摘要
以古田水库为研究对象,采用实测总磷浓度数据和2021年Sentinel-2 MSI时序影像数据,针对水质数据的非线性和时序性特征,提出了BP-SVR混合模型,实现了古田水库总磷含量的反演并分析了其时空分布特征。结果表明:BP-SVR混合模型反演精度比单一模型BP神经网络和SVR支持向量机回归模型更高,拟合能力更强,能够对总磷浓度进行较好地反演,具有一定的实际应用价值。
Taking Gutian Reservoir as the research object,using measured total phosphorus data and Sentinel-2 MSI time series image data in 2021,aiming at the non-linear characteristics and the temporal characteristics of water quality data,a BP-SVR hybrid model was proposed to achieve the inversion of total phosphorus concentration in Gutian Reservoir and analyze its spatiotemporal variation characteristics.The results show that the BP SVR hybrid model has higher inversion accuracy and stronger fitting ability than the single model BP neural network model and SVG support vector machine regression model,and can better retrieve the total phosphorus concentration,which has certain practical application value.
作者
吴瑞姣
Wu Ruijiao(Fujian Geologic Surveying And Mapping Institute,Fuzhou,350011)
出处
《福建地质》
2023年第3期224-230,共7页
Geology of Fujian
基金
福建省地质矿产勘查开发局地勘费项目“遥感技术在自然资源调查中的应用研究”和“国产卫星数据在近岸海域多时序水质污染监测中的应用研究”。
关键词
BP-SVR混合模型
总磷浓度
水质监测
时空变化分析
古田水库
BP-SVR mixed model
total phosphorus concentration
water quality monitoring
spatiotemporal change analysis
Gutian Reservoir
