›› 2008, Vol. 16 ›› Issue (3): 251-255.DOI: 10.11733/j.issn.1007-0435.2008.03.008

Previous Articles     Next Articles

Studies on Natural Succession Prediction of Typical Steppe Based on BP-NN

FENG Quan1, SHAO Xin-qing2, WANG Yun-wen 2   

  1. 1. Gansu Agricultural University, Lanzhou, Gansu Province 730070, China;
    2. Department of Grassland Science, College of Animal Science and Technology, China Agricultural University, Beijing 100193, China
  • Received:2008-01-15 Revised:2008-03-19 Online:2008-06-15 Published:2008-06-15

基于BP网络的典型草原群落自然演替预测模型研究

冯全1, 邵新庆2, 王赟文2   

  1. 1. 甘肃农业大学, 甘肃, 兰州, 730070;
    2. 中国农业大学动物科技学院, 北京, 100193
  • 通讯作者: 邵新庆
  • 作者简介:冯全(1969- ),男,四川人,博士,主要从事农业信息研究,E-mail:fquan@sina.com
  • 基金资助:
    农业部公益性行业科研专项(nyhyzx07-022)

Abstract: Focusing on the natural restoration of typical steppe community,the dynamic progress of the natural restoration and succession of the steppe was studied using BP neural network.Several representative community parameters,i.e.litter biomass,soil water content,soil bulk density,soil porosity,soil organic matter,soil microbial biomass C,soil total N,aboveground biomass,and perennial grass density,in the succession were simulate and predicted by BP-NN.The results of emulational experiment show that BP-NN had very good stability and the average prediction error was 2.04% and those indicate that BP-NN is viable to forecast the self-organized succession of typical steppe community.The advantage of artificial neural network lies on its ability to precisely simulate the vaguely understood and uncertain systemic behavior which cannot be realized by the traditional approaches.As a nonlinear approximator,artificial neural network would be an important tool complementary to the comprehensive models.

Key words: Typical steppe, Artificial neural network, Natural succession, Prediction

摘要: 以典型草原植被群落为研究对象,探讨草地生态系统自然演替恢复的动态变化,采用BP人工神经网络,以演替年度为输入量,群落的凋落物、含水量、容重、孔隙度、有机质、微生物量C、土壤N、地上生物量和多年生禾草密度为输出量,对典型草原群落自然演替进程进行模拟和预测.结果表明:BP-NN的稳定性较好,各参数预测结果的平均误差为2.04%,说明BP-NN可适用预测典型草原群落自然恢复演替,其优势在于可模拟了解较少或不确定性和模糊性较大的系统行为,这是传统数学模型所无法实现的,因而是对传统机理模型的重要补充.

关键词: 典型草原, 人工神经网络, 自然演替, 预测

CLC Number: