Evolution of Rice Cultivar Performance Across China: A Multi-Dimensional Study on Yield and Agronomic Characteristics over Three Decades

Abstract

Rice (Oryza sativa L.) is a staple food crop for over half of the world’s population, with China being the largest producer. However, the growth rate of rice yield per hectare has slowed in recent years, emphasizing the need for in-depth studies on the evolution of rice cultivar performance. This study presents a comprehensive analysis of the yield and key agronomic traits of rice cultivars across China over three decades, utilizing data from 11,811 cultivar trials conducted between 1990 and 2023. We assessed the spatial distribution and temporal evolution of rice cultivar performance, exploring regional differences and the interplay between agronomic traits and environmental factors. Our results reveal significant variations in growth duration, plant height, grains per panicle, thousand-grain weight, effective panicle number, and seed setting rate across different regions. Temporal trends showed diverse patterns of improvement, with some regions experiencing rapid advancements (up to 1.42% annual yield increase in Jiangxi Province of Central China) and others nearing yield plateaus (0.16% in Jilin Province and 0.45% in Heilongjiang Province of Northeast China). Correlation analysis between agronomic traits and grain yield highlighted the complex relationships and potential for further genetic gains through targeted breeding. This study underscores the importance of region-specific breeding strategies to optimize rice production in the face of environmental challenges and yield ceilings. The insights gained provide a scientific basis for future rice cultivar development and regional agricultural policies aimed at enhancing sustainability and efficiency in China’s diverse rice-growing regions.

1. Introduction

Rice (Oryza sativa L.) is a staple food crop for over half of the world’s population [1]. China is the largest rice-producing country in the world [2]. However, since the early 21st century, the growth rate of rice yield per hectare has slowed [3]. Compared to other crops, such as maize, where 60% of the cultivated area has not yet reached yield ceilings, 84% of China’s rice area has already plateaued [4]. This phenomenon highlights the challenges facing rice production and emphasizes the urgency for in-depth studies on the evolution of rice cultivar performance.

Rice production is influenced by multiple factors, among which soil properties play a major role, impacting 87% and 53% of the regions in eastern and southeastern China, respectively [5]. Among climate variables, temperature has the most significant impact on predicting rice phenology, especially under extreme conditions, while rainfall and solar radiation have relatively smaller effects [6]. Additionally, inter-regional differences in water, energy, nutrient, and pesticide usage contribute to yield gaps among rice production systems across different regions [7]. The climatic diversity of rice cultivation ranges from the warm subtropical climate at latitude 18° N to the cold temperate climate at latitude 50° N [8]. Such climatic diversity provides a unique opportunity to study the adaptive evolution of rice cultivars. Against the backdrop of global climate change, rising temperatures and increased CO₂ levels have significantly different effects on rice production in various regions [9,10,11]. This complex environmental interaction underscores the importance of a multi-dimensional, cross-regional investigation of rice cultivar performance. Some agronomic traits of rice are closely related to the environment. For instance, flowering time shows a latitudinal gradient [12]. Grain yield is linked to yield-related traits such as panicle number per unit area, grains per panicle, and thousand-grain weight [13,14]. Between 1981 and 2009, the introduction of new cultivars increased rice yield in China by 16–52% [15]. These findings emphasize the importance of understanding the geographic distribution of rice agronomic traits and the critical role of cultivar improvement in enhancing rice yield.

Previous research has primarily focused on temporal trends of rice traits through univariate and multivariate analyses, allowing the observation of phenotypic changes in rice germplasm [16]. However, these methods are limited in capturing the spatial differences in rice cultivars. Although some multi-year, multi-location trials have attempted to illustrate the spatial differences in yield and agronomic traits of rice cultivars [17], their spatial coverage remains insufficient to comprehensively reflect the complexity of rice cultivation areas across China. In light of these limitations, this study aims to comprehensively reveal the evolution of rice cultivar performance in China through a multi-dimensional, long-term investigation. We analyzed the yield performance and key agronomic traits of rice cultivars over the past three decades, exploring spatial and temporal changes. Specifically, we (1) mapped the spatial distribution of rice cultivar performance to uncover regional differences; (2) analyzed the temporal evolution of key agronomic traits to identify the main directions of past improvements; (3) explored the intricate relationships between agronomic traits to provide scientific insights for future breeding strategies; and (4) assessed inter-regional differences in ideal rice cultivar characteristics, providing guidance for regional breeding programs.

The insights gained from this comprehensive study of China’s diverse rice-growing regions have broad implications for global rice breeding efforts. As China’s rice cultivation spans multiple climatic zones, the methodologies and findings regarding regional adaptation strategies could be particularly valuable for major rice-producing countries across Asia, Africa, and the Americas that face similar challenges of yield plateaus and climate change. Our analysis of trait evolution and breeding strategies could help other nations optimize their breeding programs according to local environments, ultimately contributing to global food security.

2. Materials and Methods

This study analyzed information from 11,811 rice cultivar trials involving 10,987 different varieties (Table S1), conducted across 23 rice-growing provinces in China between 1990 and 2023. The data were primarily sourced from the China Rice Data Center (http://www.ricedata.cn, accessed on 25 May 2024) and the Seed Industry Management Bureau, Ministry of Agriculture and Rural Affairs of China (http://www.zzj.moa.gov.cn, accessed on 8 July 2024). The dataset included rice yield (GY) and six key agronomic traits: growth duration (GD), plant height (PH), grains per panicle (GPP), thousand-grain weight (TGW), effective panicle number (EP), and seed setting rate (SSR). Missing values were noted for EP and SSR in Heilongjiang and for GD in Guangdong. The rice production system in China was divided into six agro-ecological sub-regions based on the unique climate, topography, and management practices: Northeast (NEC), North China (NC), East China (EC), Central China (CC), South China (SC), Southwest (SWC), and Northwest (NWC) (Figure 1).

To systematically assess varietal differences in yield and agronomic traits across provinces and sub-regions, various statistical methods were employed. Descriptive statistics, including mean, standard deviation, minimum, and maximum values, were calculated for each trait within each province to capture the distribution of traits and to provide a quantitative basis for understanding regional differences. To reveal the temporal trends of rice traits, we grouped the data by year and calculated annual average values for each trait. A linear regression model was used to fit these yearly averages, evaluating the long-term trends of each trait. Regression coefficients and their significance were calculated to quantify the rate of change and assess the statistical relevance of the observed trends. To explore the relationships between GY and other agronomic traits, Pearson correlation coefficients were computed. Principal Component Analysis (PCA) was then conducted to visually present the strength and direction of associations among traits and with GY. PCA reduces the dimensionality of potentially correlated agronomic traits into a few major, independent components, highlighting underlying patterns or trends that may not be apparent in the original data. To further understand the distribution of rice cultivar characteristics and their performance across different regions, a two-stage clustering analysis was conducted. Initially, K-means clustering was used to identify cultivar groups with similar agronomic trait profiles. The optimal number of clusters was determined using the elbow method and silhouette score, involving multiple runs of the K-means algorithm with different k values and evaluating the internal metrics for each. Once the optimal cluster count was identified, K-means clustering was performed on the standardized dataset, and each rice cultivar was assigned to a specific cluster. Cluster centroids (i.e., mean values of each trait) were calculated to describe the typical characteristics of each cluster. Building upon the K-means clustering, regional hierarchical clustering was then conducted to explore inter-regional similarities in rice cultivar trait distributions. For each region, the proportion of each K-means cluster was calculated, and an Euclidean distance matrix was constructed using these distribution vectors. This two-stage clustering approach enabled a clear understanding of how rice cultivar types were distributed across different regions, providing insights into region-specific agronomic trait preferences and adaptation. All data analyses and visualizations were performed using Python (version 3.8). Data manipulation and statistical analyses were conducted using the pandas (version 2.0.3) and numpy (version 1.24.4) libraries. Principal Component Analysis (PCA) and clustering analyses were implemented using scikit-learn (version 1.3.2). Statistical tests were performed using scipy (version 1.10.1). Visualizations were created using matplotlib (version 3.7.5) and seaborn (version 0.13.2) libraries. Geographic data processing and mapping were performed using geopandas (version 0.13.2).

3. Results

3.1. Spatial Distribution of Rice Cultivars

Figure 2 illustrates the spatial distribution of agronomic traits and yield across China’s major rice-growing provinces. Overall, significant regional variations were observed in all agronomic characteristics (Table S2). GY exhibited notably high and concentrated distributions in the NEC region, particularly in Heilongjiang and Jilin provinces. Yields predominantly ranged between 8000 and 9000 kg/ha (Table S3), indicating stable and high productivity in these areas. GD also displayed distinct regional patterns, with extended periods in Heilongjiang and Jilin, reflecting the influence of cooler climatic conditions on the prolongation of growth cycles in these regions.

PH showed considerable inter-provincial variability. Rice cultivars in NEC and EC regions generally had greater stature, while those in SC and SWC were comparatively shorter. The distribution of GPP demonstrated significant heterogeneity, with EC provinces such as Jiangsu and Zhejiang displaying higher GPP values, suggesting an advantage in grain quantity for cultivars in these areas.

Regarding TGW, Jiangsu and Zhejiang showed markedly higher values. EP in NC regions, including Guangdong, Guangxi, and Fujian, exhibited lower mean values with higher variability. SSR also revealed significant regional disparities, with the CC region demonstrating higher and more concentrated SSR distributions, indicating superior seed-setting performance of rice varieties in these areas.

3.2. Temporal Dynamics of Yield and Agronomic Traits

Rice cultivars exhibited significant spatiotemporal variations in yield (Table S4) and agronomic traits across regions (Figure 3). National rice yields showed a marked upward trend, with notable regional heterogeneity. Jiangxi province stood out with the highest annual growth rate of 1.42%. Hunan (1.13%) and Zhejiang (1.06%) also demonstrated high growth rates. In contrast, the NEC region (e.g., Heilongjiang and Jilin), despite higher baseline yields, showed slower growth rates of 0.45% and 0.16%, respectively. Most provinces, such as Jiangsu (0.77%) and Anhui (0.50%), exhibited moderate growth rates, while Shaanxi showed a slight decline (−0.03%). These regional differences likely stem from variations in initial yield levels, climate, soil, and breeding strategies.

We observed coordinated evolution of key agronomic traits, reflecting breeding trade-offs and optimizations. GD showed a slight decreasing trend in most regions, with Jiangsu experiencing the most significant decline (−0.19%), possibly adapting to double-cropping systems. PH generally increased slightly, particularly in Hunan (0.69%) and Hubei (0.56%), potentially balancing higher biomass and lodging resistance. GPP increased significantly in most provinces, driving yield improvement, with Guizhou showing the largest increase (2.35%), followed by Jiangxi (1.94%) and Zhejiang (1.75%). However, TGW generally decreased slightly, most notably in Anhui (−0.78%) and Guangxi (−0.58%), possibly trading off for increased grain number. Changes in EP and SSR were relatively small with distinct regional differences, suggesting these traits may be approaching optimal levels in various areas. Overall, rice breeding has evolved from short-stature, multi-panicle types to moderately tall varieties with larger panicles.

Based on regional temporal dynamics, we categorized rice trait changes into several patterns. The high-yield stable type, represented by the NEC region, showed high baseline yields and moderate growth rates, possibly approaching yield improvement bottlenecks. The rapid growth type, exemplified by the CC region (e.g., Jiangxi, Hunan), demonstrated significant improvements in yield and agronomic traits, likely benefiting from intensive breeding programs and favorable cultivation conditions. The diversified evolution type, represented by the SWC region (e.g., Yunnan, Guizhou), reflected complex breeding objectives and unique geographical conditions through varied trait changes.

3.3. Relationships Between Agronomic Traits and Yield

To identify the primary agronomic traits driving yield variations across regions, correlation coefficients between agronomic traits and yield were calculated for each area. GD showed moderate positive correlations with GY in most regions, particularly in NEC (r = 0.51) and EC (r = 0.54). The relationship between PH and GY varied considerably across regions, with strong positive correlations in CC (r = 0.59) and NWC (r = 0.59), but weak negative correlations in NC (r = −0.19) and SWC (r = −0.11). GPP exhibited positive correlations with GY across all regions, especially in CC (r = 0.73). TGW’s relationship with GY varied significantly, ranging from strong positive correlation in NWC (r = 0.55) to weak negative correlation in CC (r = −0.11). EP showed weak correlations with GY in most regions but displayed moderate positive correlation in SWC (r = 0.41). SSR’s relationship with GY ranged from moderate positive correlation in EC (r = 0.34) to moderate negative correlation in CC (r = −0.33) (Figure 4).

To further understand the multidimensional characteristics of rice production across regions, we conducted PCA for each of the seven regions to identify patterns of influence of different agronomic traits on GY. In CC, the first two principal components explained 67.83% of the total variance, with PC1 primarily determined by PH, GPP, and GY, while PC2 was dominated by TGW. For EC, the first two components accounted for 68.18% of the total variance, with GY and GD loading heavily on PC2, indicating GD’s substantial influence on GY. In NWC, the first two components explained 53.94% of the total variance, with PH, GPP, and GY showing high loadings on PC1, while TGW exhibited strong independence on PC2, highlighting its unique role in this region (Figure 5).

For SC, the first two components explained 61.70% of the total variance, with most traits showing positive loadings on PC1, suggesting a collective influence of major agronomic traits on GY in this region. In NEC, the first two components accounted for 49.13% of the total variance, with GD and GY having high loadings on PC1, while GPP showed strong independence on PC2. PCA results for NC and SWC also revealed unique trait combination patterns, reflecting the diverse influences of agronomic traits on GY within these regions (Figure 5).

3.4. Inter-Regional Differences in Ideal Rice Characteristics

Radar charts reveal substantial overlap between average agronomic traits and the most prevalent cluster centroid characteristics within each region, indicating the representativeness of these cluster traits. For instance, in the NEC region, Cluster 2 characteristics (including long GD, relatively low PH, and high TGW) account for 93.39% of the cultivars, suggesting that rice varieties in this region generally have longer growth duration and higher thousand-grain weight to adapt to local climatic conditions. Specifically, NEC rice cultivars average 148 days for GD, 102 cm for PH, and 25.71 g for TGW, enabling high yields in the region’s short, cool growing season.

EC and CC regions display more diverse cluster distributions, indicating greater variability in rice cultivar characteristics. For example, EC’s primary clusters include Cluster 2 (36.22%), Cluster 4 (29.35%), and Cluster 1 (11.24%) (Figure 6), with significant trait differences among clusters. Cluster 2 exhibits longer GD and lower PH, while Cluster 4 has the highest GPP. In EC, Cluster 2 rice cultivars average 148 days for GD and 102 cm for PH, whereas Cluster 4 cultivars have the highest GPP at 201 grains across all clusters (Figure 7).

Hierarchical cluster analysis further reveals similarities and differences between regions. NEC and NC regions show the most similarity in rice trait distribution, while CC and SC regions cluster together, exhibiting similar rice trait distributions (Figure 8). Specifically, CC rice cultivars average 124 days for GD, 108 cm for PH, and 154 grains for GPP, while SC cultivars average 123 days for GD, 112 cm for PH, and 159 grains for GPP (Table S5), demonstrating similarities in key agronomic traits. This suggests potential similarities in climate and management practices in these regions, leading to comparable rice characteristics.

4. Discussion

4.1. Ecological Explanations for Inter-Regional Differences in Ideal Rice Characteristics

We observed that rice cultivars in the SWC, NC, and NEC regions generally exhibit longer GD, whereas cultivars in the SC region have notably shorter GD. These regional differences can be explained from an ecological adaptation perspective. The NEC and NC regions have ample sunlight but shorter growing seasons, while the SWC region is characterized by mountainous terrain with frequent clouds and rain [18]. In these areas, a longer GD allows rice to accumulate more photosynthetic products. In contrast, the SC region enjoys abundant sunlight and thermal resources year-round, but also faces multiple biotic and abiotic stresses, such as pests [19] and typhoons [20]. Under such conditions, a shorter GD helps rice avoid adverse environmental factors during critical growth stages, ensuring yield stability.

Our study also revealed inter-regional differences in ideal plant type traits. In northern regions, high SSR is a key breeding target, as low temperatures can significantly reduce GPP and TGW [21]. Therefore, a high SSR can compensate for yield losses due to lower TGW. In the SC region, lodging resistance and disease resistance are of greater importance, leading to a breeding focus on cultivars with moderate PH and strong stems [22]. Meanwhile, cultivars in the CC region exhibit traits that are intermediate between these two extremes, reflecting the transitional climate of the area. These findings emphasize the importance of breeding strategies tailored to local conditions and challenge the traditional concept of the “ideal plant type”, suggesting the need for a more flexible and diversified targeted breeding system.

4.2. Trade-Offs in Yield Components: Dynamic Balance Between Grain Number and Weight

Rice yield formation is a complex process involving interactions and trade-offs among multiple yield components. EP, GPP, and TGW are three critical factors determining final yield [23]. The number of grains per panicle is determined by the number of primary and secondary branches, while grain weight is determined by grain size (length, width, thickness) and filling rate [24]. Our observations indicate that as breeding progresses, the GPP of rice cultivars has generally increased. This trend varies across geographical regions but overall reflects the efforts of Chinese breeders to enhance yield by increasing grain number. However, an increase in GPP is often accompanied by a reduction in TGW, a negative correlation observed in multiple rice-growing regions. This trade-off between grain number and grain weight is likely due to the plant’s limited capacity for photosynthate allocation [25]. When grain number increases, the amount of nutrients and photosynthates available to each grain decreases, resulting in reduced TGW. Conversely, when the grain number is lower, each grain receives more resources, leading to increased grain weight. This dynamic balance mechanism allows rice plants to maintain relatively stable yields under varying environmental conditions. Notably, this trade-off manifests differently across rice-growing regions. For example, rice cultivars in southern China generally have higher GPP but lower TGW, whereas those in northern China tend to have fewer grains per panicle but higher grain weight. Such differences likely reflect long-term adaptation of different rice types to their native environments.

Moreover, we observed that with advancements in breeding technologies, some modern cultivars have started to break this strict negative relationship. Certain high-yield cultivars are capable of maintaining both high grain number and relatively high grain weight, suggesting that refined breeding strategies can partially overcome the trade-off between GPP and TGW. For instance, in one study, researchers knocked out maize KRN2 and rice OsKRN2, increasing grain yield by approximately 10% and 8%, respectively, without significant changes in other agronomic traits [26]. However, overcoming this trade-off is challenging. It requires a deep understanding of the physiological and biochemical mechanisms regulating grain number and weight and a holistic approach in breeding that integrates multiple traits. For example, enhancing photosynthetic capacity [27], optimizing nutrient transport and allocation [28], and improving panicle architecture [29] may all contribute to simultaneously increasing both grain number and grain weight.

4.3. Importance of Region-Specific Breeding Strategies

This study highlights the significant regional differences in rice cultivar improvement across major rice-growing regions of China by analyzing agronomic traits and yield data, emphasizing the importance of region-specific breeding strategies. Our findings reveal distinct differences in yield components and agronomic traits among rice cultivars across regions, which can be attributed to significant interactions between cultivar traits and regional environments [30]. For instance, rice cultivars in the NEC region exhibit longer GD and higher TGW, while those in the SC region have markedly shorter GD. These differences reflect the cultivars’ adaptation to specific regional climates and cultivation practices.

Currently, 84% of the global rice production areas are facing yield plateau pressures [4]. Our observations indicate regional heterogeneity in yield growth patterns. Although the NEC region has a higher baseline yield, its annual yield growth rate is only 0.16–0.45%, whereas the CC region, such as Jiangxi, has an annual growth rate as high as 1.42%. This disparity suggests that different regions are at different stages of yield enhancement. The NEC region may have already reached its yield plateau, while the CC region has not yet, highlighting the need for differentiated breeding strategies. Such strategies should aim to maintain high yields in plateaued regions and facilitate rapid yield increases in regions that have not yet reached their potential. Previous studies indicate that temperature and radiation effects on rice yield vary significantly between regions. For example, minimum and maximum temperatures have opposite effects on yield, with these impacts differing across growth stages and regions [31]. This further underscores the necessity of region-specific breeding strategies. Additionally, our results show that GD is not only decreasing over time but also varies significantly by region. Other studies have demonstrated that genetic improvements have significantly impacted rice growth durations, typically shortening the sowing to heading period while extending the heading to maturity period. This adjustment in GD could be an adaptation to regional climate change or a conscious choice by breeders [32].

However, implementing region-specific breeding strategies also presents challenges. Modern cultivars have accumulated favorable alleles, especially those related to yield, during the breeding process. At the same time, they have lost certain beneficial alleles related to stress resistance [33]. This indicates that while increasing yield, maintaining cultivar resilience and adaptability remains a critical challenge.

5. Conclusions

This study offers a comprehensive analysis of rice cultivar performance across China over the past three decades (1990–2023), revealing both spatial and temporal trends in key agronomic traits and yield. Our findings highlight significant regional variations in rice production: regions such as Northeast China exhibit stable but plateaued yields, such as 0.16% in Jilin, while Central China shows rapid growth in yield and trait evolution, particularly in Jiangxi (1.42%) and Hunan (1.13%). Meanwhile, the trade-offs between yield components highlight the need for balanced breeding strategies, while regional differences emphasize the importance of locally adapted breeding programs. Future research should focus on understanding yield component trade-offs and regional adaptations. For breeders, we recommend adopting differentiated strategies based on regional characteristics—breaking yield plateaus in Northeast China and accelerating yield improvement in Central China. These insights provide valuable guidance for future rice breeding efforts, focusing on maximizing yield potential while maintaining regional adaptability.