Monday, 8 January 2018: 9:00 AM
Salon J (Hilton) (Austin, Texas)
Forecasting natural phenomena is a way that scientific understanding can bring societal benefits. However, the link between scientific knowledge and predictability of natural phenomena is not a straightforward one. The tropospheric community's early research into forecasting led to the discovery of "deterministic chaos": high sensitivity of the physical system to initial conditions, leading to limited forecast accuracy several days into the future. The upper atmosphere community is now in the early stages of forecasting space weather and there are many lessons to learn. We have developed a tool, the Space Weather Forecast Testbed (SWFT), that enables users to explore relationships between upper atmospheric characteristics and geospace and solar wind variables, to understand the roles that such variables might play in medium-range upper atmosphere forecasts (lead times of ~1-3 days). Our initial data set for representing weather conditions in the ionosphere is the data driven Global Ionosphere Map (GIM) produced daily over that past 20 years by JPL. GIM provides a map of ionospheric Vertical Total Electron Content (VTEC) at one-degree latitude and one-degree longitude resolution every 15 minutes. We have extracted ionospheric characteristics from the GIM, such as maximum VTEC during a storm interval, or maximum separation of daytime equatorial anomaly peaks, to allow users to explore quantitatively ionospheric properties that may be forecast effectively. We have incorporated into the SWFT database post-processed solar parameters such as solar wind velocity, ion density, temperature, magnetic field, the F10.7 and sunspot indices. Geomagnetic indices such as Dst, Ap, Kp are incorporated. As part of this testbed, we are populating a database of output from first-principles coupled thermosphere-ionosphere model runs, using models currently supported at the NASA/NSF Community Coordinated Modeling Center (CCMC). These models are run in a “forecastable mode”. As currently defined, forecastable model runs use measured solar wind inputs and solar EUV proxies, to compute all needed model inputs. Such algorithms prepare us for a transition to model runs used for forecasting, once solar wind forecasts are the basis for upper atmosphere forecasts.
Multivariate linear regression is the core SWFT algorithm to characterize relationships between ionospheric properties, the solar wind and other geospace variables. SWFT rigorously and automatically segregates past data used for training from data used for evaluating forecasts. As a data-driven forecasting toolset, users can benefit from knowing the degree to which the datasets in SWFT can be effective for forecasts. Numerical physics-based approaches to forecasting thus gain a useful set of benchmarks with which to compare. Ionospheric data assimilation methods are not currently part of SWFT since the long lead times of the forecast (1-3 days) reduces somewhat the value of such methods when strong driving by the solar wind is a dominant factor. The testbed is the beginning of an effort to build a platform for the community to deposit relevant data, algorithms and validation results to evaluate and improve forecasting skills. SWFT is being developed as part of the NASA and NSF partnership for Collaborative Space Weather Modeling.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner