How well can we predict the Indian Ocean Dipole and its teleconnections? www.cawcr.gov.au Sally Langford, Li Shi, Harry Hendon and Oscar Alves. April 2011 The Centre for Australian Weather and Climate Research A partnership between CSIRO and the Bureau of Meteorology
Indian Ocean Dipole • Why look at IOD? • Most focus has been on ENSO and its impacts, it is not the whole story for regional climate forecasting - the IOD has an important role. • The IOD accounts for some of the El Nino response in South East Australia. • How well can we predict the IOD? • Conflicting reports in the literature. • Assess operational international dynamical seasonal prediction models in a consistent manner. • NINO and IOD indices from POAMA 1.5 and 2.4, NCEP CFS1 and 2, ECMWF and SINTEX-F. • How well can we predict the IOD teleconnections? • Hasn’t received much focus, but is important for regional climate forecasting. • POAMA and EU ENSEMBLES project - full fields available. The Centre for Australian Weather and Climate Research A partnership between CSIRO and the Bureau of Meteorology
Comparison of models Ocean Ensemble Fcast Atmos. Model Atmos. Initial Scheme Ocean Initial Scheme Climatology Model Members Mnths MOM2 ALI OI technique (sea temp. only; SST POAMA1.5b BAM3.0d 1982-2006 (2.0° × 0.5°; (ERA-40 nudged) relaxed to Obs. with a time scale of 3 10 9 (P15b) T47L17 1980-2005 L25) Hudson et al. (2010) days; see Smith et al. 1991); Alves et al. 2003 BAM3.1 T47L17 MOM2 PEODAS (OASIS coupler with flux POAMA Multi- ALI 1982-2006 (2.0° × 0.5°; (P24a, b) correction P24b or no flux correction 30 9 Model Ensemble (ERA-40 nudged) 1980-2005 BAM3.0d (P24c) L25) P24a & P24c) (PMTMD) OI technique (sea temp., salinity and HOPE ERA-40 direct & ECMWF Sys3 sea level anomalies; SST relaxed to 1.4° × 0.3- T159L62 singular vectors 11 7 1982-2006 (ECMWF) Obs. with a time scale of 3 days; see 1.4°; L29 perturbation Molteni et al. 2007 Balmaseda et al. 2006) SINTEX-F 2.0° × 0.5- No ocean assimilation except SST (SINTEX) T106L19 Forced by nudged SST 9 12 1982-2006 2.0° ; L31 nudged to observations Luo et al. 2005 GODAS (3D variation; Sea Temp. MOM3 only; SST relaxed to Obs. with a time NCEP CFSv1 NCEP Reanalysis-2 (1.0° × 1/3- T62L64 scale of 5 days); SST relaxed to Clim. 8 9 1982-2006 (CFSV1) direct 1.0°; L40) Synthetic salinity created by T-S Saha et al. 2006 relationship NCEP Reanalysis-2 MOM4 Same as above except the ocean NCEP CFSv2 direct, but the Atmos. (1.0° × 1/4- T126L64 model of GODAS is upgraded from 8 9 1982-2006 (CFSV2) Model is upgraded from 1/2°; L40) MOM3 to MOM4 Saha et al. 2011 T62L28 to T574L64 HOPE ERA-40 direct & Wind stress perturbations to generate ECMWF Sys3 1.4° × 0.3- T159L62 singular vectors ensemble of ocean reanalysis, SST 9 7 1980-2005 ENSEMBLES 1.4°; L29 perturbation perturbations at initial time. ERA-40 anomaly, Wind stress perturbations to generate HadGEM2 HadAM32. HadOM(1. assimilation for soil ensemble of ocean reanalysis, SST 9 7 (UK Met Office) 1980-2005 5x3.75 o T38 0°L40) ENSEMBLES moisture perturbations at initial time. The Centre for Australian Weather and Climate Research A partnership between CSIRO and the Bureau of Meteorology
Predicting SST indices Lead time Persistence The Centre for Australian Weather and Climate Research Li Shi, CAWCR A partnership between CSIRO and the Bureau of Meteorology
Predicting ENSO and IOD teleconnections Standard deviation of IOD Compare the dynamical models to a statistical model based on the persistence of the IOD and influence of ENSO. IOD p (t+ τ ) = B1 × IOD o (t) + B2 × NINO3 o (t) Statistical model Start date (months after September 1st) Persistence The Centre for Australian Weather and Climate Research Li Shi, CAWCR A partnership between CSIRO and the Bureau of Meteorology
Predicting IOD and ENSO relationship Obs Start date (months after September 1st) The Centre for Australian Weather and Climate Research Li Shi, CAWCR A partnership between CSIRO and the Bureau of Meteorology
Predicting ENSO and IOD teleconnections Include the model’s ability to predict ENSO. Statistical-dynamical model - IOD p (t+ τ ) = B1 × IOD o (t) + B2 × NINO3 p (t+ τ ) Statistical model tends towards persistence predictability of IOD at longer lead times. Statistical-dynamical model tends towards model predictability of IOD at longer lead times. Start date (months after September 1st) The Centre for Australian Weather and Climate Research Li Shi, CAWCR A partnership between CSIRO and the Bureau of Meteorology
Predicting ENSO and IOD teleconnections Skill score - correlation coefficient of IOD in dynamical models, compared to the statistical-dynamical model. Models asymptote to zero skill score at long lead time - this indicates that there is no predictability of the IOD beyond ~4 months, except from the ability to predict the ENSO and its teleconnections. The Centre for Australian Weather and Climate Research Li Shi, CAWCR A partnership between CSIRO and the Bureau of Meteorology
IOD teleconnections to Australian climate JJA lead time 1 month. IOD IOD|NINO3 NINO3|IOD NINO3 Observations ECMWF UK Met Office P15b PMTMD The Centre for Australian Weather and Climate Research A partnership between CSIRO and the Bureau of Meteorology
IOD teleconnections Regression of DMI standardised anomaly onto MSLP, 500hPa and 200hPa height anomalies - NCEP Reanalysis ECMWF MSLP 500hPa 200hPa Contour interval is 0.3 mbar, 4m and 8m per standard dev of DMI. The Centre for Australian Weather and Climate Research A partnership between CSIRO and the Bureau of Meteorology
Comparison of models Regression of DMI standardised anomaly onto MSLP, 500hPa and 200hPa height anomalies - ECMWF POAMA PMTMD UK Met Office Contour interval is 0.3 mbar, 4m and 8m per standard dev of DMI. The Centre for Australian Weather and Climate Research A partnership between CSIRO and the Bureau of Meteorology
Predicting Australian rainfall Brier Skill Score, JJA lead time 1 month - above median seasonal rainfall forecast The Centre for Australian Weather and Climate Research A partnership between CSIRO and the Bureau of Meteorology
Summary • Dynamical models show skilful forecasts of IOD are limited to 3-4 months lead time in SON. • There is scope for improvement in models by improving ENSO, IOD teleconnections. • Teleconnections are an outstanding problem in dynamical models. • Predicting convection associated with dipole - Rossby wave source. • Biases in the mean state through which the Rossby waves are propagating. The Centre for Australian Weather and Climate Research A partnership between CSIRO and the Bureau of Meteorology
The Centre for Australian Weather and Climate Research A partnership between CSIRO and the Bureau of Meteorology Sally Langford Email: S.Langford@bom.gov.au Web: www.cawcr.gov.au Thank you www.cawcr.gov.au
Detrended data – CFS1 and 2 improve. The Centre for Australian Weather and Climate Research A partnership between CSIRO and the Bureau of Meteorology
Recommend
More recommend
