Matching Statistically Downscaled Climate Projections to Northeastern U.S. Heat Application Sensitivities

One approach by which information gleaned from projections of future climate may be used to help inform adaptation planning and other decision-making processes can be described as being comprised of the following steps.
[1] 21^st century climate projections are generated by a set of climate models driven by time varying radiative forcing scenarios.
[2] Climate model projections from [1] are refined via bias correction/statistical downscaling (SD) techniques.
[3] Applied science researchers (e.g., specialists in fields sensitive to climatic variability and change, such as agriculture, human health, water resources, the energy sector, ecosystem management, etc.) select a subset of the available statistically downscaled climate projections from [2] and use the downscaled projections as input to their application models.
[4] The application models’ representations of the impact of climate variability and change on features of interest are analyzed and communicated to stakeholders.
Here we focus on the often underappreciated influence that the choice of SD technique can have on some metrics one may associate with heat-related impacts in the Northeastern United States (e.g., changes in summer mean maximum daily temperatures, counts of days above a temperature threshold, the frequency and duration of heat waves). The sensitivity of different temperature metrics to the choice of SD technique is illustrated using projections from four global climate models processed using multiple SD techniques. Though some temperature metrics are relatively insensitive to the choice of SD technique, in other cases notable differences arise from the different statistical methodologies. Results suggest that climate change impacts studies that are sensitive to the tails of the temperature distribution and to the representation of future daily weather sequences can be influenced more by the choice of SD technique than are applications whose results are linked more to shifts in time mean temperatures and central tendencies. These results demonstrate the value of strengthening the interdisciplinary linkage between steps 2 and 3 above. More specifically, determining the suitability of particular downscaled climate projection data products for use in a given climate impacts application is enhanced when consideration is given to the interplay of the characteristics of statistically downscaled climate projections and the sensitivities of a climate impacts application model to climate variable inputs. Strengthening this interdisciplinary linkage can involve bi-directional exchanges of information and knowledge beyond that typically associated with the process of downloading climate projection data products from online servers.