emissions while reducing CH 4 emissions compared with FI. For the total GWP caused by CH 4 and N 2 O under the same N application, the GWP of CI+N3, CI+N2, CI+N1, and CI+N0 decreased by 37.01%, 46.12%, 18.98%, and 27.93% compared with FI+N3, FI+N2, FI+N1, and FI+N0, respectively ( Table 2). It can be seen that GWP produced by CI was significantly less than that by FI. By comparing the proportion of GWP produced by CH 4 and N 2 O, the greenhouse effect of CH 4 accounted for more than 96% of total GWP. Therefore, CH 4 was still the main greenhouse gas produced in paddy fields in the cold region. By comparing the relationship between GWP and N application in the two irrigation modes, it was found that the trend of GWP with the increase of N application was basically the same as the trend of CH 4 with the increase of N application. This may also be because GWP produced by CH 4 accounted for a large proportion of total GWP. By comparing with the amount of total GWP in the two irrigation modes, we found that the minimum GWP between CI treatments was CI+N0 (3357.09 kg·hm −2 ), at 17.49%, 11.22%, and 14.64% lower than that of CIN1, CIN2, and CIN3, respectively. Minimum GWP among FI was FIN0 (4658.01 kg·hm −2 ), at 7.25%, 33.63%, and 25.40% lower than that of FIN1, FIN2, and FIN3, respectively. Treatments with no N application had the smallest GWP under two irrigation modes; however, in the actual production process of rice, N fertilizer need to be applied to ensure yield. Therefore, GHGI is introduced to balance the contradiction between reducing GWP and ensuring yield affected by applying N fertilizer.
GHGI is an important indicator for evaluating greenhouse-gas production per unit rice yield. According to Table 2, the GHGI of CI+N3, CI+N2, CI+N1, and CI+N0 decreased by 49.86%, 61.23%, 33.53%, and 31.71%, respectively, compared with FI+N3, FI+N2, FI+N1, and FI+N0 under the same N application. This indicated that the greenhouse gas produced by CI per unit yield was significantly less than that of FI. Among CI treatments, GHGI generally decreased with increasing N application rate. CI+N2 had the minimum GHGI (0.37 kg CO 2 -eq·kg −1 ), at 4.90%, 33.69%, and 32.47% lower than CI+N3, CI+N1, and CI+N0, respectively. Among FI treatments, GHGI first increased and then decreased with the increase of N application. FIN3 had the minimum GHGI (0.78 kg CO 2 -eq·kg −1 ), which was 18.69%, 7.56%, and 3.27% lower than FI+N2, FI+N1, and FI+N0, respectively, but it was still 40.58% larger than CI+N2. By comparing the yield and GHGI of each treatment, we can see that CI+N2 treatment has the best emission-reduction effect under the premise of ensuring yield.
According to Table 3, the WP of each treatment was compared. Under the same N application, the WP of CI+N3, CI+N2, CI+N1, and CI+N0 was 77.26%, 92.47%, 116.85%, and 92.96% higher than that of FI+N3, FI+N2, FI+N1, and FI+N0, respectively. It can be seen that CI can significantly increase the WP of rice compared with FI. By analyzing the effect of N application on WP under the two irrigation modes, we found that the WP of each treatment under CI first increased and then decreased with the increase of N application. CI+N2 had the maximum WP (2.05 kg·m −3 ), at 5.12%, 35.14%, and 40.96% higher than that of CI+N3, CI+N1, and CI+N0, respectively. Under FI, the WP of each treatment increased with the increase of N application; FI+N3 had the maximum WP (1.01 kg·hm −2 ), at 6.91%, 28.22%, and 23.18% higher than that of FI+N2, FI+N1, and FI+N0, respectively. By comparing the yield and WP of each treatment, we can see that the yield-increasing and water-saving effect of CI+N2 treatment was significantly higher than other treatments. After comprehensively analyzing the yield, GHGI, and WP of each treatment, it was found that the CI+N2 treatment has the best effect on increasing yield, reducing emissions, and saving water. Its yield, GHGI, and water productivity were 10,224.4 kg·hm −2 , 0.37 kg CO 2 -eq·kg −1 , and 2.05 kg·m −3 , respectively.

Effects of Different Water and N Management Types on CH 4 and N 2 O
Among many agricultural practices, water management has been recognized as one of the most promising approaches to reduce CH 4 emissions [23]. CI exerted a significant impact on CH 4 emission. No water layer in CI was established on paddy fields at each growth stage except for the regreening stage. Soil aeration was greatly increased compared with FI. Therefore, it increased the oxidation rate of CH 4 and inhibited the activity of methanogen, effectively reducing CH 4 emissions [24]. This result is in agreement with previous studies that showed that the frequency of soil wetting and drying determined CH 4 mitigation potential [25,26]. In the present study, later tillering drainage was responsible for the relatively lower CH 4 emission flux in the stage after tillering for both CI and FI. This is similar to the results of Tariq et al., which could be attributed to the subsequent reduced methanogenesis activity [26]. The present study indicates that the increase of soil-water content promotes CH 4 emission flux under CI, which agrees with the previous study in that CH 4 fluxes exhibit an increased trend with the increase of water filled pore space, and excessively low water content promotes soil oxidation [27]. N 2 O is the product of soil nitrification and denitrification. Frequent flooding and drainage of soil may stimulate N 2 O emissions by nitrification and denitrification [7,28]. A large number of studies have shown that there was a trade-off relationship between the emission of N 2 O and CH 4 in paddy fields [18,29]. In this experiment, the CH 4 flux sharply decreased during the drying and wetting alternation period, while an N 2 O flux peak occurred, especially in the late tillering drainage period; this was the same as the existing emission pattern [18]. Moreover, the improved diffusion of N 2 O can also be promoted by soil aeration [30]. Studies have shown that CI can significantly reduce GWP produced by N 2 O and CH 4 by 59.1% compared with FI [24]. In this paper, CI decreased GWP by 46.12% compared with FI under the conventional N application rate (110 kg·hm −2 ). This may be caused by the difference between climate, soil, and water management in different experiment areas.
At present, the results of research on the effect of N fertilizers on CH 4 emissions in paddy fields are very inconsistent [31]. It was reported that N fertilizers can promote or inhibit the emission of CH 4 [11,12]. In this study, under CI, N application could significantly increase CH 4 emissions compared with no N treatment, but there was no significant difference between different N application treatments. Under FI, the treatment of 85 and 110 kg·hm −2 N application significantly increased the emission of CH 4 compared with the treatment with no N application, but there was no significant difference between the application of 110 and 135 kg·hm −2 N. This may be because urea promoted the root development of rice, and provided a precursor matrix for the production of CH 4 in paddy fields. The competitive effect of NH 4 + hydrolyzed by urea on CH 4 promoted CH 4 emissions. However, N also promoted the activity of CH 4 -oxidizing bacteria, thus reducing CH 4 emissions [32]. The difference between N treatments was only 25 kg·hm −2 in this experiment, which was a small interval. Therefore, the difference of the N fertilizer may have no significant effect on the inhibition or promotion of CH 4 emissions. The N fertilizer not only affected soil nitrification and denitrification, but also affected the growth of rice plants and the transport of N 2 O from the soil to the atmosphere. Finally, it had a promoting effect on N 2 O emissions [33,34]. Shcherbak et al. [35] performed a meta-analysis of 78 articles with at least three N application levels in the world, and found that the amount of N 2 O emission increased exponentially with the increase of N application. By fitting the N amount and N 2 O emissions in this experiment, we found that the exponential fitting effect is better than that of the linear one, which was the same as previous research results.
This study was carried out in high latitudes and cold regions of China. Results were quite different from those in other regions because of the different climate, soil conditions, and tillage systems. The results of this study showed that CH 4 [10]. CH 4 emissions of the present study were higher than the season emissions of early rice or late rice, but lower than their annual CH 4 emissions. N 2 O emissions of the present study were much lower than the results of Li et al. They applied 165 and 180 kg N·hm −2 for early rice and late rice, respectively, to ensure the yield due to their lower total N in the soil, while our moderate N application rate was 110 kg·hm −2 . More N fertilizer promoted N 2 O emissions in southern China. On the whole, CH 4 [36], while in this experiment, CI reduced CH 4 emissions by 47%, but increased N 2 O emissions by only 9%. CI had a better CH 4 emission-reduction effect in southern China than in this experiment. However, the increase of N 2 O under CI was larger than that in the present experiment. This may be due to different soil basic properties and nitrogen-application rate. In the experiment of Yang et al., the soil organic matter and total nitrogen values were only 52% and 7% of those this study, respectively, and they applied more N fertilizer than in this study. Lower soil organic matter and higher N fertilizer application under CI inhibited CH 4 emissions, but promoted N 2 O in southern China. Although the effect of CH 4 emission reduction under CI in the paddy fields was not as good as that in the southern China, the increase of N 2 O was much smaller.

Effects of Different Water and N Management Types on Yield, WP, and GWP
FI is the most common irrigation mode for farmers. However, in this study, compared with FI, we saw that CI can effectively increase yield by 5.53~38.97%. This may be because CI could effectively control ineffective tillering, reduce N loss caused by ineffective tillering, and increase grain number per spike, seed-setting rate, 1000-grain weight (Table 3), and the development of rice-root systems [37]. FI needs to keep a thick water layer in the field, which increases the waste of water resources caused by evaporation, surface runoff, and deep seepage. The water-saving effect of CI was significant from the tillering stage until the maturity stage ( Figure 4). The management mode of a no-water layer in the field reduced the loss of water, N fertilizer, and unnecessary water pollution produced by the paddy fields [38]. The water-saving and yield-increasing characteristics of CI led to an overall increase of WP. At the same time, CI reduced the amount of irrigation water, which can effectively reduce the labor cost and fuel consumption of farmers. Promoting the popularization of CI could effectively solve the problem of water shortage in rice-growing areas in cold regions. In this experiment, the results of the comprehensive evaluation of GWP, GHGI, and WP in CIN3 and CIN2 were all good, but the higher N application rate would reduce N use efficiency, resulting in unnecessary N loss and an increase of N 2 O emissions [39,40].
CI could effectively reduce the GWP produced by N 2 O and CH 4 in paddy fields. Results showed that CI+N2 was better than other treatments on reducing greenhouse-gas emissions. According to Heilongjiang Statistical Yearbook 2016 [8], the rice-planting area in Heilongjiang is about 3.81 × 10 6 hm 2 . If the irrigation and N application mode of CI+N2 recommended in this experiment was used to replace the conventional FI+N2 treatment, it is estimated that the annual GWP could be reduced by about 1.24 × 10 10 kg CO 2 -eq, and irrigation water would be reduced by 1.07 × 10 10 m 3 during the growth period of rice. The application of CI in rice-planting areas in cold regions is of great significance for reducing greenhouse-gas emissions, alleviating the shortage of water resources and ensuring high yield of rice.

Conclusions
Under the two irrigation modes, N application could significantly increase CH 4 emissions in paddy fields compared with no N application treatments. However, there was no significant difference between 85, 110, and 135 kg·hm −2 N application treatments under CI, and 110 and 135 kg·hm −2 N application treatments under FI; N 2 O emissions in paddy fields significantly increased with the increase of N application rate under both CI and FI. Compared with FI under the same N application, CI significantly reduced CH 4 emissions by 19.42~46.94%, but at the same time increased N 2 O emissions by 5.66~11.85%, finally leading to a result of the total GWP of CI treatments being significantly less than that of FI treatments. Appropriate N application could increase yield components. Under the same N application, the spikes per unit area of CI rice were significantly smaller than those of FI, but the grain number per spikes, seed-setting rates, and 1000-grain weight were larger than those of FI, which made up for the loss of smaller spikes per unit area and increased yield by 5.53~38.97%. Compared with FI, CI significantly reduced water consumption at each growth stage (except for the regreening stage) and total water consumption. The GHGI of different CI treatments was significantly lower than that of FI treatments, and the WP of CI treatments was significantly higher than that of FI treatments. Through comprehensive comparison and analysis, CI+N2 treatment (CI with 110 kg·hm −2 N application) is recommended for rice-planting areas in cold regions because of its highest yield and WP, and lowest GHGI, meeting the purposes of water-savings, emission reductions, and high rice yield.
Author Contributions: T.N., Z.Z. and Z.Q. conceived and design the experiment; T.N. and P.C. performed the experiment; T.N. analyzed the data; T.N. wrote the paper; Z.Z., Z.Q., Y.L. and D.X. reviewed and edited the paper.