中国农业科学 ›› 2018, Vol. 51 ›› Issue (2): 233-245.doi: 10.3864/j.issn.0578-1752.2018.02.004

• 耕作栽培·生理生化·农业信息技术 • 上一篇    下一篇

交替沟灌玉米灌浆期茎流影响因子敏感性分析与模型适用性研究

杜斌1,胡笑涛1,王文娥1,马黎华2,周始威1

 
  

  1. 1西北农林科技大学旱区农业水土工程教育部重点实验室,陕西杨凌 7121002西南大学资源环境学院,重庆 400715
  • 收稿日期:2017-05-25 出版日期:2018-01-16 发布日期:2018-01-16
  • 通讯作者: 胡笑涛,E-mail:huxiaotao11@nwsuaf.edu.cn
  • 作者简介:杜斌,E-mail:287684270@qq.com
  • 基金资助:
    国家自然科学基金(51179163)

Stem Flow Influencing Factors Sensitivity Analysis and Stem Flow Model Applicability in Filling Stage of Alternate Furrow Irrigated Maize

DU Bin1, HU XiaoTao1, WANG WenE1, MA LiHua 2, ZHOU ShiWei1   

  1. 1Key Laboratory of Agricultural Soil and Water Engineering in Arid and Semiarid Area of Ministry of Education, Northwest A&F University, Yangling 712100, Shaanxi; 2college of Resources and Environment, Southwestern University, Chongqing 400715
  • Received:2017-05-25 Online:2018-01-16 Published:2018-01-16

摘要: 【目的】应用人工神经网络对不同处理玉米茎流进行精准预测,为推算玉米蒸腾耗水量以及制定合理的灌水方案提供新思路。【方法】试验研究对象为夏玉米,品种为西农985。试验设置3个处理,分别为交替沟灌高水处理(alternate furrow irrigation,AFI1)、交替沟灌低水处理(alternate furrow irrigation,AFI2)和常规沟灌处理(conventional furrow irrigation,CFI)。AFI1、AFI2每次灌水量为CFI灌水量的2/3和1/2。利用通径系数与互信息分析不同处理的影响因素与玉米茎流相关关系,基于人工神经网络理论建立了玉米茎流速率估算模型,以主成分回归模型为对比,对两个模型预测精度和稳定性进行评价。【结果】(1)不同处理对环境因子的响应有所差异,影响CFI、AFI1玉米茎流的主要因素是气象因子,影响AFI2处理玉米茎流的主要因素是土壤水分;(2)不同土层含水量对各处理茎流的影响也有所不同,研究发现10—20 cm和20—30 cm土层含水量与玉米茎流相关程度最高。利用不确定性分析法进一步分析得出,常规处理与高水处理水平下,与茎液流变化关系最密切的土壤含水层为20—30 cm,其次是10—20 cm,低水处理水平下,最敏感的土层为10—20 cm,其次是20—30 cm;(3)根据模型误差分析与模型不确定性分析,神经网络模型RMSE、d-factor值较小,R2值达到了0.9以上,说明神经网络模型预测精度更高,模型更稳定。【结论】与传统方法相比,人工神经网络模型可以快速准确地对茎流进行预测,对指导玉米灌溉具有重要的指导意义。

关键词: 交替沟灌, 玉米茎流, 人工神经网络, 分层土壤含水量

Abstract: 【Objective】The artificial neural network is used to predict the stem flow of maize with different treatments, which provides a new idea for estimating the water consumption of maize so as to make reasonable irrigation plan. 【Method】 The summer maize, variety of Xinong 985, was selected for study. The test set three treatments: alternative furrow irrigation high water treatment (alternate furrow irrigation, AFI1), alternate furrow irrigation low water treatment (alternate furrow irrigation, AFI2), and conventional furrow irrigation (conventional furrow irrigation, CFI). AFI1 and AFI2 each irrigation amount was 2/3 and 1/2 irrigation amount of CFI, respectively. In this paper, an artificial neural network is established to estimate the corn stem flow rate, and the model is compared with the principal component regression model, and the accuracy and stability of the two models are evaluated.【Result】(1) The response of different treatments to environmental factors was different. Meteorological conditions were the main factors affecting stem flow of the AFI1 treatment, and soil moisture was main factor affecting stem flow of the AFI2 treatment. (2) The influence of water content in different soil layers on stem flow was also different. It was found that the moisture contents of 10-20 cm and 20-30 cm layers were most correlated with maize stem flow. By using the uncertainty analysis method further analysis: For CFI and AFI1 treatments, 20-30 cm layer soil content was most closely related to sap flow changes, then 10-20 cm layer soil content. And for AFI2 treatment, 10-20 cm layer soil content was most closely related to sap flow changes, then 20-30 cm layer soil content; (3) After comparing the R2 , RMSE and d-factor values of two models, the ANN model was more stable and accuracy, which was found to be the best model to predict the stem flow.【Conclusion】Compared with traditional methods, artificial neural network model can greatly enhance the prediction accuracy of stem flow, and it can provide some guidance for making reasonable irrigation plan of maize.

Key words: alternate furrow irrigation, maize stem flow, artificial neural network, layered soil moisture content