摘要
针对海湾初级生产力估算与预测难题,结合大亚湾近20 a的调查资料,基于MATLAB语言编程,将NH4-N、NO3-N、NO2-N、PO4-P、SiO3-Si、N/P作为输入,叶绿素a作为输出,建立大亚湾初级生产力的人工神经网络预测模型,并进行检验,其模拟值的平均相对误差0.932%;同时应用多元回归方法进行拟合预测,其拟合结果的平均相对误差为38.970%。研究结果表明,人工神经网络方法优于传统的统计学模型,具有较好的预测能力和实用性,可进行海湾初级生产力动态的预测估算,并具有较高的精度。
According to the primary productivity estimates and forecast problems in the bay, based on the survey data, NH4-N, NO3 - N, NO2-N, PO4-P, SiO3-Si, N/P, and chlorophylla, in Daya Bay nearly 20 years, and MATLAB language programming, the establish and test of the artificial neural network forecasting model of the primary productivity in the Daya Bay were done. The average value of the simulation relative error is 0. 932% ; at the same time, application of the multiple regression methods was made, the average fitting results relative error is 38. 970%. The results show that the artificial neural network method is superior to the traditional statistical models, has good predictive capability and practicality. It could conduct the primary productivity dynamics of the gulf forecast estimates and higher precision.
出处
《海洋环境科学》
CAS
CSCD
北大核心
2009年第6期652-656,共5页
Marine Environmental Science
基金
科技部科研院所社会公益研究专项资金项目(2005DIB3J020)
中央级公益性科研院所基本科研业务费专项资金项目(中国水产科学研究院南海水产研究所)(2007ZD003)
关键词
初级生产力
人工神经网络
预测
大亚湾
artificial neural networks
primary productivity
forecast
Daya Bay
