Skill assessment of the CSIRO multi-year Climate Analysis Forecast Ensemble (CAFE) system CSIRO decadal climate forecasting project Dougie Squire, James Risbey, Carly Tozer, Thomas Moore, James Munroe
The CSIRO system • New project to understand and improve predictability on multi-year time scales • Use a variant of the GFDL CM2.1 ocean (MOM5) - atmosphere (AM2) - land (LM2) - sea ice (SIS) model • Focus on internal variability 2 Skill assessment of CSIRO’s CAFE system | Dougie Squire
The CSIRO system • New project to understand and improve predictability on multi-year time scales • Use a variant of the GFDL CM2.1 ocean (MOM5) - atmosphere (AM2) - land (LM2) - sea ice (SIS) model See O’Kane C1-03 (tomorrow) • Focus on internal variability 3 Skill assessment of CSIRO’s CAFE system | Dougie Squire
The CSIRO system • New project to understand and improve predictability on multi-year time scales • Use a variant of the GFDL CM2.1 ocean (MOM5) - atmosphere (AM2) - land (LM2) - sea ice (SIS) model • Focus on internal variability 4 Skill assessment of CSIRO’s CAFE system | Dougie Squire
The CSIRO system • New project to understand and improve predictability on multi-year time scales • Use a variant of the GFDL CM2.1 ocean (MOM5) - atmosphere (AM2) - land (LM2) - sea ice (SIS) model • Focus on internal variability Monselesan et al. 2015 GRL Fractional in-band variances of SLA 5 Skill assessment of CSIRO’s CAFE system | Dougie Squire
The CSIRO system • New project to understand and improve predictability on multi-year time scales • Use a variant of the GFDL CM2.1 ocean (MOM5) - atmosphere (AM2) - land (LM2) - sea ice (SIS) model • Focus on internal variability 1-2 years Monselesan et al. 2015 GRL Fractional in-band variances of SLA 6 Skill assessment of CSIRO’s CAFE system | Dougie Squire
The CSIRO system • New project to understand and improve predictability on multi-year time scales • Use a variant of the GFDL CM2.1 ocean (MOM5) - atmosphere (AM2) - land (LM2) - sea ice (SIS) model • Focus on internal variability 5-10 years Monselesan et al. 2015 GRL Fractional in-band variances of SLA 7 Skill assessment of CSIRO’s CAFE system | Dougie Squire
diagnostics/verification software • Leverage emerging efforts towards best practices in big data and reproducibility + James Munroe • Towards a community effort • Dataset/filetype agnostic 8 Skill assessment of CSIRO’s CAFE system | Dougie Squire
diagnostics/verification software • Leverage emerging efforts towards best practices in big data and reproducibility + James Munroe • Towards a community effort • Dataset/filetype agnostic 9 Skill assessment of CSIRO’s CAFE system | Dougie Squire
diagnostics/verification software • Leverage emerging efforts towards best practices in big data and reproducibility + James Munroe • Towards a community effort • Dataset/filetype agnostic 10 Skill assessment of CSIRO’s CAFE system | Dougie Squire
diagnostics/verification software ����������� ������ ���������� 11 Skill assessment of CSIRO’s CAFE system | Dougie Squire
diagnostics/verification software 5 x 32 CPU, 256 Gb VMs ����������� ������ ���������� 12 Skill assessment of CSIRO’s CAFE system | Dougie Squire
diagnostics/verification software 13 Skill assessment of CSIRO’s CAFÉ system | Dougie Squire
CAFE-f1 hindcasts (today’s data) • 2-year, 11-member hindcasts started monthly over 2002-2016 • Only ocean observations assimilated • Bred-vector-initialised on sub-surface ocean temperature isosurface corresponding to high in-band variance on 1-2 month time scales • Mean bias corrected (Stockdale 1997 MWR) 14 Skill assessment of CSIRO’s CAFE system | Dougie Squire
CAFE-f1 hindcasts (today’s data) • 2-year, 11-member hindcasts started monthly over 2002-2016 • Only ocean observations assimilated • Bred-vector-initialised on sub-surface ocean temperature isosurface corresponding to high in-band variance on 1-2 month time scales • Mean bias corrected (Stockdale 1997 MWR) 15 Skill assessment of CSIRO’s CAFE system | Dougie Squire
CAFE-f1 hindcasts (today’s data) • 2-year, 11-member hindcasts started monthly over 2002-2016 • Only ocean observations assimilated • Bred-vector-initialised on sub-surface ocean temperature isosurface corresponding to high in-band variance on 1-2 month time scales • Mean bias corrected (Stockdale 1997 MWR) See O’Kane C1-03 (tomorrow); and O’Kane et al. 2018 JC 16 Skill assessment of CSIRO’s CAFE system | Dougie Squire
CAFE-f1 hindcasts (today’s data) • 2-year, 11-member hindcasts started monthly over 2002-2016 • Only ocean observations assimilated • Bred-vector-initialised on sub-surface ocean temperature isosurface corresponding to high in-band variance on 1-2 month time scales • Mean bias corrected (Stockdale 1997 MWR) See O’Kane C1-03 (tomorrow); and O’Kane et al. 2018 JC 17 Skill assessment of CSIRO’s CAFE system | Dougie Squire
Temporal anomaly correlations of monthly SST 18 Skill assessment of CSIRO’s CAFE system | Dougie Squire
Ranked probability skill score of tropical T 2m 19 Skill assessment of CSIRO’s CAFE system | Dougie Squire
Ranked probability skill score of tropical T 2m Forecast skill is strongly related to ENSO variability 20 Skill assessment of CSIRO’s CAFE system | Dougie Squire
Ranked probability skill score of SE Australian T 2m 21 Skill assessment of CSIRO’s CAFE system | Dougie Squire
Ranked probability skill score of SE Australian T 2m 22 Skill assessment of CSIRO’s CAFE system | Dougie Squire
Ranked probability skill score of thermal wind 23 Skill assessment of CSIRO’s CAFE system | Dougie Squire
Ranked probability skill score of thermal wind Positive correlation with MEI (O’Kane et al. 2017 MWR) 24 Skill assessment of CSIRO’s CAFE system | Dougie Squire
Conclusions • is a diagnostics/verification software package that we are building • Early CAFE system hindcasts indicate comparable skill to other systems • Prediction skill in Australia is strongly tied to the tropical ocean (ENSO) and to the CAFE system’s ability to simulate relevant teleconnection processes Contact: Dougie.Squire@csiro.au 25 Skill assessment of CSIRO’s CAFE system | Dougie Squire
Free-running model diagnostics Longitudinal wave activity flux at 500hPa H500 EOFs Longitudinal WAF First two EOFs 26 Skill assessment of CSIRO’s CAFE system | Dougie Squire
h500 anomaly composites for heavy Tasmanian rainfall 27 Skill assessment of CSIRO’s CAFE system | Dougie Squire
Ensemble spread metrics Talagrand of SE Australian T 2m Goddard et al. ensemble spread metric, T 2m 28 Skill assessment of CSIRO’s CAFE system | Dougie Squire
Ranked probability skill score of tropical T 2m 29 Skill assessment of CSIRO’s CAFE system | Dougie Squire
Ranked probability skill score of tropical T 2m 30 Skill assessment of CSIRO’s CAFE system | Dougie Squire
Uncorrected tropical Pacific SST 31 Skill assessment of CSIRO’s CAFE system | Dougie Squire
Uncorrected tropical Pacific SST 32 Skill assessment of CSIRO’s CAFE system | Dougie Squire
Uncorrected tropical Pacific SST 33 Skill assessment of CSIRO’s CAFE system | Dougie Squire
Uncorrected tropical Pacific SST 34 Skill assessment of CSIRO’s CAFE system | Dougie Squire
Uncorrected tropical Pacific SST 35 Skill assessment of CSIRO’s CAFE system | Dougie Squire
Recommend
More recommend
