Harnessing Historical Climate Variability to Assess Multivariate Climate Changes

Abstract:

Climate is intrinsically multivariate—the collective influence of various aspects of weather at different times of year. A central challenge of climate change impact analysis is therefore to characterize changes in multiple temperature and precipitation variables simultaneously. Historical climate variability provides key context for relating climate variables to each other and assessing collective deviations from historical climate conditions. We have developed a Mahalanobian probability metric to describe spatial and temporal climatic dissimilarity in terms of local interannual climatic variability. Our approach is particularly suited to evaluation of climate analogs in space and time, but also facilitates multivariate extensions to several prominent indices of climate change. We use this metric to detect the departure of multivariate climate conditions from the historical range of local variability across North America and to identify regions that are particularly susceptible to emergence of no-analog climates. With respect to interpreting climate extremes, some critical considerations emerge from this research. In particular, we highlight the potential for temporal aggregation to exaggerate the statistical significance of extreme conditions, and the dilemma of identifying an appropriate statistical distribution for precipitation across both space and time. Despite the challenges of interpreting the specific impacts associated with multivariate climate changes and extremes, expressing these conditions relative to historical climate variability provides a useful first approximation of their ecological and socioeconomic significance.

Figure Caption: Demonstration of the use of the chi distribution to measure spatial climatic dissimilarity in terms of local interannual climatic variability.