The potential of hybrid evolutionary fuzzy intelligence model for suspended sediment concentration prediction

Abstract

Providing a robust and reliable prediction model for suspended sediment concentration (SSC) is an essential task for several environmental and geomorphology prospective including water quality, river bed engineering sustainability, and aquatic habitats. In this research, a novel hybrid intelligence approach based on evolutionary fuzzy (EF) approach is developed to predict river suspended sediment concentration. To demonstrate the modeling application, one of the highly affected rivers located in the north-western part of California is selected as a case study (i.e., Eel River). Eel River is considered as one of the most polluted river due to the streamside land sliding, owing to the highly stochastic water river discharge. Thus, the predictive model is constructed using discharge information as it is the main trigger for the SSC amount. The prediction conducted on different locations of the stream (i.e., up-stream and down-stream stations). Three different well-established integrative fuzzy models are developed for the validation purpose including adaptive neuro-fuzzy inference system coupled with subtractive clustering (ANFIS-SC), grid partition (ANFIS-GP), and fuzzy c-means (ANFIS-FCM) models. The predictive models evaluated based on several numerical indicators and two-dimension graphical diagram (i.e., Taylor diagram) that vividly exhibits the observed and predicted values. The attained results evidenced the predictability of the EF model for the SSC over the other models. The discharge information provided an excellent input attributes for the predictive models. In summary, the discovered model showed an outstanding data-intelligence model for the environmental perspective and particularly for Eel River. The methodology is highly qualified to be implemented as a real-time prediction model that can provide a brilliant approach for the river engineering sustainability.