Observational Large Ensemble

These python datasets contain the results presented in the above paper with regard to the variability in trends over North America during DJF due to sampling of internal variability. Two types of files are available. The netcdf file contains samples from the synthetic ensemble of DJF temperatures over North America from 1966-2015. The synthetic ensemble is centered on the observed trend. Recentering the ensemble on the ensemble mean trend from the NCAR CESM1 LENS will create the Observational Large Ensemble, in which each sample can be viewed as a temperature history that could have occurred given various samplings of internal variability. The synthetic ensemble can also be recentered on any other estimate of the forced response to climate change. While the dataset is both land and ocean, it has only been validated over land. The second type of file, presented as python datasets (.npz) contains the results presented in the McKinnon et al (2017) reference. In particular, it contains the 50-year trends for both the observations and the NCAR CESM1 Large Ensemble that actually occurred, and could have occurred given a different sampling of internal variability. The bootstrap results can be compared to the true spread across the NCAR CESM1 Large Ensemble for validation, as was done in the manuscript. Each of these files is named based on the observational dataset, variable, time span, and spatial domain. They contain: BETA: the empirical OLS trend BOOTSAMPLES: the OLS trends estimated after bootstrapping INTERANNUALVAR: the interannual variance in the data after modeling and removing the forced trend empiricalAR1: the empirical AR(1) coefficient estimated from the residuals around the forced trend The first dimension of all variables is 42, which is a stack of the ensemble mean behavior (index 0), the forty members of the NCAR Large Ensemble (indices 1:40), and the observations (last index, -1). The second dimension is spatial. See latlon.npz for the latitude and longitude vectors. The third dimension, when present, is the bootstrap samples. We have saved 1000 bootstrap samples.