“…This error characteristic can be represented by allowing QPP model parameters to vary within a year (but not across years). This approach is common for predictions/forecasts at monthly to seasonal timescales (e.g., Bennett et al, 2017; Li et al, 2013; Woldemeskel et al, 2018) and has also been used in daily forecasting (e.g., Kim et al, 2018); - Dynamic biases , that is, shifts in the mean of hydrological errors over longer time periods (e.g., month to year). These nonstationarities may be due to a range of factors including systematic errors in the hydrological model calibration data (e.g., Westra et al, 2014) and/or inability of hydrological model to capture changes in hydrological processes over longer time scales, for example, due to unusually wet/dry seasons, interannual climate variability, and/or changes in catchment properties (Coron et al, 2012; Fowler et al, 2016; Merz et al, 2011; Westra et al, 2014).
…”