© The Aerospace Corporation 2014

Ionospheric Aspects of COSMIC-2

Paul R. Straus The Aerospace Corporation October 2, 2014

With thanks to others: Gary Bust, Doug Brinkman, Keith Groves, Charlie Carrano

The COSMIC-2 Radio Occultation (RO) Ionospheric Enhancement

•  TGRS* RO ionospheric capability –  Sensor is able to track both GPS

(~31 SVs) & GLONASS (24 SVs) signals

–  Baseline performance of ~1000 occultations/day/COSMIC-2 satellite •  Full GPS/GLONASS capability would

be ~1450/day/C2 satellite – Might be achieved through future s/w modifications

COSMIC-2 Occultations – 3 Hrs Coverage

1,000-2,000 soundings/day

>12,000 soundings/day

Graphics courtesy University Consortium for Atmospheric Research

*TriG GNSS (Global Navigation Satellite System) RO System

COSMIC 1 Occultations–3 Hrs Coverage 90

-90

0

Latit

ude

(deg

)

0.08 0.04 RO Dens. Frac.

0

•  The baseline for COSMIC-2 (eq) translates to more than a factor of 10 enhancement in data density in the equatorial region relative to COSMIC-1 –  Low-latitude ionosphere is significantly more

variable than mid-latitude

The COSMIC-2 Spacecraft & Sensors

Graphic courtesy Surrey Satellite Technologies, Limited (SSTL)

IVM TGRS POD

Antenna (1 of 2)

TGRS RO Antenna (1 of 2)

RF Beacon Antenna

•  The Ion Velocity Meter (IVM) employs gridded electrostatic analyzers to observe & characterize the in-situ plasma"

•  Key observables: in-situ ion density, temperature, & 3D drifts (E-fields)"

•  The RF Beacon transmits three tones to ground-based receivers to enable measurement of ionospheric scintillation & total electron content (TEC)"

–  400, 960, 2200 MHz (differs from COSMIC-1)!

COSMIC-2 Studies w/ Assimilative Models

•  Although a relatively recent development compared to their tropospheric analogs, ionospheric data assimilation techniques (e.g., Kalman filters, 4D-VAR, ensembles, etc.) have begun to be employed to improve ionospheric specification over the past ~10 years –  As model accuracies improve, their utility for scientific study

increases •  Assimilative models have the potential for avoiding (or

helping to correct) errors in individual profile inversions associated with the Abel transform –  Abel errors are of particular concern in the equatorial region

•  The COSMIC-2 (equatorial) observations will provide an excellent data set for assimilative studies –  In-situ IVM observations provide horizontal information –  Upward looking TGRS TEC data also complements

occultation TEC observations •  520 km altitude results in higher relevance of “overhead” TEC

data since there is more ionospheric above the s/c than for COSMIC-2

Abel Transform Errors Climatological “Truth”: NmF2 (/cc)

NmF2 Fractional Retrieval Error

IDA4D Sensitivity Study: Effect of Different Metrics

Model accuracy is metric dependent – need to assess performance from multiple perspectives to optimize data value in models

1400 km

TEC Path

Metric #1: In-situ CHAMP density (ne) @320 km

Metric #2: Jason vertical TEC

Solar minimum, quiet conditions (days 180-190 of 2009) Inputs: Ground GPS, COSMIC-1 TEC & DORIS TEC data

Ionosphere

20,200 km

TEC1

TEC2

Plasmasphere GPS-1 GPS-2

GPS Receiver

ΔTEC = TEC1 – TEC2

Metric #3: Differential Ground GPS TEC

TEC

Metric #4: CORISS* RO TEC topside gradient & Hmax

*C/NOFS (Communications/Navigation Outage Forecasting System) Occultation Receiver for Ionospheric Sensing & Specification

Alti

tude

(km

)

TEC (TECu)

Hmax

d(TEC)/dz

Results: Percentage improvement Metric

Day Night Low Mid Low Mid

CHAMP ne 39% 34% 10% 20%

JASON TEC 40% 15% 56% 3%

GPS ΔTEC 23% 13% 17% 12%

C/NOFS Hmax 1% --- -21% ---

C/NOFS d(TEC)/dz 9% --- 4% ---

Large prereversal velocity enhancements (up to about 50m/s)occur over a broad range of longitudes near dusk duringequinox and close to dawn during June solstice. Theevening drift reversal times do not change much withlongitude during equinox, but vary considerably in theAmerican-African sector during December and June sol-stice. There are also large variations between the westernand eastern hemisphere morning drifts and reversal timesduring June solstice.[21] The local time, seasonal and longitudinal dependence

of the vertical drifts are illustrated in more detail in Figure 4.The midnight-dawn downward drifts are larger in the easternthan in the western hemisphere, and have largest magnitudesnear sunrise during June solstice. Figure 4 also show thelarge longitudinal variations in the morning and afternoondaytime upward drifts at all seasons. The December solsticedata show daytime velocity peaks near 10!W and 100!E atabout 1100 LT, and a broad longitudinal sector of enhancedupward drifts between about 170!E and 90!W, centered at

about 1000 LT. During equinox, there are four upwardvelocity peaks near 170!W, 100!W, 0!E, and 100!E atabout 1000 LT. In this season, the daytime longitudinaldrift velocity fluctuations appear to generally extend, withdecreasing amplitudes, into the evening sector. The Junesolstice drifts have moderate to large drift peaks near 90!W,0!E, and 100!E, a considerably smaller peak near 170!Wallat about 1100 LT, and also an early morning region ofenhanced upward drifts near 140!W. The latter is mostlydue to the sudden drift reversal near sunrise (see Figure 3).Figure 4 also illustrate the strong seasonal and longitudinaldependence of the prereversal velocity enhancements and ofthe evening reversal times. The premidnight downwarddrifts do not show a clear longitudinal pattern.[22] Figure 5 presents in the bottom and top panels the

seasonal and longitudinal dependence of the upward driftvelocities averaged between 0900–1200 LT and 1300–1600 LT, respectively. The equinox, June solstice and theeastern hemisphere December solstice morning and after-

Figure 4. Local time, seasonal and longitudinal dependent equatorial quiet time vertical drift velocitiesfor moderate solar flux conditions.

A05304 FEJER ET AL.: EQUATORIAL VERTICAL PLASMA DRIFTS

5 of 10

A05304

Plots from Fejer et al, JGR, 2008

•  In the past, drift meter observations from a single satellite (shown: ROCSAT) have been accumulated for many months to determine longitude/local time climatology

•  However, there is significant variability between individual orbits

•  Single orbits convolve temporal & spatial variability, making them difficult to use in models

•  Deconvolution should be possible using IVM data from the six COSMIC-2 satellites, enabling assimilation of drift data for the first time

obtained using three plasma density thresholds. This Figureindicate that the use of larger threshold plasma densitiesdecreases the scatter and the number of late night measure-ments. The upward drifts after about 1900 LT are due toequatorial plasma depletions, which are most frequent in theAtlantic-American sector. The lower plasma densities in thelate night sector are generally associated with larger down-ward velocities and also with larger scatter and errors in themeasurements, especially during low solar flux conditions.Our analysis indicated the occurrence of unrealisticallylarge late night downward velocities for plasma densitiessmaller than about 104 cm!3 due to the higher percentage oflight ions. As will be shown later, this resulted in largedepartures from the curl-free condition for the longitudinal-ly averaged zonal electric fields.[13] We obtained our best overall results using drifts

corresponding to plasma densities larger than 105 cm!3

between 0800 and 0200 LT, and larger than 5 " 104 cm!3

from 0200 to 0800 LT, which maximized both the accuracyof the measurements during the day and the number of latenight observations We note that the number and accuracy ofour late night measurements decreased with solar flux,especially during June solstice due to the higher percentagesof light ions. Finally, we tried to minimize the effects ofupward drifts associated with plasma depletions, and ex-cluded velocities with magnitudes larger than 100 m/s since,under geomagnetically quiet conditions (Kp # 3), theygenerally appear to be due to instrumental effects or, duringearly night periods, to equatorial plasma depletions. We willshow later than these constraints seem to give quite accurateaverage drifts, except for relatively low solar flux (Sa < 130)late night June solstice conditions.

3. Model Development and Accuracy

[14] Our database consists of over 560,000 quiet time(Kp # 3) measurements. These data were grouped intofour month seasonal bins representing December solstice(November–February), June solstice (May–August), andequinox (March–April, September–October) and 15! over-

lapping 30! wide longitudinal bins. We have used 1-h localtime bins, except in the 1700–2200 LT sector, where weused 30 min bins shifted every 15 min, in order to moreaccurately account for rapidly changing evening prereversaldrift enhancements. When averaging the data, we discardedpoints outside two standards deviations.[15] We have determined that the variations of the drift

velocities with solar flux were best reproduced by using bi-linear relationships. This was also the case for the verticaldrifts measured by the Jicamarca radar [Scherliess, 1997].For each local time, season, and longitude bin, we have firstgrouped the drift data into three overlapping solar flux indexbins: Sa # 130, Sa # 160, and Sa > 160. The average solarflux indices in these bins were about 110, 130, and 185,respectively, with largest values during equinox and small-est during June solstice. Other solar flux groupings werealso tried, but they resulted in inferior fits to the data. Wehave determined the variations of the drifts with solar fluxfor 100 # Sa # 120 and 140 # Sa < 210 using our binneddata and the assumption of a linear drift dependence withsolar flux. Between 0000–1700 and 2200–2400 LT sectors,we used 4 h local time bins shifted every 2 h. It is possiblethat these large bins have smoothed out too much the earlymorning flux dependence, but this was necessary since thedata are sparse in this period. For the 1700–2200 LT sector,where the drift velocities often change rapidly, we havecalculated the solar dependence using 60 min local timebins shifted by 30 min, except for a few cases that requiredthe use of 30 min bins shifted by 15 min. The derived fluxvariations for Sa # 120, and Sa $ 140 are generally quitesimilar. For 120 < Sa < 140, we have interpolated betweenour low and high flux variations. Finally, we have per-formed 3 point running averages on the longitudinal varia-tions of the derived flux dependence. The solar fluxvariations derived from our data are generally consistentwith those presented by Scherliess and Fejer [1999], exceptnear dusk where we have stronger solar flux dependence.[16] The empirical model derived using the procedure

outlined above gives the quiet time (Kp # 3) equatorialvertical drifts at an altitude of 600 km for each seasonusually in 30 min local time, 15! overlapping 30! widelongitudinal bins for Sa = 100–200. In the 1700–2200 LTsector, the drifts are specified in 15 min local time bins. Asdiscussed below, and further highlighted later, these modeldrifts are most accurate for equinox and December solstice.[17] The equatorial zonal ionosphere electric fields,

which drive F region vertical drifts, must be irrotationalon the times scales and for the quiet time conditionsconsidered here, so that their line integrals along the dipequator must be zero; i.e.,

IE d ‘ ¼

IBvzd‘ ¼ 0 ð1Þ

where B denotes the equatorial magnetic field strength(between 0.19 and 0.30 G at the height of the satellite), andvz is the vertical drift velocity. Scherliess and Fejer [1999]used this curl-free constraint in the development of theirequatorial vertical drift model. We used this condition toestimate the longitudinally averaged accuracy of ourempirical models as a function of universal time (UT). Wehave calculated the values of the above integrals by

Figure 1. Examples of scatterplots of quiet time equatorialvertical drift velocities over a 30! longitudinal sectormeasured by ROCSAT-1 for three plasma density thresholds.

A05304 FEJER ET AL.: EQUATORIAL VERTICAL PLASMA DRIFTS

3 of 10

A05304IVM E-Fields: Data for Physics-Based Assimilative Models

Low Latitude Scintillation Map Example

Occultation Tangent Point Tracks

C/NOFS Orbit Track

90° SZA 100° SZA

PLP S4 Events CORISS S4<0.025 CORISS S4>0.025

COSMIC-2 Equatorial & “Bubble” Region Specification

COSMIC-2 (equatorial) will enable characterization of the all bubble locations in the equatorial region

Scintillation Evolution Time Scale: 15-30 minutes

C/NOFS

COSMIC-2

C/NOFS

COSMIC-2

TGRS

IVM

Image courtesy AFRL

•  Large scale plasma depletions generated by ionospheric instabilities are key nightside features

• Using the fact that these irregularity regions map along the Earth’s magnetic field, IVM measurements of these “bubble” locations can be extrapolated in latitude

•  Physics-based assimilative models that use “large scale physics” to derive driving forces need to know where instability physics can affect background state –  Presence of undetected “bubbles” will result in invalid

retrieval/feedback of driving forces •  TGRS RO observations, in concert with IVM data, can

enable assessment of scintillation strength & whether or not a bubble is active or “dead”

Open Topics: Interpretation of RO Scintillation Data

•  Can irregularity region location(s) along the RO line-of-sight be accurately determined through phase screen modeling or other similar approaches?

•  What is the effective sensitivity for scintillation detection of the RO technique?

•  TGRS on COSMIC-2 will provide high-rate (50-100 Hz) amplitude and phase observations for strongly scintillated RO profiles throughout the full range of ionospheric tangent altitudes

•  RF Beacon receiver data will provide ground truth for to assist in answering these questions

Summary

•  In the COSMIC-2 era the number of RO TEC observations available for assimilative modeling will substantially grow relative to present day

•  The COSMIC-2 (equatorial) sensor complement should enable significant advances in the science of ionospheric assimilative modeling –  In particular, the IVM sensor’s abilities to measure the most important driving

force (E-fields) for the low latitude ionosphere, and to localize instability regions will be particularly significant

–  It will be important to evaluate the models from a variety of perspectives (metrics) to understand their limitations and to optimize utilization of RO (and other) data types

–  Modeling advancements driving by COSMIC-2 data should enable movement from climatological to “ionospheric weather” studies at low latitudes

•  Resolution of open issues related to RO scintillation detection might enable complete characterization of the low altitude scintillation environment with COSMIC-2