Eco-physiology and environmental impacts of newly developed 1 rice genotypes for improved yield and nitrogen use efficiency 2 coordinately

58 Significant advancements have been made in understanding the genetic regulation of 59 nitrogen use efficiency (NUE) and identifying crucial NUE genes in rice. However, the 60 development of rice genotypes that simultaneously exhibit high yield and NUE has 61 lagged behind these theoretical advancements. The grain yield, NUE, and greenhouse 62 gas (GHG) emissions of newly-bred rice genotypes under reduced nitrogen application 63 remain largely unknown. To address this knowledge gap, field experiments were 64 conducted, involving 80 indica (14 to 19 rice genotypes each year in Wuxue, Hubei) 65 and 12 japonica (8 to 12 rice genotypes each year in Yangzhou, Jiangsu). Yield, NUE, 66 agronomy, and soil parameters were assessed, and climate data were recorded. The 67 experiments aimed to assess genotypic variations in yield and NUE among these 68 genotypes and to investigate the eco-physiological basis and environmental impacts 69 of coordinating high yield and high NUE. The results showed significant variations in 70 yield and NUE among the genotypes, with 47 genotypes classified as moderate-high 71 yield with high NUE (MHY_HNUE). These genotypes demonstrated the higher yields 72 and NUE levels, with 9.6 t ha -1 , 54.4 kg kg -1 , 108.1 kg kg -1 , and 64% for yield, NUE for 73 grain and biomass production, and N harvest index, respectively. Nitrogen uptake and 74 tissue concentration were key drivers of the relationship between yield and NUE, 75 particularly N uptake at heading and N concentrations in both straw and grain at 76 maturity. Increase in pre-anthesis temperature consistently lowered yield and NUE. 77 Genotypes within the MHY_HNUE group exhibited higher methane emissions but 78 lower nitrous oxide emissions compared to those in the low to middle yield and NUE 79 group, resulting in a 12.8% reduction in the yield-scaled greenhouse gas balance. In 80 conclusion, prioritizing crop breeding efforts on yield and resource use efficiency, as 81 well as developing genotypes resilient to high temperatures with lower GHGs, can 82 mitigate planetary warming.


Graphical Abstract
Highlights: • Limited knowledge on yield, NUE, and GHG emissions in new rice genotypes with reduced N application • Genotypes showed significant yield and NUE variations, with 47 classified as MHY_HNUE (moderate-high yield, high NUE) • N uptake at heading and N concentrations in straw and grain driven the relationship between yield and NUE • Increase in pre-anthesis temperature consistently lowered yield and NUE • MHY_HNUE genotypes exhibited 12.8% lower yield-scaled greenhouse gas balance Highlights

Introduction
Nitrogen (N) is a vital component for life and is required for the survival of all organisms (Harrison et al., 2009;Christie et al., 2020).Rice, as one of the staple crops that play a crucial role in global food security (Peng et al., 2004;Zhu et al., 2018), has seen a substantial increase in yield through the application of N fertilizer.However, the excessive use of N fertilizer exacerbates environmental problems, including soil acidification, air pollution, ozone depletion, global warming and water eutrophication (Galloway et al., 2008;Van Grinsven et al., 2013;X. Chen et al., 2014;Cui et al., 2014).
The challenge of feeding a rapidly growing global population sustainably, equitably, and in a climate-resilient manner without degrading natural resources requires innovative approaches that balance yield improvement and environmental protection (Alexandratos and Bruinsma., 2012;Harrison et al., 2021).
Expanding cropping areas can lead to increased food production, but it may come at a cost to other commodities or environmental conservation (Harrison et al., 2021).On the other hand, intensifying land use can also increase food supply (Peng et al., 2009;Wang and Peng., 2017), but it must be done in a sustainable manner (Muleke et al., 2022).Over the past 60 years, rice production in China has doubled due to the development of semidwarf and hybrid rice genotypes (Peng et al., 2009;Wang and Peng., 2017), with potential yield benefits of up to 50% from genotypes with semidwarfism and heterosis (Yuan, 2017).The recent emergence of contemporary genotypes with larger sinks, such as greater panicle size, fewer tillers, higher grain filling, and greater grain weight has led to higher yields (Yang andZhang, 2010a, 2010b).However, the adoption of practices that enable higher yields, such as excessive N fertilizer use, often overlooks the potential environmental consequences of surplus or unused nitrogen (Christie et al., 2020).
Numerous studies have shown that N use efficiency (NUE) tends to decrease with increasing N fertilizer application (Huang et al., 2019;Peng et al., 2010;Zhang et al., 2013), although the magnitude and nature of N losses to environment depend on the timing, rate, type, and method of N application (Smith et al., 2021).To mitigate potential effects of excessive N application on social, economic, human and environmental capital (Shahpari et al., 2021), a wide range of research has been conducted since the end of 20 th century (Fan et al., 2016;Xu et al., 2012;Zhang, 2007).
Compared with status quo farmers' practices, reduced N application through real-time N management and site-specific N management has shown increased yield and NUE simultaneously across a number of rice production sites (Cassman et al., 1996;Peng et al., 2006Peng et al., , 2010)).At the same time, evaluation of rice genotypic variation in NUE has also been conducted and significant differences have been reported for both indica and japonica genotypes in different parts of the world, although NUE has never been set as a breeding target in rice breeding (Broadbent et al., 1987;Hamaoka et al., 2013;Koutroubas and Ntanos, 2003;Wu et al., 2016).
NUE can be separated into component indexes, including N uptake, transportation and utilization (Peng et al., 2006).NUE for grain production (NUEg, kg grain yield over total N uptake) and partial factor productivity of applied N (PFP, kg grain per kg N applied) are often applied in genotypic comparisons.Recently, our understanding of genetic regulation in N uptake, assimilation, signaling, and utilization has been greatly increased, and strategies for improving crop NUE have also been proposed (Xu and Takahashi, 2020;Yu et al., 2022).In rice, indica and japonica exhibit differential nitrateabsorption activity, with indica genotypes having higher values (Hu et al., 2015).Field evaluation using near-isogenic and transgenic lines demonstrated that the japonica variety carrying the NRT1.1B-indicaallele had higher yield and NUE compared with the wildtype (Hu et al., 2015).Sun et al. (2014) reported that DEP1 regulated panicle size and NUE simultaneously, and therefore could coordinate high yield and high NUE in rice.Liu et al. (2021) found that introgression of the OsTCP19 allele associated with a high tillering response to nitrogen in modern rice genotypes boosts grain yield and NUE under low or moderate levels of N application.Despite these breakthroughs in theoretical studies of rice NUE, breeding rice genotypes for coordinated high yield and high NUE is still at the preliminary stage.
Management and genotype combinations that are contextualized to particular environments can advance yield and NUE of rice (Ibrahim et al, 2019); notwithstanding that such systems should account for potential changes in climate warming or variability (B.Wang et al., 2020;Wang et al., 2018;W. Wang et al., 2020).Extreme weather events in which are potentially damaging to global agricultural production are increasingly common in many regions of the world (Harrison, 2021).In 2022, ambient temperatures above 38 ℃ lasted for over two weeks at booting and heading stages in main rice growing areas of China.Rice plants are highly sensitive to changes in climate conditions (Ziska and Bunce, 2007;Ziska et al., 2018).Global warming during different rice growing stages tends to decrease N uptake with an increase in the tissues' N content (Xiong et al., 2017;Wang et al., 2018).Elevated temperature decreased N recovery efficiency, N agronomic efficiency, and N physiological efficiency in early rice due to a reduction in grain yield caused by extreme temperatures, while warming increased plant N uptake and NUE in late rice as no heat stress existed (Wang et al., 2020;Zhang et al., 2022).Zhang et al. (2022) found that a previously unidentified allele of the nitrate transporter gene OsNRT2.3was required to maintain high yield and high NUE in rice under high temperatures.In addition, the heterogeneity of minimum and maximum temperature on rice yield might confer different NUE by affecting N uptake and biomass accumulation (Peng et al., 2004).
Developing green super rice (GSR) was proposed in 2007, and one aspect of the project was to improve N use efficiency while maintaining a high yield potential (Zhang, 2007).
To date, a number of new genotypes have been released, but there is still not a good understanding of the relationship between yield and NUE in rice and its ecophysiological basis.Moreover, it has been found that high-yielding rice genotypes show significantly decreased CH4 emissions, especially under continuous flooding (Jiang et al., 2017(Jiang et al., ,2019)).In double cropping systems, N2O emissions were consistently and negatively correlated with N agronomy efficiency (NAE), but no clear relationship between CH4 emissions and NAE was observed under optimal N management (Yu et al., 2021).To date, it is unclear whether improvement in NUE of different rice genotypes could affect GHG emissions.Therefore, this study aimed to understand the eco-physiological mechanisms relating to the interplay between yield and NUE and the potential environmental benefits of increased cropping with high NUE genotypes.
Nearly one hundred rice genotypes, including 80 indica rice genotypes were planted during 2014-2018 in Hubei province and 12 japonica rice genotypes during 2017-2019 in Jiangsu province, both key rice production zones in.The objectives were to: (1) determine the genotypic variations in yield and NUE among these newly-bred rice genotypes, (2) examine the eco-physiological mechanisms underlying high yield and high NUE, and (3) verify whether high yield and high NUE genotypes exhibited significantly decreased GHG emissions.

Experiment sites
In this study, Hubei and Jiangsu provinces were selected as representative areas for single-season indica and japonica rice (Fig. 1).The growing area and rice production in these provinces are 21% and 24% of the total amount in China's single-season rice growing system, respectively.The average solar radiation of Yangzhou experiment sites was higher than that of Wuxue, but the temperature and precipitation indices in Yangzhou were relatively lower than that in Wuxue during the rice growing (Fig. 1 and Supplementary Table S1).The soils of Wuxue were clay loam with a pH of 4.6-5.6,organic matter (OM) of 26.7-33.5 g kg −1 , total N (TN) of 1.83-3.01g kg −1 , available P (OP) of 4.91-33.5 mg kg −1 and available K (AK) of 105.8-176.6 mg kg −1 (Supplementary Table S2).Soils in Yangzhou were of sandy loam texture with pH of 6.0, OM of 25.5 g kg −1 , TN of 1.52 g kg −1 , OP of 34.7 mg kg −1 , and AK of 87.9 mg kg −1 on average across the experiment years.

Plant materials and experiment design
From 2014 to 2018, 14-19 newly released rice genotypes were selected each year as research subjects in Wuxue, Hubei province.Most genotypes were hybrid indica rice, and only ten rice genotypes were inbred rice (Chunliangyouyuehesimiao,Huanghuazhan,Huangguanghuazhan1hao,Huangsilizhan,Lvjian1,Lvjian8,Wushansimiao,Xiushui134,Yungeng29,Zhongzu14).Among these ten inbred rice genotypes, Lvjian1, Lvjian8, Xiushui134 and Yungeng29 were japonica rice.At the Yangzhou experiment site, 8-12 japonica rice genotypes were selected each year to evaluate their differences in yield and NUE from 2017 to 2019.Details of the selected rice genotypes were listed in the Supplementary Excel File.
The experiments in Wuxue and Yangzhou were arranged in a completely randomized block design with four or three replications, respectively.The seedlings were raised in a seedbed with a sowing date of May 23~26 and 28~29 in Wuxue and Yangzhou, respectively.After 25 days, seedlings were transplanted on June 17~20 and 22~23, respectively, with two seedlings per hill.The fertilizers were manually incorporated 1 day before transplanting for basal application (40 kg P ha −1 calcium superphosphate and 50 kg K ha −1 potassium chloride).Additional K was topdressed at panicle initiation stage at a rate of 50 kg ha −1 during the experimental period.N application of the farmer's practice is usually excessive, 180 kg ha -1 in Wuxue and 270 kg ha -1 in Yangzhou on average, respectively.To evaluate the performance of the yield and NUE under reduced N conditions, 100 kg ha -1 N fertilizer in Wuxue and 180 kg ha -1 N fertilizer in Yangzhou were adopted.The ratio of N (urea) application at basal, topdressings at midtillering, and panicle initiation to spikelet differentiation stage were 4:3:3.Pests and diseases were controlled using insecticides and fungicides.

Crop measurements
The phenology of the rice developmental stage at heading (HD) and physiological maturity (PM) were recorded.Twelve hills of rice plants were sampled from each plot.
Plant height and stem numbers were recorded.The plant samples were separated into leaf, stem, and panicle.The green leaf area was measured using a leaf area meter (LI-3100, LI-COR, Lincoln, NE, USA) and was expressed as the leaf area index (LAI).The specific leaf weight (SLW) was defined as the ratio of the leaf dry weight to the leaf area.The dry weight (DW) of different parts of a plant was determined after oven drying at 80 • C to a constant weight.
At PM stage, the panicles were hand-threshed, and the filled spikelets were separated from unfilled spikelets after submerging them in tap water.The empty spikelets were separated from the half-filled spikelets through winnowing.Three 30-g subsamples of filled spikelets, three 2-g subsamples of empty spikelets, and the total number of halffilled spikelets were obtained to quantify the number of spikelets per m 2 .The dry weights of rachis, filled, half-filled, and unfilled spikelets were determined after oven drying at 80 • C to constant weight.The spikelets per panicle and grain filling percentage were calculated.In addition, the grain yield was determined from a 5 m 2 area in each subplot and was adjusted to a standard moisture content of 0.14 g H2O g −1 fresh weight.
The tissue N concentration of each component at HD and PM was determined using an Elemental analyzer (Elementar Vario MAX CNS/CN, Elementar Trading Co., Ltd, Germany).The plant N accumulation at HD and PM was calculated as the sum of N in each of the organs.The N use efficiency for grain production (NUEg) was calculated as the grain yield per unit plant N accumulation.The N use efficiency in biomass production (NUEb) was determined as the ratio of biomass production to plant N accumulation.The N harvest index (NHI) was calculated as the percentage of accumulated N in grain to plant N accumulation.The partial factor productivity of applied N (PFP) is commonly expressed as yield per unit N input (Peng et al., 2006(Peng et al., , 2010)).

Environmental effects estimation
Estimation of the nitrous oxide (N2O) emissions was based on the empirical equations from the IPCC's Good Practice Guidance 2006 methodology (Eggleston et al., 2006), which has been widely applied (Aliyu et al., 2019;X. Yu et al., 2021;Yuan et al., 2021).
The direct emissions of N2O from agricultural fields were quantified utilizing a correlation established between N surplus and N2O emissions (Zou et al., 2010).N surplus was determined as the difference between the aboveground N uptake of the crop, including both inorganic and organic sources, and the rate of N application (Van Groenigen et al., 2010).The total emissions of N2O were computed as the aggregation of direct and indirect N2O emissions, with the latter being approximated to constitute 20% of the direct emissions (IPCC 2019).The detailed equation for estimation of N2O is as follows: The CH4MOD model was used to simulate CH4 emissions of rice genotypes.This daily step-based, semi-empirical model can simulate CH4 production and emissions in paddy fields under various conditions and agricultural practices (Huang, 2004;Huang et al., 1998) and is recommended by the IPCC (Eggleston et al., 2006).The model includes two sub-models: one for simulating the production of methanogenic substrates from root exudation and organic matter, and another for simulating CH4 production and emissions through rice plants and bubbles.It has been extensively validated and used worldwide, with minimal input data required (Bogner et al., 2000;Jiang et al., 2023).
The main input parameters for the CH4MOD model include daily air temperature, soil sand percentage, phenology of rice, organic matter addition, rice grain yield, and water management patterns, which were recorded during experiment periods.As we did not know which emission types of these rice genotypes, sensitivity analysis was empolyed to address the uncertainty and refine the simulation of CH4 emissions.In this process, we adopted a step size of 0.01, ranging from 0.5 to 1.5 of the variety index.The average simulated values were then used as representative CH4 emissions for the specific rice genotype.

Statistical Analysis
K-means clustering was initially used to clarify the rice genotypes into six groups using the package factoextra (version 1.0.7).Furthermore, to understand the relationship between yield and the NUE of the rice populations at a coarse level, the yield, NUEg, NUEb, NHI, and PFP were employed as proxies to construct a hierarchical cluster heatmap using the package pheatmap with default settings (version 1.0.12).The columns were annotated by group number (1-6), categorized by K-means clustering and the subgroups based on the variated trend of the yield and NUE among groups.
Co-linearity among independent variables will influence the stability of the final results in regression models.Because strong collinearity occurred among particular growth properties, N properties, and environmental factors (weather and soil properties), we used cluster analysis to assess the collinearity or redundancy of environmental variables by the varclus procedure in the Hmisc R package before further analyses (version 4.6.0).In addition, principal component analysis (PCA) was conducted for each of the four type variables using the FactoMineR and factoextra R packages (version 1.0.7).The important variables with low collinearity were selected by setting the threshold of pearson R 2 at 0.6 and according to loading values on the PCA dimensions.
After the variable selection procedure, we used the varpart function in vegan package (version 2.5.7) to estimate the proportion of variation of yield and NUE that was explained uniquely by the effect of growth properties, N properties and environmental factors (climate and soil properties).Non-metric multidimensional scaling (NMDS) using Bray-Curtis dissimilarity of Yield, NUE, and biotic or abiotic variables were performed with the vegan package.The Mantel test (Rossi, 1996) was carried out using the "mantel_test" function in the LinkET R package (version 0.0.3.6) to evaluate relationships among Yield, NUE, and the selected biotic and abiotic variables.
Similarities or dissimilarities in yield and NUE were calculated by the Euclidean distance via the "vegdist" function in the vegan R package, and other biotic and abiotic variables distance were also calculated by the Euclidean distance based on the matrix of the measured variable.
Co-occurrence networks were constructed for different classifications of yield and NUE groups by pearson correlations using the "corr.test"function in the psych R package (version 2.1.9).A correlation was considered statistically robust between two items with pearson's correlation coefficient (ρ) > 0.6 and the p-value <0.05 (Xu et al., 2021).
The network analysis was conducted with psych package and visualized in Gephi 0.9.2 based on the Fruchterman-Reingold algorithm (Bastian et al., 2009).The network parameters were extracted, including nodes, edges, degree, eigenvector centrality, complexity (linkage density; degree/node), diameter, transitivity, and modularity.
All data were analyzed with R software (http://www.r-project.org/,version 4.1.2).For statistical analysis, the data were first tested for normality using Kolmogorov-Smirnov method, and then were subjected to ANOVA for multiple sets of data for pairwise comparisons.A backward stepwise regression linear model was constructed to investigate the effects of biotic and abiotic variables on yield and NUE indices among the rice genotypes across year and experiment sites.Histograms, boxplots, and forest figures were generated using the package ggplot2 (version 3.3.6).

Yield, NUE, and biotic variables
The mean values of the yield and its components for these rice genotypes were as follows: 9.1 t ha -1 for yield, 224 for panicles per unit area, 200.4 for spikelets per panicle, 43.9  10 3 for spikelets per unit area, 81.3% for grain filling percentage, and 24.9 mg for individual grain weight, with the coefficient of variation (CV%) ranging from 10.7 to 20.5% (Supplementary Table S3).The parameters for NUE varied from the lowest value of 49.0 kg kg -1 for NUEg to the highest value of 100.3 kg kg -1 for NUEb, while the coefficient of variation (CV%) of partial factor productivity (PFP) was the highest among the indicators, exceeding 20.0%.At the HD stage, the average plant height was 128.5 cm, the number of stems per unit area (SN) was 254.2, the aboveground biomass (TDW) was 11.3 t ha -1 , the leaf area index (LAI) was 7.0 m 2 m -2 , the specific leaf weight (SLW) was 44.2 mg cm -2 , and the crop growth rate (CGR) was 18.3 g m -2 d -1 , the coefficient of variation (CV%) for all of these parameters was between 10.7% and 16.9%.At the PM stage, the plant height and CGR decreased compared to the values observed at the HD stage, but the TDW increased significantly.The coefficient of variation (CV%) for these parameters was around 10.0% and did not exceed 13.0%.At the HD stage, the N concentration was higher in the leaf tissues, followed by the panicle and stem organs.However, at the PM stage, the N concentration in these organs was lower compared to that at the HD stage.The CV% for N concentration indicators was mostly above 13.0%,although the CV% for N concentration in filled grains was 8.7%.The average N uptake at the HD and PM stages was 159.2 and 181.5 kg ha -1 , respectively, with a relatively larger CV% for N uptake at the HD stage.
Histograms showed that the parameters had a partially skewed distribution (Supplementary Fig. S1).To further analyze the datasets, the scaled method was used to adjust the dataset to a normal distribution.After this adjustment, the values of the Kolmogorov-Smirnov test range were mostly larger than 0.05, except for PFP, HD.SLW, HD.CGR, and HD.LeafN (Supplementary Table S3).

Yield and NUE variation among groups
Using K-means clustering, we categorized genotypes into six groups (Fig. 2 and Supplementary Fig. S2).The average yield of genotypes in group 1 was the highest at 10 t ha -1 , followed by genotypes in groups 5, 3, and 6, which had median yield levels.
The yield of genotypes in group 4 was the lowest.The NUEg of genotypes in groups 1 and 6 was the largest, followed by genotypes in groups 3, 5, and 2. NUEg of group 4 was the lowest of all groups.The variation pattern of NUEb and NHI between groups was similar to the variation in NUEg.PFP was calculated from yield values divided by the N application rate.Therefore, the varied characteristics were in line with the variation of the yield among groups, except that the value of genotypes in group 6 was the lowest, attributable to its N level.
After analyzing the variation pattern in yield and NUE indices among groups, we divided them into two distinct categories based on their change gradient.Groups 1, 3, and 6 were classified as having a middle to high yield and NUE level (MHY_HNUE), while groups 2, 4, and 5 were categorized as having a low to middle yield and NUE level (LMY_LMNUE) (Fig. 3).The MHY_HNUE group had significantly higher yield and NUE indices than the LMY_LMNUE group.Details of the clarified groups of the rice genotypes were shown in Supplementary Fig. S3.The hierarchical cluster results agreed with the K-means cluster findings, which enhanced the robustness of the classification results.The linear correlation analysis revealed a strong and positive association between yield and NUEg, NUEb, NHI, and PFP for the LMY_LMNUE group.
In contrast, for the MHY_HNUE group, only a significant correlation was observed between PFP and yield (Fig. 4).

Co-occurrence network and its stability
Yield and NUE co-occurrence networks were constructed based on pearson correlations among rice genotypes of different groups to investigate rice genotypes' interconnections along the gradient of variation in Yield and NUE (Fig. 5A).We found that node connectedness (degree), and centrality (closeness, eigenvector, and betweenness) network nodes, increased significantly by 110.9%, 21.9%, 7.1%, and 12.8%, respectively, in the LMY_LMNUE group, compared to MHY_HNUE group (Fig. 5B).Increases in these properties for nodes suggested lower network stability.This was confirmed by the changes observed in the properties characterizing the overall network structure of the LMY_LMNUE and MHY_HNUE groups, which displayed a 14.3% decrease in diameter and a 78.1% decrease in modularity with LMY_LMNUE (Fig. 5B).

Relationship of yield and NUE similarities over different variables properties.
To ensure that the effects of multicollinearity were avoided, a pearson correlation cluster was utilized to preselect the parameters for analysis.As shown in Supplementary Fig. S4 and S5, the selected parameters included yield components such as panicles per unit area, grain filling percentage, grain weight, and spikelets per panicle, growth properties such as LAI, CGR, SLW at the HD stage, and CGR, TDW, and plant height at the PM stage, N properties such as N concentration of organs at the HD stage, stem N, leaf N, and filled grain N, as well as N uptake at the PM stage.
Additionally, climate properties such as minimum and maximum temperature, solar radiation, and precipitation during pre-and post-heading stages were considered, as well as soil properties such as OP, AK, pH, and TN.The results of non-metric multidimensional scaling showed that the selected parameters could effectively classify the groups of MHY_HNUE and LMY_LMNUE (Supplementary Fig. S6).
Correlation analysis indicated that yield was linearly and positively related to the grain filling percentage in the LMY_LMNUE category, and was further significantly and positively influenced by spikelets per panicle in the MHY_HNUE category or the combined group (Supplementary Fig. S7).
To better understand the drivers of these relationships between yield, NUE, and biotic and abiotic factors distance, we conducted a Mantel test (Fig. 6).Our results revealed that specific growth properties (with the exception of HD.SN), N properties, climate elements, and soil elements were responsible for the significant changes in the relationships between yield and NUE similarities over biotic and abiotic factors distance.Furthermore, PCA was used to character the individual parameter effect (Fig. 7).The results showed that growth parameters significantly loaded on the PCA1 and PCA2 axes, explaining 37.7% and 26.4% of the variation in the changes of yield and NUE, respectively.N properties on PCA1 and PCA2 explained 41.6% and 23.0% of the changing pattern of yield and NUE, with most parameters significantly loaded on PCA1.
With respect to climate variables, most loaded considerably on PCA1, while the minimum and maximum temperature during pre-heading stage loaded on PCA2.Soil elements significantly loaded on PCA1, while OP loaded on the PCA2 axis, with PCA1 and PCA2 explaining 56.7% and 24.5% of the variation, respectively.

Biotic and abiotic factors between groups of MHY_HNUE and LMY_LMNUE
ANOVA tests between the newly classified groups of MHY_HNUE showed that the grain filling percentage and grain weight of yield components were significantly higher than the LMY_LMNUE (Table S4).The CGR at the HD stage and TDW at the PM stage were relatively higher in MHY_HNUE, compared to LMY_LMNUE, but the plant height was lower.The N concentration of organs in MHY_HNUE at both the HD and PM stages consistently showed lower values than those in LMY_LMNUE.Significant differences were also observed between the two groups in the minimum and maximum temperatures during pre-and post-heading stages, with MHY_HNUE exhibiting lower values than LMY_LMNUE.Lastly, it was found that soil properties such as TN and AK were lower in MHY_HNUE than in LMY_LMNUE, while pH was higher.

Effects of biotic and abiotic factors on yield and NUE
The variance partition analysis revealed that the N properties indices accounted for 16% of the variation in both yield and NUE, whereas growth properties, climate, and soil elements collectively explained a similar proportion of about 2% each (Supplementary Fig. S8).Backward stepwise-regression models, including all the growth, N, climate, and soil predictors, explained 69.8%, 89.2%, 99.1%, 83.4%, and 92.3% (adjusted R 2 ) of the total variances observed in yield, NUEg, NUEb, NHI, and PFP, respectively (Fig. 8).For yield, panicle N concentration at HD stage, N uptake at PM stage, the minimum temperature at the post-heading stage, and OP and TN had positive effects on yield formation, but N concentration of organs at PM stage and maximum temperature at the pre-heading stage, and AK had adverse impacts on yield variation.Regarding NUEg, plant height at the PM stage, and N concentration of stem at the HD stage, the maximum temperature at the post-heading stage and OP had positive effects, N concentration and N uptake of organs, and the minimum temperature at the post-heading stage had adverse effects.Our observations revealed that several factors negatively impacted NUEb, including SLW at HD, N uptake at PM, the minimum temperature during pre-and post-heading stages, and AK.Conversely, N concentration of organs at HD and PM (except for the N concentration of the stem and leaf at PM stage), TDW, plant height at PM, and maximum temperature during pre-heading had positive and significant contributions to the formation of NUEb.In terms of NHI, SLW at HD, N concentration of the stem and leaf at PM, maximum temperature at the pre-heading stage, and solar radiation at the PM stage had negative effects.However, N concentration of the panicle and stem, N uptake, maximum temperature at the post-heading stage, pH, and organic phosphorus (OP) positively affected NHI.The factors affecting PFP were similar to those included in the yield regression model.The effects of the concentration of organs at HD and PM stages, maximum temperature at the pre-heading stage, and N uptake were compatible with their impact on rice yield.Additionally, we observed that maximum temperature at post-heading stage had positive effects on PFP, which was opposite to the effects of minimum temperature.

Environmental effects
To address the uncertainty regarding the emission types of the rice genotypes, we conducted a sensitivity analysis to refine the simulation of CH4 emissions.Supplementary Fig. S9 illustrates that the mean values of simulated CH4 emissions were higher for individual rice genotypes in the MHY_HNUE group than those in the LMY_LMNUE group.Consequently, we observed a significant increase in the simulated CH4 emissions of the MHY_HNUE group.However, the N2O emissions in the MHY_HNUE group showed a remarkable decrease relative to those in the LMY_LMNUE group (Fig. 9).As a result, there were no noticeable differences in the greenhouse gas balance (GHGB) between the two groups.Nevertheless, the yield-scaled greenhouse gas balance (GHGBi) decreased significantly in the MHY_HNUE group compared to that in the LMY_LMNUE group.

Stability of the coordination of yield and NUE
In this study, we examined 80 indica and 12 japonica rice genotypes, including the most recent and advanced bred rice genotypes released from 2006 to 2019.The comprehensive results of our study showed that the yield ranged from 4.2 to 12.0 t ha - 1 , the most useful NUE indices of NUEg, NUEb, NHI, and PFP ranged from 24.3 to 62.0 kg kg −1 , 78.1 to 148.7 kg kg −1 , 34.1 to 75.8%, and 39.9 to 119.7 kg kg −1 , respectively (Supplementary Fig. S1).K-means and hierarchical clustering corroborated the disparities in yield and NUE, as group 4 was characterized by reduced yield and NUE, while group 1 demonstrated the highest yield and NUE in general under a reduced N rate (Fig. 2).The results of the study revealed a transparent gradient from low yield and NUE to moderate or relatively higher yield and NUE, which was confirmed in two additional groups.Notably, the rice genotypes of MHY_HNUE exhibited a higher capacity for both yield and NUE compared to the LMY_LMNUE groups (Fig. 3).
Furthermore, the MHY_HNUE genotypes demonstrated greater stability compared to the LMY_LMNUE groups, as evidenced by a decrease in the degree and centrality cocurrency network, coupled with an increase in the diameter and modularity.Prior studies indicated that increased node connectivity (Fan et al., 2018), centrality (Jordán, 2009), and complexity (May, 2019) were associated with reduced network stability.
These network analyses originate from graph theory (Pavlopoulos et al., 2011) or social network analysis (Otte and Rousseau, 2002) and have previously been utilized to explore the stability of microbial networks in response to disturbances (de Vries et al., 2018;Xiao et al., 2018).To the best of our knowledge, this was the first instance where network analysis was applied to evaluate the stability of the coordination of yield and NUE.The findings of this study suggested that specific rice genotypes held promise as sources for further increases in both yield and NUE in China.However, variation in yield and NUE was often determined by the interplay between genotype and N rate (Nehe et al., 2018;Wang et al., 2021a;Zhang et al., 2013).In the current rice breeding programs, rice genotypes are usually developed in environments where N is abundant, providing limited information on the performance of the rice genotypes when N supply is restricted.Therefore, we recommend that breeders carry out at least two N levels (moderate and reduced N rate) to reduce the uncertainty of the performance of rice genotypes under low N levels.

Effects of growth and N properties on yield and NUE
The results of the study revealed that growth and N properties played a significant role in the variations of yield and NUE (Figs. 6, 7).Individually, growth properties only explained 2% of the coordination of yield and NUE, whereas N properties contributed significantly, explaining 16% of the variation (Supplementary Fig. S8).The stepwise regression analysis confirmed these findings, indicating that growth indicators such as dry matter, plant height, and CGR at PM stage, as well as SLW at HD stage, primarily influenced the formation of NUEg, NUEb and NHI formation but did not have significant effects on the yield and PFP variation.This suggested that to achieve high yield and NUE characteristics, focus should be placed primarily on the N properties.Yield formation depends on the accumulation and translocation of dry matter to the panicle tissues (Huang et al., 2019;Yang and Zhang, 2010b).Previous linear regression results recommended increasing leaf net photosynthesis rate while reducing respiration rate to achieve higher dry matter accumulation (Li et al., 2009;Wang et al., 2018).However, a higher dry matter phenotype only resulted in a higher NUEb and did not significantly increase yield, likely due to limitations in the flux of carbohydrates from source to sink organs (Yang andZhang, 2010a, 2010b).The study also observed a significant positive contribution of N uptake to yield at the PM stage, while high N concentration in organs had a negative effect on yield formation.Thus, while high yield is partly related to dry matter accumulation, it is also dependent on the translocation and distribution of N.
Previous studies demonstrated that the yield advantage of ordinary hybrid rice over inbred rice was mainly due to higher aboveground dry matter, while for super hybrid rice, higher grain yield was attributed to both high harvest index and aboveground dry matter at 90 kg ha -1 N rate (Huang et al., 2018).In this study, while a few super hybrid rice genotypes were used and classified into 1 or 3 groups, for most rice genotypes, biomass accumulation was still the primary driver for the improvement of yield and NUE.
The effects of N concentration in tissues at HD and PM stages were controversial to the yield and NUE (NUEg, NHI, and PFP) simulation.Previous studies primarily focused on the N concentration at the PM stage while ignoring the effects of N concentration at the HD stage.For instance, Huang et al. (2018) and Wu et al. (2016) recommended reducing the N concentration in leaf, stem, and grain tissues at maturity to improve NUEg, particularly for stem N. We concur with these recommendations if the focus is solely on yield-related NUE indices.However, when considering biomass-based NUE, we found that a higher tissue N concentration should be targeted, particularly at the HD stage.This indicated that the N concentration at HD stage may have a significant influence on yield and NUE.Nehe et al. (2018) investigated 28 wheat genotypes and found that N uptake at anthesis was crucial for higher yield and NUE, as it maintained green leaves, reduced the N translocation rate, and ultimately lowered N concentration in grains while increasing yield.In this study, we found that N uptake at PM was primarily derived from the HD stage rather than the period between HD and PM stages, suggesting that higher N concentration and N uptake at HD should be targeted to increase yield and NUE simultaneously.

Effects of climate and soil properties on yield and NUE
Climate variables, such as solar radiation and temperature, can influence rice yield and NUE (Peng et al., 2004;Wang et al., 2021b).Many studies have investigated the asymmetric effects of temperature on rice yield production (Peng et al., 2004;Shi et al., 2016;Wang et al., 2016;W. Wang et al., 2020).Generally, increasing temperature at high latitudes can increase yield, but at low latitudes, it can decrease rice production (Wang et al., 2016;H. Zhang et al., 2019).Our study demonstrated that the maximum temperature at the pre-heading stage negatively affected yield, while the minimum temperature at the post-heading stage positively regulated it.These results partly agreed with previous findings, which also addressed the crucial and positive effects of the maximum temperature at the post-heading stage on rice yield (Welch et al., 2010).
While the positive impact of radiation use efficiency on rice yield has been recognized (Wang et al., 2016), we did not observe a significant influence of solar radiation on the yield.The absence of a clear observation can be attributed to the challenge of discerning the independent effects of weather variables, as they often lack independence (Deng et al., 2015).Temperature indices, especially the minimum temperature at pre-and post-heading stages, negatively affected NUEg, NUEb, and PFP.A plausible explanation for this phenomenon could be the effects of temperature on N uptake.Prior research suggested that elevated temperatures increase N uptake by increasing leaf evaporation rates (Chen et al., 2014;Wang et al., 2018).Higher temperatures also increased the capacity for adaptation, which can enhance nutrient uptake to meet or alleviate the adverse effects of an increased minimum temperature that does not exceed natural threshold values (Shi et al., 2017).The maximum temperature at the post-heading stage contributed to higher NUEg, NHI, and PFP.The physiological mechanism underlying these results remained unclear.An increase in maximum temperature at the post-heading stage can avoid the potential influence of low temperature on yield formation, particularly for grain filling quality (Arshad et al., 2017).
Total soil N has been shown to have a positive effect on rice yield, but it can also decrease NUE when the soil has ample N supply.The N concentration in plant tissues tends to increase with higher levels of indigenous N in the soil, which may lead to decreased NUE despite increased yield (Yin et al., 2021).Total soil N has a positive effect on rice yield, but it can also decrease NUE when the soil has ample N supply.
The N concentration in plant tissues tends to increase with higher levels of indigenous N in the soil, which may lead to decreased NUE despite increased yield (Duan et al., 2014), highlighting the importance of phosphorus in achieving a balance between yield and NUE.To achieve simultaneous increases in yield and NUE, adjusting the N and phosphorus rates should be considered for long-term soil amelioration, rather than focusing on short-term gains.

Environment effects
The results of our study indicated that rice genotypes with high yield and NUE tended to increase CH4 emissions but decrease N2O emissions.These findings appeared to contradict those reported by Jiang et al. (2017), which suggested that high-yielding rice genotypes could reduce CH4 emissions.The CH4 formed in the soil is released into the atmosphere through the rice plant.Up to 90% of the CH4 emission was rice plantmediated through the well-developed intracellular air spaces (aerenchyma) in leaf blades, leaf sheaths, culm, and roots (Lou et al., 2008;Schütz et al., 1989).Differences in the amount and type of aerenchyma between genotypes may contribute to variations in gas transport capabilities and CH4 emissions.Furthermore, longer growing periods for high-yielding and NUE rice genotypes may also increase CH4 emissions during the growing season.However, no significant difference in CH4 emissions between early and late-maturing rice genotypes was observed, as the emission rates during the ripening stage were relatively low and had minimal effects on the total CH4 flux in late-maturing genotypes (Gutierrez et al., 2013).A metaanalysis conducted by Chen et al. (2021) found that an increase in root biomass could enhance CH4 emissions.However, the improvement in CH4 fluxes and cumulative CH4 emissions were relatively stable at approximately 36.6% and 29.5%, respectively, regardless of the increased percentage of root biomass (Chen et al., 2021).This finding demonstrated that root biomass may had a limited effect on CH4 emissions.Other studies proposed that root exudates could increase the soil organic carbon content in the rhizosphere, providing more carbon sources for methanogens, and hence leading to higher methane emissions (Aulakh et al., 2001;Jia et al., 2006).High-yield and NUE rice genotypes are generally associated with a larger root system (Zhang et al., 2018(Zhang et al., , 2021)), which could contribute to increased CH4 emissions.
Although N2O emissions from paddy rice growing systems are generally lower than CH4 emissions, N2O has a larger global warming potential.A study by Cui et al. (2022) estimated that the annual N2O emissions per year in China's rice growing system have decreased from 1980 to 2017, mainly due to the decreased N application rate (Cui et al., 2022).Researchers have found a positive relationship between NUE and N2O emissions reduction under reduced N rates (Huang and Tang, 2010;K. Yu et al., 2021;C. Zhang et al., 2019).Higher N uptake by roots could reduce the N residual in the soil, decrease the N source for denitrifying microorganisms, and hence mitigate the emission of N2O.Rice root exudates could increase the soil carbon content and improve the soil C/N ratio, affecting the N2O emission rate.In our study, improvement in NUE significantly reduced the GHGBi, mostly driven by the N2O emission reduction.
The findings demonstrated that through NUE enhancement of rice genotypes, we can further mitigate greenhouse gas emissions and the followed adverse effects under reduced N rate.

Limitations
Our study demonstrated that an increase in N uptake at the HD stage and a decrease in N concentration of leaf, stem, and grain at the PM stage were crucial for optimizing the coordination of yield and NUE.An increase in N uptake during the HD stage can provide sufficient nutrition for yield and biomass production, promoting stay-green characteristics and sustained photosynthesis in the leaves.However, high N uptake during HD may prolong the maturity of rice plants, which could increase the risk of lodging.Additionally, reducing N concentration in the grain exacerbates hidden hunger for nutrition caused by elevated CO2 levels, which have been shown to decrease protein content (Smith et al., 2018).While the increased temperature can partly counteract or mitigate this trade-off, it can also result in yield losses, particularly in regions where warming is predicted to occur during critical developmental periods (Cai et al., 2016;Wang et al., 2018).In future climate conditions, minimum and maximum temperatures are predicted to increase by approximately 2-5 ℃ under middle or high emission scenarios (Wang et al., 2021b), which would exceed the minimum temperature levels in our study and potentially decrease rice yield and further reduce NUE.The effects of increased temperature require further evaluation through additional field experiments to refine our findings.Nonetheless, our present study, involving nearly 100 released rice genotypes, can help improve our understanding of how to simultaneously increase yield and NUE in current environmental conditions.We recommend that breeders consider temperature tolerance characteristics when screening for yield and NUE performance in near future.
Although we used the CH4MOD and IPCC procedures to estimate the impact of rice genotypes on greenhouse gas emissions, there were still uncertainties associated with the equations.For instance, the characteristics of emission types of these rice genotypes were unclear, therefore we were not clear what's the specific values of variety index should be used.To address the uncertainty and refine the simulation of CH4 emissions, a sensitivity analysis was used with a 0.01 step size from 0.5 to 1.5 of the variety index in the simulation process.However, further experimentation with the selected rice genotypes in two groups is urgent to assess their effects on CH4 and N2O emissions.

Conclusion
A comprehensive evaluation was conducted to assess the coordination of yield and NUE in newly-released rice genotypes from 2006 to 2019 and to determine the potential for reducing environmental impacts.Growth, nitrogen, climate, and soil properties were analyzed to uncover the critical factors affecting yield and NUE.There was uncertainty in selecting for higher yield and NUE in the current breeding system.
The results showed that only a portion of the rice genotypes was classified as having moderate-high yield and high NUE, and these genotypes maintained stability in the emission from Chinese croplands during 1980-2000. Environ. Pollut. 158, 631-635.S4.S4.

Fig. 1
Fig. 1 Maps of the experiment sites (Wuxue and Yangzhou) and weather conditions during rice growing period.

Fig. 2
Fig. 2 Boxplot of the yield and NUE indices of the six groups based on the K-means cluster analysis.A, yield; B, NUEg; C, NUEb; D, NHI; E, PFP.The different letter indicates the statistical significance at 0.05 level.NUEg, nitrogen use efficiency for grain production; NUEb, nitrogen use efficiency in biomass production; NHI, nitrogen harvest index; PFP, partial factor productivity.The box boundaries indicate the 25th and 75th percentiles; the black line within the box mark the median; and whiskers below and above the box indicate the 10th and 90th percentiles, respectively.

Fig. 3
Fig. 3 Boxplot of the yield and NUE indices of the two groups LMY_LMNUE and MHY_HNUE.A, yield; B, NUEg; C, NUEb; D, NHI; E, PFP.LMY_LMNUE represents the low to moderate yield and nitrogen use efficiency group; MHY_HNUE represents moderate to high yield and high nitrogen use efficiency group.Different letters indicate statistical significance at the 0.05 level.NUEg, nitrogen use efficiency for grain production; NUEb, nitrogen use efficiency in biomass production; NHI, nitrogen harvest index; PFP, partial factor productivity.The box boundaries indicate the 25th and 75th percentiles; the black line within the box mark the median; and whiskers below and above the box indicate the 10th and 90th percentiles, respectively.

Fig. 4
Fig. 4 Linear regression of the yield and NUE.LMY_LMNUE represents low to moderate yield and nitrogen use efficiency group; MHY_HNUE represents moderate to high yield and high nitrogen use efficiency group.NUEg, nitrogen use efficiency for grain production; NUEb, nitrogen use efficiency in biomass production; NHI, nitrogen harvest index; PFP, partial factor productivity.

Fig. 5
Fig. 5 Co-occurrence network analysis.LMY_LMNUE represents low to moderate yield and nitrogen use efficiency group; MHY_HNUE represents moderate to high yield and high nitrogen use efficiency group.Different letters indicate the statistical significance at the 0.05 level.

Fig. 6
Fig. 6 Correlations between growth, nitrogen, climate and soil factors and relationships between yield and NUE.Line width corresponds to the Mantel's r statistic, and line color denotes the statistical significance based on 999 permutations.Pairwise comparisons of independent factors are also shown, with a color gradient denoting Pearson's correlation coefficient, and these factors are synthesized into four groups based on attribute of data surveyed.Asterisks indicate the statistical significance (***P < 0.001; ** P < 0.01; and * P < 0.05).The variables are described in Supplementary TableS4.

Fig. 7
Fig. 7 Principal component analysis of (A) growth variables, (B) nitrogen-related variables, (C) climate variables, and (D) soil variables.Dim1 and Dim2 present the first and second principal components, respectively.The variables are described in Supplementary TableS4.

Fig. 9
Fig. 9 Environmental effects of the rice cultivars of groups LMY_LMNUE and