Team O1-02
Quantification of Uncertainties in Space Weather Forecasts
(Leaders: N. Pogorelov, C.N Arge, E. Camporeale, T. Singh, and M. Zhang)
COSPAR ISWAT Working Meeting
Cape Canaveral, FL, February 10-14, 2025
1. All models, including solar and heliospheric ones, have uncertainties associated with them. They arise from various sources including
2. Our goal is to reduce such uncertainties. However, theory and modeling efforts are considerably more advanced as compared with data availability. Exceptions are the phenomena at the boundary of applicability of each model: magnetic reconnection, instabilities, etc..
3. The average error in CME arrival time is 10-12 hours regardless of the the applied physics-based code. This uncertainty is not acceptable for forecasting.
4. There seems to be a consensus that the errors are mostly associated with uncertainties in the identification of parameters of CMEs inserted in the background solar wind.
5. Sensitivity analysis of the simulation results shows that speed, longitude, and tilt of a CME are the main sources of uncertainties.
6. Machine learning (ML) techniques are capable of reducing uncertainties in CME arrival time, and potentially Bz, forecasts. However, this is demanding procedure because it requires efficient training of ML algorithms.
7. It is important for the models to be able to simulate continuous-coverage observations in the space between the Sun and Earth, and establish correlations between the corresponding errors and the discrepancies at Earth, or other points of interest (Mars?). STEREO A and B HI time-elongation maps (J-maps) have been used successfully for this purpose, but other data sets can also be involved.
8. ML can be used to automatically quantify uncertainties in ensemble modeling and reconstruct the reasons of unsuccessful forecasts.
9. Uncertainties in the ambient solar wind may play a minor role for strong CMEs, but are important for weak ones and situations when multiple CME follow each other.
10. The Taylor diagrams can be used to evaluate uncertainties at multiple points in the same figure.
11. Solar Energetic Particles (SEPs) forecasts require data-driven simulations of CMEs in the solar corona. Shock speed, size, shape, location, and propagation affect SEP production, their spectrum, intensity, and time profile of injection. In addition to modeling uncertainties, theoretical ones are also present.
Future work
2. Enhance ML training with other data sets that provide us with a continuous coverage of different quantities in the heliosphere.
3. Perform comparative simulations of the same CME.
3. Investigate the prospects of applying ML to beacon data, particularly using software that allows for “data improvement.”
4. For SEPs, we need (i) more runs with CME shock reconstruction or generic CME shock models; (ii) reliable solar wind density model; (iii) modeling with MHD CME and background to improve understanding of the effects of SW modulation on SEPs; (iv) improvements to the estimates of the shock cutoff momentum; (v) fixing particle interaction for strong shock; (vi) Investigation of how diffusion coefficients affect the peak intensity and (vi) study SEP production ot interaction CMEs.