ABSTRACT In catchment hydrology, it is common practice to aggregate time series to annual values, for example to calculate the annual water balance or annual flood statistics. All annual statistics are potentially influenced by the start date used. Generally, the water year is used instead of the calendar year to avoid that precipitation from one year influences runoff in the following year. In snow‐dominated catchments in the northern hemisphere, such a carryover would occur if the calendar year were used to aggregate hydrological data. To ensure that the snow melts in the same year in which it fell, calculations are usually based on hydrological years, which for many snow‐dominated catchments in the northern hemisphere are defined as starting in fall (e.g., October 1 or November 1). However, worldwide, there is a wide variation in the start date of the water year. In this study, we use a modelling approach to numerically evaluate the definition of a water year and discuss how it could be defined in different hydroclimatic regions. For this purpose, almost 3000 catchments in Brazil, Germany, and the US were simulated with the HBV model, and the time series of the simulated snow, soil, and groundwater storage were analysed. The results show that the ‘optimal’ definition of the water year varies widely across regions and often differs considerably from the official start of the water year. There were many catchments for which the interannual variation in storage was smaller in spring, when storage was filled, than in fall, when refilling varied from year to year.
Seibert et al. (Mon,) studied this question.