摘要
利用2003年1月、2004年1月在珠江口海域的叶绿素浓度和辐射同步实测资料,建立了反演珠江口海域叶绿素浓度的人工神经网络模型。应用该模型由SeaWiFS资料获取珠江口海域叶绿素浓度分布图,并与SeaBAM推荐的OC2和OC4这2种统计算法的反演结果进行比较,结果表明人工神经网络模型的反演效果明显优于统计算法。人工神经网络模型的均方根差是0.289 9,可决系数是0.884 8;而统计算法的均方根差大于0.5,可决系数小于0.6。
Based on the in situ chlorophyll concentration and radiance data observed in Jan. 2003 and Jan. 2004 in the Zhujiang River estuary, an artificial neural network （ANN） algorithm was developed to retrieve the chlorophyll concentration in the Zhujiang River estuary from the SeaWiFS image of October 31, 1998, and the results were compared with those determined from the statistical algorithms OC2 and OC4. It was shown that the retrieve effect of the neural network outperformed those of the statistical algorithms. The root-mean-square error （RMS） and relation square （RSq） of ANN were 0. 289 9 and 0. 884 8, respectively, while the RMS of the statistical algorithms was over 0.5 and the RSq below 0.6.
出处
《热带海洋学报》
CAS
CSCD
北大核心
2005年第6期38-43,共6页
Journal of Tropical Oceanography
基金
国家自然科学基金项目(40276049)
国家973赤潮项目(2001CB409708)
国家863自由探索项目(2002AA639130)
国家863青年项目(2004AA639860)
关键词
人工神经网络
二类海水
叶绿素
反演算法
artificial neural network
case 2 water
chlorophyll
inversive algorithm
