Drought and weir construction impact stationarity assumption in watershed water quality modeling in South Korea

Existing watershed water quality models assume that the physio-biochemical processes affecting water quality variables are stationary. Using Hydrologic Simulation Program-Fortran (HSPF) parameterized before drought and weir construction, this study investigates if this assumption is valid in the model used for simulating water quality variables after drought and weir construction. The study area is located in South Korea, where tremendous ecological stress occurred following a drought and the completion of weir construction in 2012. To assess model performance and explore temporal changes in water quality variables, various goodness-of-fit measures and wavelet analyses are used. This study indicates that drought is the dominant stressor on water quality, and the co-occurring impact of weir construction is not discernible. Drought and weir construction cause nonstationary temporal changes in individual water quality variables and in their interrelationships, for example, dependence of dissolved oxygen on water temperature. This study demonstrated that the stationarity assumption implicit in the model parameterized under a non-drought condition is not valid during a drought condition. As a result, the model performance degrades when simulating drought-influenced physio-biochemical processes of water quality variables. The HSPF model lacks functionality for translating temporal changes in water temperature into kinetic parameters that govern biochemical processes of various water quality variables. Wavelet analysis is a good tool for elucidating unique aspects of nonstationary temporal change at various temporal scales, especially for recognizing algal succession pattern changes and identifying nonstationary relationships between any two variables.