The ER-M-climate is used in association with Extended Range ENS forecasts, primarily to highlight significant forecast anomalies in weekly averages of 2m temperature, soil temperature and sea-surface temperature, mean sea level pressure, and precipitation from the norm for a given location and time of year.
To evaluate the ER-M-climate, 3 consecutive re-forecast sets are used, based on Mondays and Thursdays, the middle one of which corresponds to the same date (i.e. day-of-month and month) as the data time of the extended ENS run itself. These 3 re-forecast sets are each created using the same calendar start dates for each of the last 20 years. So each ER-M-climate dataset consists of 3 sets of an 11-member ensemble (1 control and 10 perturbed members) run over the 46-day Extended Range ENS period. Therefore altogether 20 years x 3 runs x 11 ENS members = 660 re-forecast values are available to define the ER-M-climate for a given location and forecast lead-time. Note also that the ER-M-climate re-forecast runs are at the same resolution as the extended range runs themselves (currently ~18km up to day15 and ~36km thereafter). In fact it is essentially the same re-forecast runs that are utilised to build the M-Climate and the ER-M-Climate; the key difference is that those runs are grouped together in different ways.
ER-M-climate is updated twice a week, every Monday and Thursday, and based on 00UTC runs only (there are no 12UTC re-forecast sets). The new files start to be used from the 00UTC run that day. So if one compares, for the same lead-time, the ER-M-climate quantile plots (e.g. for a Thursday run, and a run the following Monday, they will be slightly different as the ER-M-climate changes). The impact of twice-weekly updates to the ER-M-climate is similar to that for M-climate and can be significant, particularly in spring and autumn when mean temperatures are changing most rapidly day by day. Note also that 33% of the re-forecast members making up the ER-M-Climate change each time that the ER-M-climate is updated. For the shorter range M-Climate the equivalent change is 1/9 = 11%. So the ER-M-climate is a little more prone to jumpiness related to sampling.
One might ask why ECMWF uses different length reference periods to define the M-Climate and the ER-M-Climate. The reason is that they are used in different ways. In the shorter ranges, where extremes are a priority, we want to define the climatological distribution tails as well as we can, and it has been shown that using 1980 realisations (spanning 4 weeks) achieves this much better than using 660 (spanning 1 week). Conversely for longer ranges the priority is correct representation of seasonal cycles, which can be better achieved by spanning 1 week rather than 4. And because we use week-long averaging in the extended ranges the tails should not be so prone to having a reduced sample size anyway.
Overview
Content Tools