4种机器学习模型反演太湖叶绿素a浓度的比较
|
摘要
基于太湖实测叶绿素a浓度数据以及同步HJ-1B卫星CCD多光谱影像,综合比较4种机器学习模型(随机森林,RF;支持向量回归,SVR;反向传播人工神经网络,BPANN;深度学习,DL)反演太湖叶绿素a浓度的精度、稳定性及鲁棒性。利用11种波段组合分别建立基于RF、SVR、BPANN和DL的反演模型,筛选出最佳波段组合模型用于验证和评价。结果表明,模型精度方面,DL(决定系数R2=0.91,均方根误差RMSE=3.458μg/L,相对预测偏差RPD=3.13)和SVR(R2=0.88,RMSE=3.727μg/L,RPD=2.90)具有较优的验证精度;模型稳定性方面,DL模型不易受模型校正样本数影响,稳定性较好,而RF模型稳定性较差;模型鲁棒性方面,DL模型不易受噪声影响,鲁棒性较好,其次是SVR、BPANN和RF模型。综合4种模型的验证精度、稳定性和鲁棒性,DL模型在太湖叶绿素a浓度的反演具有较大应用潜力,能为研究湖泊水色参数提供借鉴。
Lakes play a vital role in the sustainable development of human production,societies and regional economies.Most lakes in China are threatened by eutrophication,the direct cause of cyanobacteria blooms.Chlorophyll-a(Chl-a)measurements are used to indicate the degree of eutrophication and to monitor the growth and decline of harmful algal blooms(HABs).The spatial distribution of Chl-a in a water body can be used to guide the remediation and management of lake ecosystems.In this study,we compared the accuracy,stability and robustness of four machine learning models[Random Forest(RF),Support Vector Regression,(SVR),Back Propagation Artificial Neural Network(BPANN)and Deep Learning(DL)]for predicting chlorophyll-a concentration in Taihu Lake from satellite multispectral images.Calibration and validation of chlorophyll-a simulation by the four learning models were based on in-situ measurements of Chl-a concentration and synchronous HJ-1 BCCD multispectral images.The spectral reflectance of 11 wave band combinations were used as model input data and the measured Chl-a concentrations were used for model calibration.The model and spectral band combination that best fit the Chl-a field data was selected as the optimum model for evaluation and validation.The calibration and validation coefficients(Rc2 and Rv2),root mean square errors of calibration and validation(RMSEcand RMSEv),bias(Bv)and relative predictive deviation(RPD)were used to evaluate model performance.Among the four models,two displayed superior performance:DL(Rv2=0.91,RMSEv=3.458μg/L,RPD=3.13)and SVR(Rv2=0.88,RMSEv=3.727μg/L,RPD=2.90).The DL model was less sensitive to calibration sample size and displayed better stability than the RF model and the DL model was less sensitive to noise and more robust than the other three models.Based on the comprehensive comparison of the accuracy,stability and robustness,this study shows that the DL model has the best potential for predicting Chl-a in Taihu Lake based on multispectral imaging.Generally,the four machine learning methods,in combination with remote satellite imaging are practical for predicting Chl-a.However,incorporating water quality parameters(water temperature,dissolved oxygen,dissolved phosphorus and total nitrogen)into the model will improve the accuracy of Taihu Lake Chl-apredictions.
Lakes play a vital role in the sustainable development of human production,societies and regional economies.Most lakes in China are threatened by eutrophication,the direct cause of cyanobacteria blooms.Chlorophyll-a(Chl-a)measurements are used to indicate the degree of eutrophication and to monitor the growth and decline of harmful algal blooms(HABs).The spatial distribution of Chl-a in a water body can be used to guide the remediation and management of lake ecosystems.In this study,we compared the accuracy,stability and robustness of four machine learning models[Random Forest(RF),Support Vector Regression,(SVR),Back Propagation Artificial Neural Network(BPANN)and Deep Learning(DL)]for predicting chlorophyll-a concentration in Taihu Lake from satellite multispectral images.Calibration and validation of chlorophyll-a simulation by the four learning models were based on in-situ measurements of Chl-a concentration and synchronous HJ-1 BCCD multispectral images.The spectral reflectance of 11 wave band combinations were used as model input data and the measured Chl-a concentrations were used for model calibration.The model and spectral band combination that best fit the Chl-a field data was selected as the optimum model for evaluation and validation.The calibration and validation coefficients(Rc2 and Rv2),root mean square errors of calibration and validation(RMSEcand RMSEv),bias(Bv)and relative predictive deviation(RPD)were used to evaluate model performance.Among the four models,two displayed superior performance:DL(Rv2=0.91,RMSEv=3.458μg/L,RPD=3.13)and SVR(Rv2=0.88,RMSEv=3.727μg/L,RPD=2.90).The DL model was less sensitive to calibration sample size and displayed better stability than the RF model and the DL model was less sensitive to noise and more robust than the other three models.Based on the comprehensive comparison of the accuracy,stability and robustness,this study shows that the DL model has the best potential for predicting Chl-a in Taihu Lake based on multispectral imaging.Generally,the four machine learning methods,in combination with remote satellite imaging are practical for predicting Chl-a.However,incorporating water quality parameters(water temperature,dissolved oxygen,dissolved phosphorus and total nitrogen)into the model will improve the accuracy of Taihu Lake Chl-apredictions.
引文
毕京博,郑俊,沈玉凤,等,2012.南太湖入湖口叶绿素a时空变化及其与环境因子的关系[J].水生态学杂志,33(6):7-13.
边博,夏明芳,王志良,等,2012.太湖流域重污染区主要水污染物总量控制[J].湖泊科学,24(3):327-333.
陈军,周冠华,温珍河,等,2010.太湖表层悬浮泥沙遥感定量模式研究[J].光谱学与光谱分析,30(1):137-141.
高玉蓉,刘明亮,吴志旭,等,2012.应用实测光谱估算千岛湖夏季叶绿素a浓度[J].湖泊科学,24(4):553-561.
江敏,余根鼎,戴习林,等,2010.凡纳滨对虾养殖塘叶绿素a与水质因子的多元回归分析[J].水产学报,34(11):1712-1718.
金鑫,李云梅,王桥,等,2010.基于太湖气溶胶类型分区的环境一号卫星CCD大气校正[J].湖泊科学,22(4):504-512.
李文朝,陈开宁,吴庆龙,等,2001.东太湖水生植物生物质腐烂分解实验[J].湖泊科学,13(4):331-336.
刘朝相,宫兆宁,赵文吉,等,2014.基于SVM模型的妫水河叶绿素a浓度的遥感反演[J].遥感技术与应用,29(3):419-427.
汪西莉,周兆永,延军平,2009.应用GA-SVM的渭河水质参数多光谱遥感反演[J].遥感学报,13(4):735-744.
王丽平,郑丙辉,2011.大宁河叶绿素a的因子分值-多元线性回归预测模型研究[J].长江流域资源与环境,20(9):1120-1124.
王艳丽,王飞,2016.白洋淀湖泊叶绿素a对水位周期波动的响应[J].中北大学学报(自然科学版),37(2):126-132.
韦玉春,王国祥,孙华芸,2010.使用线性回归方法构建水体叶绿素a浓度高光谱估算模型的一个逻辑问题[J].数学的实践与认识,40(18):100-110.
吴阿娜,朱梦杰,汤琳,等,2011.淀山湖蓝藻水华高发期叶绿素a动态及相关环境因子分析[J].湖泊科学,23(1):67-72.
夏叡,2010.基于遥感的太湖水质参数空间分异研究[D].南京:南京师范大学:42-78.
徐朝,2016.基于机器学习的海面盐度遥感反演模型[D].北京:中国地质大学:60-67.
许云峰,马春子,霍守亮,等,2012.以程海为例用支持向量机回归算法预测叶绿素a浓度[J].环境工程技术学报,2(3):207-211.
姚付启,张振华,杨润亚,等,2010.基于主成分分析和BP神经网络的法国梧桐叶绿素含量高光谱反演研究[J].测绘科学,35(1):109-112.
张玉超,钱新,钱瑜,等,2009a.基于机器学习方法的太湖叶绿素a定量遥感研究[J].环境科学,30(5):1321-1328.
张玉超,钱新,钱瑜,等,2009b.支持向量机在太湖叶绿素a非线性反演中的应用[J].中国环境科学,29(1):78-83.
赵汉取,韦肖杭,姚伟忠,等,2011.南太湖近岸水域叶绿素a含量与氮磷浓度的关系[J].水生态学杂志,32(5),59-63.
Abdel-Rahman E M,Ahmed F B,Ismail R,2013.Random forest regression and spectral band selection for estimating sugarcane leaf nitrogen concentration using EO-1Hyperion hyperspectral data[J].International Journal of Remote Sensing,34(2):712-728.
Agodi A,2011.Comparing performance and robustness of SVM and ANN for fault diagnosis in a centrifugal pump[J].Celebrazioni Archimedee del Sec XX(Siracusa,1961),3:81-86.
Bocai G,1996.NDWI-A Normalized Difference Water Index for Remote Sensing of Vegetation Liquid Water From Space[J].Remote Sensing of Environment,58(3):257-266.
Breiman L,2001.Random forests[J].Machine Learning,45(1):5-32.
Chao Z Y,Xin Q,Yu Q,et al,2009.Application of SVM on Chl-a concentration retrievals in Taihu Lake[J].China Environmental Science,29(1):78-83.
Iverson L R,Prasad A M,Matthews S N,et al,2008.Estimating potential habitat for 134eastern US tree species under six climate scenarios[J].Forest Ecology&Management,254(3):390-406.
Lary D J,Alavi A H,Gandomi A H,et al,2016.Machine learning in geosciences and remote sensing[J].Geoscience Frontiers,7(1):3-10.
Lu F,Chen Z,Liu W,et al,2016.Modeling chlorophyll-a concentrations using an artificial neural network for precisely eco-restoring lake basin[J].Ecological Engineering,95:422-429.
Mishra S,Mishra D R,Schluchter W M,2009.A novel algorithm for predicting phycocyanin concentrations in cyanobacteria:aproximal hyperspectral remote sensing approach[J].Remote Sensing,1(4):758-775.
Moses W J,Gitelson A A,Berdnikov S,et al,2009.Estimation of chlorophyll-a concentration in case II waters using MODIS and MERIS data-successes and challenges[J].Environmental Research Letters,4(4):549-567.
Mutanga O,Adam E,Cho M A,2012.High density biomass estimation for wetland vegetation using WorldView-2imagery and random forest regression algorithm[J].International Journal of Applied Earth Observations&Geoinformation,18(1):399-406.
Saeys W,Mouazen A M,Ramon H,2005.Potential for Onsite and Online Analysis of Pig Manure using Visible and Near Infrared Reflectance Spectroscopy[J].Biosystems Engineering,91(4):393-402.
Sun D Y,Li Y M,Qiao W,2009.A unified model for remotely estimating chlorophyll a in Lake Taihu,China,based on SVM and in situ hyperspectral data[J].IEEETransactions on Geoscience&Remote Sensing,47(8):2957-2965.
Vanhellemont Q,Ruddick K,2015.Advantages of high quality SWIR bands for ocean colour processing:Examples from Landsat-8[J].Remote Sensing of Environment,161:89-106.
Vapnik V N,2000.The Nature of Statistical Learning Theory[M].Springer:988-999.
Wong F K K,Tung F,2014.Combining EO-1Hyperion and Envisat ASAR data for mangrove species classification in Mai Po Ramsar Site,Hong Kong[J].International Journal of Remote Sensing,35(23):7828-7856.
Zhang Y,Koponen S S,Pulliainen J T,et al,2003.Application of empirical neural networks to chlorophyll-a estimation in coastal waters using remote optosensors[J].IEEE Sensors Journal,3(4):376-382.
边博,夏明芳,王志良,等,2012.太湖流域重污染区主要水污染物总量控制[J].湖泊科学,24(3):327-333.
陈军,周冠华,温珍河,等,2010.太湖表层悬浮泥沙遥感定量模式研究[J].光谱学与光谱分析,30(1):137-141.
高玉蓉,刘明亮,吴志旭,等,2012.应用实测光谱估算千岛湖夏季叶绿素a浓度[J].湖泊科学,24(4):553-561.
江敏,余根鼎,戴习林,等,2010.凡纳滨对虾养殖塘叶绿素a与水质因子的多元回归分析[J].水产学报,34(11):1712-1718.
金鑫,李云梅,王桥,等,2010.基于太湖气溶胶类型分区的环境一号卫星CCD大气校正[J].湖泊科学,22(4):504-512.
李文朝,陈开宁,吴庆龙,等,2001.东太湖水生植物生物质腐烂分解实验[J].湖泊科学,13(4):331-336.
刘朝相,宫兆宁,赵文吉,等,2014.基于SVM模型的妫水河叶绿素a浓度的遥感反演[J].遥感技术与应用,29(3):419-427.
汪西莉,周兆永,延军平,2009.应用GA-SVM的渭河水质参数多光谱遥感反演[J].遥感学报,13(4):735-744.
王丽平,郑丙辉,2011.大宁河叶绿素a的因子分值-多元线性回归预测模型研究[J].长江流域资源与环境,20(9):1120-1124.
王艳丽,王飞,2016.白洋淀湖泊叶绿素a对水位周期波动的响应[J].中北大学学报(自然科学版),37(2):126-132.
韦玉春,王国祥,孙华芸,2010.使用线性回归方法构建水体叶绿素a浓度高光谱估算模型的一个逻辑问题[J].数学的实践与认识,40(18):100-110.
吴阿娜,朱梦杰,汤琳,等,2011.淀山湖蓝藻水华高发期叶绿素a动态及相关环境因子分析[J].湖泊科学,23(1):67-72.
夏叡,2010.基于遥感的太湖水质参数空间分异研究[D].南京:南京师范大学:42-78.
徐朝,2016.基于机器学习的海面盐度遥感反演模型[D].北京:中国地质大学:60-67.
许云峰,马春子,霍守亮,等,2012.以程海为例用支持向量机回归算法预测叶绿素a浓度[J].环境工程技术学报,2(3):207-211.
姚付启,张振华,杨润亚,等,2010.基于主成分分析和BP神经网络的法国梧桐叶绿素含量高光谱反演研究[J].测绘科学,35(1):109-112.
张玉超,钱新,钱瑜,等,2009a.基于机器学习方法的太湖叶绿素a定量遥感研究[J].环境科学,30(5):1321-1328.
张玉超,钱新,钱瑜,等,2009b.支持向量机在太湖叶绿素a非线性反演中的应用[J].中国环境科学,29(1):78-83.
赵汉取,韦肖杭,姚伟忠,等,2011.南太湖近岸水域叶绿素a含量与氮磷浓度的关系[J].水生态学杂志,32(5),59-63.
Abdel-Rahman E M,Ahmed F B,Ismail R,2013.Random forest regression and spectral band selection for estimating sugarcane leaf nitrogen concentration using EO-1Hyperion hyperspectral data[J].International Journal of Remote Sensing,34(2):712-728.
Agodi A,2011.Comparing performance and robustness of SVM and ANN for fault diagnosis in a centrifugal pump[J].Celebrazioni Archimedee del Sec XX(Siracusa,1961),3:81-86.
Bocai G,1996.NDWI-A Normalized Difference Water Index for Remote Sensing of Vegetation Liquid Water From Space[J].Remote Sensing of Environment,58(3):257-266.
Breiman L,2001.Random forests[J].Machine Learning,45(1):5-32.
Chao Z Y,Xin Q,Yu Q,et al,2009.Application of SVM on Chl-a concentration retrievals in Taihu Lake[J].China Environmental Science,29(1):78-83.
Iverson L R,Prasad A M,Matthews S N,et al,2008.Estimating potential habitat for 134eastern US tree species under six climate scenarios[J].Forest Ecology&Management,254(3):390-406.
Lary D J,Alavi A H,Gandomi A H,et al,2016.Machine learning in geosciences and remote sensing[J].Geoscience Frontiers,7(1):3-10.
Lu F,Chen Z,Liu W,et al,2016.Modeling chlorophyll-a concentrations using an artificial neural network for precisely eco-restoring lake basin[J].Ecological Engineering,95:422-429.
Mishra S,Mishra D R,Schluchter W M,2009.A novel algorithm for predicting phycocyanin concentrations in cyanobacteria:aproximal hyperspectral remote sensing approach[J].Remote Sensing,1(4):758-775.
Moses W J,Gitelson A A,Berdnikov S,et al,2009.Estimation of chlorophyll-a concentration in case II waters using MODIS and MERIS data-successes and challenges[J].Environmental Research Letters,4(4):549-567.
Mutanga O,Adam E,Cho M A,2012.High density biomass estimation for wetland vegetation using WorldView-2imagery and random forest regression algorithm[J].International Journal of Applied Earth Observations&Geoinformation,18(1):399-406.
Saeys W,Mouazen A M,Ramon H,2005.Potential for Onsite and Online Analysis of Pig Manure using Visible and Near Infrared Reflectance Spectroscopy[J].Biosystems Engineering,91(4):393-402.
Sun D Y,Li Y M,Qiao W,2009.A unified model for remotely estimating chlorophyll a in Lake Taihu,China,based on SVM and in situ hyperspectral data[J].IEEETransactions on Geoscience&Remote Sensing,47(8):2957-2965.
Vanhellemont Q,Ruddick K,2015.Advantages of high quality SWIR bands for ocean colour processing:Examples from Landsat-8[J].Remote Sensing of Environment,161:89-106.
Vapnik V N,2000.The Nature of Statistical Learning Theory[M].Springer:988-999.
Wong F K K,Tung F,2014.Combining EO-1Hyperion and Envisat ASAR data for mangrove species classification in Mai Po Ramsar Site,Hong Kong[J].International Journal of Remote Sensing,35(23):7828-7856.
Zhang Y,Koponen S S,Pulliainen J T,et al,2003.Application of empirical neural networks to chlorophyll-a estimation in coastal waters using remote optosensors[J].IEEE Sensors Journal,3(4):376-382.