自然资源学报 ›› 2018, Vol. 33 ›› Issue (4): 669-683.doi: 10.11849/zrzyxb.20161178

• 资源评价 • 上一篇    下一篇

基于BP神经网络的三峡库区土壤侵蚀强度模拟

刘婷1, 邵景安1,2,*   

  1. 1.重庆师范大学地理与旅游学院, 重庆 400047;
    2.三峡库区地表过程与环境遥感重庆市重点实验室,重庆 400047
  • 收稿日期:2016-10-31 修回日期:2017-07-15 出版日期:2018-04-10 发布日期:2018-04-10
  • 通讯作者: 邵景安(1976- ),男,安徽亳州人,博士,研究员,主要从事土地利用与生态过程研究。E-mail: shao_ja2003@sohu.com
  • 作者简介:刘婷(1989- ),女,重庆长寿人,主要从事水土保持研究。E-mail: liu_tt@sohu.com
  • 基金资助:
    重庆市基础科学与前沿技术研究重点专项(cstc2017jcyjB0317)

Simulation of Soil Erosion Intensity in the Three Gorges Reservoir Area Using BP Neural Network

LIU Ting1, SHAO Jing-an1,2   

  1. 1.College of Geography and Tourism, Chongqing Normal University, Chongqing 400047, China;
    2. Key Laboratory of Surface Process and Environment Remote Sensing in the Three Gorges Reservoir Area, Chongqing 400047, China
  • Received:2016-10-31 Revised:2017-07-15 Online:2018-04-10 Published:2018-04-10
  • Supported by:
    Chongqing Research Program of Basic Research and Frontier Technology, No. cstc2017jcyjB0317.

摘要: 降雨侵蚀力变化是一复杂过程,其变化存在一定的随机波动性,土壤侵蚀是三峡库区生态环境脆弱最主要的影响因素之一,查明库区土壤侵蚀强度的演化过程及未来趋势是库区生态文明建设过程中急需解决的关键科学问题。论文基于三峡库区1990年侵蚀降雨特征,利用BP神经网络对2010年75个站点降雨侵蚀力进行模拟、验证,预测2030年75个站点降雨侵蚀力。选取2030年预测结果中位于库区周围的27个站点,结合2030年库区自然增长、生态保护情景下土地利用模拟数据,使用RUSLE计算2030年土壤侵蚀强度。结果表明:1)2010年库区降雨侵蚀力模拟相对误差为15%,测试样本数据相对误差为14.67%,预测相对误差为19.65%,NE系数为0.85,说明BP神经网络对库区降雨侵蚀力具有良好模拟效果;2)2010年库区土壤侵蚀强度的Kappa指数为0.75,计算结果能满足模拟与预测需求;3)在土地利用不变情况下,2030年库区轻度、中度侵蚀面积均有所增加,微度及强烈以上侵蚀面积均呈减少趋势,且侵蚀强度转变中的58%来源于相邻侵蚀强度,跨侵蚀等级区的较少;4)在降雨侵蚀力不变情况下,自然增长、生态保护情景下未来土地利用变化所导致的土壤侵蚀均呈下降趋势,后者下降的趋势更为明显;5)在降雨侵蚀力及土地利用均变化的情况下,自然增长、生态保护情景下土壤侵蚀均呈下降趋势。

关键词: BP神经网络, RUSLE, 三峡库区, 土壤侵蚀强度

Abstract: Soil erosion is one of the most important factors affecting the fragility of the ecological environment in the Three Gorges Reservoir area. The change of rainfall erosivity is a complex process and its variation has certain stochastic fluctuation. Understanding the evolution of soil erosion intensity and its future trends are the key scientific issues, which need to be resolved in the process of ecological civilization construction in the Three Gorges Reservoir area. Moreover, it is of great significance to build an appropriate ecological production paradigm, and to formulate measures to prevent and control soil erosion. Based on the characteristics of rainfall erosion in the Three Gorges Reservoir area in 1990, this paper simulated and verified the rainfall erosivity at 75 stations in 2010 using BP neural network. On this basis, the rainfall erosivity at 75 stations in 2030 was predicted. The forecast results of rainfall erosivity at 27 stations located around the Three Gorges Reservoir area were selected and interpolated with Kriging method. Combined with the simulated land use in the Three Gorges Reservoir area in the natural growth and ecological protection scenarios in 2030, the soil erosion intensity in 2030 was calculated using the revised soil loss equation (RUSLE). The results were as follows: In 2010, the relative error of rainfall erosivity simulation was 15%, the relative error of tested samples was 14.67%, the relative error of prediction was 19.65%, and the NE coefficient was 0.85, which indicated that BP neural network had a good result of rainfall erosivity simulation in the Three Gorges Reservoir area. In 2010, the Kappa index of soil erosion intensity in the Three Gorges Reservoir area was 0.75, and the overall calculation results could meet the needs of simulation and prediction. When land use does not change in the Three Gorges Reservoir area till 2030, the areas of slight and moderate erosion will both increase, the areas of micro erosion and above intensity erosion will decrease. About 58% of the change in erosion intensity happen between adjacent erosion intensity types, and less change happen across erosion grade types. Under the condition of constant rainfall erosivity, the soil erosion caused by future land use change in both natural growth scenario and ecological protection scenario has decreasing tendency, while the tendency in the latter scenario is more obvious. If both rainfall erosivity and land use change, soil erosion in both scenarios show downward trends. However, there will be a certain degree of deterioration in areas with less erosion. Therefore, the policies of “controlling severe erosion, preventing slight erosion” should be taken. It was worth noting that simulated results only show the possibility of rainfall erosivity change which are not completely deterministic.

Key words: BP neural network, RUSLE, soil erosion intensity, the Three Gorges Reservoir area

中图分类号: 

  • S157