1. Polar Climate Prediction
Seasonal-to-Decadal Prediction at the Climate Forecasting Unit (Catalan Institute of Climate Sciences)
F.J. Doblas-Reyes
Seasonal
Forecasting
Decadal
Climate
Services
Forecast
System
Development
Research Groups
Research Lines
Data Assimilation
Polar Climate Prediction

2. Focus: seasonal-to-decadal prediction
Francisco J Doblas-Reyes
Isabel Andreu-Burillo: ocean processes
Muhammad Asif: EC-Earth
Lauriane Batté (V): stochastic parameterizations
Pierre-Antoine Bretonnière: SPECS data
Louis-Philippe Caron (V): tropical cyclones
Alberto Carrassi: data assimilation
Melanie Davis: climate services
Neven Fuckar: Arctic sea ice
Virginie Guémas: sea ice, XXIst century hiatus
Fabian Lienert: North Pacific, climate services
Domingo Manubens : autosubmit developer
Oriol Mula-Valls: system administrator
Aida Pintó: prediction of extremes
Mar Rodríguez: SPECS manager
Luis Rodrigues: seasonal climate predictability
Gabriela Tarabanoff: secretary, climate services
Verónica Torralba: climate services
Danila Volpi: initialisation, decadal prediction
Robin Weber: initialisation in simple models
Objectives:
Development of s2d prediction capability
Forecast quality assessment
Downscaling of probabilistic forecasts
Climate services
We share on request:
Autosubmit
Sea-ice restarts
R diagnostic functions
We run on:
Marenostrum (Spain)
ECMWF
Lindgren (Sweden)
HECTOR (UK)
Our local cluster

3. Autosubmit
Autosubmit acts as a wrapper to run a climate experiment on a HPC. The experiment is a sequence of jobs that it submits, manages and monitors. When a job is complete, the next one can be executed.
Divided in 3 phases: ExpID assign, experiment creation, run.
Separation experiment/autosubmit codes.
Config files for autosubmit and experiment.
Database to store experiment information.
Common templates for all platforms.
Recovery after crashes.
Dealing with a list of schedulers and communication protocols.
Each job has a colour in the monitoring tool: yellow=completed, green=running, blue=pending, etc.

4. Initial conditions: sea-ice reconstructions
Sea ice simulation constrained by ocean and atmosphere observational data
Arctic sea ice area
March and September reconstruction
Root Mean Square Error of predictions
Observation 1
Sensitivity experiment from sea-ice reconstruction
CMIP5 predictions
Reconstruction 1
Reconstruction 2
Observation 2
Sea ice reanalysis
Guemas et al. (2014)

5. Seasonal prediction: Increase in resolution
Correlation of the ensemble mean and ratio RMSE/spread of Niño3.4 SST (versus ERSST) from four-month EC-Earth3 simulations: m019 (T255/ORCA1), m01n (T511/ORCA025). May start dates over using ERA-Interim and GLORYS initial conditions and five-member ensembles. ECMWF S4 added for reference.
M. Asif, I. Andreu-Burillo (IC3)

6. Init RMSSS of ensemble mean Ratio RMSE Init/NoInit
Impact of initialisation in CMIP5
(Top row) Root mean square skill score (RMSSS) of the ensemble mean of the initialised predictions and (bottom row) ratio of the root mean square error (RMSE) of the initialised and uninitialised predictions for the near- surface temperature from the multi-model CMIP5 experiment ( ) for (left) 2-5 and (right) 6-9 forecast years. Five-year start date interval.
Forecast year 2-5
Forecast year 6-9
Init RMSSS of ensemble mean
Ratio RMSE Init/NoInit
Doblas-Reyes et al. (2013)

7. Attribution of the XXIst century hiatus
Predictions of the recent global-temperature slow down with EC-Earth 2.3. Global-mean SST from observations (ERSST) and simulations, three-year averages. The experiments suggest an important role of the internal variability, especially increased capture of heat in the ocean, in the hiatus.
Forecast years 1 to 3 from initialised climate predictions
Historical simulations starting from 1850 preindustrial control simulations
Observations
Guemas et al. (2013)

8. Attribution of the XXIst century hiatus
Predictions of the recent global-temperature slow down with EC-Earth 2.3. OHC for the top 800 m (109 J, excluding the mixed layer) for the ORAS4 reanalysis and the initialised hindcasts, three-year average at the onset of the hiatus. The hiatus is associated with an increased ocean heat absortion, especially in the Pacific. This is captured by the Init experiment.
ORAS4
Init

9. Attribution of the XXIst century hiatus
Predictions of the recent global-temperature slow down with EC-Earth 2.3. OHC for the top 800 m (109 J, excluding the mixed layer) for the ORAS4 reanalysis and the initialised hindcasts, three-year average at the onset of the hiatus. The hiatus is associated with an increased ocean heat absortion, especially in the Pacific. This is captured by the Init experiment.
ORAS4
NoInit

10. AI/FFI in a simple model
Mean error of two variables from 360 decadal predictions performed with the Lorenz model with three compartments (ocean, tropical atmosphere and extratropical atmosphere). The configurations where AI outperforms FFI are associated with a strong initial shock and a larger bias.
Carrassi et al. (2014)

11. Calibration and combination
(Left) Multi-model seasonal predictions of Sahel precipitation, including its intraseasonal variability from June to October, started in April. (Right) Correlation of the ensemble mean prediction for Guinean and Sahel precipitation. Reliability is fundamental for climate services.
Rodrigues et al. (2014)

12. Climate services: renewable energy
Climate services: renewable energy
Some initial climate analyses relevant to the RE sector have been made by the climate scientists at IC3.
Here I will show the results for temperature and then downward solar radiation.
These top 2 plots show observed and forecast temperature variation from the mean across the Mediterranean for Although the forecast represents temperatures slightly different to the observed, the climate model has accurately forecast the overall trend which shows a warmer eastern mediterranean compared to the west.
The probability plot is an example of how climate services within the CLIMRUN project can present climate information in a format that can help stakeholders to incorporate it within their decision making processes. Plot C shows that most areas across the Mediterannean demonstrate an above normal temperature between 2005 and 2010, with the majority of the areas showing a % probability of this above normal temperature trend.
Doblas-Reyes et al. (2013) 12

13. Climate services: Users should also adapt
Impact surfaces of a simple wind-energy model over the North Sea for DJF as a function of the mean seasonal wind and the wind intraseasonal variability. Power density estimates obtained using the XXth Century Reanalysis, a Rayleigh function to estimate high-frequency winds from mean daily values and a wind profile power law to obtain 100 m winds from 10 m winds.
D. MacLeod (Univ. Oxford)

14. Expected progress in climate prediction
Extratropical SSTs: the North Atlantic, North and South Pacific
Model response: representation of forced response to fast and slow forcings
Model inadequacy
Model improvement: all components
Initialisation: soil moisture, snow, sea ice, ensemble generation
Stratospheric processes, including ozone and aerosol
Calibration and combination
Documentation (follow the IPCC calibrated language), demonstration of value and outreach