China's growing population and economic development have significantly increased the demand for maize, both for food and animal feed. This surge in demand raises crucial questions about China's ability to maintain maize production self-sufficiency, especially given the observed trend of declining growth rates in global crop yields. China, the world's second-largest maize producer, currently imports a substantial amount of maize to meet its domestic needs. Continued reliance on imports poses risks to global food security and price stability. Therefore, understanding the potential for yield improvement in China is vital for ensuring both national and global food security. Crop yield is a complex interplay of genetics, environmental factors (climate, soil), and agricultural management practices. Studies in other maize-producing regions highlight the importance of plant density as a key factor in yield gains. While Chinese farmers have increased maize plant density over time, the extent to which this strategy, coupled with soil management and improved cultivars, can meet future demands remains a critical question. This study addresses this gap by employing a data-driven approach to project optimal planting density and its associated yield improvements under current and future climate scenarios. The study uses a machine learning algorithm (Random Forest) to model optimal planting density (OPD) as a function of environmental factors, management practices, and soil organic matter content. This model is then used to project OPD across different regions of China under current conditions and a future climate change scenario (SSP585). The results are validated with data from 87 field trials across China. Finally, the study assesses whether China can achieve maize self-sufficiency by 2030 given its existing cropping areas and optimal agricultural management.

Existing literature reveals a global pattern of declining growth rates in crop yields, raising concerns about future food security. Studies in regions like Nebraska emphasize the significant role of climate trends and agronomic improvements, rather than solely genetic improvements, in recent maize yield gains. Plant density is consistently identified as a critical factor influencing maize yield. In North America, optimal plant density has steadily increased, contributing significantly to yield increases. Brazil’s maize productivity improvements are also linked to increased density-tolerance in modern hybrids. Similarly, France experienced a dramatic increase in maize yields correlated with increased planting density. China has also seen increases in planting density, albeit at a slower rate compared to North America. While statistical approaches based on large datasets offer insights into yield-density relationships, uncertainties remain due to data limitations. Machine learning algorithms, with their capacity to handle complex interactions, offer a more robust approach to predicting OPD and its impact on yield. Existing studies on maize OPD prediction in China often rely on simplified models, potentially overestimating OPD due to neglecting genotype-environment-management interactions. Therefore, this study employs a more sophisticated machine learning approach to address these limitations and provide a more accurate prediction of OPD and its potential to enhance maize production in China.

This study employs a data-driven approach using a Random Forest (RF) algorithm to determine the optimal planting density (OPD) for maize across China. The RF model incorporates six key variables: daily minimum temperature (Tmin), daily maximum temperature (Tmax), precipitation (Prec), solar radiation (Radn), growing-degree days (GDD), and soil organic matter content (SOM). The model was trained and tested using a database of 2442 paired yield-density observations from 125 studies published between 2000 and 2021. The performance of the RF model was evaluated using metrics such as R-squared, RMSE, and RRMSE. The trained model was then applied to a larger dataset comprising 402 stations across major maize-producing areas in China to project OPD under current conditions. The projected OPDs were compared with results from 87 independent field trials conducted across China between 2017 and 2020. These field trials involved two treatments: a control treatment using farmers' practices and an optimum treatment using the projected OPD and appropriate hybrids. To assess the impact of future climate change, the RF model was also used to project OPD under a high-end radiative forcing scenario (SSP585) from the CMIP6 climate models for the 2030s. The impact of soil improvement (optimizing SOM to 20 g kg⁻¹) on OPD and yield was also analyzed under both current and future climate scenarios. Finally, the study estimated China's maize demand for 2035 and assessed whether the projected yield increases under the optimal management scenario would be sufficient to meet this demand.

The OPD-RF model explained 60% of the variance in OPD, demonstrating good predictive performance. Analysis of variable importance revealed that Tmin, Radn, and SOM were consistently among the top three influential factors across different regions. Spatial patterns of OPD simulated by the RF model indicated an average OPD of 7.8 × 10⁴ plants ha⁻¹ across China, with considerable variation among regions. Comparison with farmers' current practices revealed significant gaps between current planting densities and the projected OPDs, indicating substantial potential for yield improvement. Field trial results confirmed the feasibility of achieving the projected OPD and associated yield gains. On average, the optimum treatment resulted in a 21% yield increase compared to the control, nearly doubling the current farmers' yield. Genetic improvements and density increases contributed 5.9% and 7.3% to yield improvement respectively. The density × genetics interaction contributed 7.4% of yield improvement. Projections for the 2030s under the SSP585 scenario showed a 1.6% average decrease in OPD due to climate change, with larger decreases in some regions. However, soil improvement (optimizing SOM to 20 g kg⁻¹) offset these negative climate impacts. Considering both climate change and soil improvement, the study projects a 52% increase in grain yield by 2030s compared to historical trends. This projected yield increase would be sufficient to meet 100% of China's projected maize demand in 2035, based on current cropping areas.

The findings of this study highlight the significant potential for improving maize yields in China through optimal crop management, particularly through the optimization of planting density. The results challenge the prevailing view that yield stagnation has reached an attainable maximum in most global areas. The substantial yield gains projected by optimizing plant density, in combination with soil improvements and the use of appropriate hybrids, demonstrate the effectiveness of an integrated crop-soil management approach. The ability of soil improvements to offset the negative impacts of climate change on OPD emphasizes the importance of sustainable soil management practices. The study's methodology, combining machine learning with extensive field data, provides a robust framework for projecting OPD and yield under various climate scenarios. While the study focuses on optimizing planting density, the actual implementation may require additional measures such as irrigation and fertilization to address increased competition for resources at higher densities. The projected yield increases are contingent upon farmers adopting the recommended practices. The study's results suggest that achieving the projected yield gains requires not only technological advancements but also changes in farmers' practices and policies that support these changes. Further research could explore the interactions between genotype, environment, and management in more detail to further enhance yield potential.

This study demonstrates a significant potential for China to achieve maize self-sufficiency by 2030 by optimizing planting density. Combining data-driven projections with field trials, the study reveals a substantial yield improvement potential through denser planting coupled with suitable hybrid varieties and soil improvements, even under future climate change scenarios. The findings emphasize the importance of integrated crop-soil management strategies and challenge the notion of yield stagnation. Future research should focus on addressing practical limitations to implementing optimal planting densities and further exploring the complex interactions among genotype, environment, and management to maximize yield potential.

The study's projections are based on a specific climate change scenario (SSP585), and the results may vary under different scenarios. The analysis simplifies some aspects of real-world conditions, such as water stress and nutrient deficiencies which could become more pronounced at higher planting densities. The success of achieving the projected yield gains depends on farmers' adoption of the recommended practices and supportive policies. Furthermore, the model's accuracy could be improved by incorporating additional factors such as disease and pest pressure, as well as lodging risks.