基于变量重要性评分-随机森林的溶解氧预测模型——以深圳湾为例

Dissolved oxygen prediction model based on variable importance measures and random forest:A case study of Shenzhen Bay

运用Pearson相关性分析,变量重要性评分和随机森林方法构建了溶解氧(DO)实时预测模型,并以深圳湾为例采用浮标资料预测1,3,6和12h的溶解氧.模型预测结果表明,模型最优的输入条件为pH值,水温,叶绿素a,氧化还原电位和蓝绿藻5个水质指标,1h预报的相关系数在0.9以上,6h预报结果一定程度上可以满足工程要求,但对低溶解氧事件的预报必须在3h以内.

A real-time prediction model for dissolved oxygen was established by using Pearson correlation analysis,variable importance measures and random forest method.Taking Shenzhen Bay as an example,the model was used to predict the dissolved oxygen in 1h,3h,6h and 12h based on the buoy data.The results showed that the optimal input conditions of the model were pH,water temperature,chlorophyll A,redox potential and blue-green algae.The correlation coefficient of 1h prediction results was more than 0.9,and the 6h prediction results could meet the engineering requirements to a certain extent.However,the prediction of low dissolved oxygen events might be within 3h.