抚仙湖流域负荷削减的水质风险分析

A Linked EFDC-NN model for risk-based load reduction analysis of Lake Fuxian watershed

以抚仙湖为研究对象,基于三维水动力-水质模型EFDC平台,开发了EFDC-神经网络(NN)耦合模型;并选用30d移动平均值为解译方式核算抚仙湖在不同风险下的流域负荷削减(TMDL).结果表明,对于100%的达标频度,为了达到I类水质,TP允许增加14%～18%,COD允许增加9%～11%,但TN需要削减13%～14%.如果放松对达标频度的要求,污染负荷将允许相应地增加.研究结果可为流域管理依据不同的风险与管理费用偏好实施流域削减提供基础.

This study employed an approach linking the Environmental Fluid Dynamics Code （EFDC） model and a Neural Network （NN） model to conduct risk based TMDL analysis for Lake Fuxian in Southwestern China. The EFDC-NN system was developed based on a three-dimensional hydrodynamic-water quality model of the lake, and NN functional approximators of the EFDC model using the 30-day moving average concentration of target nutrients as predictors. The developed NN functional approximators were then applied to conduct risk based TMDL analysis. To reach the primary water quality standard with 100% probability, the TP and COD loadings from the watershed still had room for increase by 14%-18% and 9%-11% respectively, but TN must be reduced by 13%-14%. If the water quality compliance probability was relaxed by various levels, it was found that the load reduction requirement will be correspondingly relaxed such that in most cases further increase of watershed loadings are allowed. The results of this analysis provided decision makers with risk based load management requirement for guiding their management plan in this lake watershed.