许阳春,张明峰,苏玉萍,洪颐,苏金洙,陈晶晶
1.福建师范大学 环境科学与工程学院, 福建 福州 350007;2.福建省污染控制与资源循环重点实验室, 福建 福州 350007;3.福建师范大学 地理科学学院, 福建 福州 350117;4.法国巴黎高科路桥学院, 法国 巴黎 77455

摘要:
以福建平潭海域为研究对象、以叶绿素a浓度为输出指标，根据2009-2018年赤潮期数据规律及2013-2017年海洋监测数据主成分分析结果，对拟构建的BP模型进行输入指标筛选，选定结果包括4个气象因子和4个水质因子。基于此结果，对2013-2017年的698组海洋监测数据中叶绿素a浓度进行归一化处理并进行模型演算，随机选取80%数据作为演算模型的训练样本，其余进行模型验证。通过交叉变换输入指标，寻求最优的输入节点组合，以气温、溶解氧浓度、日照时长指标为输入参数时，BP模型误差较小（均方根误差为0.05 μg/L，平均绝对误差为0.03 μg/L），演算结果精度较高（可决系数R²=0.81）。以上结果表明，气温、溶解氧浓度和日照时长对叶绿素a浓度表征效果较好，可为平潭海域以叶绿素a浓度作为判定指标的赤潮预警研究提供参考。
关键词:BP人工神经网络赤潮叶绿素a浓度平潭海域
DOI：10.11759/hykx20190704004
分类号:X55
基金项目:国家重点研发计划（2016YFE0202100）；国家自然科学基金（41573075）；福建省科技厅高校产学合作项目（2017Y4003）；福州市科技局项目（2016-G-68）

Calculation of the Chlorophyll-a concentration of red tide in the Pingtan Coastal Zone by a BP artificial neural network model
XU Yang-chun^1,2, ZHANG Ming-feng³, SU Yu-ping^1,2, Hong Yi^4,5, SU Jin-zhu^1,2, CHEN Jing-jing¹
1.Environmental Science and Engineering College, Fujian Normal University, Fuzhou 350007, China;2.Fujian Key laboratory of Pollution Control and Resource Reuse, College of Environmental Science and Engineering, Fujian Normal University, Fuzhou 350007, China;3.Institution of Geography, Fujian Normal University, Fuzhou 350117, China;4.LEESU, MA 102, É5.cole des Ponts Paris Tech, 77420 Champs-sur-Marne, France
Abstract:
Based on the change trend of red tide data from 2009 to 2018 and a principal component analysis of ocean monitoring data of Pingtan for the 2013-2017 period, we propose a back propagation (BP) model, which was used to screen input indicators with chlorophyll a as an output indicator. Four meteorological factors and four water quality factors were obtained as input indicators. Next, the chlorophyll a concentrations in 698 sets of ocean monitoring data from 2013 to 2017 were normalized and used for model calculation. 80% of the normalized data were randomly selected for use in model training, and the rest were used for model verification. An analysis of the optimal combination of input indexes revealed that when the temperature, dissolved oxygen concentration, and sunshine duration were set as a combination index, the model accuracy was higher than other cases (R²=0.81, RMSE=0.05 μg/L, and MAE=0.03μg/L). These results indicate that temperature, dissolved oxygen, and sunshine duration are favorable factors for predicting the chlorophyll a concentration, which could be helpful for forecasting red tides in the Pingtan coastal zone.
Key words:BP neural networkred tideChlorophyll a concentrationPingtan coastal zone





PDF全文下载地址:

点我下载PDF