In Soil Taxonomy, the reference timeframe for soil temperature and moisture regime is the 30-year normal as you suggest. Specifically, a normal year (Keys to Soil Taxonomy, 12th ed., 2014) is one that has:
1. Annual precipitation that is plus or minus one standard deviation of the long-term (30 years or more) mean annual precipitation; and
2. Mean monthly precipitation that is plus or minus one standard deviation of the long-term monthly precipitation for 8 of the 12 months.
For the most part, normal years can be calculated from the mean annual precipitation; however, when catastrophic events occur during a year, the standard deviations of the monthly means should also be calculated.
So, a 30-year record is used that increments on a decadal basis and the use of normal years prevents unusual catastrophic events from causing regime shifts.
The moisture and temperature regimes used in classifying soils are modeled rather than observed or measured in most cases. The notable exception is the wet soils or Aquic regime where it may be observed or inferred from morphology. The models for soil moisture and temperature regime as represented on maps at various scales serve as an aid classifying the soil. The location of the lines or boundaries may shift on a decadal basis as the model is rerun on the 30-year record as the most recent 10 years of record are added and earliest 10 years are dropped.
Long-term climate change will shift the location of boundaries, if not the entire moisture or temperature regime. Even short-term fluctuations in the climate record can cause the modeled boundaries to shift several tens of miles horizontally and hundreds of feet vertically. These fluctuations are more readily observed in the production and possibly somewhat by the composition of plant communities than in the morphology of the soil. Shifts from mesic to frigid, or ustic to udic, and back again over a 30 year climate record would not warrant reclassifying soils or plant communities because of these are likely short-term fluctuations. In the western US, we used longer lived species (trees) as primary indictors of temperature and moisture regime, though simple presence or absence was not always the singular indicator.
If models are beginning to show shifts in soil moisture or temperature regime, it could be due to improvements in the accuracy of the model or addition of new weather observation points enabling greater precision and definition in the map. If vegetation indicates a real difference in temperature or moisture regime not shown by the model, then it is probably the earlier model was not sensitive to the causal parameter or was of too coarse scale to delineate it. Topographic and terrain factors such as rain shadows caused by mountains, lake effect enhanced precipitation or moderated temperature, and slope aspect may not been accounted for in previous models over large areas.
I do not believe the State and Transition (S&T) models used to the illustrate the pathways and drivers of changes in plant communities is sensitive to long-term climate change. That is, climate change is outside the temporal scale of the ecological site and the S&T model. Climate change would be reflected in the long term as changes from one ecological site to another ecological site, rather than from one state to another state within a single ecological site.
