Precipitation amounts totaled over several days to weeks is one of the hardest things to forecast for a specific location. I have observed that good forecasts can often be close to the mark for significant rainfall in a general region, but accuracy for a specific location is much harder to achieve. Viewing the ensemble forecast members helps to gain a better appreciation of spatial displacement variation, magnitude (total rain) variation, and pattern variation that often occurs with significant rainfall events. But it is tedious trawling through all of the ensemble members to gain that appreciation. Animations attached show the control member and each of the fifty ensemble members. 5 day, 10 day, and 15 day from the EC EPS, and 46 day from the EC extended. It's much easier to moderate expectations when you have an appreciation of the variation and difficulty in forecasting rain amounts.
IMHO, I can't look at accumulative totals (even day-to-day & ensembles) without first understanding a model's resolve. There's so many variables to skew this detail beyond 48 hours - particularly if it's including convective falls. If I am honest, accumulative precip figures are nothing but pretty colours if you can't first understand the synoptic set-up that puts it there. Match that with another global model and I might buy into it.
I respect Ben and his insightful posts on Twitter, so I am by no means taking anything away from his post below, but this type of information can be interpreted differently depending on your ability (or inability) to read a MSLP plot. i.e. I'd likely to be devastated if I was in Dalby, QLD and didn't see a drop in the first week of Jan, after reading that - but highly possible!
I totally agree with that. It aligns to my observations that precipitation is hard to forecast as an outlook of a weather system setup compared to say forecasting the context of the setup Yet it is also the precipitation outcomes that many people want to know- so if people want precipitation amounts IMO it is better to appreciate ensembles rather than single model runs. I also agree with lead time limitations and difficulty with convective amounts and placement.