Tree species’ distributions in the western United States may be driven by long term trends (or means) in climate or climatic extremes (deviations from means). It is fairly simple to model tree species’ occurrence relative to climatic means; however, it is challenging and warranted to model tree species’ occurrence with respect to climatic extremes. One grand challenge is to ascertain if the classifications of time intervals based on large-scale oscillations (such as the Pacific Decadal Oscillation) are suited to characterize extreme climatic behaviors through time and space for the purpose of species distribution modeling.  Here, we test the efficacy of various classifications of time intervals. We determine whether certain classifications are more effective than others in grouping behavior of climatic extremes. We systematically sample 25 locations across Oregon, California, and Washington. We extract time series for each location including monthly precipitation, temperature, potential evapotranspiration, and the Palmer Drought Severity Index (PDSI) from interpolated climate grids. The time series date from 1933 to 2007. We also generate classification schemes for time series. We explore all combinations of the Pacific Decadal Oscillation (PDO), Atlantic-Mulitdecadal Oscillation (AMO), and El Niño and La Niña (ENSO) as classifiers. We also examine the time interval from 1976 to present, which is marked by more frequent and longer El Niños and temperature rise in some areas of the world (we call this interval R1). Our classifications are mostly based on published works characterizing regime switches. We also rely on the Southern Oscillation Index (SOI) to characterize ENSO. We treat the 25 time series of PDSI as a multivariate response matrix. We use the Multivariate Response Permutation Procedure (MRPP) to test the significance and degree to which the classifications group behavior in climatic deviations.

Results/Conclusions

Preliminary results reveal that most classifications (e.g. AMO-PDO, AMO, PDO, PDO-ENSO etc.) group drought behavior (all p values are less than 0.0001). However, the ‘A’ statistics, which reveal the degree of separation among groups, range from 0.01 to 0.11 with the ENSO-PDO-AMO combination yielding the best separation for drought behavior (A=0.11). ENSO alone yielded the least degree of separation (A=0.01). These results concur with McCabe et al. (2004), although, we demonstrate that the inclusion of ENSO with PDO-AMO may better distinguish drought behavior across the Pacific coastal United States. This work will further explore the efficacy of classification of extreme climatic behavior for different climatic variables suited to modeling tree species’ distributions.