基于GF-1卫星遥感反演排水河沟水体溶存N_(2)O浓度模型对比研究  

Comparative Study on Retrieval Model of Dissolved N_(2)O Concentration in Drainage Ditch Based on GF-1 Satellite Remote Sensing

作　　者：嵇晶晶 白立影[2] 佘冬立[1] 管伟 阿力木·阿布来提 潘永春 Ji Jingjing;Bai Liying;She Dongli;Guan Wei;Ali mu·Abu laiti;Pan Yongchun(College of Agricultural Science and Engineering,Hohai University,Nanjing 211100,China;Zhenjiang Branch of Jiangsu Provincial Hydrology and Water Resources Survey Bureau,Zhenjiang,Jiangsu 21208,China;Jiangsu Linhai Farm Co.,Ltd.,Sheyang,Jiangsu 224353,China)

机构地区：[1]河海大学农业科学与工程学院,南京211100 [2]江苏省水文水资源勘测局镇江分局,江苏镇江212008 [3]江苏省临海农场有限公司,江苏射阳224353

出　　处：《水土保持研究》2024年第5期257-264,共8页Research of Soil and Water Conservation

基　　金：国家自然科学基金“海涂盐碱地农业汇水区反硝化和厌氧氨氧化脱氮机制与过程拟”(42177393)。

摘　　要：[目的]探究利用GF-1卫星数据反演水体溶存氧化亚氮(N_(2)O)浓度的可行性,为实现低成本、高效率的水质实时监测提供有效途径。[方法]以宁夏青铜峡灌区第1和第5排水河沟为研究对象,选取与排水河沟水体溶存N_(2)O浓度相关性高的GF-1卫星影像波段反射率和水质参数作为自变量,通过最优子集筛选法确定最优自变量组合,分别构建了多元线性回归、BP神经网络和支持向量机模型,对水体溶存N_(2)O浓度进行预测对比。[结果]水温(T)、溶解性有机碳(DOC)等是影响水体溶存N_(2)O浓度的主要因素,同时近红外(NIR)等卫星波段与水体溶存N_(2)O浓度变化趋势显著相关。当自变量包括T,NIR等7个因素时,模型预测效果最佳。在3种模型中,BP神经网络模型验证结果R^(2)为0.64,具有最高预测精度。[结论]GF-1卫星数据以及水质参数与水体溶存N_(2)O浓度存在复杂的相关性关系,且BP神经网络能够实现利用GF-1卫星数据较高精度地反演水体溶存N_(2)O浓度。[Objective] The aims of this study are to explore the feasibility of using GF-1 satellite data to retrieve the concentration of dissolved nitrous oxide(N_(2)O in water,so as to provide an effective way to realize low-cost and high-efficiency real-time monitoring of water quality.[Methods] The 1st and 5th drainage ditches in Qingtongxia Irrigation District of Ningxia were selected as the research objects,and the reflectance and water quality parameters of GF-1 satellite image band,which were highly correlated with the concentration of dissolved N_(2)O in the drainage ditches,were selected as independent variables,and the optimal combination of independent variables was determined by optimal subset screening method.Multiple linear regression,BP neural network and support vector machine models were respectively constructed to predict and compare the concentration of dissolved N_(2)O in water.[Results] The water temperature(T) and dissolved organic carbon(DOC) were the main factors affecting the concentration of dissolved N_(2)O in water,and satellite bands such as near infrared(NIR) were significantly correlated with the variation trend of dissolved N_(2)O concentration in water.When the independent variable including 7 factors such as T and NIR,the model had the best prediction effect.Among the three models,the R^(2) of BP neural network model was 0.64,which had the highest prediction accuracy.[Conclusion] There is a complex correlation between GF-1 satellite data and water quality parameters and dissolved N_(2)O concentration in water bodies,and BP neural network can use GF-1 satellite data to retrieve dissolved N_(2)O concentration in water bodies with high accuracy.

关 键 词：N_(2)O浓度 GF-1卫星 反演 机器学习 

分 类 号：X87[环境科学与工程—环境工程]