Seamless prediction systems prove potential for skillful Arctic - - PowerPoint PPT Presentation

▶
seamless prediction systems prove potential for skillful
SMART_READER_LITE
LIVE PREVIEW

Seamless prediction systems prove potential for skillful Arctic - - PowerPoint PPT Presentation

Jun 10, 2023 208 likes •463 views

Seamless prediction systems prove potential for skillful Arctic sea-ice forecasts far beyond weather time scales IUP Seminar Bremen, May 4 th , 2018 Lorenzo Zampieri Alfred Wegener Institute Outlook Predictability and ensemble forecasting

slide-1
SLIDE 1

Seamless prediction systems prove potential for skillful Arctic sea-ice forecasts far beyond weather time scales

IUP Seminar Bremen, May 4th, 2018 Lorenzo Zampieri Alfred Wegener Institute

slide-2
SLIDE 2

Outlook Predictability and ensemble forecasting Research Motivation The S2S Prediction Project Verification metrics Predictive skills Dynamical models vs Climatology

slide-3
SLIDE 3

Predictability and Ensemble Forecasting

slide-4
SLIDE 4

A simple set of equations… !" !# = −&" + &( !" !# = −") + *" − ( !) !# = "( − +)

Lo Lorenz, E. N. (1963) De Dete termi rministic no nonp nper eriodic fl flow.

slide-5
SLIDE 5

…with an interesting solution

J.

  • J. Sl

Sling ngo an and T.

  • T. Palmer (2011)
slide-6
SLIDE 6

Ensemble Forecasting

J.

  • J. Sl

Sling ngo an and T.

  • T. Palmer (2011)
slide-7
SLIDE 7

Research Motivations

slide-8
SLIDE 8

Im Images from the YOPP PP Pr Promotional Video

slide-9
SLIDE 9

Im Images from the YOPP PP Pr Promotional Video

slide-10
SLIDE 10

S2S

Sub-seasonal to Seasonal Prediction Project

slide-11
SLIDE 11

The S2S timescale

Im Image from the S2S Pr Promotional Video

slide-12
SLIDE 12

The S2S Database

  • Coupled models from operational weather

forecast centers

  • Ensemble forecasts
  • Dynamical sea ice components
  • Assimilated sea surface temperature and

sea-ice concentration

  • Long temporal coverage (25 years)

UK Met Office ECMWF KMA CMA Météo France NCEP

slide-13
SLIDE 13

Verification Metrics

slide-14
SLIDE 14

The sea-ice edge position

Sea-ice edge Observation Sea-ice edge Model

Integrated Ice Edge Error IIEE = " + $

slide-15
SLIDE 15

The Spatial Probability Score Ensemble sea-ice forecasts Probabilistic verification metric required

Spatial Probability Score !"! = $

%

('( − '*), -% Spatial Probability Skill Score !"!! = . − !"! !"!/012

slide-16
SLIDE 16

Ensemble Forecasting

J.

  • J. Sl

Sling ngo an and T.

  • T. Palmer (2011)
slide-17
SLIDE 17

Methods Summary

Ensemble S2S sea-ice forecasts

UKMO ECMWF KMA CMA MF NCEP

Verification against satellite observations through the SPS Forecast SPS compared with the climatological CSPS Assessment of the forecast predictive skills Evaluation of the forecast errors and biases

slide-18
SLIDE 18

Skills of S2S forecast systems

  • ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

Lead Time [Days] 10 20 30 40 50 60 0.25 0.75 1 SPS [106฀km2] 0.5

ECMWF UKMO KMA NCEP CMA (out of range) MF ECMWF Pres. Climatology

  • ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

Lead Time [Days] 10 20 30 40 50 60

  • 0.5

0.5 1 SPSS

Results averaged over 12 years

  • f hindcasts (1999-2010)
slide-19
SLIDE 19

Forecasting the 2007 minimum

NCEP Day 30 - 2007.09.15 MF Day 30 - 2007.09.13 CMA Day 30 - 2007.09.15 ECMWF Pres. Day 30 - 2007.09.16 ECMWF Day 30 - 2007.09.15 KMA Day 30 - 2007.09.15 UKMO Day 30 - 2007.09.15

Sea Ice Probability

15% sic OSI-SAF Climatology 2007.09.16

slide-20
SLIDE 20

The CSPS seasonal cycle

  • Time

Jan 01 Apr 01 Jul 01 Oct 01 Jan 01 0.25 0.5 0.75 1 CSPS [106฀km2]

  • Time

Jan 01 Apr 01 Jul 01 Oct 01 Jan 01 20 40 60 80 100 100 80 60 40 20 ME [%] AEE [%]

  • Time

Jan 01 Apr 01 Jul 01 Oct 01 Jan 01 20 40 60 80 100 100 80 60 40 20 U [%] O [%]

Skills of the climatological forecasts based on the previous 10 years of observations 15/09/2007 forecast is based on: 15/09/1996, 15/09/1997, … , 15/09/2006

slide-21
SLIDE 21

ECMWF & UKMO

−2.0 −1.5 −1.0 −0.5 0.0 0.5 1.0 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20 −2.0 −1.5 −1.0 −0.5 0.0 0.5 1.0 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20 −2.0 −1.5 −1.0 −0.5 0.0 0.5 1.0 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20

Initial

−2.0 −1.5 −1.0 −0.5 0.0 0.5 1.0 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20

SPSS AEE and ME O and U ECMWF UKMO

  • −2.0
−1.5 −1.0 −0.5 0.0 0.5 1.0 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 SPSS
  • Jan 01
Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20 ME [%] AEE [%]
  • Jan 01
Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20 U [%] O [%]
  • ●
  • ●
  • ●
  • ●
  • ●
  • ●
  • −2.0
−1.5 −1.0 −0.5 0.0 0.5 1.0 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 SPSS
  • ●
  • ●
  • ●
  • ●
  • ●
  • ●
  • ●
  • ●
  • ●
  • ●
  • Jan 01
Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20 ME [%] AEE [%]
  • ●
  • ●
  • ●
  • ●
  • ●
  • ●
  • ●
  • ●
  • ●
  • ●
  • Jan 01
Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20 U [%] O [%] −2.0 −1.5 −1.0 −0.5 0.0 0.5 1.0 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20 −2.0 −1.5 −1.0 −0.5 0.0 0.5 1.0 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20
slide-22
SLIDE 22

CMA & Météo France

ECMWF Pres.

100

SPSS AEE and ME O and U CMA MF

  • Jan 01
Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20 U [%] O [%] −2.0 −1.5 −1.0 −0.5 0.0 0.5 1.0
  • −2.0
−1.5 −1.0 −0.5 0.0 0.5 1.0 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 SPSS
  • Jan 01
Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20 ME [%] AEE [%]
  • −6
−5 −4 −3 −2 −1 1 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 SPSS
  • Jan 01
Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20 U [%] O [%] 100
  • Jan 01
Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20 ME [%] AEE [%] −2.0 −1.5 −1.0 −0.5 0.0 0.5 1.0 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20 −2.0 −1.5 −1.0 −0.5 0.0 0.5 1.0 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20 −2.0 −1.5 −1.0 −0.5 0.0 0.5 1.0 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20

Initial

−2.0 −1.5 −1.0 −0.5 0.0 0.5 1.0 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20
slide-23
SLIDE 23

Improvements in ECMWF system

−2.0 −1.5 −1.0 −0.5 0.0 0.5 1.0 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20 −2.0 −1.5 −1.0 −0.5 0.0 0.5 1.0 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20 −2.0 −1.5 −1.0 −0.5 0.0 0.5 1.0 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20

Initial

−2.0 −1.5 −1.0 −0.5 0.0 0.5 1.0 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20

SPSS AEE and ME O and U ECMWF

  • −2.0
−1.5 −1.0 −0.5 0.0 0.5 1.0 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 SPSS
  • Jan 01
Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20 ME [%] AEE [%]
  • Jan 01
Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20 U [%] O [%] −2.0 −1.5 −1.0 −0.5 0.0 0.5 1.0 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20 −2.0 −1.5 −1.0 −0.5 0.0 0.5 1.0 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20 −2.0 −1.5 −1.0 −0.5 0.0 0.5 1.0 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20

ECMWF Pres.

  • ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●
  • Jan 01
Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20 ME [%] AEE [%]

Target Time

  • ● ●
  • ● ● ● ● ●
  • ● ●
  • ●
  • ●
  • ● ● ●
  • ●
  • −2.0
−1.5 −1.0 −0.5 0.0 0.5 1.0 Jan 01 Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 SPSS −2.0 −1.5 −1.0 −0.5 0.0 0.5 1.0 −6 −5 −4 −3 −2 −1
  • ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●
  • Jan 01
Mar 01 May 01 Jul 01 Sep 01 Nov 01 Jan 01 20 40 60 80 100 100 80 60 40 20 U [%] O [%]

N E W ! O L D !

slide-24
SLIDE 24

Improvements in ECMWF system

slide-25
SLIDE 25

Conclusions

  • Spread of predictive skills between different

forecast systems

  • Errors during the data assimilation and

relevant model biases affect the forecasts

  • Lack of model tuning
  • The ECMWF system shows encouraging
  • results. Predictive skills after 46 days.
  • Assimilation of new sea-ice thickness
  • bservations could be beneficial