荒漠草原植物群落现存量的BP神经网络模拟与预测研究  

The research of simulation and prediction on plant community biomass in desert steppe based on BP neural network

作　　者：张爽[1] 卫智军[1] 吕世杰[2] 刘文亭[1] 白玉婷[1] 

机构地区：[1]内蒙古农业大学生态环境学院,内蒙古呼和浩特010019 [2]内蒙古农业大学理学院,内蒙古呼和浩特010018

出　　处：《草原与草业》2016年第1期53-59,共7页Grassland and Prataculture

基　　金：内蒙古自然科学基金项目(2015MS0349);国家自然科学基金项目(31460126);国家重点基础研究计划项目(2014CB138800)

摘　　要：本研究以内蒙古锡林郭勒盟苏尼特右旗短花针茅荒漠草原为研究对象,采用分层取样技术,分别对碱韭和短花针茅为主的草地植物群落进行调查(测定指标为高度、盖度、密度和地上现存量),获得禾本科为主、碱韭为主以及由二者构成的复合样本数据。通过MATLAB软件平台,把植物群落的数量特征(高度、盖度和密度)作为输入因子,以群落现存量作为输出因子来建立BP神经网络模型,并对预测结果进行检验。结果显示,禾本科为主、碱韭为主以及由二者构成的复合样本,采用BP神经网络平均预测准确率分别为94.1%,92.98%,91.01%。因此,BP神经网络可作为草地植物群落地上现存量模拟与预测的有效工具之一。采用BP神经网络对不同草地植物类群进行模拟和预测可能会存在差异,但这种差异会随着样本容量的增大或训练精度的增加而弱化。The grassland biomass of desert steppe is a very important research field of prairie research,it is not only resolve accumulation of the dry matter of desert steppe ecosystem,but also resolve its biomass so that effect the function of ecology.This paper take plant communityof poaceae and allium polyrhizum and its compound plant community indesert steppeas the main research object and the data was come from an investigation which held in Inner Mongolia Xilinguole.According to the quantitative features（natural height,relative coverage degree,relative density）of plant community in desert steppe and the practical data of biomass,we use MATLAB established a BP neural network model.We took quantitative features of plant community as input factors,and took corresponding practical biomass as output factor,so that we constructed a three layers BP neural network and simulated the biomass of plant community.The result shows that the standing crop of poaceae,allium polyrhizum and mixed plant communities dominated the average prediction accuracy rate belong to a good level.This model could predict plant community biomass and it has higher accuracy.

关 键 词：荒漠草原 现存量 BP神经网络 MATLAB 模拟与预测 

分 类 号：S812[农业科学—草业科学]