AS1.4
Biases in weather and climate models: representing uncertain sub-grid processes, understanding large-scale drivers, and paths to improvement
Co-organized as CL5.06/NP5.5/OS4.19
Convener: Hannah Christensen | Co-conveners: Stefan Sobolowski, Craig H. Bishop, Ariane Frassoni, Daniel Klocke, Erica Madonna
Orals
| Thu, 11 Apr, 16:15–18:00
 
Room 0.11
Posters
| Attendance Thu, 11 Apr, 14:00–15:45
 
Hall X5

Weather and climate models used for weather forecasts, seasonal predictions and climate projections, are essential for decision making on timescales from hours to decades. However, information about future weather and climate relies on complex, though imperfect, numerical models of the Earth-system. Systematic biases and random errors have detrimental effects on predictive skill for dynamically driven fields on weather and seasonal time scales. Biases in climate models also contribute to the high levels of uncertainty in many aspects of climate change as the biases project strongly on future changes. A large source of uncertainty and error in model simulations is unresolved processes, represented through parameterization schemes. However, these errors typically materialize at large spatial scales. Our physical understanding of the mechanical and dynamical drivers of these large-scale biases is incomplete. Incomplete mechanistic understanding hinders marked improvements in models, including identification of the parameterizations most in need of improvement.

Understanding and reducing the errors in weather and climate models due to parameterizations and poorly represented mesoscale to regional scales processes is a necessary step towards improved weather and climate prediction. This session aims to bring together these two perspectives, and unite the weather and climate communities to address this common problem and accelerate progress in this area.

This session seeks submissions that aim to quantify, understand, and reduce sources of error and bias in weather and climate models. Themes covered in this session include:

- Theory and development of parameterization. Impact on errors in mean state, model variability and physical process representation;

- Improved physical understanding of the drivers of large-scale biases including the use of process studies, idealized modeling studies and studies with strong observational components;

- Growth and propagation of error and bias in models; model errors across temporal and spatial scales; dependency of errors on model resolution and the development of scale-aware parameterization schemes;

- Use of “emergent constraints” to relate present day model biases with the climate change signal;

- Understanding and representing random model error.

Invited presentations: Felix Pithan (AWI) and Bob Plant (University of Reading)

Lead Convenors: Hannah Christensen and Stefan Sobolowski
Co-convenors: Craig Bishop, Ariane Frassoni, Daniel Klocke, Erica Madonna, Isla Simpson, Keith Williams, Giuseppe Zappa