Description
The prediction of future climate changes is one of the most complex problems undertaken by the scientific community. Although scientists have been striving to better understand the physical basis of the climate system behavior and to improve climate models, the overall uncertainty in projecting future climate has not been reduced (e.g., from the IPCC 2007 to 2013). With the rapid increase of complexity in Earth system models, reducing uncertainties and increasing reliability of climate projections becomes an extremely challenging task. 

Since uncertainties always exist in climate models, interpreting model simulations and quantifying uncertainty is key to understanding and modeling atmospheric, land, ocean, and socio-economic phenomena and processes. Meanwhile, climate change adaption and impact communities rely on climate models to provide climate change information. Such information, if not accurate, should be provided with well-quantified uncertainty.  

Uncertainty Quantification (UQ) is a fundamental challenge in numerical simulations of Earth’s weather and climate. It entails much more than attaching defensible error bars to predictions. In recent years, formal methods of verification, validation, and UQ employed in other simulation problems have been applied to climate simulations. The topics to be discussed in this workshop will include many aspects of UQ in climate modeling, such as identifying sources of uncertainty, describing uncertainty associated with input parameters, evaluating model uncertainty through validation against observations, model comparison between numerical and/or analytical solutions, and upscaling/downscaling, as well as quantifying uncertainty through both forward modeling (sensitivity analyses) and inverse modeling (optimization/calibration) in all components of climate and integrated Earth system models at various spatial and temporal scales.

The workshop is also aimed at providing both theoretical lectures and hands-on sessions on the theory and application of the various UQ methods and approaches, such as sensitivity analysis, construction of surrogate models and response surfaces, input parameter calibration studies, forward propagation of uncertainties, and assessment of model discrepancies and structural uncertainties. Supervised by the directors and lecturers, participants will be encouraged to design, complete and report on short research projects during the event.
Go to day
  • Monday, 13 July 2015
    • 08:15 - 21:00
      • 08:15 REGISTRATION 15' ( Adriatico Guest House - lower level )
         
      • 09:00 Welcome 30'
      • 09:30 Climate Modeling and sources of uncertainty in projections 1h0'
        Speaker: Chris Forest (PSU USA)
        Material: Slides Video
      • 10:30 Coffee Break 30' ( AGH Terrace )
      • 11:00 Bayesian Probabilistic Framework 1h0'
        Speaker: Gabriel Huerta (UNM USA)
        Material: Slides Video
      • 12:00 Lunch Break 1h30' ( AGH Cafeteria )
      • 13:30 Participant introductions and introduction to computer lab 30' ( AGH - Informatics Laboratories )
      • 14:00 Lab session 1h30'
        Speaker: Gabriel Huerta (UNM USA)
      • 15:30 Coffee Break 30' ( AGH Terrace )
      • 16:00 Quantification of Parametric Uncertainty of Large Complex Geophysical Models 1h0' ( AGH Informatics Laboratories )
        Speaker: Qingyun Duan (BNU China)
        Material: Slides Video
      • 19:00 Icebreaker - Reception Dinner 2h0' ( AGH Terrace )
        All participants are cordially invited!
  • Tuesday, 14 July 2015
    • 09:00 - 17:00
      • 09:00 Use of observations in uncertainty quantification 1h30'
        Speaker: Charles Jackson (Un. Texas, USA)
        Material: Slides Video
      • 10:30 Coffee Break 30' ( AGH Terrace )
      • 11:00 Gaussian Process Modeling and the use of Emulators in UQ 1h0'
        Speaker: David Higdon (Virginia Tech., USA)
        Material: Slides Video
      • 12:00 Lunch Break 1h30' ( AGH Terrace )
      • 13:30 Participant introductions and introduction to computer lab 30' ( AGH Informatcis Laboratories )
      • 14:00 GP modeling 1h30' ( AGH Informatics Laboratories )
        Speaker: David Higdon (Virginia Tech., USA)
      • 15:30 Coffee Break 30' ( AGH Terrace )
      • 16:00 Discussion: The problem of structural errors and what to do about them 1h0' ( AGH )
  • Wednesday, 15 July 2015
    • 10:00 - 23:00
      • 10:00 Interpretation of the Hadley Center probabilistic framework 1h30'
        Speaker: Ben Booth (Met Office UK)
        Material: Slides Video
      • 11:30 Coffee Break 30'
      • 12:00 Lunch Break 1h30' ( AGH )
      • 13:30 Risk assessment 1h0'
        Speaker: Chris Forest (PSU USA)
        Material: Slides Video
      • 14:30 Exploration of single and multi-model ensemble output and observational data 1h0' ( Adriatico Guest House - Informatics Laboratory )
        Speaker: C. Jackson / C. Forest (Un. of Texas, USA / PSU, USA)
      • 15:30 Coffee Break 30' ( AGH Terrace )
      • 16:00 Discussion: How to make use of multi-model ensembles within a UQ framework 1h0' ( AGH )
      • 20:00 poster presentations with refreshment 2h0' ( Adriatico Guest House - Informatics Laboratory )
  • Thursday, 16 July 2015
    • 09:00 - 17:00
      • 09:00 Calibration of hydrodynamic models 1h30'
        Speaker: Z. Jason Hou (PNNL USA)
        Material: Slides Video
      • 10:30 Coffee Break 30' ( AGH Terrace )
      • 11:00 Simpler approaches to exploring Climate Projection Uncertainty 40'
        Speaker: Ben Booth (Met Office UK)
        Material: Slides Video
      • 11:40 Model Predictability Depends on Model Fidelity 40'
        Speaker: Jagadish Shukla (COLA, George Mason Un., USA)
        Material: Slides Video
      • 12:20 Lunch Break 1h10' ( AGH )
      • 13:30 Participant introductions and introduction to computer lab 30' ( AGH )
      • 14:00 Exercises related to work on hydrodynamic modeling 1h30' ( AGH Informatcis Laboratories )
        Speaker: Z. Jason Hou (PNNL USA)
      • 15:30 Coffee Break 30' ( AGH Terrace )
      • 16:00 Discussion: The problem of specifying priors (Hou)/ How to use observations to test for model skill (Jackson) 1h0' ( AGH )
  • Friday, 17 July 2015
    • 09:00 - 16:00
      • 09:00 Sources of uncertainty inherent in regional climate projections 1h30'
        Speaker: Filippo Giorgi (ICTP)
        Material: Slides Video
      • 10:30 Coffee Break 30' ( AGH Terrace )
      • 11:00 Parametric sensitivity and autotuning in regional and global climate models 1h0'
        Speaker: Yun Qian (PNNL, USA)
        Material: Slides Video
      • 12:00 Lunch Break 1h30' ( AGH )
      • 13:30 Participant introductions and introduction to computer lab 30'
      • 14:00 Discussion: Calibration and propagation of errors within a coupled model setting. How to account for process interactions 1h30'
      • 15:30 Wrap-up 30'