Lauren Stevens and Matt Chamberlain Outline Setup of the experiment Potential predictability of ENSO Potential predictability of airsea CO2 fluxes and Net Primary Productivity Future Work 1 ID: 801370
Download The PPT/PDF document "Potential Predictability of Tropical Pac..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
Potential Predictability of Tropical Pacific Ocean
Lauren Stevens and Matt ChamberlainOutline:Set-up of the experimentPotential predictability of ENSOPotential predictability of air-sea CO2 fluxes and Net Primary ProductivityFuture Work
1 |
1
Slide2Coupled Model – MOM4 (
ACCESSo
grid and SIS) with AM2 (v1 control with 500 years of simulations)
Forecasts from January 1
st
going for 6 years with 10 ensemble members
First set is from years 315 – 319 (5 years)Small random SST perturbation in the tropicsSecond set is from years 305, 315, 320, 335, 350, 365,…,440 (11 years)Small random SST perturbation at one point in the tropics
Forecasts
2
Slide3NINO3.4 Variability
3
15 January forecasts
Example Forecast
Slide4NINO3.4 Variability
4
15 January forecasts
Example Forecast
Slide5Year 305 – Red Observed; Black and grey forecast
Potential Predictability: NINO3.4
5
11 January forecasts
Example Forecast
Slide6Potential Predictability: NINO3.4
6
11 January forecasts
Slide7Potential Predictability: Tropical Pacific
7
Seferian
et al., 2013, PNAS
1 year lead
2-5 year lead
Slide8Potential Predictability: CO2 Flux
8
12-month lead: Correlation Coefficient
forecasts
Persistence
Slide9Potential Predictability: CO2 Flux
9
24-month lead: Correlation Coefficient
forecasts
Persistence
Slide10Potential Predictability: Net Primary Productivity
10
12-month lead: Correlation Coefficient
forecasts
Persistence
Slide11Potential Predictability: Net Primary Productivity
11
24-month lead: Correlation Coefficient
forecasts
Persistence
Slide12Future work
Potential predictability experiments provide a convenient way to assess different forecasting strategiesEnsemble generation methodsImpact of different climate regimesNovel forecasting productsothers
12
12
Slide13Potential Predictability: NINO3.4
13
11 January forecasts
Slide14Potential Predictability: NINO3.4
14
11 January forecasts
Slide15Potential Predictability: NINO3.4
15
11 January forecasts
Slide16Potential Predictability: NINO3.4
16
11 January forecasts
Slide17Potential Predictability: Soil Moisture
17
15 January forecasts
20-month lead: Correlation Coefficient
forecasts
Persistence
South Australia: (south of 33°S)
Anomaly Correlation Coefficient
Lead Time (months)
Blue- forecast
Orange - persistence
Slide18Potential Predictability: NINO3.4
18
15 January forecasts
Example Forecast
Slide1919
ENSO
–
Anomaly Correlation
Zheng 2010
v0
v1
Slide2020
Nino4 Discrimination Plot:
v0
v1
3 months
6 months
9 months
Assess the ability to forecast an event (El Nino or La Nina) and no event
Not able to reliably forecast
an event
Slide2121
Nino4 Discrimination Plot:
v1
v1
3 months
6 months
9 months
Not enough events
In the forecast
Slide22CAFE System
Climate Model – MOM5 (SISE,WOMBAT) with AM2 ->MOM5 with AM3 and ACCESS-ESM 1Data Assimilation – EnOI with ocean observations ->EnKF (96 members) with ocean, sea ice and atmospheric dataEnsemble generation with Bred Vectors -> several different sets of BVs targeting different time-scales
Forecasts – new dataset of monthly forecasts to follow the couple DA development (6 months)
22
22
Slide23https://jobs.csiro.au
Postdoctoral Fellowship Sea Ice Modeller (56719)Postdoctoral Fellowship - Climate regimes (56508)Postdoctoral Fellowship - Ocean-Atmosphere dynamics (56622)CSIRO Postdoctoral Fellowship – Atmospheric DynamicsCSIRO Postdoctoral Fellowship - Ocean-Atmosphere Carbon Fluxes (49441)
Post Doctorial Positions
23 |
23
Slide2424
ROC:
4 months
6 months
9 months
v0
v1
Slide25Potential Predictability: Rainfall
25
15 January forecasts
South Australia (south of 33°S):
Anomaly Correlation Coefficient
Blue- forecast
Orange - persistence
Lead Time (months)
20-month lead: Correlation Coefficient
forecasts
Persistence
Slide26Forecasts:
DA vs Reanalysis
–
for initial state
26
January forecasts (2005
to 2016)
Slide27Decadal Climate Forecasting Project
Initial GoalsBuild the Climate Analysis Forecast Ensemble (CAFE) system and deliver multi-year to decadal climate forecasts (probabilistic problem and we will provide ensemble forecasts) Apply diagnostics tools, including ensemble verification metrics, to accurately assess the skill of the forecastsAdvance fundamental research into: where does the predictability of the climate system resides, the processes that give rise to that predictability, and the key observations that help us to realise the potential climate predictability Explore the utility of our climate forecasts for a select group of external clients (e.g. Digiscape) 27 |
27
Slide28Data Assimilation, Climate Modelling and Ensemble Generation
Develop and run a coupled ocean-atmosphere-sea ice climate model data assimilation scheme to incorporate observations into the climate model to characterise the climate stateEnsemble climate forecasting system initiated from the climate stateThis is the core of the Climate Analysis Forecast Ensemble (CAFE) system
28
Slide29Processes and Observations
Climate Processes that drive potential predictabilityPredictability StudiesObserving System Experiments and Observing System Simulation Experiments New observation for data assimilation (e.g. sea ice, ocean colour) and assessment of their impact on the climate forecasts
29
Slide30Application and Verification
• need process-based skill assessment• understand mechanisms underlying forecasts• outline deficient process representations in model • provide narrative for forecast use• document skill in public archives and over time• no magic Strong overlap with all components of CAFE System
30
Slide31Forecast Dataset - 2002 -2016
every month of length 2 years with 11 ensemble membersevery 6 months of length 6 years with 11 ensemble membersTo apply a forecast need to understand what the forecast is need to know how to use it need to evaluate how good it isneed to understand its limitationsneed to support the close collaboration between the generation of the forecast and the users
31