Motivation for accommodating needs of MEO environment including GPS/Galileo + O3B and EPOR

Satellites should last ~20 years

BAS model including transport and waves

Boundary conditions: min/max pitch angle (gradient set to 0); Emax flux set to 0 [+kp??]

Outer L* boundary based on GOES >2 MeV e- flux

Model needs drift-average differential flux spectrum with Fixed L*

Spectrum shape approximated from GOES-MAGED

Kappa distribution fitted to PSD

Very good agreement with AE9 at GEO for 100cm-2sr-1s-1keV-1 at GEO

Short-term variability shows storms and electron desert

Impenetrable barrier from Baker is not valid at all times

When slot region is enhanced it can take weeks to return to normal levels (even if recently it has been benign)

Limiting from from Meredith is very close to the BAS model output (even better if we just look at IREM)

Validation done on SREM count rates from Giove-B. [would be great to re-visit this for Galileo/EMU]

Skill scores in the range of 0.6 to 0.8

 

Output covers L*/pitch-angle/energy

 

Making use of simulation:

Long-term spatial-temporal correlations for outer zone for range of timescales up to decades.

Filling the energy range where instruments lacks channels and using PADs

Using model to get back to the uni-directional fluxes

L*=2 limits the applications of the model for EPOR or LEO satellites (600 - 1000 km) - being looked at in Rad-Sat project

Incorporate a ring current model

 

PAD included

75GB output file size in raw form at present

 

IS

Correlations with magnetospheric indices as well as data

 

Need to separate data assimilation and modelling.

Extent of need for actual modelling is limited where we have good data and can do assimilation instead.

Data assimilation allows us to match data and model results and produce a Global flux map output.

Data -> model for next time step -> Kalman filter to incorporate more data -> step forward

Peaks in PSD are better generated with data assimilation

Assimilation: Can combine range of data sets and weight based on data quality; Good for post-event reconstructions; Good for initial conditions for forecasting.

GEO/GPS/LANL/POLAR data used to blend model and data and to derive 

Have 1-d and 3-d (VERB) model used with single boundary condition and then extended to use VAP and GOES (fluxes and PSD)

VAP will stop in one year and is nominally not operational. POES can help create a good reconstruction of the environment.

Convection model of ring current electrons based on different model due to different physical equations - VAP data useful for this reconstruction.

Forecast predictions can be quite accurate for ~2 days due to time it takes to extend.

 

What is the longest simulation you presently have? Oct 2012 - Oct 2016 10s keV up to >3 MeV

Ring current model takes more time with TBytes data output

NASA interested to get CCMC to run data assimilative models for climatology for re-analysis

SSA is getting aware. They should want to put nowcast and forecasts into a longer-term context.

VERB predictions is running operationally

 

Funding by CNES and FP7 (MAARBLE)

Data assimilation using Salammbo combined with data. Running 200 runs of Salammbo in parallel and then selecting the output(s) to propagate on the basis of data check (counts or fluxes)

Set of boundary conditions to run ensemble mode based on distribution of fluxes based on Kp along with energy-energy correlations

Bw median and stdev considered.

Radial diffusions also sampled randomly with correlations considered.

Re-analysis with ensembled Kalman filter; e- 03  - 5 MeV; mean and std of PSD; 10-min resolution; 6 weeks output - up to 15 years for electrons and almost 30 years for protons.

Basis for the OZONE model and used in GREEN-e (4<L<8) and GREEN-p (1<L<6)

 

Good to use actual response and counts for the assimilation process. Need: to have the response functions and count rates!

How do you use the count-rate check when the have a range of spectra coming from the model?

Drop-out in POES trapped protons get captured by the SALAMMBO model? Seems to over-estimate fluxes.

 

What's the relation between static (data-driven) models in GREEN? AS: not decided but thinks the best will be to use the data assimilation output as the basis and check against existing models.

 

IRENE Cycle Variation (Stu)

[example POES inverted fluxes]

User request for low-altitude protons - controlled by atmosphere as much/more than magnetic field

Inside IRENE need to be able to map to some kind of reference state

How will it be used? Factor 10 differences are possible but those are in regions of low flux

How can we forecast solar cycle to use this analysis? Cycle appears at different times (lag) for different energies...

Requirement to cover: 0.1 - 2000 MeV; 0<K^0.5<3; 0 < hmin < 1000

Correlation of flux related to F10.7 (with phasing lag) - using POES/Selesnick Inner Zone Proton Model (SIZM) to derive parameters (lag, h_min and flux scaling)

SAMPEX data not used at present but there are plans to reprocess that and then make use of it.

DH: Why K^0.5 and not K? SH: Seems better results away from equatorial region.

DH: Why are you using hmin? POB: Basically we found it was a better measure of the low altitude environment.

RH: factor 10 variation in POES data inverted, do you need anything beyond the worst case? SH: Previously we were just using solar min design, user case it not clear.

 

LARB (Daniel)

[model maps, e.g.slide 17]

Reconstruct PSB97 for quick revision of mission rad. env. (previously XIPE, now THESEUS)

Then establish the new LARB model.

<1000 km & 0.1 - 1000 MeV p+

Fluxes range from 10^6 down to 10^-3

10 deg pitch angle bins on REPT not easily mapped to equatorial pitch angle - N.B. equatorial pitch angle not used for the model

RPS has

L, a_eq breaks down at these low altitudes so h_min used along with density parameter, n, as a proxy of L* 

h_min is found tracing the drift shell but claculation is heavy so parabolic fits are using

Plan to use REPT to cover part of K v n map that SAMPEX doesn't

Geographical maps starting to come together - some more cleaning to be done

 

SH: using rho instead of h_min doesn't seem to improve performance

DH: hoping to capture the density in the same fit

RH: anything else matters apart from density?

DH: Well, there are injections but it's not clear how these should be incorporated

 

VALIRENE Toolkit

[slide 5 or 8]

Validation toolkit for the Ap9/AE9/SPM models against in-situ data sets and comparisons to other ECSS-recommended models also performed

Outputs also given for effects quantities.

Data used includes SAMPEX, EPT, PROBA-1/SREM, TSX-5/CEASE and AZUR + MERLIN, GPS, etc....

IRENE outputs of perturbed mean account for the uncertainties ain flux maps whilst m-c results are needed to capture space weather impacts

INTEGRAL p+ results shows that medium L ranges at lower energies AP8 appears to do better but all high energies AP9 appears to be better

RPS p+ data is almost always better matched to AP9 than AP8

INTEGRAL e- shows that for the most part one of AE8 max/min are better than AE9

However, for CRRES/MEA e- AE9 appears to be equally out-performing of AE8

AE9 doesn't appear to capture the variabilty even in M-C mode

AE9 appears to fall between MEO upper and MEO mean

Low-altitude high inclination orbits are not well-captured by IRENE

At GEO AE9 and AE8 are higher than IGE-POLE

EPOR for <1 year should use M-C mode of operation

 

YS: How do you know that the data sets are valid 

PT: We also passed the fluxes from the model through the responses to compare counts which seems to show similar results.

 

YS: 1990-1991 CRRES/MEA should be higher than the model

PT: The comparisons don't bear this out

 

IJ: Your thinking is that if you use IRENE be cautious?

PT: There are cases fir the electrons where even 95% doesn't appear to span up to some of the data set yearly variables.

POB: This is really due to the nature of IRENE as a consensus model and data sets (especially e-) have likely much higher intrinsic errors than data owners claim

 

AE9/AP9-IRENE (Paul)

[image?]

V1.55 to US gvt contractors in 04/2019

v2.0 planned to include solar protons and sample solar cycle

Need: long-term (> cycle) assimilation (or just physics) run output (10 eV - 10 MeV e- and 10-eV - 1 GeV p+, 10 eV - 200 keV He+, O+

Like to check also the low altitude electrons (e.g. using DEMETER data)

Inside model all distributions are Weibull or lognormal. In AE9 this creates discontinuities.

Could use TREPEM-style approach but hard to capture the errors when you don't have a distribution fit

Effects kernels in V1.55 will replace SHIELDOSE2 and extend to charging

Need: thin shielding dose tools, plasma effects and ion effects

POB: for development the python version will be shared within collaboration

RH: why is the low altitude stuff so hard? It's well-understood.

POB: We need (physical) models we can use that get the numbers right

 

YS: What is plasma in this context?

POB: 40 keV cut-off between plasma and radiation

 

YS: What's the use case?

POB: Material degradation primarily

YS: What about surface charging?

POB: In the design scenario you probably take more of a worst-case scenario

 

SPM-IRENE (Paul)

Tim Guild is the main author.

Want to include MLT variation in the plasma model - nothing in there presently

Grid parameters are Energy, Lm and equatorial pitch angle

Coverage issues are being addressed including THEMIS from v1.2

Limit of 1 keV due to high scatter in conjunctions preventing any kind of cross-calibration between plasma analysers

Templates taken from data sets supplemented by models

Issues: only 1 data set including composition, no mean local time variation

Plan to include HOPE in some way even though it cannot operate that well in the inner zone due to interference

Would like ring current simulators as well.

 

YS: Can HOPE not become a gold standard?

POB: Should be able to be

 

JAS: Why does you use the Cluster CIAS data which has time-of-flight and composition?

POB: Initially it wasn't in an interesting orbit, we might have to re-look at it.

Yoshi: Has up to 200 keV data which could be included inside the model (LEPI and MEPI on Arase).

 

HOPE Plasma data analysis (Dave)

[slide 8,Plasma interactions drive high-level charging

NASA and ECSS have worst case for GEO but other orbits aren't well specified for this effect

HOPE is a top-hat ESA with time-of-flight allowing energy + mass to be determined

Data has been ingested into ODI (25 eV - 51.8 keV)

Charging events result in acceleration of protons (so we don't see the lower energies) - Can check HOPE output against the E-field and waves instruments (but that is limited to 200 V whereas HOPE measures events up to 1000 V)

As is well known the events all occur in the midnight-dawn region of the orbit

Strangely events were mainly seen in 2012 - 2013 (perhaps resulting from a change in material properties)

Rare temperature enhancements show bit enhancements which are most pronounced at L=6 - 6.5

Mean data appears to show peak at L = 3 (would need to be further investigated)

Need electron and ion Maxweillian temperature and density and to see if the data is well fit by these during charging events (data may need to be corrected)

Temperature alone doesn't seem sufficient (maybe a flux ratio at different energies)

 

POB: Can we use 10 keV e- flux instead of temperature

DR: There are many people using different proxies which is an issue.

 

POB: How can we filter events to derive the worst-case for spacecraft specification

DR: Indeed 

 

Arase (Yoshi)

1-d Fokker-Planck solver with assimilation using Particle-filter

Prediction from distributions of state (vectors) are then compared to data to give a weighting for the likelihood and then resampling of the data based on this before next set of state vectors are given.

Radial diffusion coeff, lifetime and internal source parameters are output state.

400 - 800 keV electrons

Inclusion of internal source process improved results based on MAPE proxy for time series performance.

Arase data XEP (similar to REPT) to be extended into 2022/2023

 

RH: What range of energies does the source need to cover?

YM: 300-400 keV

 

Discussion

POB: We talked about gap filling for environments but templates in IRENE has resulted in kinks in the data. Can use data assimilation in this but would require running the data assimilation inside the model

YS: Can be done but we should add derived dose which is measured(?) and the number of steps that are run is important for smoothing

POB: Should use the BAS model but we might like something which has more of a data assimilative aspect than just GEO boundary conditions.

YS: Need to work out how to use LEO data

POB: we will need to translate the outputs into the IRENE coordinates system from PSD.

We would need to investigate funding process for comparing and integrating data assimilative model outputs.

 

PJ: Can we really include modes for cycle phases due to how poor cycle predictions tend to be.

YS: For cubesats where the production time is very short one it could be reasonable. They would like to understand to fix launch dates

POB: Tends to be unlikely that someone builds a line of satellites on the basis of being able to launch at only one phase of the cycle.

PJ: If you want to use a cycle phase then it should be used to build a historic picture to capture variability and confidence

RE: We need to define what we want to use for worst cases for spacecraft designers to make use of models.

PJ: How can we work on kinks introduced with low altitude models as opposed to 

SH: Doing the templates is the present effort to line these things up.

POB: Using physics-based models at LEO with protons

AS: LEO included as well as we can right now.

POB: OPAL is broadly based on data rather than physics-based modelling.

Inner zone p+ needs to be developed including CRAND and to go down to lower energies (100 keV) whereas now there is an issue below 20 MeV

DH: Need to get people to clean data and introduce processes so we don't have to keep re-processing it.

DH: We are also duplicating effort in cleaning these things up.

PJ: There are issues of trust but no longer issues to release the factors produced under our contract because it's now approved for o/s

POB: Probable in AFRL that this is as much a process issue than something intractable