Sub-Seasonal Prediction Activities and Plans in CAWCR Debbie Hudson, Andrew Marshall, Oscar Alves, Harry Hendon, Matthew Wheeler, Yonghong Yin, Guomin Wang The Centre for.
Download ReportTranscript Sub-Seasonal Prediction Activities and Plans in CAWCR Debbie Hudson, Andrew Marshall, Oscar Alves, Harry Hendon, Matthew Wheeler, Yonghong Yin, Guomin Wang The Centre for.
Sub-Seasonal Prediction Activities and Plans in CAWCR Debbie Hudson, Andrew Marshall, Oscar Alves, Harry Hendon, Matthew Wheeler, Yonghong Yin, Guomin Wang The Centre for Australian Weather and Climate Research A partnership between CSIRO and the Bureau of Meteorology POAMA: Predictive Ocean Atmosphere Model for Australia POAMA v1.5, v2 Originally designed for seasonal prediction Atmos and Land Initial Condition Atmospheric Model BAM T47L17 3h OASIS Coupler • POAMA-1 operational 2002 • POAMA-1.5 operational 2007 (10 member ensemble) • POAMA-2 operational 2011 (33 member ensemble) Ocean Initial Condition Ocean Model ACOM2 (lat 0.5~1.5º; lon 2º; 25 lvls) Has separate sub-seasonal and seasonal systems • POAMA-3 ACCESS-based development version (UM atmospheric model; ocean model based on GFDL MOM4) POAMA-2 Systems Seasonal System Sub-seasonal System Atmosphere/land data assimilation ALI (Atmosphere Land Initialisation Scheme: nudging atmos model to ERA-40 or operational NWP) ALI (Atmosphere Land Initialisation Scheme: nudging atmos model to ERA-40 or operational NWP) Ocean data assimilation PEODAS (Multivariate pseudoEnsemble Kalman Filter) PEODAS (Multivariate pseudoEnsemble Kalman Filter) 30 members 33 members Multi-model (3 versions) Multi-model (3 versions) Ocean perturbations from PEODAS; No atmosphere perturbations Ocean and atmosphere perturbations from Coupled Ensemble Initialisation Scheme (CEIS) 30 members on the 1st of the month out to 9 months (19602010) 33 members on the 1st, 11th and 21st of the month out to 120 days (1989-2010) 30 members on the 1st and 15th of the month out to 9 months 33 members every 0z Thursday out to 120 days Ensemble generation Hindcast Operational Activities in sub-seasonal • • • • Documenting the skill of the current POAMA system Understanding predictability of Australian climate on subseasonal timescales Producing experimental products Development of POAMA-3 Two externally funded sub-seasonal R&D projects: a) Sub-seasonal prediction with POAMA System development; system skill; impact of initialisation strategy; sources of predictability (MJO, SAM, blocking); experimental products (end 2012) b) Northern Australia/Monsoon Prediction Development and delivery of climate products for northern Australia, especially to characterise aspects of the monsoon and northern wet season (end 2013) Why do the seasonal and sub-seasonal systems differ in their ensemble generation strategy? Ensemble spread is far too small in the POAMA-2 seasonal system in the first month of the forecast Ensemble Spread (stddev): 500hPa heights (Jul) Day 10 of the forecast POAMA-1.5 Ensemble from lagged atmospheric initial conditions POAMA-2 (seasonal system) Ensemble from ocean perturbations only Initial Conditions for the Sub-Seasonal System Towards Coupled Assimilation... Based on the PEODAS infrastructure Coupled Ensemble Initialisation System Coupled Model integrations Central unperturbed analyses: PEODAS and ALI Bred vectors are rescaled and centred to the central analyses 1 day Generates coupled perturbations of the atmosphere and ocean based on a breeding method u10 anom; SHEM 20º- 60ºS; JAN/JUL/AUG POAMA-1.5 POAMA-2 (seas) RMSE Ensemble spread POAMA-1.5 POAMA-2 (seas) POAMA-2 (intra) Correlation POAMA-2 (intra) Australian RAINFALL above upper tercile: all forecast start months Fortnight 1 POAMA-1.5 POAMA-2 (seas) POAMA-2 (intraseas) Fortnight 2 POAMA has good skill in predicting rainfall and TMAX over eastern Australia in the second fortnight of the forecast, particularly during spring forecast months. Precipitation TMAX ROC area of the probability that rainfall (left) and TMAX (right) for the 2nd fortnight of the forecast is in the upper tercile for spring forecast months (SON, 1989-2006). ROC areas significantly more skilful than climatology are shaded (5% significance level). Evaluating key drivers of Australian intraseasonal climate variability MJO Blocking SAM (Risbey et al. 2009) Climate drivers operating on timescales longer than intraseasonal influence prediction skill For rainfall forecast in the 2nd fortnight, there is higher skill when the IOD is strong and when ENSO is in an extreme (JJASON) El Nino / La Nina (n=48) IOD large (n=30) ENSO in neutral state Neutral (n=60) IOD small (n=30) ENSO in neutral state Rainfall correlation skill (Hudson et al 2011) Research Approach: MJO, SAM, and blocking Simulation of rainfall/temperature anomalies associated with each climate driver Prediction of each climate driver (~weeks) Prediction of rainfall/temperature anomalies associated with each climate driver (intra-seasonal) MJO – simulation of rainfall (Marshall et al 2011) MJO – Prediction of Index Wheeler and Hendon (2004) RMM Index RMSE & correlation between observed and POAMA RMM indices (over all start months) RMSE COR Climatological forecast skill POAMA-2 skill exceeds POAMA-1.5 (Rashid et al 2010, Marshall et al 2011) MJO – prediction of rainfall in weeks 3-4 (Nov-Apr) ROC area: rainfall in the upper tercile (POAMA-2) Real-time products on POAMA web http://poama.bom.gov.au Operational configuration: • 33 member ensemble • Updated once per week • 120 day forecast Also include skill MJO Example Products for hindcasts also available Plans (2012) • • • • • Continue research in understanding predictability of Australian climate on sub-seasonal timescales Extend sub-seasonal hindcast set back to 1980 Produce experimental products Re-contribute POAMA-2 MJO forecasts to MJO task force Development of POAMA-3 ACCESS model: UM atmos (N96L38 – N216L80) and MOM4 ocean (0.25º tropics) Coupled breeding + coupled assimilation Maybe stochastic physics 2012 two additional externally funded R&D projects: a) Predictions of heat extremes over Australia Large-scale climate drivers/processes; prediction skill (drivers and heat event); potentially skilful products b) Un-coupled predictions with ACCESS d0-45; N96L80 atmosphere-only model; sensitivity experiments to explore optimal model configurations (e.g. horizontal and stratospheric resolution)