Algorithmic optimisation of key parameters of OpenIFS. Implications on ensemble forecasts

Numerical weather prediction models contain physical parameters describing various small-scale phenomena as a part of parameterization schemes. These parameters are uncertain and can be tuned manually, or more efficiently, using algorithmic methods. Algorithmic tuning is an appealing approach to increase transparency and repeatability of the tuning process. Often, the focus of model tuning is on deterministic forecasts and the effect of model tuning on ensemble forecasts receives little to no attention. This presentation exemplifies how a superficially justifiable choice of activating initial state perturbations in algorithmic tuning of model parameters can have a systematic (and potentially detrimental) effect on the spread-skill relationship of ensemble forecasts.

This presentation continues directly from the poster presented last year in EGU2023 (Tuppi et al. 2023). This time, the objective is to understand how algorithmic optimization of a weather model affects the skill of ensemble forecasts. This presentation focuses on ensemble forecasting-based verification of the tuned model versions using root-mean squared error/spread relationship, continuous ranked probability score, and filter likelihood score. The headline results show that ensemble forecasts run with tuned model parameters experience a significant reduction of spread when initial state perturbations are active during the tuning of the parameters. However, both choices of tuning the model with initial state perturbations activated and deactivated lead to optimal deterministic forecasts. This behavior likely arises from conflicting interests between the method to generate initial state perturbations and the method of determining goodness of the parameter values during tuning.

Tuppi, L., Ekblom, M., Ollinaho, P., and Järvinen, H.: Algorithmic optimisation of key parameters of OpenIFS, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-4817, https://doi.org/10.5194/egusphere-egu23-4817, 2023.