3D electron density estimation in the ionosphere by using IRI-Plas model and GPS-TEC measurements HAKAN TUNA, ORHAN ARIKAN, FEZA ARIKAN Bilkent University, Ankara, Turkey htuna@bilkent.edu.tr, oarikan@ee.bilkent.edu.tr IONOLAB GROUP, Hacettepe University, Ankara, Turkey arikan@hacettepe.edu.tr IRI Workshop KMITL, Bangkok, Thailand 9-13 November, 2015
OUTLINE INTRODUCTION GPS-TEC MEASUREMENTS IRI-PLAS MODEL PROBLEM DEFINITION OPTIMIZATION PARAMETERS IONOLAB-CIT RESULTS CONCLUSION 2
INTRODUCTION Estimation of 3D electron density in the ionosphere is a crucial problem for investigating the ionospheric effects on electromagnetic propagation. Two important tools are generally used for investigating the ionosphere: GPS-TEC measurements widely used in ionospheric studies very sparse and non-uniform for employing 3D tomography methods (illconditioned problem) Ionospheric models like IRI-Plas can estimate monthly averages of 3D electron density distributions not generally compliant to the real measurements obtained from GPS receivers. In this study, a novel method for estimating the 3D electron density distribution in the ionosphere by using both GPS measurements and IRI-Plas model is presented. Proposed method perturbs default ionospheric parameters used in IRI-Plas model over a region of interest by using parametric perturbation surfaces, and iteratively searches for the best physically feasible 3D electron density distribution, which is compliant with the GPS-TEC measurements. 3
GPS-TEC GPS systems can give a good approximation of the Slant Total Electron Content (STEC) in a cylindrical path between the satellite and the receiver. Turkish National Permanent GPS Network (TNPGN) consists of 147 settled GPS receiver stations in Turkey. The pseudo range and phase data obtained from the network are used for calculating GPS-TEC values by utilizing IONOLAB- STEC method. 4
IRI-Plas Model (1) International Reference Ionosphere (IRI) is a physical and empirical model of the ionosphere. IRI model can estimate the electron density profile in the ionosphere for any given location and date. International Reference Ionosphere extended to Plasmasphere (IRI-Plas) is an extended version of IRI and increased the coverage to the extent of 20000 km. 600 500 Height (km) 400 300 200 100 0 0.5 1 1.5 2 2.5 3 3.5 Elektron Density (el/m 3 ) x 10 11 5
IRI-Plas Model (2) IRI-Plas is a parametric model, i.e. user can input known ionosphere parameters. f 0 F 2 and electron density profile relation h m F 2 and electron density profile relation 600 600 500 500 h m F 2 increases Elevation (km) 400 300 f 0 F 2 increases Elevation (km) 400 300 200 200 100 0 2 4 6 8 10 12 14 Electron Density (el/m 3 ) x 10 11 100 0 0.5 1 1.5 2 2.5 3 3.5 4 Electron Density (el/m 3 ) x 10 11 6
Problem Definition TNPGN pseudo range and phase data IONOLAB-STEC Real STEC measurements Perturbed ionospheric parameters in region Regional IRI-Plas IRI-Plas Synthetic STEC calculation (IRI-Plas-STEC) Synthetic STEC measurements Find the optimum perturbation values on the selected ionospheric parameters in a region, such that, the resultant 3D electron density distribution generates synthetic STEC values similar to the real GPS-TEC measurements. 7
Optimization Parameters (1) Perturbation surfaces f λ,φ = m 1 λ + m 2 φ + m 3 h λ,φ = m 4 λ + m 5 φ + m 6 Perturbed ionospheric parameters fof2 λ,φ = fof2 (λ,φ) IRI Plas + f(λ,φ) hmf2 λ,φ = hmf2(λ,φ) IRI Plas + h(λ,φ) λ : latitude, λ : normalized latitude [-1, 1] φ : longitude, φ : normalized longitude [-1, 1] fof2 λ,φ Default IRI-Plas hmf2 values are obtained for given perturbed fof2 Regional IRI-Plas hmf2(λ,φ) IRI-Plas IRI Plas A synthetic 3D electron density distribution (N) is computed from perturbed fof2 and hmf2 values, then any synthetic STEC value (T) is calculated from this matrix by multiplying it with the corresponding B vector. fof2 λ,φ hmf2 λ,φ Regional IRI-Plas IRI-Plas N Synthetic STEC calculation (IRI-Plas-STEC) B T 8
Optimization Parameters (2) Difference between measured and calculated STEC values Objective: Minimize the following cost function C = 2 M T M + ρ H H H 2 Deviation from physical relation between fof2 and hmf2 values Relaxation parameter (used as 3 in simulations) M: The array containing measured GPS-TEC (IONOLAB-STEC). T: The array containing synthetic GPS-TEC (IRI-Plas-STEC). H: The array containing perturbed hmf2 values. H : The array containing default hmf2 values for given perturbed fof2 values from IRI-Plas. 9
IONOLAB-CIT Real Synthetic 10
Optimization Techniques Three optimization techniques are used in this study. Gradient Descent (GD) Basic optimization tool using the 1st order derivative of the optimization function. Broyden Fletcher Goldfarb Shanno (BFGS) Quasi-Newton method using both the 1st and 2nd order derivatives of the optimization function. Particle Swarm Optimization (PSO) Stochastic optimization method which does not depend on the gradient of the optimization function. 11
RESULTS (1) Two days are selected for demonstrating the results of IONOLAB-CIT technique: 1 September 2011, 12:00 GMT quiet, Wp = 2.0 10 March 2011, 12:00 GMT positively disturbed, Wp = 5.2 Start Day UT Peak Day UT Wp-max End Day UT Hrs Wp-mean Power ----------------------------------------------------------------------- 2011/03/10 21 2011/03/12 15 5.2 2011/03/13 7 59 4.6 306.8 G1-TYPE GEOMAGNETIC STORM CME on March 7, 2015 12
RESULTS (2) 1 September 2011, 12:00 GMT, Quiet Day Cost function, C, obtained by IRI-Plas 0.1711 Cost function, C, obtained by IONOLAB-CIT 0.0496 Chosen Optimization Method!! Cost function with respect to iteration number IRI-Plas VTEC Map (TECU) IONOLAB-CIT VTEC Map (TECU) 13
RESULTS (3) 1 September 2011, 12:00 GMT, Quiet Day Perturbation on fof2 (MHz) Optimized fof2 (MHz) Perturbation on hmf2 (km) Optimized hmf2 (km) 14
RESULTS (4) 1 September 2011, 12:00 GMT, Quiet Day IRI-Plas electron density slices (10 11 el/m 2 ) Fixed Latitude (39 N) Fixed Longitude (35 E) Fixed Height (250 km) IONOLAB-CIT electron density slices (10 11 el/m 2 ) Fixed Latitude (39 N) Fixed Longitude (35 E) Fixed Height (250 km) 15
RESULTS (5) 10 March 2011, 12:00 GMT, Positively Disturbed Day Cost function, C, obtained by IRI-Plas 0.4281 Cost function, C, obtained by IONOLAB-CIT 0.0278 Chosen Optimization Method!! Cost function with respect to iteration number IRI-Plas VTEC Map (TECU) IONOLAB-CIT VTEC Map (TECU) 16
RESULTS (6) 10 March 2011, 12:00 GMT, Positively Disturbed Day Perturbation on fof2 (MHz) Optimized fof2 (MHz) Perturbation on hmf2 (km) Optimized hmf2 (km) 17
RESULTS (7) 10 March 2011, 12:00 GMT, Positively Disturbed Day IRI-Plas electron density slices (10 11 el/m 2 ) Fixed Latitude (39 N) Fixed Longitude (35 E) Fixed Height (250 km) IONOLAB-CIT electron density slices (10 11 el/m 2 ) Fixed Latitude (39 N) Fixed Longitude (35 E) Fixed Height (250 km) 18
RESULTS (8) Comparison with Ionosonde Measurements Comparison of plasma frequencies obtained by using IRI-Plas model and proposed IONOLAB-CIT technique at Cesme (38.3 N, 26.4 E), with the plasma frequencies obtained by using ionosonde measurements and two automatic ionogram scaling techniques ARTIST and POLAN, at Athens (38.0 N, 23.5 E) 1 September 2011, 12:00 GMT 10 March 2011, 12:00 GMT 19
1 September, 2011 RESULTS (9) Quiet day Electron density slices (10 11 el/m 2 ) 20
RESULTS (10) 10 March, 2011 Positively Disturbed day Electron density slices (10 11 el/m 2 ) 21
CONCLUSION Input parameters of IRI-Plas model are tuned in a way that the resulting 3D electron density profile is in compliance with GPS based STEC measurements and the input parameters are in compliance with each other. fof2 and hmf2 values over Turkey are both represented with additive surface models with 3 parameters. The problem is reduced to a 6 parameter optimization problem. GD, BFGS and PSO optimization methods are used for solving the optimization problem. BFGS method provided the best results. Results show that the proposed methodology provides 3D electron density distributions compliant with both real GPS STEC measurements, and ionosonde measurements. The proposed technique only takes the spatial variability of ionosphere parameters into account and does not consider temporal correlation of the optimization parameters. Future work will introduce the time dimension into the proposed method. 22
THANKS Joint TÜBİTAK 112E568 and RFBR 13-02-91370-Cta Joint TÜBİTAK 114E092 and AS CR 14/001 TÜBİTAK 114E541 You are all invited to COSPAR 2016 in Istanbul, Turkey!! https://www.cospar-assembly.org/ http://cospar2016.tubitak.gov.tr 41st COSPAR Scientific Assembly - C1.4 Session Regions of the Enhanced Risk for the Ionospheric Weather 30 July 7 August 2016, Istanbul, Turkey, MSO Feza Arikan, DO Tamara Gulyaeva. Your Contributions are Welcome! Limited funding may be available for young scientists and researchers!! 23