For the last two decades Earth System Models (ESMs) have been applied to predict the future climate in ranges of decades to century timescales. More recently, advanced technology has been allowing the growth of computing power and the development of new techniques, like Machine Learning (ML). Enhanced computing capacity and techniques enable more complex modeling systems and larger ensembles. Complex modeling systems like ESMs are increasingly being used in weather and climate applications in a more seamless way in the last few years, setting the frame for a new generation of operational weather and climate prediction systems. This has brought more accurate and reliable forecasts, making weather and climate information more valuable for stakeholders and policymakers. However, in addition to errors within individual ESM components, complex inter-component interactions can lead to the growth of errors whose root causes are difficult to identify and correct. Despite the evolution of ESMs and the increase in the understanding of physical, dynamical and biogeochemical processes of the Earth System, the observational network does not provide enough information to constrain ESMs in ways that allow fully understanding and resolution of model errors. The efforts to assess, test, and enhance models have reduced major systematic errors, but some persist, and new ones have emerged. Therefore, there is a continued need to improve the representation of different processes in ESMs by identifying and correcting systematic errors.
This session invites contributions that help to increase understanding of the nature and cause of systematic errors in ESMs. Of particular interest are studies that consider:
-model errors across space and time scales; the use of hierarchies of models, including single column models and constrained ESM components;
-physics-dynamics and physics-physics cross-component coupling;
-initialized predictions;
-climatology of weather prediction models;
-data assimilation methodologies to identify systematic errors and constrain parameters;
-use of ML to identify systematic errors and/or to detect causal connections between seemingly disparate parameters; stochastic parameterization to represent uncertainty.
Verifying diagnostics and metrics to identify and characterize systematic errors and process understanding across different modeling communities (regional and global km-scale modeling) are also welcomed.