modeling particle precipitation in the earth’s atmosphere · 2013-11-22 · modeling particle...
TRANSCRIPT
Modeling Particle Precipitation in the Earth’sAtmosphere
and how precise it is10th European Space Weather Week, Antwerpen
Jan Maik Wissing
Department of Physicse-mail: [email protected]
November 21, 2013
Motivation
[Jackman et al., 2001]
Why are we interested inprecipitating particles?
Impacts:ionizationchemical processeselectric conductivity
Measurement of1 input (needs in-situ
particlemeasurements)
2 effects (TEC: radars,GPS or limp-sounding:MIPAS)
Jan Maik Wissing (University of Osnabruck) Modeling Particle Precipitation November 21, 2013 2 / 20
Outline
Outline1 How do particle precipitation models work?2 How precise are these models?
satellite datamodel assumptions
How does an ionization model work?
Scheme
@@R
optional
?
1 5 10 50 100 500
1
10
100
1000
104
proton energy @MeVD
part
icle
flux
@Hm2 s
srM
eVL-
1 D
spectrum fit northern polar cap: 2006 doy 305 16 -18 h
Power Law
Maxwell
kind of fit function
@@R
?
pitch angle assumption
��
ASSUMPTIONS AND LIMITATIONS IN MODELING ATMOSPHERICIONIZATION BY PRECIPITATING PARTICLES
WHICH PARTS OF IONIZATION MODELS AFFECTED BY UNCERTAINTIES? AND HOW MUCH?
ASSUMPTIONS AND LIMITATIONS INMODELING ATMOSPHERIC IONIZATIONBY PRECIPITATING PARTICLESJAN MAIK WISSING AND MAY-BRITT KALLENRODE
MOTIVATIONWe model atmospheric ionization by parti-cles because they impact chemical processesand change the electric conductivity.But are ionization models reliable? Whichassumtions are used and are they realistic?Which parts of the ionization modeling stillneed improvements?
IONIZATION MODEL SCHEME
R
optional?
1 5 10 50 100 500
1
10
100
1000
104
proton energy @MeVD
parti
cle
flux
@Hm2 s
srM
eVL-
1 D
spectrum fit northern polar cap: 2006 doy 305 16 -18 h
Power Law
Maxwell
kind of fit function
R
?pitch angle assumption
An ionization model consists of twobranches, left side: particle measurementand spectrum fit, and right side: simulationof ionization by single particles and theassumtion of a pitch angle distribution.Estimated errors will now be discussed.
DETECTOR DEGENERATIONImpinging particles cause structure defects,that effectively shift in the channel’s energythresholds to higher energies.
1.0 1.2 1.4 1.6 1.8 2.0 2.2
2
4
6
8
10
12
N15N16�N17N18
2
4
6
8
10
12
Α factor Henergy shift in detector thresholdsL
ratio
corr
ecte
d�unc
orre
cted
flux
Α factors for for mid 2011 from Asikainen et al. 2012Detector degeneration in mep0P1 channels
Γ=-1
Γ=-2
Γ=-3
Γ=-4
steepness ofthe spectra
PARTICLE DATADifferent satellites = different ionization
10 20 50 100 200 500
1000
50001´104
5´1041´105
5´1051´106
proton energy @MeVD
parti
cle
flux
@Hm2 s
srM
eVL-
1 D
spectrum fit northern polar cap: 2003 doy 303 8-10 h
GOES 11 : 3
GOES 10 : 3
satellite : number of fits
107 108 109
104
103
102
101
100
10-1
10-2
10-3
10-4
16
31
49
66
81
94
108
144
233
ionization rate @m-3s-1D
pres
sure
@PaD
altit
ude
appr
ox.@k
mD
ionization northern polar cap: 2006 doy 303 8-10 h
GOES 11 : 3
GOES 10 : 3
range : number of fits
Measurements may differ because of
• initial characteristics (construction)• degradation, crosstalk• local flux variations
REFERENCES
[Asikainen et al.] Asikainen,T. , K. Mursula , V. Ma-liniemi, Correction of detector noise and recalibration ofNOAA/MEPED energetic proton fluxes, JGR 2012, inpress
[Yando et al.] K. Yando, R.M. Millan, J.C. Green andD.S. Evans, A Monte Carlo simulation of the NOAA POESMedium Energy Proton and Electron Detector instrument,JGR, 116, A10231, 2011
[AIMOS] Atmospheric Ionization ModuleOSnabrück
CROSSTALK (POES MEPED)Contributions by out of view particles:<10% in the isotropic case. However, above2 MeV the telescopes become increasinglysusceptible to out-of-field electrons. Andflux is not necessarily isotropic (e.g. radia-tion belts).
Crosstalk by different particle energiesProton detectors:
our simulation results by comparison with known char-acteristics of the MEPED instrument response, we concludethat the generally excellent agreement validates the modelgeometries and simulation physics.
4.1. Proton Telescope Response
[28] Figure 3 illustrates energy deposition by protons (red,orange) and electrons (blue) as a function of incidentenergy. In the range of 20 to 4900 keV, protons exhibit aneffectively linear mapping between energy incident and totalenergy deposited. This linearity corresponds to completeabsorption of the incident particle in the detector silicon.Qualitatively, this behavior approaches the ideal of a one‐to‐one mapping between incident energy and energydeposited desirable in a particle detector. Above ∼4.9 MeV,it is found that protons may penetrate the first detector(“D1”; deposits mapped in red) to deposit energy in thesecond detector (“D2”; deposits mapped in orange). Toenergies of 7 MeV, however, the sum of these depositsremains linear. Above this energy, protons pass throughboth front and back detectors and are not stopped; suchprotons do not deposit the entirety of their energy, and theirmapping is not linear.
[29] It is apparent that electrons (blue) show a morecomplicated behavior. For energies ≥500 keV, their pres-ence becomes increasingly obvious, yet electrons do notexhibit the linear energy deposition characteristic of protons.An electron’s tendency to undergo range straggling meansthat it will typically deposit only a fraction of its total energyin a thin detector. This fraction of energy will in general beinconsistent, and characterization must be done on the basisof averages. Thus quantifying the true extent of electroncontamination in the proton telescope is done throughgeometric factor, as previously introduced in Section 3.4.[30] The concept is illustrated for the proton telescope in
Figure 4: shown in Figure 4a are two stepwise functionswhich correspond to the total number of simulated hits(events depositing energy in detector silicon) per energy bin.Bins may be of arbitrary fixed width, and herein correspondto one‐quarter minor tick interval (e.g., 36 bins per ordermagnitude). The “Counts (n)” scale (at far right) is equivalentto geometric factor when properly normalized, and a measureof the 95% confidence interval is superposed on each stepfunction. These functions represent the “raw” response of theproton instrument to protons (at G ∼ 10−2 cm2 sr) and elec-trons (with G rising from 10−5 cm2 sr). In Figures 4b and 4c,the proton and electron responses receive treatment by energy
Figure 4. Geometric factor in the MEPED proton telescope for simulated protons and electrons.(a) Stepwise functions corresponding to the instrument’s raw geometric factor (all hits), before applicationof energy level logic. Per‐channel geometric factor for incident (b) protons and (c) electrons; channels P1through P6 in color.
YANDO ET AL.: TECHNIQUE A10231A10231
6 of 13
1
• original detector condition: negligible crosstalk• degraded detector: significant energy shift as already shown
Electron detectors:
level logic, and are broken into the 6 channels of proton dataP1–P6. The nominal 10−2 cm2 sr geometric factor of theproton telescope is overlaid for reference (dashed line). Notethat the sum of these six per‐channel responses need not equalthe instrument’s raw response–events depositing less than21.4 keV in the front detector (corresponding to the firstenergy threshold) are excluded from consideration by thelogic circuitry and do not appear in any of the instrument’s sixcount channels.[31] Considering the proton response (Figure 4b), one
observes clean and well‐defined channel separation, witheach channel closely corresponding to its nominal energyrange. The geometric factor is near the expected 0.01 cm2 srthroughout the interval 30 keV–10 MeV. In contrast, theelectron response (Figure 4c) exhibits significant overlapbetween the six energy channels. An electron of givenenergy may fall into any of several channels, and we say theresponse demonstrates “competition” between the channels.[32] The P4 and P5 energy channels appear uncontaminated
by incident electrons, with the P4 channel recording few hits inthe energies considered, and the P5 channel recording none atall. The P1, P2, P3, and P6 energy channels, in contrast, eachdisplay some sensitivity to incident electron fluxes. As can beseen from both the raw electron response in Figure 4a and its
per‐channel equivalent in Figure 4c, the proton telescopeexhibits only a slight sensitivity (<10% of the proton response)to electrons of energy≤450 keV,which is a feature attributed tothe effectiveness of the cross‐aperture magnetic field insweeping aside electrons at these energies. The effectiveness ofthis broom magnet decreases with increasing electron energy,and parity with the proton response is reached near 1MeV. Forintermediate energies, it is the case that electrons have reducedaccess to the instrument.[33] Below 100 keV, any electron hits are recorded in the
P1 energy channel. For 150–500 keV, the bulk of thiselectron contamination is confined to the P2 energy channel.Above 500 keV, incident electrons are found in the P6, P2,P3 and P1 channels with increasing probability, such thatthe P6 channel eventually comes to record the majority ofelectron hits. The P3 channel fails to ever capture a pluralityof the total count.[34] Total admittance in all channels reaches G ∼ 10−3 cm2
sr near 460 keV and G ∼ 10−2 cm2 sr by 1400 keV, and thetelescope’s P6 channel exhibits a sensitivity ofG ∼ 1.9 × 10−3
cm2 sr at 800 keV, and G ∼ 0.009 cm2 sr near 2000 keV.Thus anomalously high particle counts in the P6 channel(relative to the P4 and P5 energy channels) suggest contam-
Figure 5. Geometric factor in the MEPED electron telescope for simulated protons and electrons.(a) Stepwise functions corresponding to the instrument’s raw geometric factor (all hits), before applicationof energy level logic. Per‐channel geometric factor for incident (b) protons and (c) electrons; channels E1through E3 in color.
YANDO ET AL.: TECHNIQUE A10231A10231
7 of 13
1
• no significant crosstalk is known so far
Crosstalk by different particle speciesProton detectors affected by electrons:
our simulation results by comparison with known char-acteristics of the MEPED instrument response, we concludethat the generally excellent agreement validates the modelgeometries and simulation physics.
4.1. Proton Telescope Response
[28] Figure 3 illustrates energy deposition by protons (red,orange) and electrons (blue) as a function of incidentenergy. In the range of 20 to 4900 keV, protons exhibit aneffectively linear mapping between energy incident and totalenergy deposited. This linearity corresponds to completeabsorption of the incident particle in the detector silicon.Qualitatively, this behavior approaches the ideal of a one‐to‐one mapping between incident energy and energydeposited desirable in a particle detector. Above ∼4.9 MeV,it is found that protons may penetrate the first detector(“D1”; deposits mapped in red) to deposit energy in thesecond detector (“D2”; deposits mapped in orange). Toenergies of 7 MeV, however, the sum of these depositsremains linear. Above this energy, protons pass throughboth front and back detectors and are not stopped; suchprotons do not deposit the entirety of their energy, and theirmapping is not linear.
[29] It is apparent that electrons (blue) show a morecomplicated behavior. For energies ≥500 keV, their pres-ence becomes increasingly obvious, yet electrons do notexhibit the linear energy deposition characteristic of protons.An electron’s tendency to undergo range straggling meansthat it will typically deposit only a fraction of its total energyin a thin detector. This fraction of energy will in general beinconsistent, and characterization must be done on the basisof averages. Thus quantifying the true extent of electroncontamination in the proton telescope is done throughgeometric factor, as previously introduced in Section 3.4.[30] The concept is illustrated for the proton telescope in
Figure 4: shown in Figure 4a are two stepwise functionswhich correspond to the total number of simulated hits(events depositing energy in detector silicon) per energy bin.Bins may be of arbitrary fixed width, and herein correspondto one‐quarter minor tick interval (e.g., 36 bins per ordermagnitude). The “Counts (n)” scale (at far right) is equivalentto geometric factor when properly normalized, and a measureof the 95% confidence interval is superposed on each stepfunction. These functions represent the “raw” response of theproton instrument to protons (at G ∼ 10−2 cm2 sr) and elec-trons (with G rising from 10−5 cm2 sr). In Figures 4b and 4c,the proton and electron responses receive treatment by energy
Figure 4. Geometric factor in the MEPED proton telescope for simulated protons and electrons.(a) Stepwise functions corresponding to the instrument’s raw geometric factor (all hits), before applicationof energy level logic. Per‐channel geometric factor for incident (b) protons and (c) electrons; channels P1through P6 in color.
YANDO ET AL.: TECHNIQUE A10231A10231
6 of 13
1
• close to no contamination (P4 and P5)• potentially contaminated (P1,P2 and P3)• channels that can be used as reliable indicator for high-energy electron fluxes
(P6)
Electron detectors affected by protons:
level logic, and are broken into the 6 channels of proton dataP1–P6. The nominal 10−2 cm2 sr geometric factor of theproton telescope is overlaid for reference (dashed line). Notethat the sum of these six per‐channel responses need not equalthe instrument’s raw response–events depositing less than21.4 keV in the front detector (corresponding to the firstenergy threshold) are excluded from consideration by thelogic circuitry and do not appear in any of the instrument’s sixcount channels.[31] Considering the proton response (Figure 4b), one
observes clean and well‐defined channel separation, witheach channel closely corresponding to its nominal energyrange. The geometric factor is near the expected 0.01 cm2 srthroughout the interval 30 keV–10 MeV. In contrast, theelectron response (Figure 4c) exhibits significant overlapbetween the six energy channels. An electron of givenenergy may fall into any of several channels, and we say theresponse demonstrates “competition” between the channels.[32] The P4 and P5 energy channels appear uncontaminated
by incident electrons, with the P4 channel recording few hits inthe energies considered, and the P5 channel recording none atall. The P1, P2, P3, and P6 energy channels, in contrast, eachdisplay some sensitivity to incident electron fluxes. As can beseen from both the raw electron response in Figure 4a and its
per‐channel equivalent in Figure 4c, the proton telescopeexhibits only a slight sensitivity (<10% of the proton response)to electrons of energy≤450 keV,which is a feature attributed tothe effectiveness of the cross‐aperture magnetic field insweeping aside electrons at these energies. The effectiveness ofthis broom magnet decreases with increasing electron energy,and parity with the proton response is reached near 1MeV. Forintermediate energies, it is the case that electrons have reducedaccess to the instrument.[33] Below 100 keV, any electron hits are recorded in the
P1 energy channel. For 150–500 keV, the bulk of thiselectron contamination is confined to the P2 energy channel.Above 500 keV, incident electrons are found in the P6, P2,P3 and P1 channels with increasing probability, such thatthe P6 channel eventually comes to record the majority ofelectron hits. The P3 channel fails to ever capture a pluralityof the total count.[34] Total admittance in all channels reaches G ∼ 10−3 cm2
sr near 460 keV and G ∼ 10−2 cm2 sr by 1400 keV, and thetelescope’s P6 channel exhibits a sensitivity ofG ∼ 1.9 × 10−3
cm2 sr at 800 keV, and G ∼ 0.009 cm2 sr near 2000 keV.Thus anomalously high particle counts in the P6 channel(relative to the P4 and P5 energy channels) suggest contam-
Figure 5. Geometric factor in the MEPED electron telescope for simulated protons and electrons.(a) Stepwise functions corresponding to the instrument’s raw geometric factor (all hits), before applicationof energy level logic. Per‐channel geometric factor for incident (b) protons and (c) electrons; channels E1through E3 in color.
YANDO ET AL.: TECHNIQUE A10231A10231
7 of 13
1
• contamination is confined to protons of energies 200–2700keV
(Figures from [Yando et al.])
DIFFERENT ENERGY RANGE
0.01 1 100
1
100
104
106
108
1010
proton energy @MeVD
parti
cle
flux
@Hm2 s
srM
eVL-
1 D
spectrum fit northern polar cap: 2006 doy 344 14 -16 h
800keV-500MeV : 2
80keV-500MeV : 3
150eV-500MeV : 5
range : number of fits
104 106 108
104
102
100
10-2
10-4
16
31
49
66
81
94
108
144
233
ionization rate @m-3s-1D
pres
sure
@PaD
altit
ude
appr
ox.@k
mD
ionization northern polar cap: 2006 doy 344 14 -16 h
< 0.08 MeV
< 0.8 MeV
800keV-500MeV : 2
80keV-500MeV : 3
150eV-500MeV : 5
range : number of fits
Same quality of fit but still significant varia-tions just because of the energy range.
Nov
–Dez
2006
800keV-500MeV:2 vs. ext. r.
80keV-500MeV:3 vs. ext. r.720 profiles, Nov-Dec 2006
1.000.50 2.000.30 1.500.7020
40
60
80
100
ratio of ionization rates: range�Hextended rangeL
altit
ude
@kmD
ionization depending on model specific energy range
25
50
75
90
100
amou
ntof
com
paris
ons
arou
ndm
edia
n@%
D
FIXED/VARIABLE INTERSECT.
1 5 10 50 100 500
1
10
100
1000
104
105
106
proton energy @MeVD
parti
cle
flux
@Hm2 s
srM
eVL-
1 D
spectrum fit northern polar cap: 2006 doy 344 22 -24 h
fixed : 3
variable : 3
intersections : number of fits
104 106 108
104
102
100
10-2
10-4
16
31
49
66
81
94
108
144
233
ionization rate @m-3s-1D
pres
sure
@PaD
altit
ude
appr
ox.@k
mD
ionization northern polar cap: 2006 doy 344 22 -24 h
< 0.8 MeV
fixed : 3
variable : 3
intersections : number of fits
Nov
–Dez
2006
720 profiles, day 305 to 364, 2006
0.6 0.7 0.8 0.9 1 1.2 1.4 1.6 2
20
40
60
80
ratio of ionization rates: fixed�variable positions of intersections
altit
ude
@kmD
ionization depending on type of intersections
25
50
75
90
100
amou
ntof
com
paris
ons
arou
ndm
edia
n@%
D
kind of in-tersectionmatters
SUMMARY/CONCLUSIONSreason maximum expected error possible improvements
parti
clem
easu
rem
ents
degeneration, energy shift depending on age and steepness of spectra up to 1000% correction e.g. [Asikainen et al.], or newer satellitedegeneration, noise effects up to 1300% (NOAA-12, including energy shift) correction e.g. [Asikainen et al.], or newer satellitecrosstalk by out-of-field contrib. 10% contribution by out-of-view particles [Yando et al.] not neededcrosstalk by different energies apart from degradation: negligible not neededcrosstalk by different species depending on channel, partly severe different detector setup or coincidence correction neededsatellite selection up to 300% proved - more possible, (but >90% are within a
range of ±50%)-
spec
trum
fit
quality of fit (number of functions) orders of magnitude limited by channels and theoryselection of energy range ±20% (in the center of the range significantly less) -fixed or variable intersections up to 60% use variable intersectionsfit function 30%, 70% at end of energy range theory suggestes power-law
ioni
zatio
nsim
ulat
ion atmospheric conditions few % without tropopause not needed
kind of deposition algorithm some % not neededstatistics in e.g. Monte-Carlo <5% not neededtransformation to ion pair prod. <10% not needed
rem
aini
ngpr
oblem
s non-isotropic pitch angle distrib. -40% to +60% needs new detector setupSAA ? -regional resolution strongly depending on model, but without regional resolution
is not betterlimited: a) temporal resolution vs. spatial resolution, b) moresatellites needed
PITCH ANGLEThe pitch angle distribution is not mea-sured regularly. Most common assumptionis isotropic. What happens if this assump-tion is wrong?
• Ionization will be shifted in altitude.• Particle measurement not reliable due
to contamination by out-of-view parti-cles.
0.01 0.05 0.10 0.50 1.00 5.0010.00
105
110
120
140
170
200
250
10-5
10-4
10-3
10-2
energy deposition @eV�Hm particleLD
altit
ude
@kmD
pres
sure
@PaD
monoenergetic proton incident: 141 keV
max 0°Hclose to verticalL
max 90°Hbig pitch anglesL
isotropic
QUALITY OF FITThe amount of fit functions per magnitudedefines quality of fit.
0.01 1 100
100
104
106
108
1010
1012
proton energy @MeVD
parti
cle
flux
@Hm2 s
srM
eVL-
1 D
spectrum fit northern polar cap: 2006 doy 342 4-6h
150eV-500MeV : 1
150eV-500MeV : 3
150eV-500MeV : 5
range : number of fits
107 108 109
104
103
102
101
100
10-1
10-2
10-3
10-4
16
31
49
66
81
94
108
144
233
ionization rate @m-3s-1D
pres
sure
@PaD
altit
ude
appr
ox.@k
mD
ionization northern polar cap: 2006 doy 342 4-6h
150eV-500MeV : 1
150eV-500MeV : 3
150eV-500MeV : 5
range : number of fits
PARTICLE MEASUREMENTS
• degeneration, energy shift• degeneration, noise effects• crosstalk by out-of-field contributions• crosstalk by different energies• crosstalk by different species• satellite selection
REMAINING PROBLEMS
• non-isotropic pitch angle distribution• SAA• regional resolution
SPECTRUM FIT
• quality of fit (number of functions)• selection of energy range• fixed or variable intersections• kind of fit function
MODELING IONIZATION
• atmospheric conditions• kind of deposition algorithm• statistics in e.g. Monte-Carlo• transformation to ion pair production
Particle measurements
Particle measurements
Jan Maik Wissing (University of Osnabruck) Modeling Particle Precipitation November 21, 2013 6 / 20
Detector degeneration - energy shift
Reasonimpinging particlescause structuredefects
Effect
shifts channel’senergy thresholds tohigher energies
1.0 1.2 1.4 1.6 1.8 2.0 2.2
2
4
6
8
10
12
N15N16�N17N18
2
4
6
8
10
12
Α factor Henergy shift in detector thresholdsL
rati
oco
rrec
ted
�un
corr
ecte
dfl
ux
Α factors for for mid 2011 from Asikainen et al. 2012Detector degeneration in mep0P1 channels
Γ=-1
Γ=-2
Γ=-3
Γ=-4
steepness ofthe spectra
SolutionUsing new satellites or applying correction factors[Asikainen et al., 2012] may improve data.
Crosstalk
e.g. in MEPED electron detectors [Yando et al., 2011]
Electron sensitivity:
level logic, and are broken into the 6 channels of proton dataP1–P6. The nominal 10−2 cm2 sr geometric factor of theproton telescope is overlaid for reference (dashed line). Notethat the sum of these six per‐channel responses need not equalthe instrument’s raw response–events depositing less than21.4 keV in the front detector (corresponding to the firstenergy threshold) are excluded from consideration by thelogic circuitry and do not appear in any of the instrument’s sixcount channels.[31] Considering the proton response (Figure 4b), one
observes clean and well‐defined channel separation, witheach channel closely corresponding to its nominal energyrange. The geometric factor is near the expected 0.01 cm2 srthroughout the interval 30 keV–10 MeV. In contrast, theelectron response (Figure 4c) exhibits significant overlapbetween the six energy channels. An electron of givenenergy may fall into any of several channels, and we say theresponse demonstrates “competition” between the channels.[32] The P4 and P5 energy channels appear uncontaminated
by incident electrons, with the P4 channel recording few hits inthe energies considered, and the P5 channel recording none atall. The P1, P2, P3, and P6 energy channels, in contrast, eachdisplay some sensitivity to incident electron fluxes. As can beseen from both the raw electron response in Figure 4a and its
per‐channel equivalent in Figure 4c, the proton telescopeexhibits only a slight sensitivity (<10% of the proton response)to electrons of energy≤450 keV,which is a feature attributed tothe effectiveness of the cross‐aperture magnetic field insweeping aside electrons at these energies. The effectiveness ofthis broom magnet decreases with increasing electron energy,and parity with the proton response is reached near 1MeV. Forintermediate energies, it is the case that electrons have reducedaccess to the instrument.[33] Below 100 keV, any electron hits are recorded in the
P1 energy channel. For 150–500 keV, the bulk of thiselectron contamination is confined to the P2 energy channel.Above 500 keV, incident electrons are found in the P6, P2,P3 and P1 channels with increasing probability, such thatthe P6 channel eventually comes to record the majority ofelectron hits. The P3 channel fails to ever capture a pluralityof the total count.[34] Total admittance in all channels reaches G ∼ 10−3 cm2
sr near 460 keV and G ∼ 10−2 cm2 sr by 1400 keV, and thetelescope’s P6 channel exhibits a sensitivity ofG ∼ 1.9 × 10−3
cm2 sr at 800 keV, and G ∼ 0.009 cm2 sr near 2000 keV.Thus anomalously high particle counts in the P6 channel(relative to the P4 and P5 energy channels) suggest contam-
Figure 5. Geometric factor in the MEPED electron telescope for simulated protons and electrons.(a) Stepwise functions corresponding to the instrument’s raw geometric factor (all hits), before applicationof energy level logic. Per‐channel geometric factor for incident (b) protons and (c) electrons; channels E1through E3 in color.
YANDO ET AL.: TECHNIQUE A10231A10231
7 of 13
1Sensitivity for proton contamination:
level logic, and are broken into the 6 channels of proton dataP1–P6. The nominal 10−2 cm2 sr geometric factor of theproton telescope is overlaid for reference (dashed line). Notethat the sum of these six per‐channel responses need not equalthe instrument’s raw response–events depositing less than21.4 keV in the front detector (corresponding to the firstenergy threshold) are excluded from consideration by thelogic circuitry and do not appear in any of the instrument’s sixcount channels.[31] Considering the proton response (Figure 4b), one
observes clean and well‐defined channel separation, witheach channel closely corresponding to its nominal energyrange. The geometric factor is near the expected 0.01 cm2 srthroughout the interval 30 keV–10 MeV. In contrast, theelectron response (Figure 4c) exhibits significant overlapbetween the six energy channels. An electron of givenenergy may fall into any of several channels, and we say theresponse demonstrates “competition” between the channels.[32] The P4 and P5 energy channels appear uncontaminated
by incident electrons, with the P4 channel recording few hits inthe energies considered, and the P5 channel recording none atall. The P1, P2, P3, and P6 energy channels, in contrast, eachdisplay some sensitivity to incident electron fluxes. As can beseen from both the raw electron response in Figure 4a and its
per‐channel equivalent in Figure 4c, the proton telescopeexhibits only a slight sensitivity (<10% of the proton response)to electrons of energy≤450 keV,which is a feature attributed tothe effectiveness of the cross‐aperture magnetic field insweeping aside electrons at these energies. The effectiveness ofthis broom magnet decreases with increasing electron energy,and parity with the proton response is reached near 1MeV. Forintermediate energies, it is the case that electrons have reducedaccess to the instrument.[33] Below 100 keV, any electron hits are recorded in the
P1 energy channel. For 150–500 keV, the bulk of thiselectron contamination is confined to the P2 energy channel.Above 500 keV, incident electrons are found in the P6, P2,P3 and P1 channels with increasing probability, such thatthe P6 channel eventually comes to record the majority ofelectron hits. The P3 channel fails to ever capture a pluralityof the total count.[34] Total admittance in all channels reaches G ∼ 10−3 cm2
sr near 460 keV and G ∼ 10−2 cm2 sr by 1400 keV, and thetelescope’s P6 channel exhibits a sensitivity ofG ∼ 1.9 × 10−3
cm2 sr at 800 keV, and G ∼ 0.009 cm2 sr near 2000 keV.Thus anomalously high particle counts in the P6 channel(relative to the P4 and P5 energy channels) suggest contam-
Figure 5. Geometric factor in the MEPED electron telescope for simulated protons and electrons.(a) Stepwise functions corresponding to the instrument’s raw geometric factor (all hits), before applicationof energy level logic. Per‐channel geometric factor for incident (b) protons and (c) electrons; channels E1through E3 in color.
YANDO ET AL.: TECHNIQUE A10231A10231
7 of 13
1
contamination is confined to protons of energies 200–2700keV
Particle measurements Crosstalk
Effect of crosstalk on quite day ionization profile (electrons)
mep0e3
104 106 108 1010
104
102
100
10-2
10-4
0
16
31
49
66
81
94
108
144
233
444
ionisation rate @m-3s-1D
Pre
ssu
re@Pa
D
alti
tud
eap
pro
x.@km
D
proton and electron HdashedL ionization at geomagn. pole
quiet timeAIMOS v1.1
275:2h
270:2h
265:2h
250:2h
time
SOHO EPHIN
104 106 108 1010
104
102
100
10-2
10-4
0
16
31
49
66
81
94
108
144
233
444
ionisation rate @m-3s-1D
Pre
ssu
re@Pa
D
alti
tud
eap
pro
x.@km
D
proton and electron HdashedL ionization at geomagn. pole
quiet timeAIMOS v1.3
275:2h
270:2h
265:2h
250:2h
time
SolutionProbably solved/improved by new correction algorithms. [Janet Green]
Jan Maik Wissing (University of Osnabruck) Modeling Particle Precipitation November 21, 2013 9 / 20
Variation among (same) satellites
306 307 308 309 310 311
day of year
10-2
10-1
100
101
102
flux@i
onsHc
m2
ssr
MeVL-
1D GOES10 vs. GOES11 HdashedL
p+ 40-80
p+ 15-40
p+ 9-15
p+ 4-9
Α 21-60
Α 10-21
Α 4-10
energy @MeVD
measurements may differ because of
different initial characteristics (construction)degradation, crosstalklocal flux variations
Variation among satellites
672 profiles, day 280 to 335, 2003
0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.420
30
40
50
60
70
80
ratio of ionization rates: GOES10�GOES11
altit
ude
@kmD
ionization depending on satellite
10-3 10-2 10-1 1 101 102 10320
30
40
50
60
70
80
25
50
75
90
100
amou
ntof
com
paris
ons
arou
ndm
edia
n@%
D
SolutionNone
Spectrum fit
Spectrum fit
1 5 10 50 100 500
1
10
100
1000
104
proton energy @MeVD
part
icle
flux
@Hm2 s
srM
eVL-
1 D
spectrum fit northern polar cap: 2006 doy 305 16 -18 h
Power Law
Maxwell
kind of fit function
Jan Maik Wissing (University of Osnabruck) Modeling Particle Precipitation November 21, 2013 12 / 20
Quality of fit (number of functions)
0.01 1 100
100
104
106
108
1010
1012
proton energy @MeVD
part
icle
flux
@Hm2 s
srM
eVL-
1 D
spectrum fit northern polar cap: 2006 doy 342 4-6h
150eV-500MeV : 1
150eV-500MeV : 3
150eV-500MeV : 5
range : number of fits
107 108 109
104
103
102
101
100
10-1
10-2
10-3
10-4
16
31
49
66
81
94
108
144
233
ionization rate @m-3s-1Dpr
essu
re@P
aD
altit
ude
appr
ox.@
kmD
ionization northern polar cap: 2006 doy 342 4-6h
150eV-500MeV : 1
150eV-500MeV : 3
150eV-500MeV : 5
range : number of fits
Different energy range of the model
History
1970 1980 1990 2000
1
100
104
106
108
year
ener
gyra
nge
ofm
odel
Halls
peci
esL@
eVD
Berger
Hardy
Callis
JackmanSchröter
Wissing
Different energy range of the model
0.01 1 100
1
100
104
106
108
1010
proton energy @MeVD
part
icle
flux
@Hm2 s
srM
eVL-
1 D
spectrum fit northern polar cap: 2006 doy 344 14 -16 h
800keV-500MeV : 2
80keV-500MeV : 3
150eV-500MeV : 5
range : number of fits
104 106 108
104
102
100
10-2
10-4
16
31
49
66
81
94
108
144
233
ionization rate @m-3s-1D
pres
sure
@PaD
altit
ude
appr
ox.@
kmD
ionization northern polar cap: 2006 doy 344 14 -16 h
< 0.08 MeV
< 0.8 MeV
800keV-500MeV : 2
80keV-500MeV : 3
150eV-500MeV : 5
range : number of fits
The ratio of magnitudes in energy range divided by the number of fitfunction here is the same.
Different energy range - results for Nov–Dec 2006
800keV-500MeV:2 vs. ext. r.
80keV-500MeV:3 vs. ext. r.720 profiles, Nov-Dec 2006
1.000.50 2.000.30 1.500.7020
40
60
80
100
ratio of ionization rates: range�Hextended rangeL
altit
ude
@kmD
ionization depending on model specific energy range
25
50
75
90
100
amou
ntof
com
paris
ons
arou
ndm
edia
n@%
D
SolutionNone
Summary of the uncertainties in modeling particleprecipitation
reason maximum expected error possible improvements
degeneration, energy shift depending on age and steepness ofspectra up to 1000%
correction e.g. [Asikainen et al., 2012],or newer satellite
degeneration, noise effects up to 1300% (NOAA-12, including en-ergy shift)
correction e.g. [Asikainen et al., 2012],or newer satellite
crosstalk by out-of-field contrib. 10% contribution by out-of-view parti-cles [Yando et al., 2011]
not needed
crosstalk by different energies apart from degradation: negligible not neededcrosstalk by different species depending on channel, partly severe different detector setup or coincidence
correction neededsatellite selection up to 300% proved - more possi-
ble, (but >90% are within a range of±50%)
-
quality of fit (number of functions) orders of magnitude limited by channels and theoryselection of energy range ±20% (in the center of the range sig-
nificantly less)-
fixed or variable intersections up to 60% use variable intersectionsfit function 30%, 70% at end of energy range theory suggestes power-law
atmospheric conditions few % without tropopause not neededkind of deposition algorithm some % not neededstatistics in e.g. Monte-Carlo <5% not neededtransformation to ion pair prod. <10% not needed
non-isotropic pitch angle distrib. -40% to +60% needs new detector setupSAA ? proton crosstalk in e-channels may be
correctedregional resolution strongly depending on model, but with-
out regional resolution is not betterlimited: a) temporal resolution vs.spatial resolution, b) more satellitesneeded
Regarding these uncertainties, should particle precipitationmodels be used?
AIMOS
HAMMONIA
electron density
IS radar observation
electron density
ionization rate ��
��
fit to theoretical model
��ks +3
1312 EISCAT profilesOct 2003 – April 2004event & quiet
Yes!
[Wissing et al., 2011]
10-4 10-3 10-2 10-1 100
electron density: Hsimulation without particle ionizationL�measurement
100
120
140
160
180
200
altitude@kmD
dependence on local time
25
50
75
amountofcomparisonsaroundmedian@%D
local day: 6h-18h
177 profiles
local night: 18h-6h
135 profiles
10-4 10-3 10-2 10-1 100
electron density: Hsimulation including particle ionizationL�measurement
100
120
140
160
180
200
altitude@kmD
dependence on local time
25
50
75
amountofcomparisonsaroundmedian@%D
local day: 6h-18h
177 profiles
local night: 18h-6h
135 profiles
1
Outlook: AIMOS Update v1.7
removed proton contamination
work in MLT rather than LT
reduced impact of spikes
improved spatial resolution inmain precipitation regions
current status: verification againstAARDDVARK
http://aimos.physik.uos.de
104 106 108 1010
ionisation rate @m-3s-1D
104
102
100
10-2
10-4
Pre
ssu
re@P
aD
0
16
31
49
66
81
94
108
144
233
444
altitu
de
ap
pro
x.@k
mD
proton and electron HdashedL ionization at geomagn. pole
October Event 2003AIMOS v1.1
302:12h
302:4h
301:12h
301:10h
time
[Wissing and Kallenrode, 2009]
[Wissing et al., 2010]
Agostinelli, S. et al., GEANT4-a simulation toolkit, Nucl. Instr. Meth. 506, p. 250–303, 2003
Asikainen,T. , K. Mursula , V. Maliniemi, Correction of detector noise and recalibration of NOAA/MEPED energetic protonfluxes, submitted 2012 to JGR
Berger, M.J., S.M. Seltzer and K. Maeda, Energy deposition by auroral electrons in the atmosphere, J. Atm. Solar-Terr.Phys. 32, p. 1015–1045, 1970
Callis, L.B., Odd nitrogen formed by energetic electron precipitation as calculated from TIROS data, Geophys. Res.Letters, vol 24, No 24, p. 3237–3240, 1997
Callis, L.B., M. Natarajan, D. S. Evans and J. D, Lambeth, Solar atmospheric coupling by electrons (SOLACE) 1. Effects ofthe May 12, 1997 solar event on the middle atmosphere, J. Geophys. Res. 103, p. 28 405–28 419, 1998
Callis, L.B. and J. D, Lambeth, NOy formed by precipitating electron events in 1991 and 1992: Descent into thestratosphere as observed by ISAMS, Geophys. Res. Let., Vol. 25, No. 11, p. 1875–1878, 1998.
Fang, X., M.W. Liemohn, J.U. Kozyra, D.S. Evans, A.D. DeJong and B.A. Emery, Global 30-240 keV proton precipitation inthe 17-18 April 2002 geomagnetic storms: 1. Patterns, J. Geophys. Res., 112, A05301, doi:10.1029/2006JA011867, 2007
Funke, B., A. Baumgaertner, M. Calisto, T. Egorova, C.H. Jackman, J. Kieser, A. Krivolutsky, M. Lpez-Puertas, D.R. Marsh,T. Reddmann, E. Rozanov, S.-M. Salmi, M. Sinnhuber, G. P. Stiller, P. T. Verronen, S. Versick, T. von Clarmann, T. Y.Vyushkova, N. Wieters and J. M. Wissing, Composition changes after the ”Halloween” solar proton event: the High-EnergyParticle Precipitation in the Atmosphere (HEPPA) model versus MIPAS data inter-comparison study, AtmosphericChemistry and Physics, 11, p. 9089-9139, acp-2010-1001, 2011
Hardy, D.A., M.S. Gussenhoven and D. Brautigam, A statistical model of the auroral ion precipitation, J. Geophys. Res.,94, p. 370, 1989
Jackman, C.H., R.D. McPeters, G.J. Labow, E.L. Fleming, C.J. Praderas and J.M. Russell, Northern hemisphereatmospheric effects due to the July 2000 Solar Proton Event, Geophys. Res. Lett., 28(15), p. 2883–2886, 2001
Jackman, C.H., M.T. DeLand, G.J. Labow, E.L. Fleming, D.K. Weisenstein, M.K.W. Ko, M. Sinnhuber, and J.M. Russell,Neutral atmospheric influences of the solar proton events in October–November 2003, J. Geophys. Res., 110, A09S27,doi:10.1029/2004JA010888, 2005a
Roble, R.G. and E.C. Ridley, An auroral model for the NCAR thermospheric general circulation model (TGCM), Ann.Geophys., 5A, p. 369–382, 1987
Schroter J., B. Heber, F. Steinhilber and M.-B. Kallenrode, Energetic particles in the atmosphere: a Monte Carlo approach,Adv. Space Res. 37, (8), p. 1597–1601, 2006.
Wissing, J.M., J.P. Bornebusch and M.-B. Kallenrode Adv. Space Res. 41, p. 1274–1278, 2008
Wissing, J.M. and M.-B. Kallenrode, Atmospheric Ionization Module OSnabruck (AIMOS) 1: A 3-D model to determineatmospheric ionization by energetic charged particles from different populations, J. Geophys. Res., 114, A06104,doi:10.1029/2008JA013884, 2009
Wissing, J.M., M.-B. Kallenrode, N. Wieters, H. Winkler and M. Sinnhuber, Atmospheric Ionization Module Osnabruck(AIMOS): 2. Total particle inventory in the October/November 2003 event and ozone, J. Geophys. Res., 115, A02308,doi:10.1029/2009JA014419, 2010
Wissing, J.M., M.-B. Kallenrode, J. Kieser, H. Schmidt, M.T. Rietveld, A. Strømme and P.J. Erickson, AtmosphericIonization Module OSnabruck (AIMOS) 3: Comparison of electron density simulations by AIMOS/HAMMONIA andincoherent scatter radar measurements, J. Geophys. Res., 116, A08305, doi:10.1029/2010JA01630, 2011
K. Yando, R.M. Millan, J.C. Green and D.S. Evans: A Monte Carlo simulation of the NOAA POES Medium Energy Protonand Electron Detector instrument, JGR, vol. 116, A10231, doi:10.1029/2011JA015571, 2011