NASA/NOAA MOU Annex Final Report: Evaluating Model Advancements for Predicting CME Arrival Time

The purpose of this project was to assess improvements in CME arrival time forecasts at Earth using the Air Force Data Assimilative Photospheric Flux Transport (ADAPT) model driven by data from the Global Oscillation Network Group (GONG) ground observatories. These outputs are then fed into the coupled Wang-Sheeley-Arge (WSA) - ENLIL model and compared to an operational version of WSA-ENLIL (without ADAPT). SWPC selected a set of 38 historical events over the period of five years from 2012–2014 (33 events) and 2017–2019 (5 events). The overall three-year project consisted of multiple simulation validation studies for the entire event set (1292 simulations): (a) benchmark single map (operational version prior to May 2019) (b) time-dependent sequence of GONG maps driving WSA-ENLIL with 4 different model settings (c) single test simulation of a time-dependent sequence of GONG maps driving ADAPT-WSA-ENLIL (d) single GONG map driving ADAPT-WSA-ENLIL (e) time-dependent sequence of GONG maps driving ADAPT-WSA-ENLIL. We report that for all 38 events, within each model version/settings combination, the CME arrival time error decreased by 0.2 to 0.9 hours when using a sequence of time-dependent zero-point corrected magnetograms compared to using single magnetogram input. Overall, for all events, when using the older uncorrected magnetograms, the CME arrival time error increased for all new model versions/settings combination compared to the benchmark. Notably for the 5 events in the period 2017–2019 when more reliable zero-point corrected magnetograms were available, the ADAPT-WSA-ENLIL (median arrival realization) CME arrival time error decreased against all benchmarks. In this report we also discuss replicating the operational model, challenges in detecting CME arrival in simulations, and comparing zero-point corrected and uncorrected magnetogram inputs.

💡 Research Summary

The NASA‑NOAA MOU Annex Final Report evaluates whether incorporating the Air Force Data‑Assimilative Photospheric Flux Transport (ADAPT) model, driven by Global Oscillation Network Group (GONG) magnetograms, can improve the prediction of coronal mass ejection (CME) arrival times at Earth when coupled with the Wang‑Sheeley‑Arge (WSA) solar‑wind model and the ENLIL heliospheric MHD model. The study spans three years and focuses on 38 historical CME events—33 from 2012‑2014 and 5 from 2017‑2019—selected by the Space Weather Prediction Center (SWPC). A total of 1,292 simulations were performed across five experimental configurations: (a) a benchmark using a single daily‑updated, zero‑point‑uncorrected GONG map (GONGb) with WSA v2.2 and ENLIL v2.6.2 (replicated with ENLIL v2.9e); (b) a time‑dependent sequence of GONGb maps with the same model versions; (c) a single‑event test using a time‑dependent GONG sequence to drive ADAPT‑WSA‑ENLIL (WSA v4.5, ENLIL v2.9e); (d) a single‑map ADAPT‑WSA‑ENLIL run; and (e) a full time‑dependent GONG sequence driving ADAPT‑WSA‑ENLIL.

ADAPT generates an ensemble of twelve realizations from the GONG observations; the study uses the ensemble median (central realization) for performance assessment. Model versions were systematically varied so that only one change (e.g., WSA version, ENLIL version, magnetogram correction) occurred at a time, allowing clear attribution of any performance differences.

Key findings are: (1) Within any given model configuration, replacing a single magnetogram with a time‑dependent sequence of zero‑point‑corrected GONGz maps reduces CME arrival‑time error by 0.2–0.9 h on average. This improvement is attributed to more accurate representation of far‑side active‑region emergence, which the ADAPT flux‑transport scheme captures through differential rotation, meridional flow, supergranular diffusion, and stochastic flux emergence. (2) When using the older, uncorrected GONGb maps, all newer model versions (WSA v4.5, ENLIL v2.9e) and ADAPT‑driven runs actually increase the arrival‑time error relative to the benchmark, underscoring the dominant role of input magnetogram quality.

The five events from 2017‑2019, for which reliable zero‑point corrections are available, show the most pronounced benefits. Compared with the operational benchmark, the ADAPT‑WSA‑ENLIL median‑realization yields a mean error reduction of 3.1 ± 4.0 h for a single‑map ADAPT run, 4.4 ± 7.2 h for a time‑dependent ADAPT run, and 5.8 + 8.6 − 7.6 h for the time‑dependent ADAPT run relative to the single‑map benchmark. These reductions are statistically significant given the limited sample size and demonstrate that the ADAPT flux‑transport model materially improves the background solar‑wind environment used by ENLIL.

The report also discusses practical challenges: (i) detecting CME arrival in ENLIL output, especially when multiple CMEs are simulated simultaneously; (ii) handling events where ancillary CMEs, not intended for analysis, influence the primary CME’s propagation; and (iii) replicating the operational model settings (ambient parameters, grid resolution) to ensure a fair comparison. A resolution test confirms that the chosen ENLIL grid (0.25 R⊙) is sufficient for the studied events.

In conclusion, the study provides three actionable insights for operational space‑weather forecasting: (a) upgrading to ADAPT‑driven, time‑dependent, zero‑point‑corrected magnetogram sequences should be prioritized over merely updating WSA or ENLIL versions; (b) a real‑time pipeline for zero‑point correction of GONG data is essential to realize the full benefit of ADAPT; and (c) the ensemble approach of ADAPT can be leveraged to quantify forecast uncertainty, offering a path toward probabilistic CME arrival predictions. The authors recommend integrating ADAPT‑WSA‑ENLIL into the routine SWPC workflow and establishing a continuous validation framework using the same 38‑event dataset, which can serve as a community benchmark for future heliospheric modeling efforts.