重庆山地丘陵区紫色土饱和导水率传递函数研究

发布时间：2018-02-25 18:15

本文关键词： 紫色土饱和导水率影响因素多元非线性回归 BP神经网络　出处：《西南大学》2017年硕士论文　论文类型：学位论文

【摘要】：紫色土广泛分布于我国西南和南方山地丘陵区,川渝丘陵区及低山区分布最广。三峡库区中紫色土耕地面积占70%以上,并且库区内耕地以坡耕地为主。紫色因其特殊的成土过程导致抗蚀性差,而且土壤的质地松软,紫色土分布的地区也是水土流失高发区。同时,三峡库区因地势起伏变化大,虽然年降雨量大但时空分布不均,致使库区内土壤养分和水土流失异常严重。因此,本文以三峡库区腹地的云阳县、奉节县、开州区、万州区、梁平区、忠县和丰都县7个渝东北区县为研究区域,在研究区内设置土壤样品采样点,挖取土壤剖面,分别在0-10 cm、10-20 cm及20-30 cm处采集土样,通过试验测定饱和导水率及相关的土壤理化性质,分析研究区饱和导水率及土壤理化性质的空间分布特性,以及各土壤理化性质参数对土壤饱和导水率的影响,并建立不同土壤层次饱和导水率与土壤理化性质参数之间的定量模型。研究传统传递函数模型在本研究区的适应性,并应用多元非线性回归法和BP神经网络技术,分别构建不同土壤层次的土壤饱和导水率传递函数模型,为山地丘陵区土壤中物质的运移、区域耗水规律、水土保持治理、农业面源污染治理等提供了一定的方法参考和决策支持,主要结论如下:(1)随着土壤深度增加,土壤饱和导水率、饱和含水量和土壤有机质均呈现出逐渐递减的规律,而土壤容重却随土壤深度增加而增加,不同深度的土壤颗粒均以粉粒为主,其次为砂粒、粘粒,土壤颗粒以0.05~0.002 mm居多。各层次土壤的饱和导水率与有机质、饱和含水量联系较为紧密,而与土壤容重和土壤质地的相关性不显著。(2)三个土层的饱和导水率均随有机质含量的增加而增加,它们之间的相关性较高,并且饱和导水率与有机质含量呈指数函数关系变化;三个土层的饱和导水率随土壤容重的增大在减小,并呈指数函数关系,且土壤容重对饱和导水率的影响不显著;浅层土壤(0-10 cm、10-20 cm)的饱和导水率与饱和含水量均呈现显著正相关,并且呈二次曲线关系,而20-30 cm土层饱和导水率与饱和含水量的相关性不显著;研究区饱和导水率受土壤质地的影响较小,仅20-30cm土层粘粒含量对饱和导水率有显著负相关关系,其它土层的相关性不显著。(3)运用Campell、Cosby、Saxton、Wosten1997、Wosten1999和Puckett 6种传递函数对不同深度土壤饱和导水率进行估算的结果都不理想,表明前人建立的传递函数模型已经不能适用于本研究区饱和导水率的预测。(4)采用多元非线性回归方法建立的不同深度土壤的饱和导水率传递函数模型估算效果较好,模型的预测值与实测值基本相当,模型的拟合效果能够满足估算的要求,表明利用有机质、饱和含水量、土壤质地等通过多元非线性传递函数模型进行不同土壤层次的饱和导水率的预报是可行的。(5)利用MATLAB工作平台建立的0-10 cm、10-20 cm和20-30 cm三个土壤层次的饱和导水率BP神经网络传递模型,其饱和导水率的预测值与实测值误差最小,是本文研究中预测精度最高的模型。(6)在本文的研究中,构建的多元非线性回归模型和BP神经网络模型预测土壤饱和导水率的精度均较高,并且BP神经网络传递模型的误差更小,但其构建过程繁杂,不如多元非线性回归模型简单易于操作,在实际生产应用中,应综合考虑实际情况合理选择预报模型。
[Abstract]:Purple soil is widely distributed in Southwest China and southern hilly area, Sichuan Hilly and low mountainous areas. The most widely distributed in purple soil area of cultivated land in Three Gorges area accounted for more than 70%, and the area of cultivated land in purple slope land. Because of the special soil processes lead to poor corrosion resistance, and the soil texture is soft, the purple soil distribution area and soil erosion in Three Gorges Reservoir area. At the same time, because of the terrain changes, although the annual rainfall is large but uneven distribution, the area of soil nutrient and soil erosion is very serious. Therefore, the hinterland of the Three Gorges Reservoir Area in Yunyang County, Fengjie County, the State District, Wanzhou District, Liangping area Fengdu County, Zhongxian and 7 northeast of Chongqing county as the study area in the study area is arranged in the soil sample sampling points, digging soil profile, respectively at 0-10 cm, 10-20 cm and 20-30 cm soil samples were collected, through test determination of saturated water The physicochemical properties of soil and the related rate, analysis of the spatial distribution characteristics of saturated hydraulic conductivity and soil physical and chemical properties, and the physicochemical properties of soil parameters on the effect of soil saturated hydraulic conductivity, and the establishment of different soil layers of saturated hydraulic conductivity and soil physicochemical properties of the quantitative model between the parameters of traditional. Transfer function model in the study area and adaptability, using multiple nonlinear regression method and BP neural network technology, were constructed in different soil layers and soil saturated hydraulic conductivity transfer function model for migration of substances in the soil hilly area in the region, water consumption, soil and water conservation, provides some reference and decision making method support the agricultural non-point source pollution control, the main conclusions are as follows: (1) with the increase of soil depth, soil saturated hydraulic conductivity, saturated water content and soil organic matter showed the law of diminishing, and soil But the soil bulk density increased with soil depth, soil particles with different depth were dominated by silt, followed by sand, clay, soil particles with 0.05~0.002 mm. The saturated hydraulic conductivity and soil organic matter, saturated water content closely, no significant correlation between volume and weight and soil texture and soil. (2) the three soil saturated hydraulic conductivity increased with the content of organic matter, high correlation between them, and the saturated hydraulic conductivity and organic matter content changed in the relation of exponential function; three soil saturated hydraulic conductivity with the increase in soil bulk density decreases, and the exponential function the relationship, and the effect of soil bulk density on saturated hydraulic conductivity is not significant; the shallow soil (0-10 cm, 10-20 cm) of the saturated hydraulic conductivity and saturated water content showed a significant positive correlation, and a two curve, and 20-30 cm soil saturated hydraulic conductivity and No significant correlation between the saturated water; the study area saturated hydraulic conductivity is less affected by soil texture, soil clay content of only 20-30cm significant negative correlation between saturated hydraulic conductivity, no significant correlation between other soil layers. (3) the use of Campell, Cosby, Saxton, Wosten1997, Wosten1999 and Puckett 6 kinds of transfer function different depth of soil saturated hydraulic conductivity were estimated. The results are not ideal, that established the transfer function model is not suitable for the study area. The prediction of water saturated rate (4) using multiple nonlinear regression method to establish the different depth of soil saturated hydraulic conductivity transfer function estimation model better, prediction model the value and the measured value is the result of fitting of the model can meet the requirements that estimate, using organic matter, saturated water content, soil texture through multivariate nonlinear transfer function model. For different levels of soil saturated hydraulic conductivity prediction is feasible. (5) using the MATLAB platform built 0-10 cm, 10-20 cm and 20-30 cm three level of soil saturated hydraulic conductivity BP neural network transfer model, the saturated water values of the minimum error rate prediction and the measured value, the accuracy of the model is the highest prediction in this study. (6) in this study, the prediction of soil saturated hydraulic conductivity of the accuracy of multivariate nonlinear regression model and BP neural network model, BP neural network and error transfer model, but its construction process is complicated, as the multivariate nonlinear regression model is simple and easy to operate in. In practical application, should consider the reasonable selection of prediction model of the actual situation.

【学位授予单位】：西南大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：S152.7

【参考文献】