High yields of hybrid rice do not require more nitrogen fertilizer than inbred rice: A meta‐analysis

Abstract

Hybrid rice has been planted throughout China to ensure food security owing to its higher yield potential than inbred rice. Meanwhile, substantial nitrogen (N) fertilizer has been applied to feed hybrid varieties for maximizing grain yield. However, to what extent the higher yield of hybrid than inbred rice depends on N fertilizer input remains unclear. A meta‐analysis was conducted in this study to (1) quantify the difference in N uptake and utilization between hybrid and inbred rice; (2) determine whether hybrid rice requires more N than inbred rice for producing per unit grain yield; and (3) evaluate the impact of the difference in crop growth duration between hybrid and inbred rice on their yield performance. The results showed that, overall, hybrid rice achieved a 10.1% higher grain yield, and 9.7% higher daily grain yield than inbred rice. This grain yield advantage increased from 6.1% to 11.9% depending on whether hybrids have longer crop growth duration than inbred rice. The yield advantage of hybrid rice was explained by higher total N uptake and internal N use efficiency, but not explained by N fertilizer input. Moreover, the yield advantage of hybrid rice with N fertilizer was driven by higher yield without N fertilizer instead of yield response to N fertilizer. These results suggest that hybrid rice does not necessarily require more N fertilizer to achieve higher yield than inbred rice. Therefore, hybrid rice could be planted with fewer external N to ensure food security and reduce environmental costs.

INTRODUCTION

Rice is a staple food crop for more than 3.5 billion people (Khush, 2013). Rice yield is required to increase at least 1% annually to meet the growing global rice demand which results from population growth and economic development (Normile, 2008). To achieve this goal, breeding new rice varieties with higher yield potential is the major strategy to improve on‐farm yield (Foley et al., 2011). Over the past decades, rice yield potential had been much improved with the symbolic development of semi‐dwarf varieties in 1950s, hybrid varieties in 1970s, and super hybrid varieties in 1990s (Cheng et al., 2007; Peng et al., 1999). It is well documented that hybrid rice varieties have a 9–20% yield advantage over inbred rice varieties (Peng et al., 1999; Yuan et al., 1994). This yield advantage was mainly attributed to higher biomass production and lodging resistance especially under high nitrogen (N) fertilizer environment. But precisely owing to improved lodging resistance and N tolerance, farmers in China usually tend to overuse N fertilizer to maximize grain yield (Huang et al., 2008; Peng et al., 2002). In China, the planting area of hybrid rice has been expanding since the late 1970 s and recently accounts for almost half of total rice‐planting area (Peng, 2016). Simultaneously, N consumption in rice has continuously increased, and the average N fertilizer input has exceeded 180 kg ha⁻¹ which is 75% higher than the world average (Peng et al., 2006).

The overuse of N fertilizer leads to rapid N loss from rice agroecosystem, which impose serious environmental pollution including soil acidification (Guo et al., 2010), water eutrophication (Hamilton et al., 2018), and increased greenhouse gas emission (Cassman et al., 2003). Moreover, under excessive N environment, high foliar N content and moisture retention within rice canopy is more prone to pests and diseases damage which consequently could cause significant reduction in rice yield and economic benefit (Cu et al., 1996). Therefore, it is urgent for China to reduce the current N input in rice production without yield penalty (Peng et al., 2006).

Most of hybrid varieties were selected for better yield performance and strong lodging resistance by rice breeders in an excessive N fertilizer environment. These hybrid varieties tend to be more adaptive to high N environment than inbred varieties and produce more grains as previous yield trials reported (Corbin et al., 2016; Yang, 2015). Moreover, many high‐yielding records above 15 t ha⁻¹ were achieved through 280–350 kg ha⁻¹ N input using hybrid varieties (Gu et al., 2017; Li et al., 2014). These high yields of hybrid varieties are often achieved under optimum growing environments where substantial N fertilizer is provided, which give rise to the perception that hybrid rice perform better than inbred rice only under high N input conditions (Islam et al., 2007; Yuan et al., 2017). However, limited information is available about the yield performance of inbred and hybrid varieties under reduced N input. Therefore, a long‐standing question need to be answered: to what extent the high yield of hybrid rice depends on N fertilizer input.

Crop growth duration determines the total amount of incident solar radiation and N uptake which significantly impact on biomass accumulation and crop yield (Yoshida, 1981). Yet, it has not received much attention when comparing the yield performance of hybrid and inbred varieties. It is common that hybrid varieties had longer growth duration than inbred varieties which might contribute to the high yield of hybrids (Laza et al., 2004; Zhang et al., 2009b). There were also few cases that hybrid variety with shorter crop growth duration produced lower yield than inbred variety (Wei et al., 2016; Yang et al., 2007). The differences in growth duration were clearly a source of yield difference between inbred and hybrid rice, which was nonetheless not directly associated with intrinsic ability of the plant to produce grain (Crisantas & Tanguy, 2009). However, few researches to date have quantified the contribution of long crop growth duration on the yield superiority of hybrid over inbred varieties.

Meta‐analysis can provide formal statistical techniques for summarizing the published results of independent experiments to quantify the yield performance of hybrid and inbred rice with wide range of N fertilizer input. In this study, a comprehensive meta‐analysis was conducted with the following objectives: to (1) quantify the difference in N uptake and utilization between hybrid and inbred rice; (2) determine whether hybrid rice requires more N than inbred rice for producing per unit grain yield; and (3) evaluate the impact of the difference in crop growth duration between hybrid and inbred rice on their yield performance.

MATERIALS AND METHODS

Data collection

The literature search was conducted on Web of Science, Science Direct, Google Scholar, and China Knowledge Resource Integrated Database (CNKI) for collecting articles published from 1970 to 2019 using four search terms: “inbred”, “hybrid”, “rice”, and “yield”. An initial search resulted in 30,926 publication records from four databases. Duplicate studies were removed from these records by comparing article title, journal name, author list, and publication year. Then, the abstract and full text of publications were checked to determine the eligibility for this meta‐analysis based on the following requirements: (1) experiment must be conducted under field conditions and not in pot or greenhouse experiments, (2) inbred and hybrid rice must be included and compared side‐by‐side in field experiments, and all other management practices kept the same, and (3) N fertilizer input, crop growth duration, and yield‐related data must be available for both hybrid and inbred rice. The multi‐factorial studies (i.e., in which hybrid and inbred rice were combined with other treatments in a factorial design) and studies that reported results of multiple years would contribute more than one paired observation to our dataset. A total of 134 articles containing 3569 of paired observations met our criteria and were incorporated into this meta‐analysis, in which 11 articles came from our research group. The reference list for included articles was provided as Supplementary Materials appendix 1.

Crop growth duration and yield‐related data in each included article were recorded by extracting directly from tables and text, or indirectly from figures using GetData Graph Digitizer 2.26 (http://getdata‐graph‐digitizer.com/). Daily yields (grain yield/crop growth duration) were calculated for both hybrid and inbred rice. Besides, N fertilizer input, grain yield without (Y₀) and with N fertilizer (Y_N), as well as N uptake in rice plants at physiology maturity were also collected. Then, the increment of yield with N fertilizer (△Y = Y_N − Y₀), and internal N use efficiency (IE = grain yield/N uptake) were calculated. To evaluate the confounding impact of rice growth duration on the effect sizes, growth duration difference (GDD, growth days of tested hybrid rice variety minus inbred rice variety) was used as the moderating variable, and it was categorized as: GDD ≤ −5, −5 < GDD < 5, and GDD ≥ 5 for subgroup analysis.

Data analysis

Following our previous work (Xu et al., 2019), the random‐effects model was applied to compute the summary effect using Comprehensive Meta‐Analysis software (Biostat, Inc.). First, the natural log of the response ratio (R) for each paired observation was calculated as the effect size to compare hybrid and inbred rice yield performance, N uptake, and IE in our meta‐analysis (Equation 1)[Image Omitted. See PDF]where X_hybrid and X_inbred are response variables (the yield, N uptake or IE) of hybrid and inbred rice, respectively. And the variance of the ln R for study (i) (V_i) was approximated using the following formula:[Image Omitted. See PDF]where SD₁ and SD₂ are the standard deviation for hybrid and inbred rice in study (i), respectively; n₁ and n₂ are the sample sizes for hybrid and inbred rice, respectively. Next, a weight was assigned to each study under the inverse scheme (Equation (3)) as follows:[Image Omitted. See PDF]where W_i is the weight assigned to study (i), V_i is the within‐study variance for study (i), and T² is the between‐study variance that is common to all studies. The T² estimation is made using the DerSimonian and Laird method (DerSimonian & Laird, 1986). Given that some studies fail to provide standard deviation (SD) for their outcomes, this missed information was imputed by using a well‐established method to increase the precision of the overall effect (Furukawa et al., 2006).

The weighted summary effect size was then computed as follows:[Image Omitted. See PDF]where M is the weighted summary effect size; R_i and W_i are the response ratio and the weight for study (i), respectively. Finally, the variance of the summary effect was estimated as the reciprocal of the sum of the weights, and the 95% confidence intervals (CI) for the effect was calculated as follows:[Image Omitted. See PDF][Image Omitted. See PDF]where LL_M and UL_M are the 95% lower and upper limits for the summary effect, respectively; M and V_M are the summary effect and its variance, respectively. To simplify interpretation, the response effects were expressed as the percentage change of hybrid rice relative to inbred rice, using the equation A = $\left(e^{\text{lnM}}‐1\right)\times 100\%$. Positive percentage indicates an increase hybrid over inbred rice, while negative percentage indicates a decrease.

To test whether the effect size was significantly different between growth duration subgroups, the homogeneity was examined using a Q‐test based on analysis of variance where total heterogeneity (Q_Total) was divided into within‐group (Q_W) and between‐group (Q_B) heterogeneity. Under the null hypothesis that the effect size is the same for all groups, 1 to p, Q_B would be distributed as chi‐squared with degree of freedom equal to p − 1. Q_B rather than Q_W can be of considerable scientific interest (Gurevitch & Hedges, 1999). The significance of Q_B was tested by comparing it against the critical value of the χ² distribution. And a significant Q_B denotes that the cumulative effect size is not the same for different groups. Differences were considered as statistically significant when p < 0.05.

Moreover, a multivariate meta‐regression analysis using standard random‐effects model was performed to assess the extent to which one or more potential factors explained heterogeneity of comparative yield performance between hybrid and inbred rice. Nitrogen fertilizer input, N uptake, IE, and GDD were incorporated to construct meta‐regression model. The Z‐test was used to test the statistical significance of the covariate slope (Equation 6).[Image Omitted. See PDF]where B and SE_B are the coefficient for covariate and its standard error, respectively.

In meta‐regression analysis, a common approach to describing the impact of a covariate is to report the proportion of variance explained by that covariate. In the present study, that index, S², was defined as the ratio of explained variance to total variance (Equation 7)[Image Omitted. See PDF]where T²_explained and T²_total represent the true explained variance between‐study and total variance between‐study, respectively.

Publication bias

Publication bias is a common problem in meta‐analysis because that the studies reported relatively high effect sizes are more likely to be published than studies that reported insignificant effect sizes. Publication bias in this study was assessed with symmetry in Funnel Plots (Egger et al., 1997) and fail‐safe numbers according to Rosenthal's method (Rosenthal, 1979). To estimate how much impact the bias had, an unbiased funnel plot was generated and re‐computed the imputed effect size followed Duval and Tweedie's Trim and Fill method (Duval & Tweedie, 2000) to determine whether the imputed observations could deny the summary effect size. Rosenthal's fail‐safe number means the number of additional non‐significant unpublished observations that would be required to nullify the summary effect. 5n + 10 (n, total number of paired observations in this study) could be given as a reasonable critical lower limit for fail‐safe number.

RESULTS

Most of the observations in this meta‐analysis were located in Asia, with 49.5% of observations from China and 45.1% from Philippines. A small portion of observations came from India, Japan, United States, and other countries (Table 1). N fertilizer input applied in field experiments ranged from 0 to 360 kg ha⁻¹ with an average of 144 kg ha⁻¹(Figure 1).

1 TABLELocations and observation number of the studies included in meta‐analysis

Region	Observation number	Proportion (%)
China	1767	49.5
Philippines	1609	45.1
India	60	1.7
Japan	47	1.3
USA	41	1.1
Bangladesh	23	˂1.0
The rest	22	˂1.0
Total	3569	100.0

1 FIGURE. Frequency distribution of nitrogen fertilizer input based on the paired observations incorporated into this meta‐analysis

Overall, hybrid rice significantly increased grain yield by 10.1% compared to inbred rice with quite narrow confidence intervals (CI: 9.7%–10.6%) (Figure 2). And daily yield of hybrid rice was also 9.7% (CI: 9.1%–10.2%) higher than that of inbred rice. The response of yield and daily yield between hybrid and inbred rice varied significantly among crop growth duration subgroups, although in all cases hybrid rice was higher than inbred rice in yield and daily yield. Yield advantage of hybrid rice over inbred rice displayed an increasing trend whereas daily yield advantage displayed a decreasing trend with longer GDD. From GDD ≤ −5 to −5 < GDD < 5, yield advantage increased from 6.1% (CI: 5.3%–7.0%) to 10.2% (CI: 9.4%–11.0%). When GDD ≥ 5, grain yield of hybrid rice was 11.9% (CI: 11.0%–12.9%) higher than that of inbred rice (Figure 2). In contrast, daily yield advantages of hybrid over inbred rice were 16.3% (CI: 15.3%–17.2%), 10.3% (CI: 9.5%–11.1%), and 4.4% (CI: 3.5%–5.4%) for GDD ≤ −5, −5 < GDD < 5, and GDD ≥ 5, respectively.

2 FIGURE. Comparison of yield daily yield between hybrid and inbred rice in relation to the differences in crop growth duration. GDD represents growth days of tested hybrid rice variety minus inbred rice variety. The number of paired observations/number of studies included in each dataset is presented in parenthesis

Multivariate meta‐regression analysis was performed to determine whether yield comparison between hybrid and inbred rice depends on N fertilizer input. As shown in Table 2, the p value was 0.89 and less than 0.01% heterogeneity was explained by N fertilizer input, meaning that N fertilizer input was not significantly related to the comparative yield performance between hybrid and inbred rice. By contrast, the most of heterogeneity in yield comparison between hybrid and inbred rice was explained by IE (48.5%), N uptake (47.2%), and GDD (3.1%).

2 TABLEMultivariate meta‐regression analysis evaluate the effect of nitrogen (N) uptake, internal N use efficiency (IE), growth duration difference (GDD), and N fertilizer input on yield comparison between hybrid and inbred rice

Covariate	Coefficient	Confidence intervals	p value	Heterogeneity explained by covariate (%)
N uptake	0.807	0.769–0.845	<0.001	47.2
IE	0.848	0.805–0.891	<0.001	48.5
GDD	0.002	0.001–0.003	<0.001	3.14
N fertilizer input	<0.001	0.000–0.000	0.890	0.00

^aGDD: Growth duration (days) of tested hybrid rice variety minus inbred rice variety.

Some field experiments involved in this meta‐analysis included zero‐N control and N fertilization treatments. Therefore, we compared yield with N fertilizer (Y_N) and without N fertilizer (Y₀), and also the yield increase attained with N fertilizer (△Y) between hybrid and inbred rice to further confirm whether increasing N input would improve yield advantage of hybrid over inbred rice. As presented in Figure 3a, the overall Y₀ for hybrid rice was 7.2% (CI: 5.9%–8.5%) higher than inbred rice. For three crop growth duration subgroups, the relative Y₀ was the highest (9.5%, CI: 7.6%–11.4%) for GDD ≥ 5, followed by −5 < GDD < 5 (7.2%, CI: 5.2%–9.2%), the lowest (1.8%, CI: 0.0%–3.6%) for GDD ≤ −5. This was consistent with the trend of yield response among three subgroups (Figure 2). In contrast, the overall △Y for hybrid rice was 2.3% (CI: −4.1% to −0.5%) lower than inbred rice (Figure 3b). There was no significant difference in △Y performance among three crop growth duration subgroups Therefore, yield advantage of hybrid over inbred rice mainly resulted from higher Y₀ rather than △Y, which supported our meta‐regression result that increasing N fertilizer could not expand the yield advantage of hybrid rice.

3 FIGURE. Comparison of grain yield without N fertilizer (Y0) and increase in grain yield with N fertilizer (△Y) between hybrid and inbred rice in relation to the differences in crop growth duration. GDD represents growth days of tested hybrid rice variety minus inbred rice variety. The number of paired observations/number of studies included in each dataset is presented in parenthesis

Total N uptake at maturity and IE were also compared to answer why grain yield for hybrid rice was more than inbred rice. As shown in Figure 4a, the overall N uptake for hybrid rice was 7.2% (CI: 5.9%–8.5%) higher than inbred rice. When GDD ≤ −5, N uptake of hybrid rice was 4.0% (CI: −8.9% to 1.1%) lower than inbred rice. But as for −5 < GDD < 5 and GDD ≥ 5, hybrid rice increased in N uptake by an average of 7.3% compared to inbred rice. There was a positive curvilinear relationship between rice yield and N uptake, meaning that the increase in yield was driven by N uptake. But as yield approaching the yield ceiling, the return for plant N convert to grain was diminishing (Figure 5). On the other hand, overall IE for hybrid rice was 5.9% (CI: 4.4%–7.3%) higher than inbred rice (Figure 4b). There was no significant difference in IE among three crop growth duration subgroups. This was consistent with information in Figure 5 that hybrid rice had a larger increase in yield per unit N uptake than inbred rice. Besides, the IE advantage of hybrid over inbred rice tended to be larger as N uptake increasing (Figure 5).

4 FIGURE. Comparison of total nitrogen (N) uptake and internal N use efficiency (IE) between hybrid and inbred rice in relation to the differences in crop growth duration. GDD represents growth days of tested hybrid rice variety minus inbred rice variety. The number of paired observations/number of studies included in each dataset is presented in parenthesis

5 FIGURE. Relationship between total N uptake and grain yield for hybrid and inbred rice

Funnel plot indicated that the studies appearing in the bottom of the plot were distributed asymmetrically about the summary effect size (Figure S1). Observed effect size under the random‐effects model was the same as the imputed effect size estimated by the Trill and Fill method. Moreover, the fail‐safe number (7,774,211) was much greater than threshold value (5 × 3569 + 10 = 17,855). These results revealed that the impact of publication bias was trivial, and the effect size was valid.

DISCUSSION

This study included a total of 3569 paired observations into meta‐analysis, depicting the yield comparison between hybrid and inbred rice across wide genetic background and crop management conditions. Overall, hybrid rice achieved 10.1% higher grain yield and 9.7% higher daily yield than inbred rice (Figure 2). This was consistent with previous studies in which 9–20% yield advantage was reported (Peng et al., 1999; Yuan et al., 1994). However, only a few of these previous studies considered the impact of varietal phenology difference on the yield comparison between hybrid and inbred rice (Crisantas & Tanguy, 2009). Crop phenology is closely linked to solar radiation, water, and nutrient resources. Hence, varietal difference in crop growth duration would definitely cause unfair yield comparison. Many studies attributed higher yield of hybrid rice to its longer growth duration compared to inbred rice (Katsura et al., 2007; Zhang et al., 2009b). On the contrary, it was also reported few times that hybrid rice produced lower yield than inbred rice due to shorter growth duration (Wei et al., 2016; Yang et al., 2007). In this study, this confounding effect of varietal difference in crop growth duration was detected and quantified on yield comparison between hybrid and inbred rice. Our results suggested that yield advantage of hybrid over inbred rice ranged from 6.1% to 11.9%, depending on whether crop growth duration of hybrid rice was longer than inbred rice (Figure 2).

Comparison between inbred and hybrid rice is often conducted under sufficient N fertilizer environment, but their performance in reduced N environment remains an open issue (Islam et al., 2007; Katsura et al., 2007). The popular perception is that hybrid varieties maximize its yield advantage over inbred varieties only when massive resources including N fertilizer are provided (Huang et al., 2017; Zhang et al., 2009a). However, meta‐regression results revealed that yield advantage of hybrid over inbred rice was not explained by N fertilizer input (Table 2). It was consistent with the findings of our previous researches that hybrid variety still had higher rice yield than inbred variety under low or moderate N input (Huang et al., 2018; Yuan et al., 2017). Moreover, as shown in Figure 3, yield advantage of hybrid rice over the inbred was mainly due to the higher Y₀, but not △Y. These suggested that applying more N fertilizer could improve rice yield, but not increase the magnitude of hybrid yield advantage relative to inbred rice. Chen et al. (2011) quantified optimum N rate for rice production in south China by using a set of statistical models and found that the optimal N rate required for maximizing yield was not much different, even less than 10 kg N ha⁻¹ between inbred and hybrid variety. Increasing N input beyond that threshold is not necessary for hybrid rice to realize their yield potential. Under excessive N environment, yield advantage of hybrid variety may be partially as a consequence of strong lodging resistance and high N tolerance (Huang et al., 2008; Peng et al., 2002). Therefore, the perception mentioned above should be denied, recognizing limited yield returns and serious ecological problems caused by the overuse of N fertilizer for hybrid rice in China.

Rice yield was positively related with plant N uptake (Figure 5). Yield advantage of hybrid over inbred rice was mainly due to higher N uptake and IE, while it was less obviously explained by GDD (Table 2). Overall, hybrid rice absorbed 7.2% more N than the inbred to produce more grain (Figure 4a). This can be ascribed to the fact that hybrid rice plant is more vigorous in leaf expansion, tiller emergence, and root system compared to inbred rice especially during early growth stage, consequently leading to more N absorption and biomass accumulation (Dobermann & Fairhurst, 2000; Norman et al., 2013; Zhang et al., 2009a). Furthermore, IE of hybrid rice was significantly higher than inbred rice, suggesting N in hybrid plant was more efficiently converted to grain yield (Figure 4b). This was confirmed in Figure 5 in which hybrid rice produced more grain yield than inbred rice with the same plant N uptake. Most hybrid rice varieties are characterized by erect, dark green, thick, and delay‐senesced leaves; few unproductive tillers; large panicle size; and high harvest index (Peng et al., 2008). These improved leaf morphological traits contributed to higher photosynthetic N use efficiency and biomass production of hybrid rice compared with inbred rice (Chang et al., 2016; Zeng et al., 2006). The increased photosynthetic N use efficiency and biomass production coupled with high harvest index resulted in greater IE of hybrid rice (Huang et al., 2016). Moreover, few unproductive tillers and large panicle size in hybrid rice ensured N to be efficiently translocated and used for grain formation, leading to higher IE. Because some parts of N in unproductive tillers could not be mobilized to productive tillers before they die that caused decrease in IE (Peng & Bouman, 2007). In particular, recently released indica–japonica hybrid varieties that were developed by cross‐breeding japonica and indica rice have showed strong heterosis and further improved yield potential especially under high N environment (Wei et al., 2018), resulting in the yield of hybrid rice did not reach the ceiling at N uptake over 200 kg ha⁻¹ (Figure 5). When the observations involved indica–japonica hybrid rice were removed, hybrid rice yield would reach the ceiling at N uptake around 200 kg ha⁻¹ and its yield was still higher than inbred at same N uptake (Fig. S2). In summary, hybrid rice is more effective in enhancing grain yield under side‐by‐side comparison with inbred rice, indicating higher resources use efficiency.

CONCLUSION

This study offered a quantitative analysis of the available literatures to investigate to what extent the high yield of hybrid rice depends on N fertilizer. Major efforts of this study were made to disclose yield difference between hybrid and inbred from the perspective of N absorption and utilization, while limited information was provided about morphological and other agronomic traits. Our results showed that hybrid rice produced significantly higher grain yield than inbred rice in all cases, although the difference in crop growth duration between hybrid and inbred varieties changed the magnitude of yield advantage. Total N uptake and internal N use efficiency were mainly responsible for the yield advantage of hybrid rice. The higher yield of hybrid rice with N fertilizer was driven by Y₀ instead of △Y. Moreover, the meta‐regression results demonstrated that less than 0.01% of the variance can be explained by N fertilizer input. These outcomes suggest that the yield superiority of hybrid over inbred rice does not depend on N fertilizer rate. Further researches should optimize N management of hybrid rice in pursuits of high yield with reduced environmental cost.

ACKNOWLEDGEMENTS

This work was supported by the National Natural Science Foundation of China (No. 31971845), the earmarked fund for China Agriculture Research System (CARS‐01‐20), the Program of Introducing Talents of Discipline to Universities in China (the 111 Project no. B14032), and the Program for Changjiang Scholars and Innovative Research Team at the University of China (IRT1247).