Robustness of hydroclimate metrics for climate change impact research

Metrics based on streamflow and/or climate variables are used in water management for monitoring and evaluating available resources. To reflect future change in the hydrological regime, metrics are estimated using climate change information from Global Climate Models or from stochastic time series representing future climates. Whilst often simple to calculate, many metrics implicitly represent complex physical process. We evaluate the scientific validity of metrics used in a climate change context, demonstrating their use to reflect aspects of timing, magnitude, extreme values, variability, duration, state, system services, and performance. We raise awareness about the following generic issues (a) formulation: metrics often assume stationarity of the input data, which is invalid under climate change; and do not always consider potential changes to seasonality and the relevance of the temporal window used for analysis; (b) climate change input data: how well are the physical processes relevant to the metric represented in the climate change input data; what is the impact of bias correction on relevant spatial and temporal scale dependencies and relevant intervariable dependencies; how realistic are the data in representing sequencing of events and natural variability in large-scale ocean-atmosphere systems; and (c) decision-making context: are rules and values that frame the decision-making process likely to remain constant or change in a future world. If critical climate or hydrological processes are not well represented by the metric constituents, these indices can be misleading about plausible future change. However, knowledge of how to construct a robust metric can safeguard against misleading interpretations about future change.