Comparison of runoff modelled using rainfall from different downscaling methods for historical and future climates

This paper: (i) assesses the rainfall downscaled from three global climate models (GCMs) using five downscaling models, (ii) assesses the runoff modelled by the SIMHYD rainfall–runoff model using the downscaled daily rainfall, and (iii) compares the modelled changes in future rainfall and runoff characteristics. The modelling study is carried out using rainfall and streamflow data from eight unimpaired catchments near the headwaters of the Murray River in south-east Australia. The downscaling models used, in increasing order of complexity, are a daily scaling model, an analogue statistical downscaling model, GLIMCLIM and NHMM parametric statistical downscaling models, and CCAM dynamic downscaling model. All the downscaling models can generally reproduce the observed historical rainfall characteristics. The rainfall–runoff modelling using downscaled rainfall also generally reproduces the observed historical runoff characteristics. The future simulations are most similar between the daily scaling, analogue and NHMM models, all of them simulating a drier future. The GLIMCLIM and CCAM models simulate a smaller decrease in future rainfall. The differences between the modelled future runoff using the different downscaled rainfall can be significant, and this needs to be further investigated in the context of projections from a large range of GCMs and different hydrological models and applications. The simpler to apply daily scaling and analogue models (they also directly provide gridded rainfall inputs) can be relatively easily used for impact assessments over very large regions. The parametric downscaling models offer potential improvements as they capture a fuller range of daily rainfall characteristics.