These files contain 1000 members of the Observational Large Ensemble.

The code to produce the files, as well as more documentation, can be found at

https://github.com/karenamckinnon/observational_large_ensemble

Each climate variable is modeled as a linear combination of
- the climatology
- a forced component
- the response to ENSO, PDO, and AMO, and
- residual climate noise

The ensemble is produced through randomizing the latter two components. See McKinnon and Deser (2018) JClim.

The provided files include all components, although all validation is focused on the variability alone (not the forced
component). As such, the user should proceed with caution when analyzing forced changes in the ensemble. The forced
component is estimated through regressing the global mean-ensemble mean time series of either temperature or
precipitation onto the gridbox level time series. This method has been validated somewhat for temperature, but much
less so for precipitation. (The forced trend in sea level pressure is assumed to be zero).

As such, the user may want to remove the forced trend from each member of the ensemble, and replace it with their own
best estimate (from e.g. a multi model ensemble, dynamical adustment, etc.) To allow for this, the files tas_forced.nc
and tas_climatology.nc (as well as the same for pr_) are provided and can be linearly removed from each ensemble
member. The result will simply be different sequences of the _variability_, which is the focus of the methodology.

Please contact Karen McKinnon (kmckinnon@ucla.edu) with questions or if you are using the data in a publication.