In the following discussion @cbengel discussed cycling and free-running approaches. I mentioned I’d sometimes seen step changes in near-surface variables when taking a cycling approach (e.g. in BARRA, see below) which might affect land-sea breeze dynamics when analysed in aggregate. On the other hand, @pjs548 warned that a free-running soil moisture might drift over time. I looked into this, from our 13-month BARRA-driven free-running ACCESS simulations, and I don’t see any drift of soil moisture compared with flux tower locations in SE Australia. @pjs548, can you detail what drift you expected to see?
There are three types of long-running simulations in my opinion:
“climate” runs, where data is used only for statistical properties,
“re-analysis” runs, that continually incorporate observational data, and
“pseudo-forecast” runs, where you emulate the type of information that would have been available if an operational forecast centre had been running your experiment/domain, etc and end up with a long collection of short-term forecasts.
In the talk on Friday I was asking @andrewb1 if he was wanting to do a “pseudo-forecast” rather than a pure climate-like run. “Pseudo-forecast” runs would have jumps in parameters but the “forecast” data would be more accurate because it would be less hours into a forecast (starting from reanalysis data).
All choices are valid but affect the type of analysis that can be done.
