中文摘要：

基于吸收光谱相关关系建立了一种海水总颗粒物吸收光谱（ap（λ））分解模型,利用有约束的非线性优化算法实现了浮游植物（aph（λ））和非藻类颗粒物吸收光谱（aNAP（λ））的分离。采用2004年南海北部航次调查水体的实测数据分析发现,在短波波段（aph（λ））可以较好地表示为aph（443）的二次多项式形式,aNAP（λ）遵循普遍的指数衰减规律。基于这种光谱变化规律本文建立了ap（λ）分解模型,并采用2005年同一海区实测数据进行验证,发现模型计算光谱与实测数据有较好的相似性,在主要的吸收波段如443,490和683 nm处的平均相对偏差值在17%以内,两者之间线性关系拟合的决定系数（R2）均在0.97以上,斜率接近于1.0。与现有的分解方法相比,该模型具有明显的区域优势,这种基于光谱相关关系的分解思想可进一步应用于水色遥感信息的反演和海水总吸收系数的分解。

英文摘要：

A model for estimating the contributions of phytoplankton and nonalgal particles to the total particulate absorption coefficient was developed based on their separate spectral relationships,and a constrained nonlinear optimization code was used to realize the spectral decomposition. The spectral absorption of total particulate matter including phytoplankton and nonalgal particles was measured using the filter-pad method during two cruises in autumn in Northern South China Sea.Using the dataset collected in 2004,the spectral relationships of particle absorption coefficients were examined and the results showed that the phytoplankton absorption coefficients at various wavebands could be well expressed by aph（443） as the second-order quadratic equations;and the nonalgal particle absorption（aNAP（λ）） could be successfully modeled with the simple exponential function.Based on these spectral relationships,we developed this partition model.The model was tested using the independently measured absorption by phytoplankton and nonalgal materials which were obtained in 2005 from the same area.The test results showed that the computed spectral absorption coefficients of phytoplankton and nonalgal particles were consistent with in situ measurement.Good correlations were found between the computed phytoplankton absorption coefficient and the measured value,with the determination coefficients（r^2） being higher than 0.97 and slopes being around 1.0;and the RMSE values could be controlled within 17% over the main absorption wavebands such as 443,490 and 683 nm.Compared with the other two existing models from Bricaud et al.and Oubelkheir et al.,this method shows many advantages for local applications.Moreover,this model does not need any information about pigment concentrations and the selected spectral bands are consistent with the ocean color satellite sensor.This method could also be used in the total absorption coefficient decomposition which provides much insight into the phytoplankton absorption retrieval fro