【作者】 魏占民；

【导师】 陈亚新； R.A.Feddes；

【作者基本信息】 内蒙古农业大学 ， 农业水土工程， 2003， 博士

【摘要】 本文针对我国西部干旱区的特大型灌区—河套灌区所具有的独特水文水资源条件，选择两个典型区域分别进行了作物-水分关系和基于SWAP模型模拟的田间灌溉水有效性研究，以揭示和确立水分对干旱区作物产量的影响和量化表达与浅地下水位灌区农田水分运移转化规律和灌溉水对作物生长利用效率的评价方法，为河套灌区以节水为中心的技术改造和可持续发展提供理论基础和技术支撑。研究成果也可为自然环境和农业生产条件类似地区所参考。研究中注重借鉴和采用先进研究手段，在节水灌溉基础理论方面进行新的探索。对于节水灌溉理论的丰富与完善以及我国节水灌溉这一革命性措施的实施和农业的可持续发展具有重要意义。 在河套灌区开展了春小麦、玉米和葵花三种主要作物的非充分灌溉试验，以研究和确立作物产量与水分的量化关系。针对灌区未来可能的缺水趋势，按我国较少采用的控制灌水定额方法设计缺水试验处理。该方法参考灌水经验减少灌溉定额，与常采用的土壤湿度控制法相比更符合当地灌溉技术水平和管理现状，有利于非充分灌溉的推广应用。 对国外具有代表性的Jensen等6个作物水分响应模型从按多元线性回归原理推求模型参数出发进行了评述，深入分析和进一步明确了敏感指标的意义。结果表明，敏感指标是以作物水分响应模型所表达的统一回归方程的偏回归系数，反映回归自变量（水分）对因变量（产量）的作用效应，故可度量作物对水分的敏感程度。所选择的6个模型作物对水分的敏感程度均随敏感指标绝对值的增大而增大。受最小二乘法求优数学解的约束，在求解作物水分响应模型的敏感指标时常出现从作物生理和物理上难以科学解释的情况，这除与试验处理数量多少及缺水水平有关外，也与试验数据的统计分布有关。若试验数据为非正态或非对数正态分布，最小二乘法求优会受到限制。作物水分响应模型均为按非充分灌溉试验得到的经验模型，按数理统计和应用要求使用时应限制在非充分灌溉试验处理的水分下限值以内，不宜随意外推，否则预测或模拟的作物产量可能会有较大误差。研究结果有助于完善作物水分响应模型确认方法和避免在应用中出现偏差。 采用春小麦、玉米两种主要作物的非充分灌溉资料，选择上述6个作物水分响应模型分析和确立了河套灌区这两种主要作物的产量与全生育期和生育阶段水分之间的量化关系。结果表明，两种作物相对减产量与相对亏水量的线性相关系数偏低，显示除缺水程度外缺水时间对产量也有显著影响，线性模型仅比较粗略地表达这两种作物的产量与全生育期水分的关系。5个生育阶段的作物水分响应模型由于数学结构和敏感指标的表达不同对于同一种作物的回归效果存在很大差异，表现出不同的适应性。研究结果表明Minhas模型均能较好的表达本地区这两种作物的产量与生育阶段水分之间的量化关系。这一结果可能是由于Minhas模型的自变量采用b₀和λ_i双重幂指数的数学结构，而能够较好反映干旱区作物对水分敏感的特性。国内外对Minhas模型注意较少，应引起同行的重视。 由于现有作物水分响应模型的数学结构和敏感指标的表达对某些作物品种、受旱形式及生长环境等存在着不适性，传统建模方法仍有诸多不便之处。本文依据人工神经网络的基本原理在充分吸收其最新理论研究成果的基础上，以非充分灌溉试验得到的实测春小麦产量和各生育阶段水分资料为样本建立了基于BP神经网络的春小麦作物水分响应模型。BP神经网络的训练采用了附加动量因子和自适应学习速率相结合的方法参与连接权值的修正，与常规训练方法相比有助于减少网络陷入误差曲面局部极小值的可能性和增大了初始学习速率的任意性。由于训练样本实际存在的量测误差，在网络的训练过程中为保证模型的预测效果不必追求过小的拟合误差。在不同供水量下对春小麦产量的模拟预测表明，该模型能够正确反映作物产量与水分关系的一般规律。通过与当地条件下拟合效果较好的Minhas模型比较，该模型所预测的产量随备生育阶段水分的变化规律与Mi讪as模型敏感指标所反映的敏感程度变化规律相一致，且两者产量预测值接近。BP神经网络在作物水分响应模型的建模和产量模拟预测中是值得借鉴的理论和有效工具。本文开拓了作物水分响应模型建模理论的新思路。 针对一般水平的试验站用水量平衡法计算作物实际腾发量时考虑深层渗漏量和地下水补给的困难，以河套灌区曙光试验站春小麦非充分灌溉试验为例，引进SWAP模型探讨了其在不同水分处理条件下作物实际腾发量计算的应用问题。研究表明，该模型可以可靠地揭示田间水量平衡分量的动态变化过程，与传统水量平衡法相比提供了作物根系层水量平衡分量运动过程的清晰图景和较容易准确确定各水分平衡分量，是获得不同缺水试验处理作物实际腾发量的有效和实用简便的工具。根据各，试验处理实际腾发量模拟结果和实测产量也用上述6个作物水分响应模型分析了春小麦的作物产量与水分的量化关系。由于两种方法得到的春小麦各试验处理的实际腾发量差别不大，因此所得作物水分响应模型类似。 通过对传统的田间灌溉水利用率评估方法的更多还原

【Abstract】 Crop-water relationship and availability of field irrigation water based on SWAP model simulation were studied in two areas of a largest-sized irrigation district-Hetao Irrigation District in the western arid area of China respectively according to its unique characteristics of hydrology and water resources so as to provide a theoretical basis and technical support for its water-saving transformation and agricultural sustainable development. The results can also be applied to other areas with similar natural and agricultural conditions. The study will contribute to the improvement of water-saving irrigation theory and the implementation of water-saving measures as well as agricultural sustainable development of China.The physical meaning of sensitive index for six worldwide representative models of crop response to water (Stewart a, Jensen, Minhas, Blank, Stewart band Singh) was analyzed based on multi-variable linear regression theory and was verified by a case study. Results show that sensitivity of drop increases as the absolute value of sensitive index become larger for all the 6 models. The sensitive index obtained by least-squared method sometimes can not be scientifically explained in terms of crop physiology. The phenomenon is related to the statistical distribution of experiment data in addition to the number of experiment treatment and water deficit level. When the data is not in normal or logarithmic normal distribution, the least-squared method is not applicable to obtaining the sensitive index. Because the models are all empirical ones established based on deficit experiment, the application of them should be limited above the low soil moisture of the water deficit experiment according to statistical theory, otherwise the prediction of crop yield by the models may produce considerable error .In order to establish relationship between crop yield and water use, a deficit irrigation experiment was conducted each for 3 main crops- spring wheat, maize and sunflower in Hetao Irrigation District. According to the manner of possible water shortage in future, the water deficit treatment was designed with controlled water application method (reduce water application with irrigation date remaining unchanged according to the irrigation schedule in use) which is more suitable to the irrigation management of the local area. The relationship between crop yield and water use of spring wheat and maize was studied using the data respectively. The results show that crop yields of spring wheat and maize hold an roughly good linear relationship with seasonal evapotranspirations respectively. Statistical analysis shows that Minhas model is capable to express therelationship of crop yields of spring wheat and maize with their evapotranspirations at different growth stages accurately. The sensitivity indices at different growth stages of the two models differ significantly, which indicates that the time of occurrence of water deficit has a great influence on the reduction of the yield of spring wheat an maize. The evapotranspirations of spring wheat for the field deficit irrigation experiment were calculated using the agro-hydrological simulation model SWAP 2.0. Comparison show the SWAP 2.0 is an attractive and effective tool to obtain evapotranspiration for the study of relationship between crop yield and water use.A model of crop response to water based on BP neural network for spring wheat was developed using deficit irrigation experiment data. Analysis of simulation results of the model shows that the model is able to express correctly the relation between the yield and water use of spring wheat and this method has some unique merit. The comparison with Minhas model, which fits well the experiment data, indicated that the spring wheat’s sensibility to water expressed by these two models is identical and their results of yield prediction agree with each other for same water application. It can be concluded that the BP neural network is a new method suitable to simulate the crop response to water.A eval更多还原