In March 2016 the horizontal resolution of the IFS was increased, to about 9km for the high resolution deterministic forecast (HRES), and to about 18km for the ensemble (ENS). Also the ensemble now runs at the same resolution to 15 days; previously there had been a step change to lower resolution at day 10. Of themselves these changes bring some noteworthy benefits: for example better representation of topography and topographic effects such as orographic rainfall, better representation of coastal and island features, and no step changes in site specific forecasts at day 10. The resolution of the assimilation cycles has also been improved, allowing a better use of observations. Together with changes to the model dynamics and physics, this upgrade has resulted in a significant increase in the forecast accuracy throughout the medium range. Modifications to radiation code now help provide better forecast of 2m temperature near to skin temperature discontinuities. Meanwhile the sunshine duration diagnostic has also been dramatically improved. Unrealistic extreme rainfall events sometimes generated near certain mountains have been eradicated by changes to the numerics, and the representation of tropical cyclone structure has also been improved.
A new cycle due to be introduced in late 2016 will bring further improvements. For example a dynamic sea ice model will be introduced and the resolution of the ocean model component used in the ENS will be increased. A slant path representation will be introduced for radiative transfer calculations for satellite data assimilation, there will be better, more localised SST perturbations in the ENS, and cloud physics changes will improve the representation of boundary layer clouds in certain parts of the world. We will also start to assimilate NEXRAD snowfall data.
All the key model changes will be described and illustrated during this presentation, showing evidence of forecast improvements from case studies (commonly relating to severe weather events) and from more systematic verification of parameters representing both broadscale patterns (such as 500hPa height) and surface weather. Also included will be illustrations of some of the changes ECMWF is making to model output fields, and products for forecasters. Insights will also be given into ECMWF’s new 10-year strategy, published in 2016, together with an outline of forthcoming model cycles planned for 2017.