| 摘要: |
| 为了更好的评估湖库富营养化及其引发藻类爆发风险性的程度,取福州市山仔水库水质监测数据作为训练样本,基于BP神经网络建立预测藻类浓度的水体富营养化模型,采用GLUE方法对水体富营养化模型的相关水质参数进行不确定性分析,评估分析各参数对水体富营养化程度的影响,并给出各参数90%置信区间作为风险范围以供管理。结果表明：BP神经网络可以较好地预测藻类浓度变化,纳什系数(NSE)及均方根误差(RMSE)均较小；Ph、DO、CODmn、TN、TP、Chl跟水华风险有着较为明显的正相关趋势,SSD存在负相关趋势,温度T则不明显；Ph、TN、TP、Chl、SSD的高、低风险特征在区间内表现较为明显；T、DO、CODmn的风险特征则不明显；各水质参数90%置信区间可用作其藻类爆发的风险评估范围。 |
| 关键词: GLUE;BP神经网络;富营养化;风险评估 |
| DOI： |
| 分类号:X822 |
| 基金项目: |
|
| Evaluation of Lacustrine Eutrophication Model Based on GLUE |
|
ZHENG Zheng
|
|
Fuzhou Research Academy of Environmental Sciences
|
| Abstract: |
| In order to accurately assess the eutrophication degree of water and the risk interval of various water quality indexes related to algae outbreak, a water eutrophication model was established based on BP neural network based on the water quality monitoring data at the monitoring point of shanzai reservoir in fujian province. Using GLUE method, the uncertainty analysis of water quality parameters related to eutrophication model was carried out, and 90% confidence interval of each parameter was statistically evaluated. The results showed that the BP neural network could predict the change of algae concentration, and the NSE and RMSE were small. Ph, DO, CODmn, TN, TP and Chl have a relatively obvious positive correlation with the risk of water bloom, SSD has a negative correlation trend, and T was not conspicuous. The high and low risk characteristics of Ph, TN, TP, Chl and SSD were conspicuous in the parameters of the interval and T and others were the opposite. The 90% confidence interval parameter evaluation range of each water quality parameter can be used as a risk assessment index. |
| Key words: GLUE; BP neural network; eutrophication; risk assessment |
