Disaggregated monthly hydrological models can outperform daily models in providing daily flow statistics and extrapolate well to a drying climate

Abstract

Daily timescale hydrological information is important for many purposes such as flood estimation, predicting the consequences of catchment management and meeting the needs of freshwater ecology. In hydrological assessments, daily timestep modelling is typically used because of the availability of daily data and many of the processes governing impacts occur at this timescale. However, daily models can suffer from poor performance in certain contexts (e.g. drier climates), and their computational requirements can make it difficult to efficiently explore many sources of uncertainty in some situations, such as understanding the impacts of climate change in larger water resource systems. Here, we test an alternate approach based on monthly modelling with a post-processing step involving monthly-to-daily disaggregation using historic flow patterns conditioned on soil moisture estimates. We apply our approach to 214 catchments across Australia representing a wide range of climate and hydrological conditions, and assess outcomes for multiple objectives spanning water supply, flood magnitude and freshwater ecological outcomes, and validate performance over an extreme multi-year drought with substantially different rainfall and streamflow characteristics. Our results show that for many metrics including sustained low flows, annual flow maxima, and high and low flow spells, the results based on monthly hydrologic modelling with daily disaggregation are generally better than those based on daily hydrological modelling. This was especially true for ecologically relevant flow metrics. In addition, the disaggregation approach fared better than the daily model when extrapolating to the multi-year dry period. Our approach also has the potential to greatly reduce the effort required to explore uncertainty in large river systems.