基于BP神经网络算法的密云水库水质参数反演研究

密云水库是北京市重要的地表饮用水源地,在保障首都水源安全方面起着重要作用,而密云水库水质参数的区域监测已成为当前亟待解决的问题。为了实现密云水库水质参数大范围、实时获取,该文基于遥感和GIS技术,采用BP神经网络算法,结合地面监测数据和Landsat 8遥感影像,分别建立了反演总磷、总氮、氨氮和COD(化学需氧量)4个水质参数的BP神经网络模型,并反演了密云水库2013-2018年非结冰期主要水质参数,分析了密云水库主要水质参数的年际变化特征、季节变化特征和空间分异特征。结果表明,(1)水质参数的Landsat 8敏感波段分别为:总氮为1、4波段,氨氮为1-7波段,总磷为1、3-7波段,COD为2-5波段。(2)密云水库主要水质参数在2013-2018年总体呈下降趋势,氨氮和COD为Ⅰ类水质,总磷为Ⅱ类水质,总氮为Ⅲ类水质。(3)4个水质参数指标春季最高、秋季次之、夏季最低,总氮、总磷、氨氮和COD的春季值分别是夏季值的1.08、1.36、1.6、1.45倍。(4)密云水库不同水质参数的空间差异性较大,总体来看,水库北部和东部的4个水质参数含量相对较高,这种分布与北部和东部村庄密集以及密云水库两大入库河流有关。综上所述,基于BP神经网络算法的密云水库水质反演研究是可行的,且得到了较为可信的研究结果,该研究可为密云水库水质管理与政策制定提供重要的科学依据。

Retrieval of Water Quality Parameters Based on BP Neural Network Algorithm in Miyun Reservoir

Miyun Reservoir is a significant source of surface drinking water in Beijing,and plays an important role in protecting the safety of water resources in the capital.Therefore,the regional monitoring of water quality parameters of Miyun Reservoir has become an urgent issue.In order to realize the real and large-scale monitoring of water quality parameters in Miyun Reservoir,this study applied the BP neural network algorithm combined with field monitoring data of water quality parameters and Landsat8 image to establish BP neural network models for the retrieval of total nitrogen,ammonia nitrogen,total phosphorus and COD,respectively.And then the four main water quality parameters during the non-icing period from 2013 to 2018 were retrieved to reveal their interannual variation,seasonal variation and spatial variation in Miyun Reservoir.The results show that:(1)The Landsat8 band sensitive to total nitrogen is band 1 and 4,that sensitive to the ammonia nitrogen is band 1?7,that sensitive to total phosphorus is band 1 and 3?7,and that sensitive to COD is band 2?5.(2)The four main water quality parameters showed a downward trend in Miyun Reservoir from 2013 to 2018.Ammonia nitrogen and COD was Class Ⅰ,total phosphorus was Class Ⅱ,and total nitrogen was Class Ⅲ.(3)As for the seasonal analysis,it was found that four water quality parameters were highest in Spring,followed by the Autumn and lowest in Summer.The values of total nitrogen,total phosphorus,ammonia nitrogen and COD in Spring were 1.08,1.36,1.6 and 1.45 times of those in Summer,respectively.And(4)the spatial variation of different water quality parameters were obvious,and the content of four main water quality parameters in the north and east of the reservoir were relative higher.This may be caused by the concentrated villages in the north and east,as well as the two major rivers in Miyun Reservoir.In summary,Water quality retrieval based on BP neural network algorithm in Miyun Reservoir is feasible,and it has obtained creadible research results.Those conclusions will provide significant and scientific basis for water quality management and policy development in Miyun Reservoir.