论文部分内容阅读
近年来新发展起来的多波段准分析算法(QAA)是一种有广泛应用前景的海水固有光学特性反演算法.与传统的半分析算法不同,QAA无需叶绿素吸收光谱模型的先验参数化,但算法包含的若干经验公式影响了其普适性和抗扰能力.本文利用贝叶斯反演理论改进QAA.新算法以QAA的估算结果作为先验知识,结合一个分别考虑水分子与悬浮颗粒后向散射作用的前向光学模型、Akaike贝叶斯信息准则(ABIC)和遗传算法,提取参考波段的总吸收与总后向散射系数,然后利用后向散射系数光谱形状的经验公式将估算结果推广至其他波段.以国际海洋水色协调组(IOCCG)的模拟数据作反演试验,结果显示,使用QAA-555、QAA-640与QAA混合算法的反演结果作为先验估计,新算法对440nm的总吸收与总后向散射系数的反演值与真值的均方根差(RMSE)分别为0.367与0.023m?1,0.092与0.005m?1,0.096与0.005m?1,较之三种QAA算法先验估计的0.918与0.039m?1,0.205与0.007m?1,0.207与0.007m?1有一定的降低,特别是对高叶绿素样本的反演精度有较大的提高;敏感性分析结果表明,新算法的抗扰能力优于QAA算法.
In recent years, a new multi-band quasi-analytical algorithm (QAA) has been developed, which is a promising inversion method for seawater intrinsic optical properties. Unlike conventional semi-analytical algorithms, QAA does not require prior parameterization of the chlorophyll absorption spectrum model, However, some empirical formulas contained in the algorithm affect its universality and anti-jamming ability.In this paper, Bayesian inversion theory is used to improve QAA.The new algorithm takes the QAA estimation as the prior knowledge, combined with a method that separately considers water molecules and suspended particles Backscattering forward optical model, Akaike Bayesian Information Criterion (ABIC) and genetic algorithm to extract the total absorption and total backscatter coefficient of the reference band, and then use the empirical formula of the backscatter spectral shape to extract the estimation results (IOCCG). The results show that the inversion results using QAA-555, QAA-640 and QAA hybrid algorithms are priori estimates, and the new algorithm is accurate for 440nm The root mean square error (RMSE) of the total and total backscatter coefficients between the inversion and the true value were 0.367 and 0.023 m -1, 1.092 and 0.005 m -1, 1.096 and 0.005 m -1, respectively, QAA algorithm The prior estimates of 0.918 and 0.039m? 1,0.205 and 0.007m? 1,0.207 and 0.007m? 1 have a certain reduction, especially for high chlorophyll sample inversion accuracy has been greatly improved; sensitivity analysis showed that The new algorithm has better anti-interference ability than QAA algorithm.