7th OpenIFS User Meeting

The output volumes from high-resolution general circulation models are increasing exponentially but data storage, access, and analysis methods have not kept up. There is thus an increasing demand for data compression. Lossy compression methods like ZFP, SZ3, or BitRound promise higher compression ratios than bit-reproducible lossless methods. Since they lose some information (e.g. precision or resolution), care must be taken to only discard noise but no real information that impacts the scientific research performed on the data. What level of lossy compression is safe to use when the application of the model outputs is not known in advance?

Model restart files can be seen as a form of lossless compression, as they allow the full model outputs to be reproduced by resuming the model from a particular system state. By using lossy compression on the restart model states, we investigate what level of lossy compression can be safely applied without affecting the model dynamics, which may provide a safe lower bound on the compression that is possible.

We present preliminary results from our ongoing research with several toy models. Taking the concept of model restarts to its extreme, we explore the effect of lossy-compressing the model’s inner state on every timestep and analyse the effect of compressing less frequently. In particular, we investigate the impact of numerical stability in shallow water models. We also explore the relationship between compression errors and initial state perturbations in Lorenz’63 or stochastic physics in Lorenz’96 models, and the effect of compressing members of an ensemble individually.

We show how lossy compression of toy model states can be safely applied within chaotic model ensembles, but that it only provides minor compression ratio improvements over lossless methods. We plan to extend our research beyond toy- and towards operational models like OpenIFS in future work.

Extra-tropical cyclones (ETCs) are an important part of the climate system and are also responsible for much of the sensible weather in the mid-latitudes. The most extreme of these weather systems can cause damaging winds and heavy precipitation and therefore can have adverse impacts on society. The overall aim of our research is to better understand how these weather systems and their associated hazards will change in the future. In this presentation, I will give an overview of three ongoing projects. The first uses ERA5 reanalysis data combined with an objective cyclone tracking tool, TRACK, to study the intensity of extra-tropical cyclones in Europe and the North-Atlantic in the current climate by computing a range of intensity metrics. A key result from this study is that more traditional meteorological metrics of extra-tropical cyclone intensity, such as maximum vorticity, are poorly correlated with more impact relevant metrics such as precipitation or storm severity index. The second study is part of a model inter-comparison study within the H2020 project, CRiceS, where we investigate the impact of changing sea ice and sea surface temperatures on the North Atlantic jet stream and extra-tropical cyclones using OpenIFS and three other models. A notable result of this study is that while the three other models show an extension of the jet and storm track into central and Eastern Europe in the future, OpenIFS does not due to the effect of sea ice and SSTs changes acting to counteract each other in this region. The final study is an idealised modelling study where we have performed a large ensemble of baroclinic life cycles with OpenIFS@home where the initial conditions differ to reflect potential different climates. This study shows that realistic weather systems can be simulated in a highly controlled way which enables detailed analysis of their dynamics and likely impacts to be conducted, but also provides an ideal testbed for new dynamical cores and for testing new parameterizations, for example, for wind gusts

Baroclinic Wave Simulation (BWS) are used to study idealised extra-tropical cyclones. Our previous work implemented on OpenIFS 43r3 has shown the possibility to create stable and flexible background states able to run with moisture and full physics parametrisation. Moreover, 7 parameters can be controlled to produce a vast array of different background states. By varying these parameters, an ensemble of 6,500 baroclinic waves are simulated using OpenIFS@home. In this cases, the developing cyclones are physically realistic with a poleward motion, upstream and downstream developments and sensible minimum mean sea level pressure (MSLP). However, with 0 to 4 cyclones developing in each case, using a synoptic analysis approach on each cyclone is not feasible. Thus, automatic tools have to be developed to characterise each cyclone and to link the different background states to the resulting cyclones' intensities.

A distributed workflow combining an objective cyclone tracking algorithm and a Python application has been implemented. This workflow is able to (1) pre-process the OpenIFS variables, (2) track all ETC developing from each individual background state and (3) compute 75 features to characterise the tracked cyclones. The extracted features ranges from cyclone tilt, cyclones' lifespan to intensity measures such as maximum 10-m wind gust, minimum MSLP, precipitation, wind footprint and storm severity index. One ensemble member is processed in 15 minutes on average and on 1 core. Using the distributed nature of the workflow, the whole ensemble is processed within 1.5 days on 40 cores and the time is reducing linearly with the number of cores. Finally, the extracted features are saved in a data lake structure to ease statistical and machine learning processes. Further studies will use Multi-Layered Perceptron and Random Forests to analyse this large ensemble and understand the relationship between the background state and the resulting extra-tropical cyclones intensities.

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.

In this presentation, we will present how to simultaneously tune parameters of the stochastically perturbed parameterizations (SPP parameters; Ollinaho et al., 2017) in OpenIFS 43r3 using a hierarchical statistical algorithm called EPPES (Laine et al., 2012). The results will focus on how model tuning of the deterministic model affects both deterministic and ensemble forecasts. This work is based on Tuppi et al. (2023) and extends the work by showing the results related to the effect on the spread-skill relationship of ensemble forecasts. The presentation contains two subjects: (1) simultaneous algoritmic tuning of 19 SPP parameters of OpenIFS with different ensemble settings and (2) verification of deterministic and ensemble forecasts using the resulting tuned model. The results show that (1) algorithmic tuning is an efficient method for balancing subgrid-scale physics, and (2) using or not using initial state perturbations in the tuning process can have a systematic (and potentially detrimental) effect on the spread-skill relationship of ensemble forecasts.

References:
Laine, M. et al. (2012). Ensemble prediction and parameter estimation system: The method. DOI:10.1002/qj.922

Ollinaho, P. et al. (2017). Towards process‐level representation of model uncertainties: stochastically perturbed parametrizations in the ECMWF ensemble. DOI:10.1002/qj.2931

Tuppi, L. et al. (2023). Simultaneous Optimization of 20 Key Parameters of the Integrated Forecasting System of ECMWF Using OpenIFS. Part I: Effect on Deterministic Forecasts. DOI: 10.1175/MWR-D-22-0209.1