The M-Climate is derived from a set of medium range re-forecasts. These are created using the same calendar start dates over several years for data times either side of the time of the ensemble run itself. The re-forecast runs are at the same resolution as the medium range ensemble (currently 9km) and run over the 15-day medium range ensemble period.
There is merit in examining the real-time performance of a forecasting system. But the sample sizes created for one system are far too small to conclude anything about its true performance levels. Re-forecasts are used to increase the available data to produce a model climate. The results of forecast system may be compared with this model climate.
Re-forecasts are a fundamental component of all seasonal forecasting system; they have two applications:
The set of re-forecasts is made up from:
In total, each set of re-forecasts consists of 20 years x 9 runs x 11 ensemble members = 1980 re-forecast values. These are available for each forecast parameter, forecast lead-time, calendar start date, location, at forecast intervals of 6 hours.
The M-climate is used in association with the ensemble forecast:
The same M-climate set is used for 00UTC and 12UTC ensemble runs. This is to avoid inconsistencies between the validity period of the ensemble and M-climate. So, for example:
M-climate is updated twice per week on Mondays and Thursdays. So M-climate quantile plots for the same lead-time from two forecast runs on e.g. 00UTC Thursday and 00UTC Friday will be slightly different. This limitation of twice weekly updates to the M-climate can be significant. It can be particularly evident in spring and autumn when mean temperatures are changing most rapidly day by day.
ECMWF uses different reference periods but essentially the same re-forecast runs to build the M-climate and the ER-M-climate. The key difference is that those runs are grouped and used in different ways:
Note before Cy41r1 in spring 2015, the M-climate was constructed from only 500 re-forecasts was more prone to sampling errors and as a result.