| 基于支持向量机理论的地下水动态遥感监测模型与应用 |
| |
| 引用本文： | 付俊娥,苏庆珣,潘世兵,路京选.基于支持向量机理论的地下水动态遥感监测模型与应用[J].地球信息科学,2010,12(4):466-472. |
| |
| 作者姓名： | 付俊娥 苏庆珣 潘世兵 路京选 |
| |
| 作者单位： | 中国水利水电科学研究院, 北京 100048 |
| |
| 基金项目： | 国家自然科学基金项目“西北干旱区浅埋地下水水位动态遥感监测模型”（50679086） |
| |
| 摘 要： | 地下水是我国内陆干旱地区水资源的重要组成部分,也是极为敏感的生态环境因素之一。地下水动态变化影响着绿洲和湿地的演化,以及土地资源的开发。西北地区地下水监测网尚不完善,动态资料相对缺乏。遥感技术可以弥补传统地下水位监测手段的不足。由于降水极少,西北干旱区地表反射率与地下水水位埋深关系极其密切。选用归一化植被指数(NDVI)、地表温度(LST)数据,应用支持向量机回归方法,建立西北干旱地区地下水位遥感监测模型。提取MODIS影像中的NDVI和LST产品上的地表温度和植被指数信息,作为模型的输入,通过合理选择核函数进行支持向量机的回归分析,从而建立地表植被指数、地表温度与地下水位的相关数学模型,并分析了不同核函数所拟合结果。在河西走廊疏勒河流域的研究成果表明,运用MODIS数据开发地下水动态模型反演水位变化是可行的,模型拟合的结果比较符合实际情况,尤其是对于细土平原地下水浅埋地区模型应用效果更为理想。一次多项式核函数适合模拟埋深小于3m浅埋地下水,径向基函数RBF核函数和三次多项式核函数法则更适合模拟较大埋深情况。开发的地下水位遥感监测模型可为西北干旱区水循环研究和流域水资源管理提供技术手段。
|
| 关 键 词： | 支持向量机 MODIS 地下水位 监测模型 |
| 收稿时间： | 2009-12-12; |
Support Vector Machine Based Groundwater Level Monitoring Model by Using Remote Sensing Images |
| |
| FU Jun＇e,SU Qingxun,PAN Shibing,LU Jingxuan.Support Vector Machine Based Groundwater Level Monitoring Model by Using Remote Sensing Images[J].Geo-information Science,2010,12(4):466-472. |
| |
| Authors: | FU Jun＇e SU Qingxun PAN Shibing LU Jingxuan |
| |
| Institution: | Inst. of Water Resources and Hydropower Research of China,Beijing 100048,China |
| |
| Abstract: | Groundwater is not only an essential component of water resources,but also an extremely sensitive factor of the environment in Northwestern inner land basin.The groundwater dynamic could affect the evolvement of oasis,vegetation growth and development.The remote sensing technique,for its large-scale covering and cost-effective feature,may pay an important role in obtaining information of groundwater level in inner land basin where there is a paucity of basic hydrological data for backward developed monitoring systems.Ground surface reflection is significantly related to the depth of groundwater level for the very few precipitations in arid region.This paper aims to develop the groundwater level monitoring model for in arid region based support vector machine（SVM） regression method by using remote sensing.The spatial distribution and changes of ground-water level in time-series can be monitored and analyzed with vegetation and surface temperature information from the image radiation.The inputs of the model are the Normal Differential Vegetation Index（NDVI） and Land Surface Temperature（LST） data extracted from the MODIS images products,as well as the field observations of wells.Various kernel functions including the one-order polynomial,cubic polynomial and RBF function for SVM regression method are tested and simulated to find appropriate one for modeling.The research results in Shule River basin of Hexi corridor,Gansu Province illustrate that the model approach proposed is effective especially for the shallow groundwater level monitoring in arid region.The shallow groundwater levels with depth less than 3 meters fit the one-order polynomial kernel function better,and for greater groundwater depth more than 5 meters,it is even more suitable for selection of RBF kernel function or a cubic polynomial nuclear function method to simulate groundwater level.The research results could provide the basic approaches for the water cycle module and hydrology research,as well as support the sustainable water resources development and management in Northwest regions of China. |
| |
| Keywords: | support vector machine MODIS ground water level monitoring model |
| 本文献已被 维普 等数据库收录！ |
| 点击此处可从《地球信息科学》浏览原始摘要信息 |
| 点击此处可从《地球信息科学》下载免费的PDF全文 |
|
