Of Fire and Rain: Regional climate change detection and attribution

<p><strong>While the global fingerprint of climate change is becoming increasingly clear, the patterns of regional climate change remain blurred and uncertain. Climate change detection and attribution - the study of distinguishing long-term changes from internal variability and estimating the contributions of external factors – is particularly challenging at the regional scale. Regional detection and attribution is a historically challenging low signal-to-noise problem as the signals (responses to external forcings) are small compared to the internal variability (“noise”). The predominant method, optimal fingerprinting, compares observations against climate model-simulated response patterns and internal variability to identify spatiotemporal fingerprints (unique patterns across space and time) of anthropogenic and natural climate change. In previous studies using fully-coupled models, significant temporal or spatially averaging is required to compensate for inadequate samplings of internal variability. However, such averaging risks losing valuable spatial and temporal information essential for regional detection.</strong></p><p>In this dissertation, I wield the power of large ensembles of atmosphere-only model simulations to apply optimal fingerprinting with minimal spatial or temporal averaging.</p><p>These ensembles provide unprecedented samplings of internal variability, allowing the full spatiotemporal complexity of the observations and model outputs to be utilized in the optimal fingerprinting analysis. I apply this method to two case studies – western North American fire weather and New Zealand hydroclimate – which introduce multivariate and hydroclimate processes in topographically diverse regions to the regional detection problem. Both anthropogenic and natural fingerprints are robustly detected in western North American fire weather and New Zealand temperature. Anthropogenic fingerprints in New Zealand precipitation and streamflow trends are not detected, though natural fingerprints are detected when wetter areas are given more weight. In addition to deepening the understanding of the effects of anthropogenic climate change in these regions, the wider impact of this thesis is the demonstration of the benefits – and caveats – of using atmosphere-only models for the detection and attribution of regional climate change.</p>