基于GRNN模型的悬浮物遥感反演研究——以长江城陵矶段为例  

Remote Sensing retrieval of suspended matters based on GRNN model——a case of the Yangtze River nearby Chenglingji

作　　者：何报寅[1] 张文[1,2] 乔晓景[3,2] 苏振华[1,2] 

机构地区：[1]中国科学院测量与地球物理研究所环境与灾害监测评估湖北省重点实验室,武汉430077 [2]中国科学院大学,北京100049 [3]山西师范大学城市与环境科学学院,山西临汾041000

出　　处：《华中师范大学学报（自然科学版）》2014年第5期737-742,共6页Journal of Central China Normal University：Natural Sciences

基　　金：国家自然科学基金项目(51079137);武汉市科技局重大科技专项

摘　　要：遥感反演是监测水体表层悬浮物浓度的有效手段之一.广义回归神经网络(GRNN)较其它神经网络具有更强的非线性拟合能力,在小样本情况下有更好的推广性能,适用于遥感反演模型.使用长江中游城陵矶段HJ-1BCCD2遥感影像结合实地同步采样数据分别建立悬浮物GRNN及BP神经网络(BPNN)遥感反演模型,分析对比模型的精度,并使用GRNN模型预测了区域水体悬浮物分布信息.结果表明,相对于BPNN模型,GRNN模型具有较强的非线性拟合能力和较高的反演精度;长江干流的悬浮泥沙浓度总体上明显小于洞庭湖,这主要是三峡工程下泄泥沙大幅减少造成的;洞庭湖浑浊的湖水汇入长江后,在城陵矶至洪湖之间形成明显的混合带;而洞庭湖湖口悬浮物浓度明显高于其他湖区,这可能是该区域采砂活动的强烈扰动引起的.Remote Sensing retrieval is one of the effective ways to monitor suspended sediment concentration （SSC） in surface water.General Regression Neural Network （GRNN） has better nonlinear fitting ability and generalization ability than other Artificial Neural Networks,and is suitable for remote sensing retrieval modeling.In this paper,a GRNN model for SSC retrieval is established based on the satellite images of HJ-1B CCD2 and the synchronous SSC data in situ in the middle reach of the Yangtze River around Chenglingji,then its accuracy is compared to the Back Propagation Neural Network （BPNN） models.Finally,the well trained GRNN model was used to retrieve SSC and its spatial distribution features was analyzed in the study area.The results show that：1） GRNN model has greater nonlinear fitting ability and higher accuracy than BPNN model; 2） the SSC is significantly smaller in the mainstream of the Yangtze River than in the Dongting Lake,probably due to the significant decrease of sediment discharge because of the Three Gorges Project; as a result,a blending zone is formed in the reach between Chenglingji and Honghu city after the muddy waters of the Dongting Lake flows into the river; 3） the SSC is significantly higher in the outlet channel than the other area of the Dongting Lake,which may be caused by the strong disturbance of the sand mining activity in the region.

关 键 词：悬浮物 遥感反演模型 广义回归神经网络 BP神经网络 长江中游 

分 类 号：X87[环境科学与工程—环境工程]