Forecasting ionospheric Total Electron Content at global level

Forecasting ionospheric Total Electron Content at global level one day in advance Cesaroni C. 1, Spogli L. 1, 2, Aragon-Angel A. 3, Fiocca M. 4, Dear V. 5, De Franceschi G. 1, Romano V. 1, 2 1 Istituto Nazionale di Geofisica e Vulcanologia, Italy 2 Spac. Earth Technology, Italy 3 European Commission, Joint Research Centre, Italy 4 Independent Researcher, Italy 5 National Institute of Aeronautics and Space, Indonesia luca. spogli@ingv. it

Total Electron Content model Validation Features q q q Testing against independent TEC measurements q Statistical assessment q Dependence on geospace conditions q “Frozen ionosphere” q Figures of merit are provided Empirical Forecasting horizon: 24 hrs Planetary scale Time res: 3 hrs Ingredients q Global Iono Map maps by IGS q Nonlinear autoregressive neural network with external input (NARX) - supervised q Effective Sunspot Number R 12 eff q Ne. Quick 2 q Kp forecasting by NOAA Cesaroni et al. Implementation q Ionospheric Prediction Service q PECASUS Forecasting ionospheric Total Electron Content at global level one day in advance

Model principles Step 1 Single point TEC forecasting Step 2 Global TEC forecasting Cesaroni et al. Forecasting ionospheric Total Electron Content at global level one day in advance

Model principles Step 1 Single point TEC forecasting q For each grid point, a different NARX is trained by using 11 years long (2005 -2015) TEC and Kp time series divided into training (70%), validation (15%) and testing (15%) datasets. q All TEC time series are from Global Ionospheric Maps (final products) provided by IGS (time res is 2 hours). q Kp time series is obtained by interpolating 3 -hours resolution data to obtain a value of the Kp every 2 hours. q Output of the step is TEC(t+24 hrs) for each grid point Cesaroni et al. Forecasting ionospheric Total Electron Content at global level one day in advance

Model principles Retrieving of R 12 eff (Olwendo and Cesaroni, 2016) Step 2 Global TEC forecasting R 12 eff Cesaroni et al. Ne. Quick 2 [Nava et al. 2006] on a regular grid 2. 5° lat x 5°lon Forecasting ionospheric Total Electron Content at global level one day in advance

Examples of output Cesaroni et al. Forecasting ionospheric Total Electron Content at global level one day in advance

Model validation q TEC provided as daily IONEX files by UPC as part of the International GNSS Service (IGS) Final Products* q v. TEC measured by the dual-frequency altimeter instrument on board the JASON-3 spacecraft Validation q q q Testing against independent TEC measurements Statistical assessment Dependence on geospace conditions “Frozen ionosphere” Figures of merit are provided DTEC = TEC difference between the forecasted TEC and the corresponding IGS TEC map in the grid point (i, j). N grid points of the maps - (i, j) coordinates. * accuracy of about 5 TECu [Hernández-Pajares et al. , 2009], provided weekly with a latency of about 11 days Cesaroni et al. Forecasting ionospheric Total Electron Content at global level one day in advance

Model validation: statistical assessment June 2017 to May 2018 Cesaroni et al. Forecasting ionospheric Total Electron Content at global level one day in advance

Model validation: case events Date 08 Feb 2018 13 -14 Jan 2018 07 -12 Nov 2017 26 Sep – 03 Oct 2017 27 -31 May 2017 07 -11 Sep 2017 Max Kp 1+ 56+ 77 8+ Min Dst (n. T) -1 -14 -72 -75 -125 -142 Note G 0 – quiet time G 1 storm - CH G 2 storm – CH G 3 storm - CME G 4 storm - CME Cesaroni et al. Figures of merit (TECu) RMSE m s -0. 6 4. 2 -0. 6 3. 3 3. 4 0. 04 5. 1 -0. 9 4. 0 4. 1 0. 3 3. 6 0. 4 3. 8 Forecasting ionospheric Total Electron Content at global level one day in advance

Model validation: frozen ionosphere Frozen ionosphere Cesaroni et al. Model Forecasting ionospheric Total Electron Content at global level one day in advance

Model Implementation in Real-Time services http: //pecasus. eu/ http: //ips. gsc-europa. eu/ Cesaroni et al. ! ! PECASUS has started 24/7 operations of the space weather services for civil aviation on November 7 th 2019. Forecasting ionospheric Total Electron Content at global level one day in advance

Remarks • TEC Forecasting (24 Hrs) Model is in place and working in real-time • At statistical level, model accuracy is 0. 1 TECu and precision is 3. 5 TECu • Capability to model TEC variations during storms • Accuracy 0. 4 to 0. 9 TECU, precision below 5. 1 TECu • Independently on storm intensity • Performance worsens at EIA crests • Good prediction capability over the oceans (Jason 3 data) Session G 12 Long term ionosphere forecasting conveners: M. Hernandez Pajares, A. Beleaki, L. Spogli • Paper under revision in Journal of Space Weather and Space Climate Thanks! luca. spogli@ingv. it Cesaroni et al. Space Weather workshops has been proposed: stay tuned! Forecasting ionospheric Total Electron Content at global level one day in advance

Backup slides Cesaroni et al. Forecasting ionospheric Total Electron Content at global level one day in advance

Model validation: frozen ionosphere Frozen ionosphere Cesaroni et al. Model Forecasting ionospheric Total Electron Content at global level one day in advance

Model validation: Jason-3 Height = 1330 km. Coverage over the oceans between latitudes of 66°N and 66°S WARNING: JASON-3 v. TEC measurement is slightly affected by the missing topsideplasmaspheric component in the altimeter measurements with respect to the GNSS v. TEC Cesaroni et al. Forecasting ionospheric Total Electron Content at global level one day in advance