典型草原栗钙土层厚度模型的研究-农业生物环境边与能源工程专业毕业论文.pdf.docxVIP

摘 摘 要 针对典型草原土壤结构特征及JxL蚀退化的现状，在内蒙古自治区乌珠穆沁典型草 原选取约400平方公罩面积的研究区，采用GPS、GIS技术结合地面调查取样测量的 方法，获取典型草原地表特征、采样点的三维坐标数据及各实验点栗钙土层厚度，共 获得实验数据点565个。利用SuperMap软件对基础数据进行预处理，利用SQL Server 数据库对数据进行管理，建立克里格插值模型和基于MatLab的BP神经网络模型，对 模型结果进行研究分析。 研究结果表明：乌珠穆沁典型草原平坦区域克罩格插值预测最大相对误差为 2．816％，最小相对误差为0．952％。BP神经网络预测平坦地区最大相对误差为2．747 ％，最小相对误差为0．195％；山坡地带最大相对误差为6．366％，最小相对误差为 1．897％；低山丘陵地带最大相对误差为13．172％，最小相对误差为0．693％。模型 预测精度基本满足要求，BP神经网络与克罩格插值的预测精度均可以作为典型草原 栗钙土层厚度预测的～种有效的方法。 典型草原退化与荒漠化的实质是栗钙土层逐渐被风蚀的结果，利用栗钙土层预测 模型进行典型草原土壤与植被空问变异的研究，可以为典型草原荒漠化防治与科学管 理提供基础数据。 关键词：乌珠穆沁典型草原；栗钙土层；土壤厚度；克里格插值：BP神经网络模型 Study Study on Chestnut Soil Tier Thickness Model in the Typical Grassland Abstract About 400 km‘areas of WuZhuMuQin typical grassland in Inner Mongolia was selected as the research subject which was against the phenomenon of the soil structural characteristics and wind erosion degradation．The surface characteristi cs of the typical grassland．three．dimensional coordinate data of the sampling points and chestnut soil tier thickness were obtained by means of GPS、GIS technologies combined with ground survey and sampling measurement methods．565 experimental data points were obtained．The basic data were dealt with by SuperMap software beforehand and managed by SQL Server database．The kriging method model and the BP neural network model based on Matlab was set up．The resuR of the model was analyzed． The result shows that the largest relative error for the kriging method forecasting in the flat grassland iS 2．8 l 6％．the smallest one iS 0．952％．The result of the BP neural network model for forecasting indicates that the largest relative error in the flat grassland iS 2．747％．the smallest one iS 0．1 95％；the largest relative error in the hillside of grassland iS 6．366％，the smallest one iS 1．897％；the largest relative error in low mountain hill of grassland iS l 3．1 72％．the smallest one iS 0．693％．The forecast accuracy of the model meets the basic requirements．The forecast accuracy of the BP neural network and the krigi