基于多源数据融合的大豆||玉米带状复合种植土壤水分动态模拟与预测OA
Dynamic simulation and prediction of soil moisture in soybean-maize strip intercropping based on multi-source data fusion
精准模拟土壤水分动态并据此制定合理的灌溉管理策略,是提升大豆||玉米带状复合种植系统生产效益、作物水分利用效率的关键前提.以黄淮海地区典型大豆||玉米带状复合种植系统为研究对象,于 2024-2025 年在山东省泰安市开展田间试验,设置不同灌溉阈值与种植带型处理,构建了基于 NDVI 动态反演作物系数(Kc)的土壤水分运移机理模型,以及长短期记忆网络(Long Short-Term Memory Network,LSTM)土壤水分预测模型.结果表明,基于NDVI 动态反演 Kc 的机理模型能够准确刻画土壤水分动态变化过程,对大豆和玉米土壤水分的模拟决定系数(R2)分别达 0.942 和 0.947;LSTM 模型通过融合多源数据实现土壤水分预测,测试集 R2 为 0.789.所建立的机理模型与数据驱动模型协同框架,兼顾了模型的可解释性与预测性能,可为大豆||玉米带状复合种植系统的精准灌溉管理提供理论依据与技术支撑.
Accurate simulation of soil moisture dynamics and the consequent formulation of reasonable irrigation management strategies are identified as the key prerequisites for improving the production efficiency and crop water use efficiency of the soybean-maize strip intercropping system.A typical soybean-maize strip intercropping system in the Huang-Huai-Hai region was selected as the research object.Field experiments were conducted in Tai'an City,Shandong Province,during 2024-2025,with different irrigation thresholds and planting strip configurations being set.A soil water transport mechanism model was constructed based on the dynamic inversion of the crop coefficient Kc using NDVI,and a long short-term memory(LSTM)network-based soil moisture prediction model was also developed.The results indicated that the mechanism model based on NDVI-derived Kc could accurately characterize the dynamic changes in soil moisture,with the coefficients of determination(R2)for soil moisture simulation being 0.942 and 0.947 for soybean and maize,respectively.The LSTM model,which integrated multi-source data for soil moisture prediction,achieved an R2 of 0.789 on the test set.Through the established collaborative framework combining the mechanism model and the data-driven model,both interpretability and predictive performance were taken into account.Thereby,a theoretical basis and technical support were provided for precision irrigation management in the soybean-maize strip intercropping system.
李金叶;王旭冲;王帅;丁文浩;张金诺;于胜男;陈国庆
山东农业大学 农学院,山东 泰安 271018山东农业大学 农学院,山东 泰安 271018山东农业大学 农学院,山东 泰安 271018山东农业大学 农学院,山东 泰安 271018山东农业大学 农学院,山东 泰安 271018山东农业大学 农学院,山东 泰安 271018山东农业大学 农学院,山东 泰安 271018
农业科技
大豆||玉米带状复合种植作物系数土壤水分运移模型长短期记忆网络多源数据融合
soybean-maize strip intercroppingcrop coefficientsoil water movement modellong short-term memory networkmulti-source data fusion
《农业装备与车辆工程》 2026 (8)
1-11,11
山东省重点研发计划(重大科技创新工程)项目(2023CXGC010703)黄河流域粮油复合种植深度节水模型构建及应用示范山东省重点研发计划(科技型中小企业创新能力提升工程)项目:"关键技术研发与示范以及多源数据驱动的作物水分精准管控技术创新支持及应用示范"(2024TSGC0038)
评论