Physicochemical parameters data assimilation for efficient improvement of water quality index prediction: comparative assessment of a noise suppression hybridization approach

Water quality has a crucial impact on human health; therefore, water quality index modeling is one of the challenging issues in the water sector. The accurate prediction of water quality index is an essential requisite for water quality management, human health, public consumption, and domestic uses. A comprehensive review as an initial attempt is conducted on existing solutions through data-driven models. In addition, the ensemble Kalman filter is found to be a suitable data assimilation method, which is successfully applied in hydrological variables modeling and other complexes, nonlinear, and chaotic problems. In this study, a new application of ensemble Kalman filter-artificial neural network is proposed to predict water quality index using physicochemical parameters for two commonly pollutant rivers, namely Klang and Langat, in Malaysia. As a further attempt, in order to improve the models’ performance, a new preprocessing technique is adopted as the newly constructed assimilated model. The results confirm that ensemble hybrid based intrinsic time-scale decomposition has reduced root mean square error by 24% for Klang and 34% for Langat, respectively, compared with the intrinsic time-scale decomposition-conventional neural network model. Overall, the developed assimilated methodology shows the robustness of the proposed ensemble hybrid model in analyzing water quality index over monthly horizons that experts could evaluate the water quality of rivers more efficiently.