用波段比值参数提升水体悬浮颗粒物浓度反演模型稳健性的分析

为了得到更加稳健的水体悬浮颗粒物(Suspended Particular Matters, SPM)的反演模型，减少建模数据误差对算法精度的影响，将单波段和对应波段比值的六种经验模型和半经验、半分析模型分别应用于长江口高浊度水域不同航次不同测量仪器测得的实测数据集中，分析了波段比值参数对于提高SPM反演模型稳健性和稳定性的作用。结果表明，采用波段比值参数的模型的精度都远高于相应的单波段模型。在2014年5月和2014年12月航次的数据集中，有的单波段模型失效，总体的最高精度不超过0.5。而相对应的波段比值的建模精度都在0.8左右。另外，将各模型应用于Landsat 8卫星的OLI传感器的遥感数据时，同样，波段比值模型的反演精度高于单波段模型。结果证明，在高浊度的水体中反演SPM浓度时，波段比值遥感参数的应用可以加强反演模型的稳健性。

Analysis of Robustness Improvement of SPM Inversion Models Using Band Ratio Parameters

To obtain more robust inversion models of Suspended Particular Matters (SPM) in water and reduce the influence of the modeling data error on the accuracy of an algorithm, six empirical, semi-empirical and semi-analysis models of single-band and corresponding band ratio parameters are used in the data sets measured in high turbidity water at Yangtze River estuary by different instruments in different cruises. The role of band ratio parameters in robustness improvement of the SPM inversion model is analyzed. The results show that the models using band ratio parameters have much higher accuracy than the corresponding single-band model. In the data sets obtained in May and December 2014, some single-band models are invalid with the highest accuracy no more than 0.5. However, the modeling accuracy of the corresponding band ratio models is about 0.8. In addition, when the models are used in the OLI images from Landsat 8, the validation accuracy of band ratio models is higher than that of the single-band models as well. The results confirm that the use of the band ratio parameters can enhance the robustness of inversion models when the SPM concentration is inverted in high turbidity water.