Journal of Integrative Agriculture ›› 2024, Vol. 23 ›› Issue (9): 2970-2988.DOI: 10.1016/j.jia.2023.10.005

• • 上一篇    下一篇

基于机器学习构建智能决策系统提升中国西南地区稻-油、稻-麦和稻-蒜轮作系统综合效益的策略

  

  • 收稿日期:2023-05-22 接受日期:2023-08-31 出版日期:2024-09-20 发布日期:2024-08-20

Strategies for improving crop comprehensive benefits via a decision-making system based on machine learning in the rice‒rape, rice‒wheat and rice‒garlic rotation systems in Southwest China

Xinrui Li1*, Xiafei Li1*, Tao Liu1, Huilai Yin1, Hao Fu1, Yongheng Luo1, Yanfu Bai2, Hongkun Yang3, Zhiyuan Yang1, Yongjian Sun1, Jun Ma1, Zongkui Chen1#

  

  1. 1 Crop Ecophysiology and Cultivation Key Laboratory of Sichuan Province/Rice Research Institute/State Key Laboratory of Crop Gene Exploration and Utilization in Southwest China/Sichuan Agricultural University, Chengdu 611130, China
    2 College of Grassland Science and Technology, Sichuan Agricultural University, Chengdu 611130, China
    3 College of Agronomy, Sichuan Agricultural University, Chengdu 611130, China
  • Received:2023-05-22 Accepted:2023-08-31 Online:2024-09-20 Published:2024-08-20
  • About author:Xinrui Li, E-mail: 3014804562@qq.com; #Correspondence Zongkui Chen, E-mail: chenzongkui@sicau.edu.cn * These authors contributed equally to this study.
  • Supported by:
    This work was supported by the China Postdoctoral Science Foundation (2022M722301), the Sichuan Province Innovative Talent Funding Project for Postdoctoral Fellows, China (BX202207) and the Natural Science Foundation of Sichuan Province, China (2023NSFC0014 and 2024NSFSC1225).

PDF

摘要：

稻-油、稻-麦和稻-蒜轮作是中国西南地区常见的轮作模式，在确保该地区生态和经济效益以及应对全国粮食安全等问题方面发挥了重要的作用，但该地区的此轮作系统的稻谷产量和其他作物产量等比全国平均水平低1.25%-14.7%。利用机器学习构建智能决策系统分析各轮作系统投入产出特征，有利于获得更好的综合效益，但相关研究较少。因此，该研究利用data-intensive approach法，基于机器学习构建智能决策系统，以期为提升西南地区稻-油、稻-麦、稻-蒜轮作的综合效益提供依据。研究结果表明，在稻-蒜系统中，高肥料投入的基础上增加种子投入可提高产量和肥料偏生产力，最终实现最佳综合效益的概率为10%；在稻-油系统中，增加氮肥用量并且减少钾肥用量可以获得更高的产量但易增加温室气体排放，导致该系统仅实现次优效益且概率为8%；在稻-麦系统中，减少氮肥和磷肥的施用可以增加产量和肥料偏生产力，但实现最佳综合效益的概率仅为8%。基于随机森林模型的预测分析，在稻-蒜系统中减少25%的氮肥、增加8%的磷肥和74%的钾肥；在稻-油系统中减少54%和36%的氮肥和钾肥，增加38%的磷肥；在稻-麦系统中减少4%的氮肥并增加65%的磷肥和23%的钾肥，这些措施可以在不同程度上进一步提高轮作系统的综合效益。因此，该研究结果为中国西南地区稻-油、稻-麦和稻-蒜系统中通过优化农业投入以获得更高的综合效益提供了新见解。

Abstract:

Rice‒rape, rice‒wheat and rice‒garlic rotations are common cropping systems in Southwest China, and they have played a significant role in ensuring ecological and economic benefits (EB) and addressing the challenges of China’s food security in the region.  However, the crop yields in these rotation systems are 1.25‒14.73% lower in this region than the national averages.  Intelligent decision-making with machine learning can analyze the key factors for obtaining better benefits, but it has rarely been used to enhance the probability of obtaining such benefits from rotations in Southwest China.  Thus, we used a data-intensive approach to construct an intelligent decision‒making system with machine learning to provide strategies for improving the benefits of rice–rape, rice–wheat, and rice–garlic rotations in Southwest China.  The results show that raising the yield and partial fertilizer productivity (PFP) by increasing seed input under high fertilizer application provided the optimal benefits with a 10% probability in the rice–garlic system.  Obtaining high yields and greenhouse gas (GHG) emissions by increasing the N application and reducing the K application provided suboptimal benefits with an 8% probability in the rice–rape system.  Reducing N and P to enhance PFP and yield provided optimal benefits with the lowest probability (8%) in the rice‒wheat system.  Based on the predictive analysis of a random forest model, the optimal benefits were obtained with fertilization regimes by reducing N by 25% and increasing P and K by 8 and 74%, respectively, in the rice–garlic system,  reducing N and K by 54 and by 36%, respectively, and increasing P by 38% in rice–rape system, and reducing N by 4% and increasing P and K by 65 and 23% in rice–wheat system.  These strategies could be further optimized by 17‒34% for different benefits, and all of these measures can improve the effectiveness of the crop rotation systems to varying degrees.  Overall, these findings provide insights into optimal agricultural inputs for higher benefits through an intelligent decision-making system with machine learning analysis in the rice–rape, rice‒wheat, and rice–garlic systems.

Key words: rice rotation ,  agricultural management ,  greenhouse gas emissions ,  comprehensive benefits ,  fertilizer management