基于神经网络模型的湛江湾水体有色溶解有机物的遥感估算
|
摘要
有色溶解有机物(Colored Dissolved Organic Matter, CDOM)是水体中重要的水质参数之一,是水色遥感的重要研究对象,如何构建适合特定区域的近海二类水体CDOM反演模型一直是国内外研究难点。本文利用2017年5月26～29日对南海西北部海域湛江湾20个站位采集的水样和测量的光谱资料,分析归一化遥感反射率与CDOM浓度a_g(400)的相关性,发现最大负相关系数出现在586nm处,选择580、585、590、595nm这四个波段处的归一化遥感反射率与a_g(400)建立了多元线性回归模型、BP(Back-Propagation)神经网络模型和RBF(Radial-Basis Function)神经网络模型,并与其他算法模型进行对比分析。结果发现, BP和RBF神经网络模型的平均相对误差和均方根误差均远小于多元线性回归模型和其他算法模型,神经网络模型的预测值与实测值拟合效果要优于多元线性回归模型。研究表明,神经网络模型更适合于湛江湾有色溶解有机物的遥感估算。
Colored dissolved organic matter(CDOM), one of the important water quality parameters, is an important research object of ocean color remote sensing. How to construct CDOM inversion model in Case-Ⅱ Waters of coastal which is applied to particular locations has always been a research challenge at home and abroad. In this study we use the water sample extracted from 20 stations of the Zhanjiang Bay, which is the northwestern part of the South China Sea, and live spectral measurements of these stations from May 26 to 29, 2017 to analysis the correlation between normalized remote sensing reflectance and CDOM concentration [ag(400)], the maximum negative correlation coefficient appears at 586 nm. The multiple linear regression model, BP neural network model and RBF neural network model were built with ag(400) and the normalized remote sensing reflectivity at this four bands of 580 nm, 585 nm, 590 nm, 595 nm, and then compared with other CDOM algorithm models. The results show that the average relative error and root mean square error of the BP and RBF neural network model were far less than other models and that the fitting effect between the predicted values of the neural network model and the measured values was better than the effect came from the multiple linear regression model. That is to say the neural network model is more suitable for the remote sensing estimation of colored dissolved organic matter in Zhanjiang Bay.
Colored dissolved organic matter(CDOM), one of the important water quality parameters, is an important research object of ocean color remote sensing. How to construct CDOM inversion model in Case-Ⅱ Waters of coastal which is applied to particular locations has always been a research challenge at home and abroad. In this study we use the water sample extracted from 20 stations of the Zhanjiang Bay, which is the northwestern part of the South China Sea, and live spectral measurements of these stations from May 26 to 29, 2017 to analysis the correlation between normalized remote sensing reflectance and CDOM concentration [ag(400)], the maximum negative correlation coefficient appears at 586 nm. The multiple linear regression model, BP neural network model and RBF neural network model were built with ag(400) and the normalized remote sensing reflectivity at this four bands of 580 nm, 585 nm, 590 nm, 595 nm, and then compared with other CDOM algorithm models. The results show that the average relative error and root mean square error of the BP and RBF neural network model were far less than other models and that the fitting effect between the predicted values of the neural network model and the measured values was better than the effect came from the multiple linear regression model. That is to say the neural network model is more suitable for the remote sensing estimation of colored dissolved organic matter in Zhanjiang Bay.
引文
[1]Whitehead R F,De Mora S,Demers S,et al.Interactions of ultraviolet-B radiation,mixing,and biological activity on photobleaching of natural chromophoric dissolved organic matter:A mesocosm study[J].Limnology and Oceanography,2000,45(2):278-291.
[2]陈楚群,潘志林,施平.海水光谱模拟及其在黄色物质遥感反演中的应用[J].热带海洋学报,2003,22(5):33-39.Chen Chuqun,Pan Zhilin,Shi Ping.Simulation of sea water reflectance and its application in retrieval of yellow substance by remote sensing data[J].Journal of Tropical Oceanography,2003,22(5):33-39.
[3]Kirk J T O.Light and photosynthesis in aquatic ecosystems[M].Cambridge:Cambridge University Press,1994:50-92.
[4]扶丹丹,张立燕,梁志林.水体中CDOM遥感反演研究综述[J].首都师范大学学报(自然科学版),2017,38(2):84-90.Fu Dandan,Zhang Liyan,Liang Zhilin.The review of CDOM remote sensing inversion in water[J].Journal of Capital Normal University(Natural Science Edition),2017,38(2):84-90.
[5]Mannino A,Russ M E,Hooker S B.Algorithm development and validation for satellite-derived distributions of DOC and CDOM in the U.S.Middle Atlantic Bight[J].Journal of Geophysical Research:Oceans,2008,113(C7):827-830.
[6]廖珊,余果,付东洋,等.湛江湾海域春季黄色物质光学吸收特性及遥感探测[J].海洋技术学报,2018,37(2):56-62.Liao Shan,Yu Guo,Fu Dongyang,et al.Optical absorption characteristics of yellow substances in the Zhanjiang Bay in spring:research and remote sensing detection[J].Journal of Ocean Technology,2018,37(2):56-62.
[7]Lee Z,Carder K L,Arnone R A.Deriving inherent optical properties from water color:a multiband quasi-analytical algorithm for optically deep waters.[J].Applied optics,2002,41(27):5755-5772.
[8]Garver S A,Siegel D A.Inherent optical property inversion of ocean color spectra and its biogeochemical interpretation:1.Time series from the Sargasso Sea[J].Journal of Geophysical Research:Oceans,1997,102(C8):18607-18625.
[9]Maritorena S,Siegel D A,Peterson A R.Optimization of a semianalytical ocean color model for global-scale applications[J].Applied optics,2002,41(15):2705-2714.
[10]Gordon H R,Brown O B,Evans R H,et al.A semianalytic radiance model of ocean color[J].Journal of Geophysical Research:Atmospheres,1988,93(D9):10909-10924.
[11]Werdell P J,Franz B A,Bailey S W,et al.Generalized ocean color inversion model for retrieving marine inherent optical properties[J].Applied optics,2013,52(10):2019-2037.
[12]吕恒,李新国,曹凯.基于BP神经网络模型的太湖悬浮物浓度遥感定量提取研究[J].武汉大学学报(信息科学版),2006,31(8):683-686.Lv Heng,Li Xinguo,Cao Kai.Quantitative retrieval of suspended solid concentration in lake Taihu based on BP neural net[J].Geomatics and Information Science of Wuhan University,2006,31(8):683-686.
[13]Sun D Y,Li Y M,Wang Q,et al.A neural-network model to retrieve CDOM absorption from in situ measured hyperspectral data in an optically complex lake:Lake Taihu case study[J].International Journal of Remote Sensing,2011,32(14):4005-4022.
[14]丛丕福,牛铮,曲丽梅,等.基于神经网络和TM图像的大连湾海域悬浮物质量浓度的反演[J].海洋科学,2005,29(4):31-35.Cong Pifu,Niu Zheng,Qu Limei,et al.Retrieval of suspended sediment concentration in Dalian Bay on basis of neural network model and TM imagery[J].Marine Sciences,2005,29(4):31-35.
[15]孙德勇,李云梅,王桥,等.利用高光谱数据估算太湖水体CDOM浓度的神经网络模型[J].武汉大学学报:信息科学版,2009,34(7):851-855.Sun Deyong,Li Yunmei,Wang Qiao,et al.Remote sensing retrieval of CDOM concentration in Lake Taihu with hyper-spectral data and neural network model[J].Geomatics&Information Science of Wuhan University,2009,34(7):851-855.
[16]Carder K L,Chen F R,Lee Z,et al.MODIS ocean science team algorithm theoretical basis document[J].MODISOcean Science Team:Algorithm Theoretical Basis Document ATBD,2003,19:7-28.
[17]Chen C,Shi P,Zhan H,et al.Retrieval of gelbstoff absorption coefficient in Pearl River estuary using remotely-sensed ocean color data[C]//Geoscience and Remote Sensing Symposium,Proceeding of SPIE,2005,4:2511-2514.
[18]Kowalczuk P,Olszewski J,Darecki M,et al.Empirical relationships between coloured dissolved organic matter(CDOM)absorption and apparent optical properties in Baltic Sea waters[J].International Journal of Remote Sensing,2005,26(2):345-370.
[19]国家海洋局908专项办公室.908项目海洋光学调查技术规程[S].北京:海洋出版社,2006:1-16.908 Special Office Of The State Oceanic Administration.Marine optical investigation procedures of project 908[S].Beijing:China Ocean Press,2006:1-16.
[20]唐军武,田国良,汪小勇,等.水体光谱测量与分析Ⅰ:水面以上测量法[J].遥感学报,2004,8(1):37-44.Tang Junwu,Tian Guoliang,Wang Xiaoyong,et al.The methods of water spectra measurement and analysisⅠ:above-water method[J].Journal of Remote Sensing,2004,8(1):37-44.
[21]杨婧茹,臧淑英,惠洪宽.松花江哈尔滨段总悬浮物浓度高光谱估测模型研究[J].安全与环境学报,2014,14(2):239-243.Yang Jingru,Zang Shuying,Hui Hongkuan.An hyperspectral estimation model for testing the total suspended solid concentration in Harbin section of Songhua River[J].Journal of Safety&Environment,2014,14(2):239-243.
[22]Olmanson L G,Brezonik P L,Finlay J C,et al.Comparison of Landsat 8 and Landsat 7 for regional measurements of CDOM and water clarity in lakes[J].Remote Sensing of Environment,2016,185:119-128.
[23]Hornik K,Stinchcombe M,White H.Multilayer feedforward networks are universal approximators[J].Neural Networks,1989,2(5):359-366.
[24]赵健,刘展.基于灵敏度分析的海洋油气资源BP神经网络预测模型的优化[J].海洋科学,2016,40(5):103-108.Zhao Jian,Liu Zhan.Structure optimization of ocean oil and gas resources via BP neural network prediction model based on sensitivity analysis[J].Marine Sciences,2016,40(5):103-108.
[25]崔东文.几种神经网络模型在湖库富营养化程度评价中的应用[J].水资源保护,2012,28(6):12-18.Cui Dongwen.Applications of several neural network models to eutrophication evaluation of lakes and reservoirs[J].Water Resources Protection,2012,28(6):12-18.
[2]陈楚群,潘志林,施平.海水光谱模拟及其在黄色物质遥感反演中的应用[J].热带海洋学报,2003,22(5):33-39.Chen Chuqun,Pan Zhilin,Shi Ping.Simulation of sea water reflectance and its application in retrieval of yellow substance by remote sensing data[J].Journal of Tropical Oceanography,2003,22(5):33-39.
[3]Kirk J T O.Light and photosynthesis in aquatic ecosystems[M].Cambridge:Cambridge University Press,1994:50-92.
[4]扶丹丹,张立燕,梁志林.水体中CDOM遥感反演研究综述[J].首都师范大学学报(自然科学版),2017,38(2):84-90.Fu Dandan,Zhang Liyan,Liang Zhilin.The review of CDOM remote sensing inversion in water[J].Journal of Capital Normal University(Natural Science Edition),2017,38(2):84-90.
[5]Mannino A,Russ M E,Hooker S B.Algorithm development and validation for satellite-derived distributions of DOC and CDOM in the U.S.Middle Atlantic Bight[J].Journal of Geophysical Research:Oceans,2008,113(C7):827-830.
[6]廖珊,余果,付东洋,等.湛江湾海域春季黄色物质光学吸收特性及遥感探测[J].海洋技术学报,2018,37(2):56-62.Liao Shan,Yu Guo,Fu Dongyang,et al.Optical absorption characteristics of yellow substances in the Zhanjiang Bay in spring:research and remote sensing detection[J].Journal of Ocean Technology,2018,37(2):56-62.
[7]Lee Z,Carder K L,Arnone R A.Deriving inherent optical properties from water color:a multiband quasi-analytical algorithm for optically deep waters.[J].Applied optics,2002,41(27):5755-5772.
[8]Garver S A,Siegel D A.Inherent optical property inversion of ocean color spectra and its biogeochemical interpretation:1.Time series from the Sargasso Sea[J].Journal of Geophysical Research:Oceans,1997,102(C8):18607-18625.
[9]Maritorena S,Siegel D A,Peterson A R.Optimization of a semianalytical ocean color model for global-scale applications[J].Applied optics,2002,41(15):2705-2714.
[10]Gordon H R,Brown O B,Evans R H,et al.A semianalytic radiance model of ocean color[J].Journal of Geophysical Research:Atmospheres,1988,93(D9):10909-10924.
[11]Werdell P J,Franz B A,Bailey S W,et al.Generalized ocean color inversion model for retrieving marine inherent optical properties[J].Applied optics,2013,52(10):2019-2037.
[12]吕恒,李新国,曹凯.基于BP神经网络模型的太湖悬浮物浓度遥感定量提取研究[J].武汉大学学报(信息科学版),2006,31(8):683-686.Lv Heng,Li Xinguo,Cao Kai.Quantitative retrieval of suspended solid concentration in lake Taihu based on BP neural net[J].Geomatics and Information Science of Wuhan University,2006,31(8):683-686.
[13]Sun D Y,Li Y M,Wang Q,et al.A neural-network model to retrieve CDOM absorption from in situ measured hyperspectral data in an optically complex lake:Lake Taihu case study[J].International Journal of Remote Sensing,2011,32(14):4005-4022.
[14]丛丕福,牛铮,曲丽梅,等.基于神经网络和TM图像的大连湾海域悬浮物质量浓度的反演[J].海洋科学,2005,29(4):31-35.Cong Pifu,Niu Zheng,Qu Limei,et al.Retrieval of suspended sediment concentration in Dalian Bay on basis of neural network model and TM imagery[J].Marine Sciences,2005,29(4):31-35.
[15]孙德勇,李云梅,王桥,等.利用高光谱数据估算太湖水体CDOM浓度的神经网络模型[J].武汉大学学报:信息科学版,2009,34(7):851-855.Sun Deyong,Li Yunmei,Wang Qiao,et al.Remote sensing retrieval of CDOM concentration in Lake Taihu with hyper-spectral data and neural network model[J].Geomatics&Information Science of Wuhan University,2009,34(7):851-855.
[16]Carder K L,Chen F R,Lee Z,et al.MODIS ocean science team algorithm theoretical basis document[J].MODISOcean Science Team:Algorithm Theoretical Basis Document ATBD,2003,19:7-28.
[17]Chen C,Shi P,Zhan H,et al.Retrieval of gelbstoff absorption coefficient in Pearl River estuary using remotely-sensed ocean color data[C]//Geoscience and Remote Sensing Symposium,Proceeding of SPIE,2005,4:2511-2514.
[18]Kowalczuk P,Olszewski J,Darecki M,et al.Empirical relationships between coloured dissolved organic matter(CDOM)absorption and apparent optical properties in Baltic Sea waters[J].International Journal of Remote Sensing,2005,26(2):345-370.
[19]国家海洋局908专项办公室.908项目海洋光学调查技术规程[S].北京:海洋出版社,2006:1-16.908 Special Office Of The State Oceanic Administration.Marine optical investigation procedures of project 908[S].Beijing:China Ocean Press,2006:1-16.
[20]唐军武,田国良,汪小勇,等.水体光谱测量与分析Ⅰ:水面以上测量法[J].遥感学报,2004,8(1):37-44.Tang Junwu,Tian Guoliang,Wang Xiaoyong,et al.The methods of water spectra measurement and analysisⅠ:above-water method[J].Journal of Remote Sensing,2004,8(1):37-44.
[21]杨婧茹,臧淑英,惠洪宽.松花江哈尔滨段总悬浮物浓度高光谱估测模型研究[J].安全与环境学报,2014,14(2):239-243.Yang Jingru,Zang Shuying,Hui Hongkuan.An hyperspectral estimation model for testing the total suspended solid concentration in Harbin section of Songhua River[J].Journal of Safety&Environment,2014,14(2):239-243.
[22]Olmanson L G,Brezonik P L,Finlay J C,et al.Comparison of Landsat 8 and Landsat 7 for regional measurements of CDOM and water clarity in lakes[J].Remote Sensing of Environment,2016,185:119-128.
[23]Hornik K,Stinchcombe M,White H.Multilayer feedforward networks are universal approximators[J].Neural Networks,1989,2(5):359-366.
[24]赵健,刘展.基于灵敏度分析的海洋油气资源BP神经网络预测模型的优化[J].海洋科学,2016,40(5):103-108.Zhao Jian,Liu Zhan.Structure optimization of ocean oil and gas resources via BP neural network prediction model based on sensitivity analysis[J].Marine Sciences,2016,40(5):103-108.
[25]崔东文.几种神经网络模型在湖库富营养化程度评价中的应用[J].水资源保护,2012,28(6):12-18.Cui Dongwen.Applications of several neural network models to eutrophication evaluation of lakes and reservoirs[J].Water Resources Protection,2012,28(6):12-18.