Multivariate Statistical Modeling of Seasonal River Water Quality Using Limited Hydrological and Climatic Data

Abstract

Effective water resource management requires an understanding of the interactions between water and environmental parameters, especially in regions with limited data availability. This study used generalized additive models (GAMs) to investigate the relationship between climatic and hydrological factors, namely river flow, rainfall, air temperature, and physicochemical water quality parameters in the Kiso River, Japan. Seasonal and non-seasonal GAMs models were developed for each water quality parameter, resulting in 7 non-seasonal models and 28 seasonal models based on Japan’s meteorological seasons (winter, spring, summer, fall). The findings demonstrated how seasonal models captured seasonal variability, significantly outperforming the non-seasonal models. For example, turbidity in winter (R2 = 0.5030) showed significant improvement compared with non-seasonal models (R2 = 0.1470), and organic pollution in fall (R2 = 0.4099) increased compared with non-seasonal models (R2 = 0.2509). Beyond assessing the influence of environmental drivers on water quality, these findings are crucial in regions with limited data, emphasizing the role of model–based seasonal analysis in identifying high-risk contamination periods, and supporting targeted and effective water management and early warning systems.