| 摘要: |
| 【目的】更好地评估湖库富营养化及其引发藻类暴发风险的程度。【方法】基于BP神经网络，建立预测藻类浓度的水体富营养化模型，结合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，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】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-a与水华风险具有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-a、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。【结论】BP神经网络可建立水库藻类及其他类似复杂机制系统的模拟，各水质参数90%置信区间可用作其藻类暴发的风险评估范围。 |
| 关键词: GLUE；BP神经网络；富营养化；风险评估 |
| DOI：10.13522/j.cnki.ggps.2019372 |
| 分类号: |
| 基金项目: |
|
| Using GLUE to Evaluate Lacustrine Eutrophication |
|
ZHENG Zhen
|
|
Fuzhou Research Academy of Environmental Sciences, Fuzhou 350000, China
|
| Abstract: |
| 【Objective】Leaching of nitrogen and phosphorus from soil could result in eutrophication in surface water, and quantitatively assessing eutrophication and evaluating its risks is important in water quality classification and improving water and fertilization management. This paper aims to test the feasibility of using CLUE to evaluate lacustrine eutrophication.【Method】The water eutrophication model was developed based on the BP neural network using water quality data measured from a monitoring point at the Shanzai reservoir in Fujian province. Uncertainty of all parameters in the eutrophication model and their 90% confidence intervals were analyzed using the GLUE method.【Result】The BP neural network was able to predict the change in algae concentration, with small NSE and RMSE. The risk of water bloom was correlated positively to pH, DO, CODMn, TN, TP and Chl-a, while negatively to SSD. Temperature did not have noticeable impact on algae bloom. The high and low risks of pH, TN, TP, Chl-a and SSD were noticeable, while others were not.【Conclusion】The BP neural network model is able to predict algae bloom using some key parameters that are easy to measure from reservoir and lakes. The 90% confidence interval of all parameter in the model can be used as risk assessment indexes. |
| Key words: GLUE; BP neural network; eutrophication; risk assessment |
