We chose to focus on the 30-year future from 2041 through 2070, since we did not receive sufficient funding to simulate the entire 21st century, and we wanted to choose a time period that was far enough in the future so that the climate change signal would be clear, but close enough to the current so that the results would be useful for impacts calculations and planning purposes. It is now becoming common to look at even an earlier period in the future for adaptation studies (e.g., the 2020s and 2030s), but we believe our mid-range future time period will still prove useful. We may try to extend the program to modeling of those earlier future decades. The current period (1971–2000) is a standard current period from which to view the change in the future.

The NCEP-2 reanalsyis used for driving the RCMs in Phase I does not have high-quality data until 1979. Allowing 1 year for spin-up of the regional model, the Phase I runs cover 25 years, from 1980 through 2004.

Caveat: Spin-Up

Because different elements of the models equilibrate at different speeds, there is a 'spin-up' period at the beginning of each model run during which data for some variables will be unusable. For the benefit of experts, we have made data available for the entire run, but for general use, we recommend using data ONLY from the following periods:

Note that these periods start and end one month off from the year boundaries so that seasonal averages will not be thrown off by the presence of a fractional winter at the beginning and end of the run.

Missing Data

Some of the model runs may not cover the full usable period. For example, the driving CCSM run ends on 12/31/1999, and so the embedded RCMs cannot run past 12/30/1999. More discussion of missing data can be found on the Data Status page.

Calendars

In addition, there will be differences from model to model in the total number of timesteps in a run because different driving AOGCMs use different calendars. The 365-day or "noleap" calendar is the same as the standard (Gregorian) calendar, but does not have leap-years. The 360-day calendar has 12 months of 30 days each. The calendar used for each run is recorded in the calendar attribute of the time dimension.