Predicting the Nitrogen Quota Application Rate in a Double Rice Cropping System Based on Rice–Soil Nitrogen Balance and ¹⁵N Labelling Analysis

Abstract

Excessive nitrogen (N) fertilization, low use efficiency, and heavy pollution are the dominant issues that exist in intensively cultivated double rice cropping systems in China. Two-year field and ¹⁵N microregion experiments were conducted to evaluate the N fate in a soil-rice system under a series of different N rate treatments from 2020 to 2021. The economic N application rate that simultaneously improved rice yield and N use efficiency in the rotation system was also investigated. Results demonstrated that soil residues and mineralized N accounted for more than 58.0% and 53.2% of the total N input in the early and late rice seasons, respectively. Similarly, most of the total N input was absorbed by rice, ranging from 43.7% to 55.6% in early rice and from 36.8% to 54.7% in late rice. Rice N use efficiency significantly decreased with increasing N application, while rice grain yield and its N uptake increased when the N application rate was below 150 kg ha⁻¹ in early rice and 200 kg ha⁻¹ in late rice. Exceeding this point limited rice N uptake and yield formation. The apparent N recovery rate, N residual rate, and N loss rate were 23.5–34.4%, 17.0–47.1%, and 26.0–47.8% for the early rice, and 32.8–37.3%, 74.2–87.0%, and 71.5–92.1% for the late rice. The linear plateau analysis further indicated that the recommended N application rate (118.5–152.8 kg ha⁻¹ for early rice and 169.9–186.2 kg ha⁻¹ for late rice) can not only maintain a relatively higher rice yield and N utilization but also significantly reduce soil N residue. Our results provide theoretical guidance for improving N management in double-cropping rice fields in southern China.

1. Introduction

Double rice rotation is one of the most important cropping systems in South China, accounting for more than 40% of the whole rice cultivation area [1,2]. Intensive rice cultivation is characterized by high N input, low N use efficiency, and poor soil and crop management. The average chemical N input ranges from 550 to 600 kg ha⁻¹ y⁻¹ [3], which is much higher than the rice demand [4,5]. Overfertilization is detrimental to plant development, reproduction, and grain quality [6]. Excessive N fertilizer input also changes the soil N balance and leads to most of the N loss, leaching, or volatilization, thus causing a series of environmental impacts, such as reactive N emissions of greenhouse gases, groundwater pollution, and soil acidification [7,8,9]. Therefore, how to make full use of N fertilizer to increase rice yield, reduce N fertilizer loss, and coordinate N fertilizer’s agronomic and environmental effects is still a major challenge for double-cropping rice production.

Generally, rice yields cannot continue to increase or be maintained by increasing investments in fertilizer resources [10]. Therefore, applying N fertilizer should aim to obtain a higher yield and a better N use efficiency [11]. Some results showed that rice yield, N uptake, and its translocation to panicles increased significantly when the N application rate was below 180 kg ha⁻¹ in early rice and 240 kg ha⁻¹ in late rice; exceeding these values still promoted rice N absorption but limited N transport to grains and the N agronomic utilization rate [12]. Among the N output variables, the apparent N recovery rate, residual rate, and loss rate in double-cropping rice were 35.4–58.6%, 5.9–33.0%, and 18.0–55.3%, respectively [13,14]. Ammonia volatilization and denitrification were the dominant sources of N loss, accounting for 31.1% and 67.9%, respectively, of the soil’s total N loss. For a long time, ¹⁵N isotope labeling and subtraction methods have been widely adopted to calculate plant N utilization efficiency. The ¹⁵N tracing method can be used to study the N use efficiency in a season, as well as the soil residual rate and loss rate, and even the effect of residual N on the next crop [15]. The traditional difference method mainly considers the effect of N fertilizer on yield but cannot clarify the source of N in plants, i.e., whether it is derived from soil or fertilizer.

To coordinate the contradiction between the N fertilizer management and the environment, many optimized N-saving agronomic practices including multisplit topdressing [16], site-specific N management (SSNM) [5], combination of inorganic and organic fertilizers, such as biochar or legume residues [17], and an integrated and optimized crop management [18] have been applied in rice cultivation. These strategies have shown great benefits in addressing the source-sink balance in rice and have increased rice yield and nitrogen use efficiency (NUE) with reduced N inputs in irrigated rice systems. Peng demonstrated that reducing 30% of current N fertilizer input can not only maintain the relative higher rice yield and soil fertility of double-cropping paddy fields in the long run but also increase the partial productivity of nitrogen fertilizers [19]. Based on the four-year field experiment, the Nutrient Expert system (NE) has been proposed to improve yield and nitrogen (N) use efficiency (NUE) in the double rice cropping systems in China, especially when soil testing is not available or timely for smallholders [10]. Some developed countries, e.g., Poland, have issued legislation on the recommended application rate of chemical fertilizers, which suggests that N, P, and K fertilizers for the highest melon yields are 120 kg ha⁻¹, 100 kg ha⁻¹, and 150 kg ha⁻¹ under drip irrigation conditions [20]. In 2018, Zhejiang Province first proposed the concept of quota fertilization, and farmers’ fertilization must not exceed this standard.

However, majority of the work has focused on the characteristics of N uptake and utilization based on one or more methods in a single-crop season and has not yet considered the fate of N in soil-crop systems and the residual effect of N in the rotation cycle [21]. Quantitative monitoring of the fate and residual effect of N in double-cropping rice areas can enable researchers to accurately understand the characteristics of the N balance and then evaluate its agronomic and environmental effects more scientifically and comprehensively. Here, the objectives of this study were (1) to evaluate the rice yield and N absorption responses to different N application rates, (2) to study the characteristics of N absorption and utilization, loss, and residue in rice-soil systems under different N application rates, and (3) to calculate the N quota application rate to obtain both a higher yield and better N use efficiency in double rice cropping systems using the apparent N balance and ¹⁵N labeling analysis. We did so by determining the grain yield, N uptake, and N use efficiency of rice and the soil N surplus under various N application rates based on a two-year experiment with early and late rice planting, thus encompassing four growing seasons in China.

2. Materials and Methods

2.1. Experimental Site Description

A two-year field experiment was carried out in a double-cropping rice system on the research farm of the China National Rice Research Institute from 2020–2021 in Fuyang City, Zhejiang Province, China (30°03′ N, 119°57′ E). According to the Chinese Soil Taxonomy, the soil is classified as blue purple clay soil. The basic soil properties were as follows: pH 6.2, organic matter 34.9 g kg⁻¹, alkali-hydrolysable N 134.5 mg kg⁻¹, Olsen-P 18.7 mg kg⁻¹, and exchangeable K 89.4 mg kg⁻¹.

The field has a subtropical monsoon climate with a daily average air temperature of 17.8 °C and an annual average precipitation of 1454 mm. The amounts of rainfall recorded from the transplanting to maturity stages for early rice were 890.0 mm in 2020 and 451.2 mm in 2021, and were 363.0 mm in 2020 and 425.4 mm in 2021 for late rice (Figure 1). In the early rice season, rainfall was distributed from the panicle differentiation to maturity stages in 2020 but concentrated mainly at the grain-filling stage in 2021. In the late rice season, rainfall was concentrated mainly at the panicle differentiation stage in 2020 and at the tillering and panicle differentiation stages in 2021.

2.2. Experiment Design

A field experiment with three replicates was used in this study, and two seasons of early rice and two seasons of late rice were conducted in 2020 and 2021. Commercial rice varieties, Zhongzao 39 (early rice) and Tianyouhuazhan (late rice), suitable for the local environment with high yield potential, were planted. The early rice season included six treatments with different N application rates: no N fertilizer (N₀), 90 kg N ha⁻¹ (N₉₀), 120 kg N ha⁻¹ (N₁₂₀), 150 kg N ha⁻¹ (N₁₅₀), 180 kg N ha⁻¹ (N₁₈₀), and 210 kg N ha⁻¹ (N₂₁₀), whereas the late rice seasons included no N fertilizer (N₀), 120 kg N ha⁻¹ (N₁₂₀), 160 kg N ha⁻¹ (N₁₆₀), 200 kg N ha⁻¹ (N₂₀₀), 240 kg N ha⁻¹ (N₂₄₀), and 280 kg N ha⁻¹ (N₂₈₀). The individual plot area was 20 m², with a length of 4 m and a width of 5 m. Each site was separated from adjacent plots by a 0.5 m wide plant-free border. Nitrogen was applied as basal, tillering, and panicle fertilizer at 40%, 30%, and 30%, respectively. Phosphorus (75 kg P₂O₅ ha⁻¹ for early rice and 90 kg P₂O₅ ha⁻¹ for late rice, as superphosphate) was applied as the basal fertilizer, whereas potassium (60 kg K₂O ha⁻¹ for early rice and 120 kg P₂O₅ ha⁻¹ for late rice, as potassium chloride) was applied in two equal applications (50:50) as the basal and panicle fertilizer in all treatments (Table S1). All field experiments followed common local water-management and pest-control methods.

2.3. Monitoring of Environmental Nitrogen Input

The atmospheric dry and wet N deposition was collected by a Φ150 mm dustfall cylinder according to the method described by the national standard [22] (GB/T15265-94). The solution in the dustfall cylinder was sampled every month. The irrigation water was controlled and monitored using a flow meter (LXSG-50 Flow meter, Shanghai Water Meter Manufacturing Factory, Shanghai, China) installed in the irrigation pipelines. The seasonal total water input was the sum of all irrigation water from the transplanting to harvesting stages. The contents of ammonium and nitrate in the dustfall cylinder and irrigation solution were determined by a continuous flow analyzer, and the total N was determined by the potassium persulfate digestion method [23].

2.4. Isotope ¹⁵N Tracer Test

To reveal the fate of the ¹⁵N fertilizer in the rice-soil system, a microzone experiment using ¹⁵N labeling was conducted in the early and late rice seasons of 2021 [24]. The ¹⁵N isotope microzone was designed with a PVC plastic pipe with a 50-cm diameter and 60-cm height. Before transplantation, all pipes in each plot were buried, leaving 0.20 m above ground to prevent the flow and exchange of N sources from the micro area. Three rice seedlings were transplanted in the middle of the microzone. The amounts and application methods of ¹⁵N-urea, phosphorus, and potassium fertilizer were consistent with the field experiment. To ensure that the ¹⁵N-urea spread uniformly throughout the microcosm, each pipe was injected in a 10-point hexagonal pattern with a total of 100 mL of the aforementioned solution using a 10-cm double-sideport spinal syringe (Popper and Sons, Hyde Park, NY, USA) [23]. Each treatment had three replicates. In addition, three blank pipes were supplied with equal amounts of unlabeled urea for each treatment. The ¹⁵N abundance of labeled urea was 20.32%, and this urea was produced by the Shanghai Chemical Research Institute.

2.5. Sampling and Analysis

2.5.1. Rice and Soil Samples

At the physiological maturity stage, rice plants from each microzone were sampled and divided into stem-leaves and panicles. Meanwhile, three individual plants were also randomly sampled from each plot to determine the rice yield components [18]. The numbers of filled and unfilled spikelets were counted with the water-float method. The thousand-grain weight, spikelet number per panicle, and percentage of filled spikelets were determined. The grain yield was determined from a 10 m² harvesting area and normalized to 13.5% moisture. The dried samples of the grain and stem-leaf were ground into powder, and their N content was analyzed by the micro-Kjeldahl method.

Soil samples from 0–100 cm layers (one layer per 20 cm) were collected from each plot and ¹⁵N isotope microzone in the early and late rice seasons. The soils were sampled at three points and mixed with the same soil layer. All soil samples were air-dried and passed through a 1 mm sieve for chemical analysis.

2.5.2. Analysis of Samples in the ¹⁵N Microregion

Soil samples of the 0–20, 20–40, 40–60, 60–80, and 80–100 cm layers were used to determine the total N, NO₃⁻-N, NH₄⁺-N, and ¹⁵N abundances according to the methods described in [25,26]. Three portions of 10 g soil samples were weighed, extracted with 1 mol/L KCl solution (soil-water ratio of 1:5), and filtered with filter paper after oscillation for 30 min. The contents of NO₃⁻ and NH₄⁺ were simultaneously determined by continuous flow analysis. The ¹⁵N abundance of soil and rice samples was determined by a MAT 2250 mass spectrometer at Zhejiang University.

2.6. Calculation and Analysis

2.6.1. Apparent N Balance in the Soil-Plant System

The calculations of soil’s N input and output were based on the apparent N balance principle [15]. The inorganic N accumulation (N_IN) was calculated for each soil layer and used to estimate apparent N mineralization and loss according to the Equation (1):

$$N_{\mathit{IN}}={\displaystyle {\sum }_{i=1}^n{S}_T}\times S_{BD}\times N_C/10$$

(1)

where S_T is the soil layer thickness (cm); S_BD is the soil bulk density (g cm⁻³), using 1.32 g cm⁻³ at 0 to 20 cm, 1.35 g cm⁻³ at 20 to 40 cm, 1.39 g cm⁻³ at 40 to 60 cm, 1.41 g cm⁻³ at 60 to 80 cm, and 1.43 g cm⁻³ at 80 to 100 cm; and N_C represents the soil NO₃⁻-N and NH₄⁺-N (mg kg⁻¹) contents in different soil layers. n represents the five soil layers of the 0–20, 20–40, 40–60, 60–80, and 80–100 cm profiles.

Apparent N mineralization (N_mine) = N_uptake + N_residual − N_initial

(2)

where apparent N mineralization (N_mine) is the amount of N mineralizations estimated in the N-untreated plots. N_uptake is the N uptake by the above-ground part measured at harvest, and N_residual and N_initial are the total amounts of N_IN in the 0–100 cm soil layer after harvest and before sowing, respectively.

N input (N_input) = N_fertilizer + N_initial + N_mine + N_deposition + N_irrigation

(3)

where N_input is the total amount of different N resources in the early or late rice season; N_fertilizer is the amount of N applied; N_deposition is the N resource from atmospheric dry and wet N deposition; and N_irrigation is the N from irrigation water.

Apparent N loss (N_loss) = N_input − N_uptake − N_residual

(4)

Apparent N residual rate (%) = (N_residual in N_N − N_residual in N₀)/N_N × 100

(5)

Apparent N loss rate (%) = (N_loss in N_N − N_loss in N₀)/N_N × 100

(6)

Apparent N recovery rate (%) = (U_N − U₀)/N_N × 100

(7)

where N_loss is the N loss in the early or late season; N_N is the N application rate; U_N and U₀ are the N uptakes by above-ground rice parts in N-treated and N-untreated plots, respectively; and N_input, N_uptake and N_residual were calculated from the abovementioned equations.

2.6.2. Measurements of N Use Efficiency

The various parameters of NUE, i.e., N recovery efficiency (NRE), N agronomic efficiency (NAE), N partial factor productivity (NPFP), and N physiological efficiency (NPE), were all evaluated according to Pan et al. [27].

NRE (%) = (G_N − G₀)/N_N × 100

(8)

NAE (kg kg⁻¹) = (Y_N − Y₀)/N_N × 100

(9)

NPFP (kg kg⁻¹) = Y_N/N_N × 100

(10)

NPE (kg kg⁻¹) = (Y_N − Y₀)/(N_N − N₀) × 100

(11)

where G_N and G₀ are the N uptake by rice grain in N-treated and N-untreated plots, respectively, and N_N (kg N ha⁻¹) is the N fertilization rate. Y_N and Y₀ (kg ha⁻¹) are the rice grain yields at harvest in N-treated and N-untreated plots, respectively.

2.6.3. Measurements of ¹⁵N-Labeled Fertilizer Nitrogen

The proportion of rice N uptake, derived from the N fertilizer and soil, was calculated by using the method reported by Hauck and Bremner [28].

N_dff (kg N hm⁻²) = N_T × (A_rice − A₀)/(A_F − A₀)

(12)

N_dfs (kg N hm⁻²) = N_uptake − N_dff

(13)

¹⁵N recovery efficiency (%) = N_dff/N_rate × 100

(14)

¹⁵N loss rate (%) = 100 − ¹⁵N recovery efficiency − ¹⁵N residual rate

(15)

N_dff contribution (%) = (A_s − A₀)/(A_F − A₀) × 100

(16)

N_dfs contribution (%) = 100 − N_dff contribution

(17)

where N_dff and N_dfs are the amounts of rice’s absolute N uptake derived from fertilizer and soil; N_T is the N content in the grain or above-ground part of rice; W_soil and N_soil represent the soil weight in the 0–100 cm layers and their corresponding soil N content; A_rice and A_soil are the ¹⁵N atom% values of the rice plant and soil, A₀ is the average ¹⁵N atom% in plants in the N-untreated plots, which is 0.366 atom% in our experiment; and A_F is the ¹⁵N atom% of the ¹⁵N-urea used in our experiment (20.32%).

2.7. Data Statistics and Analysis

The relationship between the N application rate and different N use indicators was simulated through quadratic and linear plateau models to calculate the optimum N rate [10,29]. The quadratic model equation is as follows:

y = a · x² +b · x + c

(18)

where y is the rice yield (or N use efficiency, N uptake, soil N loss rate, N residual rate, ¹⁵N contribution), and x is the N application rate (kg N ha⁻¹).

The linear plateau model equation is as follows:

$$\mathrm{y}=\left\{{}_{\mathrm{P} (\mathrm{x}>\mathrm{C})}^{\mathrm{a}·\mathrm{x}+\mathrm{b}(\mathrm{x}<\mathrm{C})}\right.$$

(19)

where y is the rice yield (or N uptake, soil N loss rate, N residual rate, ¹⁵N contribution), x is the N application rate (kg ha⁻¹), a is the slope, b is the intercept, C is the intersection of the straight line and the platform, and P is the platform indicator. The parameters C and P represent the minimum N rate required to produce the maximum grain yield (or N uptake, soil N loss rate, N residual rate, ¹⁵N contribution) and their corresponding values.

All statistical data, including the analysis of variance (ANOVA) results, the least significant differences (LSDs), and the sources of variation due to year, N application rate, and their interaction, were derived using the SPSS system for Windows, version 14.0 (SPSS Inc., IBM Corp., Armonk, NY, USA). Significance was determined using Tukey’s test, and multiple comparisons of means were performed based on the LSD test at the 0.05 probability level. Figures were constructed using Origin v. 8.0 (OriginLab Corp., Northampton, MA, USA).

3. Results

3.1. Rice Yield and Its Yield Components

Rice grain yield was significantly affected by the N application rates (

Table 1. Rice yield and its yield components in early rice and late rice.

Treatments			Yielda	Effective Panicles	Spikelets	1000 Grain Weight	Grain Filling Rate
			kg ha−1	104		g	%
2020	Early rice	N0	6256.5 ± 198.0 c	9.2 ± 0.7 c	142.1+6.7 b	27.3 ± 0.1 a	84.5 ± 2.5 a
		N90	7461.0 ± 178.5 b	11.2 ± 0.9 b	152.5 ± 16.4 b	28.4 ± 0.8 a	84.5 ± 2.1 a
		N120	7510.5 ± 177.0 b	11.0 ± 0.3 b	144.3+10.0 b	28.5 ± 0.2 a	88.9 ± 1.8 a
		N150	7518.0 ± 19.5 b	11.3 ± 0.7 b	180.5 ± 16.0 a	27.6 ± 0.3 a	87.3 ± 2.8 a
		N180	7753.5 ± 91.5 a	11.4 ± 0.6 b	177.5 ± 17.6 a	27.3 ± 0.6 a	85.6 ± 5.4 a
		N210	7726.5 ± 231.0 a	13.2 ± 1.2 a	177.9 ± 10.7 a	27.3 ± 1.1 a	84.3 ± 4.6 a
	Late rice	N0	6001.5 ± 258.0 c	9.8 ± 0.4 c	198.6 ± 10.1 c	23.0 ± 0.5 a	83.1 ± 3.7 a
		N120	6901.5 ± 300.0 b	11.4 ± 0.2 b	194.2 ± 4.9 c	22.2 ± 0.7 bc	82.6 ± 1.9 a
		N160	7122.0 ± 174.0 b	11.7 ± 0.2 b	214.5 ± 6.2 a	22.2 ± 0.7 bc	83.1 ± 2.7 a
		N200	7456.5 ± 63.0 a	12.0 ± 0.4 ab	212.8 ± 1.9 a	22.6 ± 0.3 b	82.9 ± 2.4 a
		N240	7477.5 ± 127.5 a	12.7 ± 0.4 a	208.3 ± 11.0 ab	21.8 ± 0.4 c	78.5 ± 2.1 b
		N280	7551.0 ± 87.0 a	12.1 ± 0.2 ab	206.2 ± 3.7 b	22.7 ± 0.1 ab	78.6 ± 2.6 b
2021	Early rice	N0	6523.5 ± 169.5 c	10.4 ± 0.6 c	150.9 ± 5.8 b	28.0 ± 0.1 ab	85.8 ± 2.1 a
		N90	7761.0 ± 157.5 b	12.6 ± 0.8 ab	161.9 ± 14.2 b	29.1 ± 0.7 a	85.7 ± 1.7 a
		N120	7797.0 ± 147.0 b	12.4 ± 0.3 b	153.3 ± 8.6 b	29.0 ± 0.1 ab	90.2 ± 1.4 a
		N150	7794.0 ± 166.5 b	12.7 ± 0.7 ab	191.7 ± 13.8 a	28.2 ± 0.2 ab	86.8 ± 4.5 a
		N180	8044.5 ± 76.5 a	12.8 ± 0.6 a	188.5 ± 15.2 a	27.9 ± 0.5 b	88.6 ± 2.3 a
		N210	8031.0 ± 195.0 a	13.1 ± 1.0 a	188.9 ± 9.2 a	27.9 ± 0.9 b	85.6 ± 3.8 a
	Late rice	N0	6723.0 ± 294.0 c	10.9 ± 0.4 c	210.8 ± 11.2 b	22.8 ± 0.4 a	81.5 ± 7.6 a
		N120	7522.5 ± 177.0 b	13.0 ± 0.2 b	205.4 ± 5.5 b	22.3 ± 0.5 a	79.3 ± 3.7 ab
		N160	8049.0 ± 198.0 b	12.8 ± 0.3 b	227.2 ± 6.9 a	22.4 ± 0.3 a	75.1 ± 4.0 b
		N200	8277.0 ± 70.5 a	13.5 ± 0.5 ab	226.7 ± 4.3 a	22.6 ± 0.4 a	79.9 ± 5.3 ab
		N240	8149.5 ± 138.0 ab	14.3 ± 0.5 a	221.1 ± 12.2 a	22.2 ± 0.2 a	75.8 ± 5.1 b
		N280	8382.0 ± 96.0 a	13.6 ± 0.2 ab	218.5 ± 8.2 ab	22.6 ± 0.1 a	80.5 ± 4.6 a
ANOVA	N		*	**	*	ns	ns
	Y		ns	ns	*	ns	ns
	N × Y		ns	*	*	ns	ns

Note: The data in each column represent the average value plus the standard deviation (n = 3); N, Nitrogen; Y, Year. ns represents non-significant, and * and ** represent a significant F-value at p < 0.05 and p < 0.01, respectively. Values denoted by different lowercase letters indicate statistically significant differences between treatments for each cultivated season (p < 0.05).

, p < 0.05). The yield of early rice ranged from 6256.5 to 7726.5 kg ha⁻¹ in 2020 and from 6523.5 to 8044.5 kg ha⁻¹ in 2021, while those of late rice were 6001.5 to 7551.0 kg ha⁻¹ and 6723.0 to 8382.0 kg ha⁻¹, respectively. Rice yield and its effective panicles increased with increasing N application rates from 0 to 180 kg ha⁻¹ in the early rice season and from 0 to 200 kg ha⁻¹ in the late rice season, but no further significant increase was observed beyond these points. Meanwhile, the year and its interaction with the N application rate had no significant effect on the rice grain yield, 1000-grain weight, or grain filling rate.

3.2. N Fate in the Soil-Rice System

In 2020 and 2021, the inputs of dry-wet deposition N, irrigation N, and mineralization of organic N in the early rice season were 9.3 and 9.5 kg ha⁻¹, 6.4 and 6.0 kg ha⁻¹, and 97.2 and 124.8 kg ha⁻¹, respectively, and their values were 9.9 and 9.6 kg ha⁻¹, 6.8 and 7.4 kg ha⁻¹, and 140.9 and 127.1 kg ha⁻¹ in the late rice season (Tables S2 and S3). Soil total N input, residual N, and N loss increased with increasing N application rate, but rice N uptake did not further increase when the N application rate exceeded 180 kg ha⁻¹ for early rice and 200 kg ha⁻¹ for late rice (

Table 2. Apparent N balance in the double-rice growing system.

Treatments		N Input (kg N hm−2)	N Fate (kg N hm−2)			Ratio of N Fate (%)			NApp Recovery Rate (%)	NApp Residue Rate (%)	NApp Loss Rate (%)
			Nuptake	Nresidue	Nloss	Nuptake	Nresidue	Nloss
Early rice	N0	218.2 ± 5.2 f	135.5 ± 2.7 e	67.1 ± 3.2 f	15.6 ± 1.1 e	62.1 ± 0.8 a	30.8 ± 0.9 a	7.2 ± 0.2 c	-	-	-
	N90	318.4 ± 4.0 e	177.1 ± 4.3 d	80.8 ± 3.4 e	60.5 ± 3.3 d	55.6 ± 0.8 b	25.4 ± 0.8 c	19.0 ± 1.3 b	23.5 ± 1.8 c	17.0 ± 3.7 d	74.2 ± 3.7 a
	N120	350.4 ± 4.8 d	183.6 ± 2.8 c	91.9 ± 2.1 d	74.8 ± 5.4 cd	52.4 ± 1.5 c	26.2 ± 0.3 bc	21.3 ± 1.3 ab	26.2 ± 1.4 bc	27.0 ± 1.8 c	79.1 ± 4.5 a
	N150	389.0 ± 4.5 c	189.0 ± 1.5 b	106.9 ± 5.4 c	92.4 ± 9.1 bc	48.7 ± 0.8 d	27.5 ± 1.5 bc	23.8 ± 2.2 ab	28.3 ± 1.0 b	37.2 ± 3.6 ab	83.1 ± 6.0 a
	N180	424.4 ± 3.8 b	197.3 ± 1.5 a	116.0 ± 3.3 b	98.1 ± 13.9 b	46.5 ± 0.2 e	27.3 ± 0.8 b c	23.1 ± 3.3 ab	31.3 ± 0.8 ab	42.2 ± 1.8 a	84.1 ± 7.7 a
	N210	460.1 ± 3.8 a	206.4 ± 3.7 a	126.8 ± 4.5 a	120.4 ± 10.6 a	43.7 ± 0.8 f	27.6 ± 1.2 b	26.2 ± 4.5 a	34.4 ± 1.4 a	47.1 ± 2.1 a	87.0 ± 10.1 a
Late rice	N0	218.0 ± 8.1 f	129.1 ± 4.7 c	72.0 ± 3.7 e	16.9 ± 1.4 f	59.2 ± 0.5 a	33.0 ± 0.6 a	7.7 ± 0.3 e	-	-	-
	N120	351.6 ± 8.4 e	192.22 ± 4.7 b	97.3 ± 2.5 d	59.3 ± 5.2 e	54.7 ± 2.4 b	27.7 ± 1.0 b	16.8 ± 4.0 d	32.8 ± 1.9 b	26.0 ± 2.1 c	71.5 ± 6.2 c
	N160	402.8 ± 7.2 d	198.8 ± 3.4 ab	101.5 ± 2.3 d	102.4 ± 7.9 d	49.4 ± 0.3 c	25.2 ± 1.0 c	25.4 ± 0.8 c	35.1 ± 1.1 ab	29.1 ± 1.4 c	83.5 ± 1.2 b
	N200	457.7 ± 6.3 c	205.8 ± 3.9 a	118.0 ± 3.7 c	133.9 ± 1.8 c	45.0 ± 0.4 d	25.8 ± 0.5 c	29.3 ± 0.7 c	37.3 ± 1.0 a	39.0 ± 1.9 b	87.4 ± 2.9 ab
	N240	506.8 ± 4.7 b	205.1 ± 6.0 a	128.2 ± 4.0 b	173.5 ± 7.0 b	40.5 ± 1.5 e	25.3 ± 0.6 c	34.2 ± 1.1 b	37.1 ± 2.5 a	43.8 ± 1.6 ab	90.3 ± 2.9 a
	N280	557.7 ± 5.1 a	205.0 ± 9.1 a	137.9 ± 3.6 a	214.8 ± 1.0 a	36.8 ± 1.3 f	24.7 ± 0.8 c	38.5 ± 0.5 a	37.1 ± 2.3 a	47.8 ± 1.3 a	92.1 ± 3.4 a

Note: The data in each column represents the average value plus the standard deviation (n = 3) Values denoted by different lowercase letters indicate statistically significant differences between treatments for each cultivated season (p < 0.05).

).

The proportions of rice N uptake, soil residual N and N loss accounting for total N input were 43.7–62.1%, 25.4–30.8%, and 7.2–26.2% for the early rice season, and 36.8–59.2%, 24.7–27.2%, and 7.7–38.5% for the late rice season. The proportion of N uptake by early rice was higher than that by late rice, while their N loss showed the opposite trend, specifically under the higher N treatments. The apparent N recovery rate of early rice and the apparent N residual rate in both rice seasons significantly increased with increasing N application, but their values no longer increased when the N application rate exceeded 150 kg N ha⁻¹ for early rice and 200 kg N ha⁻¹ for late rice.

3.3. Rice N Use Efficiency

Rice NRE ranged from 19.0% to 42.3% and from 17.9% to 36.3%, NAE ranged from 7.0 to 13.9 kg kg⁻¹ and from 5.5 to 11.0 kg kg⁻¹; NPFP ranged from 36.8 to 87.0 kg kg⁻¹ and from 27.0 to 62.7 kg kg⁻¹. The average NRE, NAE, NPFP, and NPE of early rice were significantly higher than those of late rice, with averages of 30.4% vs. 25.2%, 9.9 vs. 7.5 kg kg⁻¹, 42.9 vs. 30.9 kg kg⁻¹, and 57.0 vs. 41.6 kg kg⁻¹, respectively. The regression analyses showed that rice NRE, NAE, and NPFP decreased with increasing N application rates (Figure 2, p < 0.05), but the NRE of early rice in 2020 was higher than that in 2021. NPE varied little with the N application rate in both the early and late rice seasons.

3.4. Soil Residual Mineral N

Soil residual N in the 0–100 cm soil layers significantly increased with increasing N application rates, ranging from 67.1 to 126.8 kg N ha⁻¹ in early rice and from 73.3 to 138.6 kg N ha⁻¹ in late rice (

Table 3. Average contents of residual mineral N in the 0–100 cm soil profile after rice harvest in 2020 and 2021.

Treatments		0–20 cm (kg N ha−1)		20–40 cm (kg N ha−1)		40–60 cm (kg N ha−1)		60–80 cm (kg N ha−1)		80–100 cm (kg N ha−1)		0–100 cm (kg N ha−1)
		NH4+	NO3−	NH4+	NO3−	NH4+	NO3−	NH4+	NO3−	NH4+	NO3−	NH4+ + NO3−
Early rice	N0	21.0 ± 1.7 d	14.3 ± 1.6 d	10.4 ± 0.6 c	6.6 ± 0.5 d	3.4 ± 0.6 b	2.2 ± 0.3 b	3.7 ± 0.4 a	1.7 ± 0.1 d	2.9 ± 0.2 b	1.0 ± 0.2 b	67.1 ± 3.2 f
	N90	26.0 ± 2.3 c	16.3 ± 1.4 d	13.8 ± 1.2 b	6.9 ± 0.4 cd	4.0 ± 0.6 a	2.7 ± 0.4 ab	4.4 ± 0.5 a	2.1 ± 0.1 c	3.4 ± 0.3 a	1.2 ± 0.2 b	80.8 ± 3.3 e
	N120	28.8 ± 3.5 c	23.9 ± 3.1 c	15.3 ± 1.4 a b	7.7 ± 0.5 c	3.5 ± 0.4 b	2.9 ± 0.4 a b	3.8 ± 0.3 a	2.2 ± 0.2 c	2.9 ± 0.3 b	1.0 ± 0.1 b	91.9 ± 2.1 d
	N150	34.5 ± 1.9 b	31.2 ± 2.2 b	15.7 ± 1.3 ab	8.7 ± 0.7 b c	3.7 ± 0.6 ab	3.1 ± 0.4 a	3.8 ± 0.4 a	2.4 ± 0.1 b c	2.9 ± 0.2 b	1.2 ± 0.3 b	106.9 ± 5.4 c
	N180	38.3 ± 2.2 ab	34.5 ± 2.4 ab	15.6 ± 1.1 ab	9.0 ± 0.4 b	3.9 ± 0.4 a	3.2 ± 0.4 a	4.1 ± 0.3 a	2.6 ± 0.2 a b	3.1 ± 0.2 ab	1.4 ± 0.1 ab	116.0 ± 3.3 b
	N210	40.7 ± 2.4 a	38.6 ± 2.8 a	17.3 ± 2.0 a	10.7 ± 0.6 a	4.2 ± 0.4 a	3.3 ± 0.3 a	4.2 ± 0.1 a	2.9 ± 0.3 a	3.2 ± 0.2 ab	1.7 ± 0.2 a	126.8 ± 4.5 a
Late rice	N0	22.4 ± 2.7 d	12.9 ± 1.7 d	10.9 ± 0.5 d	7.3 ± 0.4 d	4.3 ± 0.6 b	2.8 ± 0.4 c	4.5 ± 0.5 b	2.1 ± 0.1 d	3.5 ± 0.2 b	1.3 ± 0.2 c	73.3 ± 4.4 e
	N120	28.9 ± 3.3 c	16.5 ± 2.0 d	13.8 ± 0.5 c	9.3 ± 1.0 c	5.5 ± 0.6 ab	3.4 ± 0.5 bc	5.6 ± 0.6 a	2.6 ± 0.1 c	4.3 ± 0.2 a	1.6 ± 0.2 bc	97.3 ± 2.5 d
	N160	32.9 ± 3.1 bc	21.1 ± 3.0 c	14.5 ± 0.5 c	9.5 ± 1.0 c	4.9 ± 0.5 ab	3.7 ± 0.5 abc	4.7 ± 0.5 b	2.8 ± 0.1 b c	3.6 ± 0.2 a	1.3 ± 0.1 c	101.5 ± 2.3 d
	N200	35.0 ± 2.8 ab	31.5 ± 2.4 b	15.7 ± 0.5 bc	10.8 ± 1.0 c	5.1 ± 0.6 a	3.9 ± 0.5 ab	4.9 ± 0.5 a b	3.1 ± 0.2 b	3.8 ± 0.3 b	1.4 ± 0.2 c	118.2 ± 3.7 c
	N240	37.3 ± 3.0 ab	34.8 ± 2.5 ab	17.5 ± 1.7 a b	13.0 ± 0.8 b	5.5 ± 0.4 a	4.3 ± 0.5 ab	5.3 ± 0.6 a	3.4 ± 0.2 a	4.0 ± 0.3 a b	1.8 ± 0.2 ab	128.5 ± 4.0 b
	N280	39.9 ± 1.3 a	37.5 ± 1.9 a	18.8 ± 1.6 a	14.9 ± 0.8 a	5.8 ± 0.4 a	4.6 ± 0.5 a	5.7 ± 0.5 a	3.7 ± 0.2 a	4.3 ± 0.3 a	2.0 ± 0.2 a	138.6 ± 3.9 a

Note: The data in each column represents the average value plus the standard deviation (n = 3). Values denoted by different lowercase letters indicate statistically significant differences between treatments for each cultivated season (p < 0.05).

). Soil residual N contents were mainly dominant in the 0–40 cm layer, which accounted for 81.9% and 78.4% of the soil total residual mineral N in the early and late rice seasons, respectively. The soil NH₄⁺ contents were higher than those of NO₃⁻ in the different soil layers. Similarly, the residual ¹⁵N derived from the ¹⁵N fertilizer also increased with increasing N application rates, which mainly dominated in the 0–40 cm layer (Figure 3).

3.5. Rice ¹⁵N Uptake Derived from Fertilizer and Soil N Sources

The ¹⁵N uptake in the above-ground parts ranged from 23.3 to 38.9 kg ha⁻¹ for early rice and from 16.2 to 29.9 kg ha⁻¹ for late rice (

Table 4. Uptake of ¹⁵N fertilizer in the aboveg-round parts of early and late rice in 2021.

Treatments		Nuptake	Ndff	Ndfs	N Residue Rate	Nrecovery Efficiency	N Loss Rate	Ndff Contribution	Ndfs Contribution
		kg N hm−2			%
Early rice	N0	125.5 ± 4.4 b	--	125.5 ± 4.4 b	--	--	--	--	--
	N90	197.9 ± 7.8 a	23.3 ± 0.6 c	174.6 ± 2.9 a	25.2 ± 1.7 a	11.8 ± 0.5 c	63.0 ± 3.4 a	19.4 ± 0.6 a	80.6 ± 0.6 a
	N120	205.3 ± 4.1 a	28.8 ± 0.3 b	176.5 ± 2.9 a	21.7 ± 1.3 a	14.0 ± 0.8 b	64.3 ± 4.1 a	18.0 ± 2.1 a	82.0 ± 2.1 a
	N150	209.5 ± 7.6 a	36.1 ± 0.8 a	173.4 ± 4.6 a	25.2 ± 1.2 a	17.2 ± 1.0 a	57.6 ± 4.6 a	18.1 ± 0.8 a	82.0 ± 0.8 a
	N180	206.1 ± 9.8 a	38.9 ± 0.9 a	167.2 ± 5.3 a	25.0 ± 0.9 a	18.9 ± 1.5 a	56.1 ± 3.4 a	16.1 ± 1.0 a	83.8 ± 1.0 a
	N210	207.3 ± 8.7 a	35.6 ± 0.8 a	171.7 ± 2.8 a	24.2 ± 1.3 a	17.2 ± 3.3 a	58.6 ± 3.9 a	12.7 ± 2.6 b	87.3 ± 2.6 a
Late rice	N0	134.2 ± 2.7 d	-	134.2 ± 2.7 b	-	-		-	-
	N120	175.1 ± 4.8 c	16.2 ± 0.9 c	158.9 ± 3.8 a	35.2 ± 1.1 a	9.3 ± 0.5 c	55.6 ± 3.6 a	18.0 ± 0.9 a	82.0 ± 0.9 a
	N160	182.3 ± 2.6 bc	20.9 ± 0.6 b	161.4 ± 5.4 a	30.2 ± 1.0 b	11.5 ± 1.2 b	58.4 ± 4.1 a	17.4 ± 1.3 ab	82.6 ± 1.3 a
	N200	192.8 ± 5.4 ab	24.7 ± 1.2 ab	168.1 ± 5.3 a	29.6 ± 1.1 b	12.8 ± 1.6 ab	57.6 ± 3.1 a	16.5 ± 0.4 ab	83.5 ± 1.2 a
	N240	193.6 ± 6.9 ab	27.7 ± 1.6 a	165.9 ± 5.6 a	29.6 ± 1.6 b	14.3 ± 2.4 a	56.1 ± 2.9 a	15.4 ± 1.2 b	84.6 ± 1.4 a
	N280	209.5 ± 8.8 a	29.9 ± 1.5 a	179.6 ± 1.7 a	30.9 ± 0.8 b	14.3 ± 4.3 a	54.9 ± 3.6 a	14.2 ± 2.6 b	85.8 ± 2.6 a

Note: The data in each column represents the average value plus the standard deviation (n = 3); N_uptake represents the N uptake by the above-ground part measured at harvest; N_dff and N_dfs represent the amounts of N uptake of rice that derive from fertilizer and soil, respectively. Values denoted by different lowercase letters indicate statistically significant differences between treatments for each cultivated season (p < 0.05).

). Rice ¹⁵N uptake derived from fertilizer N and ¹⁵N recovery efficiency, in particular, increased with increasing N application rate but plateaued when the N application rate reached 150 and 200 kg ha⁻¹ in early rice and late rice, respectively. In contrast, rice ¹⁵N uptake in the above-ground parts was derived from soil N, and the ¹⁵N residual rate and ¹⁵N loss rate varied little among the N application rates (except for N₀ for N uptake derived from soil). Meanwhile, the ¹⁵N loss rate ranged from 56.1 to 63.0% for early rice and from 54.9 to 58.4% for early rice, and these values were significantly higher than the proportions that resided in the soil and were taken up by rice.

4. Discussion

4.1. Rice Yield and Nitrogen Use Efficiency

As one of the key elements influencing rice growth and development, an appropriate increase in N fertilizer can not only ensure high yield but can also promote plant N uptake and transport [30,31]. Based on the results of the National N fertilizer Cooperative Network, some authors demonstrated that optimized N management could achieve 10 to 50% N input reductions without sacrificing the grain yield [32]. Considering the yield effects of double-cropping rice and the fertilizer utilization efficiencies, Zhu demonstrated that the suitable N application rate should be between 105 kg N ha⁻¹ and 146 kg N ha⁻¹ [33]. In our two-year field experiment, rice grain yield and N uptake increased significantly when the N application rate was below 150 kg ha⁻¹ in early rice and 200kg ha⁻¹ in late rice; exceeding these values limited N transport to grains and yield formation. Our results are consistent with other previous results [33,34]. This may be attributed to excessive N input disrupting the physiological metabolism of crops, reducing N remobilization and the grain filling rate, and eventually reducing yield [35,36]. Qiu demonstrated that the relationship between the N application rate and grain yield followed a quadratic regression or a linear plateau model [37]. Our results showed that the recommended N application rates to produce the maximum grain yield and N residual rate described by the quadratic regression model were 163.4 and 209.2 kg ha⁻¹ and 197.6 and 236.4 kg ha⁻¹ in early and late rice, respectively, but those described by the linear plateau model were 118.5 and 169.9 kg ha⁻¹ and 152.8 and 186.2 kg ha⁻¹, respectively (Figure 4). Although there were differences between the optimized N application rates identified by the two models, no significant differences were observed between their yields in the two growing seasons. This indicates that 118.5 to 152.8 kg ha⁻¹ for early rice and 169.9 to 186.2 kg ha⁻¹ for late rice, calculated by the linear plateau model, can serve as the suitable N application rate to obtain a relatively high rice yield and low N loss in the double rice system.

In general, rice NRE, NAE, and NPFP decreased with increasing N application rates, according to the regression analyses. The average N recovery efficiency was 30.1% for early rice and 24.3% for late rice, which were lower than the average N fertilizer utilization rate of crops in China [14,38]. In this study, the grain yield and N uptake of N₀ were high, which might be attributed to an enormous amount of residual N in the cropland soils [21]. This N may be released in the form of mineralized organic N and absorbed by crops. In addition, the higher temperature and rain in the rice growing season were characterized by higher N volatilization and leaching, likely causing serious N fertilizer loss and a low N fertilizer utilization rate. Our study also demonstrated that more than 27.5% of the applied N was left in the soil in the early rice season and then used in the late rice season, which is consistent with previous conclusions [32,39,40]. On the other hand, N surplus in soil correspondingly increases N loss via ammonia volatilization, leaching, and denitrification, which have a strong impact on the agricultural ecological environment and decrease N recovery. These results partly explain the conclusion that rice NRE, NAE, and NPFP in early rice were significantly higher than those in late rice. However, some authors demonstrated that the relationship between the N application rate and REN or AEN changed, with a general trend of increasing initially and declining subsequently [37,41]. We speculated that the variation may be attributed to the differences in N application rate, environment, and soil fertility.

4.2. N Fate in the Soil-Rice System

Nowadays, the top priority of N balance research is ensuring higher N uptake and N use efficiency of crops while reducing N fertilizer input. The N budget analysis can quantitatively reflect the N input and output and the N surplus or deficit, as well as N utilization and loss in the soil-rice systems, which can provide theoretical guidance for the development of N fertilizer management practices [15]. Our results demonstrated that, in addition to N fertilizer, the amounts of soil residual N and mineralized N also served as important N sources for rice growth, accounting for more than 58.0% and 53.2% of the total N input in the early and late rice seasons, respectively. Most of the N was absorbed by rice, ranging from 43.7% to 55.6% in early rice and from 36.8% to 54.7% in late rice. These results were partly consistent with some related studies on N balance in double-cropping rice fields [42]. Rice N uptake showed an upward trend with an increase in the N application rate but reached a plateau when the N application rate reached 124.2 and 141.8 kg ha⁻¹ in early rice and late rice, respectively. This indicated that the optimal N application rate for early rice and late rice should not be less than 124.2–141.8 kg ha⁻¹. If the N application rate is less than these values, N would be deficient, whereas a continuous increase in the N application rate did not result in significant luxury N uptake.

In the single rice system, the proportions of fertilizer N absorbed by rice, remaining in soil, and being lost accounted for 21.7 to 33.0%, 5.3 to 16.0%, and 56.0 to 72.0%, respectively, of the N fertilizer applied in paddy fields [43,44]. Similarly, our study showed that the apparent N recovery efficiency of rice was 23.5 to 34.4% in early rice and 32.8 to 37.3% in late rice. However, the ¹⁵N recovery rates were 11.8 to 18.9% and 9.3 to 14.3% for the early and late rice, respectively, which were significantly lower than those calculated by the subtraction method in our study and a previous study [45]. One of the main reasons for this is the “priming effect” of N fertilizer, as plants given N absorbed more soil N than those not given N [46]. Cao demonstrated that the cumulative absorption rate of residual ¹⁵N was 8.1 to 9.3% in the late crop season [47]. Taking into account the soil residual N effect, the actual utilization rate of ¹⁵N fertilizer in the late rice season ranged from 21.7% to 26.7%, which approaches that calculated by the subtraction method. However, the apparent loss rates were 74.2 to 87.0% in the early rice season and 71.5 to 92.1% in the late rice season, which were significantly higher than the previous results. Wang [13] hypothesized that the main loss method for fertilizer N was ammonia volatilization due to the high N application rate and N surplus. The higher temperature and rainfall amount in the rice-growing season are beneficial for soil ammonia volatilization, thus causing the relative N loss rate. This different conclusion may be because the fate and loss pathways of N fertilizer are affected by many factors, such as soil type, climate, crop variety, and fertilization management [48]. However, the fertilizer N contribution to rice and the ¹⁵N loss rate varied little with the N application rates in both the early and late rice seasons. The proportion of N absorbed by rice from soil in this study ranged from 80.6% to 87.3% for early rice and from 82.0% to 85.8% for late rice, which partly explains the above conclusion. The results indicate that rice N management should fully consider the soil N supply capacity to determine the appropriate amount of N fertilizer to ensure higher N use efficiency, specifically for a double-cropping rice system.

4.3. Nitrogen Accumulation in Soil

When excessive N fertilizer accumulates in the soil profile, part of it may enter the environment through leaching, runoff, NH₃ volatilization, nitrification, or denitrification [49,50,51]. Fan demonstrated that reducing nitrogen by 20~25% below the farmer’s conventional application amount, such as 120 kg N ha⁻¹ for early rice and 135 kg N ha⁻¹ for late rice, could not only maintain the higher N absorption and N use efficiency in double-cropping rice but also reduce the N loss and surplus [2]. The ¹⁵N residual rate in the 0–100 cm soil profile was approximately 21.7 to 25.2% for early rice and 29.6 to 35.2% for late rice. Soil residual ¹⁵N was mainly retained in the 0–40 cm soil layer, accounting for 65.9% and 62.6% of the soil total residual N in the early and late rice seasons, respectively. Their NH₄⁺ contents were significantly higher than their NO₃⁻ contents. This conclusion is basically consistent with previous results [48,52], indicating that the residual N from fertilizer mainly exists in the soil tillage layer. In contrast, residual soil N below the plough layer accounted for approximately 20%, and residual N decreased significantly with increasing soil depth. The amount of residual ¹⁵N in the 40–100 cm soil layer was less, indicating that less ¹⁵N leached downwards and that there was no runoff loss of N in the microzone experiment. Therefore, it was speculated that the main loss pathway of residual N was gaseous loss dominated by ammonia volatilization and nitrification-denitrification. Due to technical limitations, this study failed to quantitatively monitor the loss of soil N due to nitrification and denitrification. In the future, the specific loss and proportion of residual N in double-cropping rice fields need to be studied.

5. Conclusions

According to the N balance in the soil-rice system, soil residual N and mineralized N also serve as important N sources for rice in addition to fertilizer N. The ¹⁵N labeling experiment showed a similar conclusion, demonstrating that more than 80% of rice N uptake was derived from soil N sources in both the early and late rice seasons. Most of the N input was absorbed by rice, ranging from 43.7% to 55.6% in early rice and from 36.8% to 54.7% in late rice. The proportions of soil residual N and N loss were 25.4 to 30.8% and 24.7 to 27.2% in the early rice season and 7.2 to 26.2% and 7.7 to 38.5% in the late rice season, respectively. Rice N use efficiency decreased significantly as N application increased. However, rice grain yield and N uptake showed an upward trend with an increase in the N application rate but reached a plateau when the N application rate reached 150 and 200 kg ha⁻¹ in early rice and late rice, respectively. Regression analysis demonstrated that 118.5 to 152.8 kg ha⁻¹ for early rice and 169.9 to 186.2 kg ha⁻¹ for late rice, calculated by the linear plateau model, can serve as a suitable N application rate in the double rice system to maintain a relatively high rice yield and N use efficiency and a low soil N supply. Our results provide theoretical guidance for improving N management in double-cropping rice fields in southern China.