FOCRAII-9, Beijing 10 th April 2013 ECMWF ECMWF long range
forecast systems Dr. Tim Stockdale European Centre for Medium-Range
Weather Forecasts Slide 2 FOCRAII-9, Beijing 10 th April 2013 ECMWF
Outline Overview of System 4 Some recent research results EUROSIP
multi-model forecasts Forecasts for JJA 2013 Slide 3 FOCRAII-9,
Beijing 10 th April 2013 ECMWF System 4 seasonal forecast model IFS
(atmosphere) T L 255L91 Cy36r4, 0.7 deg grid for physics
(operational in Dec 2010) Full stratosphere, enhanced stratospheric
physics Singular vectors from EPS system to perturb atmosphere
initial conditions Ocean currents coupled to atmosphere boundary
layer calculations NEMO (ocean) Global ocean model, 1x1 resolution,
0.3 meridional near equator NEMOVAR (3D-Var) analyses, newly
developed. Coupling Fully coupled, no flux adjustments Sea-ice
based on sampling previous five years Slide 4 FOCRAII-9, Beijing 10
th April 2013 ECMWF Reduced mean state errors S4 S3 T850U50 Slide 5
FOCRAII-9, Beijing 10 th April 2013 ECMWF Tropospheric scores
Spatially averaged grid-point temporal ACC One month leadFour month
lead Slide 6 FOCRAII-9, Beijing 10 th April 2013 ECMWF S4 extended
hindcast set Scores are smoother and systematically higher with 51
member hindcasts Slide 7 FOCRAII-9, Beijing 10 th April 2013 ECMWF
S4 extended hindcast set Gain over S3 is now stronger and more
robust Slide 8 FOCRAII-9, Beijing 10 th April 2013 ECMWF More
recent ENSO forecasts are better.... 1981-19951996-2010 Slide 9
FOCRAII-9, Beijing 10 th April 2013 ECMWF QBO 50hPa 30hPa System 3
System 4 Slide 10 FOCRAII-9, Beijing 10 th April 2013 ECMWF
Problematic ozone analyses Slide 11 FOCRAII-9, Beijing 10 th April
2013 ECMWF Land surface Snow depth limits, 1 st April Slide 12
FOCRAII-9, Beijing 10 th April 2013 ECMWF Sea ice Slide 13
FOCRAII-9, Beijing 10 th April 2013 ECMWF Tropical storm forecasts
Slide 14 FOCRAII-9, Beijing 10 th April 2013 ECMWF Recent Research
Slide 15 FOCRAII-9, Beijing 10 th April 2013 ECMWF QBO Period and
downward penetration match observations Semi-annual oscillation
still poorly represented A big reduction in vertical diffusion, and
a further tuning of non-orographic GWD, has given a big additional
improvement in the QBO compared to S4. Slide 16 FOCRAII-9, Beijing
10 th April 2013 ECMWF QBO forecasts S3 S4 New Slide 17 FOCRAII-9,
Beijing 10 th April 2013 ECMWF NH winter forecasts 0.319 0.371
Slide 18 FOCRAII-9, Beijing 10 th April 2013 ECMWF NH winter
forecasts Even with 101 members, ensemble mean signal not always
well defined Slide 19 FOCRAII-9, Beijing 10 th April 2013 ECMWF NH
winter forecasts New version has weaker signal, more noise Slide 20
FOCRAII-9, Beijing 10 th April 2013 ECMWF NH winter forecasts
Forecast skill is above perfect model predictability limit Slide 21
FOCRAII-9, Beijing 10 th April 2013 ECMWF EUROSIP A European
multi-model seasonal forecast system Operational since 2005 Data
archive and real-time forecast products Initial partners: ECMWF,
Met Office, Mto-France NCEP an Associate Partner; forecasts
included since 2012 Products released at 12Z on the 15 th of each
month Aim is a high quality operational system Data policy issues
are always a factor in Europe Slide 22 FOCRAII-9, Beijing 10 th
April 2013 ECMWF Recent changes: variance scaling Robust
implementation Limit to maximum scaling (1.4) Weakened upscaling
for very large anomalies Improves every individual model Improves
consistency between models Improves accuracy of multi-model
ensemble mean Slide 23 FOCRAII-9, Beijing 10 th April 2013 ECMWF
Revised Nino plumes Slide 24 FOCRAII-9, Beijing 10 th April 2013
ECMWF Error vs spread (uncalibrated) Slide 25 FOCRAII-9, Beijing 10
th April 2013 ECMWF Calibrated p.d.f. ENSO forecasts have good past
performance data We can calibrate forecast spread based on past
performance We can also allow varying weights for models We have to
be very careful not to overfit data at any point. Represent
forecast with a p.d.f. This is the natural output of our
calibration procedure Easier visual interpretation by user
Calibration and combination in general case Ideally apply similar
techniques to all forecast values (T2m maps etc) More difficult
because less information on past (higher noise levels) Hope to get
there eventually ... Slide 26 FOCRAII-9, Beijing 10 th April 2013
ECMWF Nino 3.4 plume and p.d.f. Slide 27 FOCRAII-9, Beijing 10 th
April 2013 ECMWF P.d.f. interpretation P.d.f. based on past errors
The risk of a real-time forecast having a new category of error is
not accounted for. E.g. Tambora volcanic eruption. We plot 2% and
98%ile. Would not go beyond this in tails. Risk of change in bias
in real-time forecast relative to re-forecast. Bayesian p.d.f.
Explicitly models uncertainty coming from errors in forecasting
system Two different systems will calculate different pdfs both are
correct Validation Rank histograms show pdfs are remarkably
accurate (cross-validated) Verifying different periods shows
relative bias of different periods can distort pdf sampling issue
in our validation data. Slide 28 FOCRAII-9, Beijing 10 th April
2013 ECMWF Forecasts for JJA 2013 Slide 29 FOCRAII-9, Beijing 10 th
April 2013 ECMWF ECMWF forecast: ENSO Past performance Slide 30
FOCRAII-9, Beijing 10 th April 2013 ECMWF EUROSIP forecast: ENSO
Past performance Slide 31 FOCRAII-9, Beijing 10 th April 2013 ECMWF
ECMWF forecast: JJA 2mT Tercile probabilities ACC skill (1981-2010)
Slide 32 FOCRAII-9, Beijing 10 th April 2013 ECMWF ECMWF forecast:
JJA precip Tercile probabilities ACC skill (1981-2010)
