Anthropogenic CO2 emissions drive atmospheric and ocean warming, along with deoxygenation and acidification.  Global Climate Models (GCMs) lack the spatial resolution needed to represent the fine-scale climate information required for resource management.  Therefore, GCM outputs must be downscaled to an appropriate resolution. The first part of the presentation will focus on the generation of actionable climate data to support climate adaptation. 

Superimposed on these long-term trends are episodic extremes of temperature, precipitation, pH and oxygen, but these are not well understood.  The second part of this presentation will focus on innovative methods for assessing single and compound stressor extremes. I will detail our recent work using machine learning where we assume that organisms are adapted to local environmental conditions and use an unsupervised clustering approach to isolate regions with similar habitat characteristics.  Extreme thresholds are defined seasonally using a fixed baseline (1996-2020). We quantify the fraction of waters that are extreme in each cluster in the recent past for both single and compound stressors and examine the relationship of local extremes to large scale Pacific climate variability.