Change-signal impacts in downscaled data and its influence on hydroclimate projections

Abstract

Downscaling is the process by which output from global climate models is translated to finer resolution regional scale projections often used in impact studies. Many fundamentally different techniques can be used, each with different capabilities of resolving or representing sub-gridscale processes. The different formulations can lead to variations in the downscaled output with consequential impacts in the interpretation of future change. Here, future runoff is estimated for six catchments in the Australian state of Victoria using five different regional rainfall projection products. We investigate how differences in rainfall input manifest in a selection of regionally important hydroclimate metrics. Overall, annual runoff is projected to decline under most methods, but seasonal changes are more uncertain reflecting differences in the rainfall change signal for different downscaled products. Whilst change in flow metrics are mostly consistent with rainfall change factors, changes in low flow (e.g. 7-day minimum flow) show considerable uncertainty, especially for drier, ephemeral catchments. Results from empirical (simple) scaling of climate observations generally lie within the range of more complex downscaling methods. However, empirical scaling is unable to provide meaningful information on spatial heterogeneity in the change signals, as well as for several metrics of rainfall and runoff. Other downscaling methods can potentially provide information on these, but the large uncertainty remains a problem, as well as our currently poor understanding of method-related biases.