基于多方法遴选的VIKOR综合指数型城市水体返黑返臭概率预测模型构建OA

The development of prediction models for water blackening and odorization recurrence has become a core driver and foundational element in the refined management of urban water environments.Based on extensive water quality monitoring data from 16 cities in the Yangtze River Basin between 2020 and 2023,this paper systematically evaluates six typical comprehensive index calculation methods,and proposes and validates a probability prediction model for water blackening and odorization recurrence centered on the VIKOR comprehensive index.Through sampling analysis and literature review,a power function relationship between water transparency(y)and turbidity(x)in the Yangtze River Basin was established(y=3.12x-0.66),with the critical turbidity for blackening and odorization determined as 46.9 NTU.Using variance analysis,recursive feature elimination,and random forest triple screening,five indicators—"turbidity,DO,TP,CODMn,and NH3-N"—were identified as the index system for predicting the probability of water blackening and odorization.Their importance ranking is as follows:turbidity>DO>TP>CODMn>NH3-N.The mapping relationship between the VIKOR comprehensive index and the probability of water blackening and odorization is the most robust.The established model demonstrates both high accuracy(RMSE=0.029 and MAE=0.020)and high consistency(Nash-Sutcliffe efficiency coefficient=0.918 and R2=0.918),while prediction results based on other comprehensive indices all have R2 values below 0.84.The model exhibits strong predictive performance in the Yangtze River Basin,Pearl River Basin,Haihe River Basin,and Yellow River Basin.The findings are expected to be integrated into urban water management smart platforms,providing a universal decision-making tool for the precise identification,early warning,and targeted management of water blackening and odorization recurrence.