Electron density profile retrieval from RO data Xin’an Yue, Bill Schreiner  Abel inversion error of Ne  Data Assimilation test

2 Assumptions used in CDAAC Abel: 1. Straight-line signal propagation 2. Circular satellite orbit 3. First-order estimation of electron density at the orbit altitude 4. Spherical symmetry of electron density Abel Ne Error Sources:

3 Orbit Ne estimation evaluation: Comparing CHAMP RO estimated with PLP observed Ne on the orbit

4 Current CDAAC estimation method tends to overestimate the orbit Ne. [Since we use the constant Ne assumption around top most points, while the Ne decrease with the altitude increase]

5 Abel 1: First order estimation; Abel 2: Given by on orbit observation (here is the model result) Abel 3: No orbit Ne 1, Different orbit Ne estimation method will mainly influence the top most points. 2, Adding an on orbit Ne observation seems like not necessary for the Abel inversion.

6 Abel Error distribution versus latitude and altitude: Modeling results Noon time LT=13 Unit: 1×10 11 /m 3 ‘True’ Ne Retrieved Ne Absolute Deviation Relative Deviation

7 110 km220 km COSMIC observations (same time/duration as simulation): Unit: 1×10 11 /m 3 COSMIC Abel retrieval from NeQuick model

8 Artificial Wave Number 4 structure of Ne below F2 region made by the Abel inversion: Simulation

9 COSMIC Observations, LT

10 The orbit Ne is overestimated by ~10% in the current CDAAC method from CHAMP observations. The orbit altitude Ne only influences the topside of the EDP. Adding an on orbit observations has no positive effect on the EDP retrieval. The Abel inversion can result in some systematic errors. e. g.: plasma cave, three peak of the E layer along the latitude, artificial wave number 4 structure in the E and F1 layer. Conclusion:

11 2.Data assimilation retrieval

12 Assimilation method and parameter choice P ij =r1X i b X j b *e -d ij /L r1=0.01 R ij =r2y 2 δ ij r2=0.001 Grid division Latitude:2.5 degree; Longitude: 5 degree; Altitude: 2.5 km For one occultation event, there are ~6000 grid points that GPS rays pass through

13 Ionospheric Correlation Length

14 Step 1: Retrieval Real Data Step 2: Simulation 1:Abel Retrieval (Spherical Symmetry) 2:Data Assimilation (DA) Retrieval Background Model: IRI Input: Real F107 BackgroundF107 DA1NeQuickreal DA2IRIreal DA3IRIReal (very quiet) is chosen to do the test 1:Simulate the Occultation side slant TEC by NeQuick Model (input real F107) 2: Abel retrieval simulated sTEC 3: Data assimilation Retrieval the simulated sTEC and obtain the EDP along the tangent point

15 Occultation event distribution during the selected day ( ). Also shown is the co-located Ionosondes and profile number(22 stations, 72 profiles in total)

16 An example of DA retrieval of simulated TEC by three different backgrounds Results1:

17 Comparison of NmF2 & hmF2 of 3 DA retrievals from simulated TEC Results2:

18 Comparison of all Ne and error statistical of Abel retrieval and DA2 retrieval from simulated TEC Results3:

19 Comparison of EDP retrieved by Abel and DA method between two co-located cases in the same time. Validation1: data assimilation retrieval is less influenced than Abel method by the ionospheric inhomogeneity

20 Comparison of the latitude and altitude Ne and its retrieval error from simulation LT=13 Manmade plasma cave disappears in DA retrieval. Validation2: (Simulation)

21 Comparison of Ne retrieved from real data by Abel and DA;LT=13 Cont.(real data)

22 Comparison of Ne and its error from simulation Validation3:(Simulation)

23 LT-MLat variation of Ne from real data Cont.(real data)

24 An example of retrieved EDPs in comparison with co- located Ionosonde EDP Validation4: (With Ionosonde)

25 Statistical comparison with co-located Ionosonde data (below hmF2 Ne) Cont.

26 Data assimilation Retrieval has a better performance than Abel retrieval from simulation study DA method can improve the retrieval especially around and below F2 peak region. The climate features such as manmade plasma cave (also call 3 E layer peaks) are improved by Data assimilation retrieval. Comparison with Ionosonde data confirms the validation of DA retrieval methond. Conclusion:

27 Data assimilation retrieval can improve the retrieval by both the simulation and real data test, but it has following disadvantages: 1, Big computation. 2,Include the model information. 3, If the relative TEC (which is very accurate) is assimilated, it assumes the background model is unbiased, which is not usually satisfied. 4, If the absolute TEC is assimilated, the uncertainty of the absolute TEC calculation will influence the retrieval quality. 5, Might have degraded performance during disturbed condition.

28 World Map by COSMIC, for fun