Thermosphere Parameters Extracted from Incoherent Scatter Observations

1
Short-Term (1-24 h) foF2 Forecast
Present day State of Art
Andrei Mikhailov, Victor Depuev, Anna Depueva
IZMIRAN Russian Academy of Sciences
2
Disturbed F2-layer short-term forecast is still
unsolved and very challenging problem despite
long history and many attempts being undertaken.
This is due to objective reasons
3
Physical mechanisms forming both negative and
positive F2-layer disturbances are well
established by now
4
Mid-latitude F2-layer

• Negative daytime O/N2 decrease, Teff
  increase
• Negative nighttime O/N2 decrease, wind (Vnx)
  diurnal variations, plasmaspheric O flux
  decrease
• Positive daytime equatorward Vnx increase,
  absolute O increase
• Positive nighttime wind (Vnx) diurnal
  variations, plasmaspheric O flux increase,
  absolute O increase

5
High latitude and equatorial F2-layer

• Auroral zone
• Negative disturbances mainly O/N2 decrease,
  magnetospheric convection E field and Tn increase
• (Teff increase), upward plasma outflow
• Positive disturbances mainly due to particle
  precipitation and horizontal plasma ExB transfer
• Equatorial zone
• Both Positive and Negative disturbances are
  mainly due to zonal Ey electric field (Eyx B)
  drift Vnx variations (low geomagnetic latitudes)

6
Main Approaches in the Ionosphere Forecast
Practice

• Theoretical
• (First principle 1D-3D models)
• Empirical
• (Statistical, Neural networks)
• Semi-Empirical
• (A combination of the two first)

7
Upper Atmosphere is an Open System with Many
Uncontrolled Inputs
Particles Precipitation
Solar EUV
Magnetospheric Electric Fields
Upper Atmosphere
Thermosphere Ionosphere
Internal Gravity Waves
Planetary Waves
Dynamo and Tropospheric Electric Fields
8
Depending on prehistory and current state of the
magnetosphere and thermosphere, the reaction will
be different to the same impact from above
however No thermosphere and magnetosphere
monitoring is made at present and is not expected
in an observable future
9
Thus the intensity of each particular process
controlling the F2-region magnetospheric
electric fields, zones and characteristics of
particle precipitation controlling Joule heating,
global thermospheric circulation resulting in
neutral composition and temperature variations,
is known pretty poor for each particular
geomagnetic storm
10
The impact from belowthe intensity of
gravity waves resulting in eddy diffusion in the
100-120 km height range which strongly controls
thermospheric neutral composition, planetary
waves, penetrating tropospheric electric
fieldsis not controlled at all
11
So at present there is not much hope to obtain a
deliberate short-term forecast of the F2-layer
parameters
12
Theoretical Approach

• A comparison by Fuller-Rowell et al. (2000) for
  disturbed conditions has demonstrated more
  visual success of the model predictions than
  quantitative correlation coefficients between 3D
  CTIM model and observations are typically
  0.3-0.65, depending on how the data are selected
  and smoothed. Negative F2-layer storm effects
  which are the most crucial for HF radio-wave
  communication cannot be satisfactory modelled
  without special fitting of aeronomic parameters
  for each particular ionospheric storm (e.g.
  Richards, et al., 1989,1994 Buonsanto, 1999).
• Theoretical modelling may be considered as a tool
  for physical analyses rather than practical
  applications

13
Empirical Approach Based On

• Statistical methods for the foF2 short-term
  prediction
• (Zevakina, 1990 Wu and Wilkinson, 1995 Muhtarov
  et al., 1998 Kutiev et al., 1999 Muhtarov and
  Kutiev, 1999 Marin et al., 2000 Kutiev and
  Muhtarov, 2001 Araujo-Pradere et al., 2002,
  2003 Tsagouri and Belehaki, 2005 Liu et al.,
  2005)
• Neural networks
• (Cander et al., 1998 Cander and Mihajlovic,
  1998 Francis et al., 2000, 2001 Wintoft and
  Cander, 2000 Chan and Cannon, 2002 McKinnell and
  Poole, 2004)
• In principle can provide an acceptable accuracy
  and so is widely used in practice

14
Problems on this way

• There is no an effective geophysical index to
  predict the ionospheric storm onset, its
  magnitude and duration.
• The correlation with currently available
  planetary indices is not very high.
• According to Short-Term Prediction Manual by
  Zevakina et.al.(1990) depending on latitude the
  correlation coefficient for ?foF2 are
• (0.86-0.52) with AE (0.71-0.46) with Dst
• (0.86-0.69) with Bz (0.77-0.33) with Kp

15
Problems on this way

• Time weighed accumulation indices such as ap(?)
  proposed by Wrenn, 1987, Wrenn et al., 1987 seem
  to increase the correlation with ?foF2, but the
  improvement is not significantly larger than for
  instantaneous indices (aa, ap, Kp, Dst).
  Correlation coefficients r lt 0.7.
• So time-weighted accumulation indices may have
  limited use in a forecasting environment (Wu and
  Wilkinson, 1995)

16
Problems on this way

• Next step was made by Araujo-Pradere,
  Fuller-Rowell and Codrescu who proposed a
  correction model STORM (2002) based on a new
  index - the integral of 3-hour ap index over the
  previous 33 hours weighted by a filter obtained
  by the method of singular value decomposition.
• ?foF2a0a1X(t0)a2X2(t0)a3X3(t0)
• where
• X(t0)?F(?)P(t0-?)d?, and F(?) is the filter
  weighting function of the ap index over the 33
  previous hours.
• STORM model is a part of IRI2000 now

17
Correlation ?foF2 with the IRI2000 index for
severe storms
18
So no miracle with geomagnetic activity indices
either direct or transformed!But there is no
much choice as

• 1. Only geomagnetic indices (aa, ap, kp) are
  available for the whole period of ionospheric
  observations - this is important for forecast
  methods development.
• 2. Only daily Ap is predicted currently 1-3 days
  in advance. Prediction of a controlling index is
  necessary for any forecast method functioning.

19
Additional problems with global geomagnetic
indices

• 1. During severe geomagnetic storms magnetometric
  stations are out of the auroral zone
  underestimating index values.
• 2. High latitude energy deposition (heating) is
  not uniform in longitude while global indices do
  not reflect this.
• 3. Ionospheric storm onset depends on LT, season
  and prehistory (state of the magnetosphere and
  thermosphere).
• Items 2,3 result in large scatter for delays
  between geomagnetic and ionospheric storm onsets.

20
Estimates of the time delay between geomagnetic
and ionospheric storm onsets

• 1. 0-6 h for positive disturbances (Zevakina and
  Kiseleva, 1978)
• 2. 12 h (Wrenn et al., 1987)
• 3. 15 h (Wu and Wilkinson, 1995)
• 4. 6-12 h (Forbes et al., 2000)
• 5. 16-18 h (Kutiev and Muhtarov, 2001)
• 6. 8-20 h depending on season (Pant and
  Sridharan, 2001)
• 7. 3-20 h depending on LT sector (Tsagouri and
  Belehaki, 2005)
• 8. No time delay is considered in IRI2000
  (Araujo-Pradere at al.,
• 2002)
• No global geomagnetic index can provide an
  efficient
• F2-layer forecast under such conditions

21
Despite all the problems the majority of the
ionosphere forecast methods are based on the
geomagnetic indices
22
Some estimates of an improvement achieved over
median prediction for storm conditions(a
statistical approach)

• 1. A 34 in the Northern and 20 in the Southern
• Hemispheres. The best results are for Summer
• (up to 50) and no improvement in Winter
• (STORM model, Araujo-Pradere et al., 2003)
• 2. A 29 gain over climatology
• (Kutiev and
  Muhtarov, 2001)
• 3. A 44 gain obtained over 15 impulse storm
  events
• (Tsagouri and
  Belehaki, 2005)

23
Some estimates of an improvement achieved over
median prediction using a neural networks
approach(no special data selection)

• 1. An up to 50 improvement for 1-hour ahead
• foF2 forecast (Chan and Cannon, 2002)
• 2. About 40-45 gain in foF2 RMS for noonday
• (Fransis et al., 2000)
• Severe storm condition cases study was made by
  Wintoft and Cander, 2000 and problems on this way
  were discussed.

24
After analysis in the whole the situation with
the empirical approach to the foF2 short-term
(1-24 h ahead) forecast, a method for practical
use has been developed and implemented at IZMIRAN

25
Main features of the Method

• The Method is designed to predict foF2 for
  various geophysical conditions. Negative storm
  effect as the most important for HF radio
  communication is the main concern.
• Input
• a) hourly foF2 for previous 28 (one solar
  rotation) days and current hourly foF2
  observations
• b) 3-hour ap index for previous 30 days current
  data daily Ap forecast for the next day.
• Output
• 24 foF2 forecasts per day with 1-24 h lead times
• (00-23 UT) for a given station (ionosonde
  location), so the forecast is renovated each
  hour.

26
Main features of the Method

• The Method is not designed to predict Positive
  and Quiet time F2-layer disturbances with lead
  time gt 2-6 hours as no reliable precursors are
  known.
• The forecast is completely automatic

27
The idea of the Method

• The regression is used
• ?foF2(UTn)C0C1? foF2(UT)C2 AI(UTn)
• where
• ?foF2foF2/foF2med, foF2med - running median over
  the 28-day training period
• AI - aeronomic index for the (UTn) moment
• n - lead time (1-24 h)

28
The Idea of the Method (Aeronomic Index AI)
29
The idea of the Method

• Unlike global direct solar and geomagnetic
  indices which exhibit only UT dependence, the
  proposed index AI, in principle, should
  demonstrate (via thermospheric parameters
  variations) the dependence on UT, LT, latitude
  and longitude, season, level of solar activity
  etc.

30
The idea of the Method

• The above mentioned method is use for quiet and
  moderately disturbed conditions.
• An approach is different for severe storm
  periods.
• Specially selected foF2 strong disturbances
  observed at a given station were used for ?foF2
  versus AI regressions for each month of the year.
  
• The thresholds for the ionospheric storm onset
  were specified for each months as well. When the
  threshold is exeeded, the method switches from
  usual mode to a corresponding regression.

31
Training and Testing the Method

• The Method was tested using all severe storms
  observed at Slough (Chilton) during 1949-2004.
• A comparison was also made with
• a) median forecast
• b) IRI2000 storm corrections
• c) empirical model by Shubin and Anakuliev (1995)

32
Summer (Chilton, 22 storm events)
33
Equinox (Chilton, 21 storm events)
34
Winter (Chilton, 21 storm events)
35
Results of Testing for Storm Events

• 1. The prediction accuracy (MRD) decreases and
  scatter (SDR) increases from Summer to Equinox
  and Winter.
• 2. MRD ranges from 6 to 24 depending on lead
  time and season. For quiet time and moderately
  disturbed conditions typical MRD?10-15 for all
  lead times.
• 3. Median forecast is the worst under all
  conditions.

36
Results of Testing for Storm Events

• 4. The IRI2000 and Shubins models provide less
  accurate forecast, but both models are not linked
  to any current foF2 observations and, in
  principle, can be used globally and this is a
  great merit of the two models.
• 5. Both models provide close results in summer
  and equinox, but the Shubins model is more
  efficient in winter. This is a very important
  result keeping in mind the IRI2000 problems for
  winter season when no improvement over median
  forecast can be demonstrated (Araujo-Pradere et
  al., 2002).

37
Visual comparisons for some storms events
38
June 4-6, 1991 Storm Event (Chilton)
39
April 6-8, 1973 Positive Q-disturbance Event
(St.Petersburg)
40
April 21-23, 1980 Negative Q-disturbance Event
(Moscow)
41
Conclusions

• 1. A deliberate high accuracy foF2 forecast is
  impossible at present due to objective reasons.
• 2. A statistical approach can provide an
  acceptable (MRD 6 -24) short-term (1-24 h)
  foF2 forecast for various geophysical conditions
  (including severe storm periods).
• 3. The IRI2000 storm time correction of median
  foF2 may be recommended for foF2 forecast where
  current ionospheric observations are absent.
  IRI2000 and Shubins models provide close
  prediction accuracy during summer and equinoxes
  while in winter the Shubins model is more
  efficient. Both models can be used globally as
  they are based on easy-accessible solar and
  geomagnetic indices.

42
Some Unsolved Problems (Empirical Approach)

• 1. Absence of an efficient geophysical index(es)
  for ionospheric F2-layer storms forecast.
• 2. Prediction of the ionospheric storm onset
  moment as well as the storm duration.
• 3. Positive F2-layer storm effect prediction (its
  magnitude and duration) for a particular storm
  event (however this is not crucial for HF
  communication as the working band becomes broader
  under such conditions).
• 4. Absence a precursor to predict quiet time both
  positive and negative F2-layer disturbances
  (Q-disturbances).

43
T H A N Y o u
K
44
References Araujo-Pradere, E.A., Fuller-Rowell,
T.J., and Codrescu, M.V., STORM An empirical
storm-time ionospheric correction model 1. Model
description, Radio Sci., 37, 1070,
doi10.1029/2001RS002467, 2002. Araujo-Pradere,
E.A. and Fuller-Rowell, T.J., Validation of the
STORM response in IRI2000, J. Geophys. Res., 108,
A3, 1120, doi10.1029/2002JA009720,
2003. Buonsanto, M.J., Ionospheric storms - a
review, Space Sci. Rev., 88, 563-601,
1999. Cander, Lj.R. and Mihajlovic, S.J.,
Forecasting ionospheric structure during the
great geomagnetic storms, J. Geophys. Res., 103,
391-398, 1998. Cander, Lj.R., Milosavljevic,
M.M., Stankovic, S.S., and Tomasevic, S.,
Ionospheric forecasting technique by artificial
neural network, Electron. Lett., 34, 1573-1574,
1998. Chan, A.H.Y. and Cannon, P.S., Nonlinear
forecast of foF2 variation of model prediction
accuracy over time, Ann. Geophysicae, 20,
1031-1038, 2002. Forbes, J.M., Palo, S.E., and
Zhang, X., Variability of the ionosphere, J.
Atmos. Solar-Terr. Phys.,62, 685-693,
2000. Francis, N.M., Cannon, P.S., Brown, A.G.,
and Broomhead, D.S., Nonlinear prediction of the
ionospheric parameter foF2 on hourly, daily, and
monthly timecales, J. Geophys. Res., 105,
12,839-12,849, 2000. Francis, N.M., Brown, A.G.,
Cannon, P.S., and Broomhead, D.S., Prediction of
the hourly ionospheric parameter foF2 using a
novel nonlinear interpolation technique to cope
with missing data points, J. Geophys. Res., 106,
30,077-30,083, 2001. Fuller-Rowell, T.J.,
Codrescu, M.V., and Wilkinson, P., Quantitative
modeling of the ionospheric response to
geomagnetic activity, Ann. Geophysicae, 18,
766-781, 2000. Kutiev, I. and Muhtarov, P.,
Cander, L.R., and Levy, M.F., Short-term
prediction of ionospheric parameters based on
autocorelation analysis, Ann. Geofis., 42,
121-127, 1999. Kutiev, I. and Muhtarov, P.,
Modelling of midlatitude F region response to
geomagnetic activity, J. Geophys. Res., 106,
15,501-15,509, 2001. Marin, D., Miro, G., and
Mikhailov, A.V., A method for foF2 short-term
prediction, Phys. Chem. Earth (C), 25, 327-332,
2000. McKinnell, L.-A. and Poole, A.W.V.,
Predicting the ionospheric F layer using neural
networks, J. Geophys. Res., 109, A08308,
doi10.1029/2004JA010445, 2004. Liu, R., Xu, Z.,
Wu, J., Liu, S., Zhang, B., Wang, G., Preliminary
studies on ionospheric forecasting in China and
its surrounding area, J. Atmos. Solar-Terr.
Phys., 67, 1129-1136, 2005. Muhtarov, P. and
Kutiev, I., Autocorrelation method for temporal
interpolation and short-term prediction of
ionospheric data, Radio Sci., 34, 459-464,
1999. Muhtarov, P., Cander, L., Levy, M., and
Kutiev, I., Application of the geomagnetically
correlated statistical model to short-term
forecast of foF2, Proc. of the 2nd COST 251
Workshop, 30-13 March 1998 Side, Turkey, 241-245,
1998. Pant, T.K. and Sridharan, R., Seasonal
dependence of the response of the low latitude
thermosphere for external forcing, J. Atmos.
Solar-Terr. Phys., 63, 987-992,
2001. Richards, P.G., Torr, D.G.,. Buonsanto,
M.J., and Miller, K.L., The behaviour of the
electron density and temperature at Millstone
Hill during the equinox transition study
September 1984, J. Geophys. Res., 94,
16.969-16.975, 1989. Richards, P.G., Torr, D.G.,
Buonsanto, M.J., and Sipler, D., Ionospheric
effects of the March 1990 magnetic storm
comparison of theory and measurements, J.
Geophys. Res., 99, 23,359-23,365, 1994.
45
Picone, J.M., Hedin, A.E., Drob, D.P., and Aikin,
A.C., NRLMSISE-00 empirical model of the
atmosphere Statistical comparison and
scientific issues, J. Geophys. Res., 107, A12,
1468, doi10.1029/2002JA009430, 2002. Richards,
P.G., Torr, D.G.,. Buonsanto, M.J., and Miller,
K.L., The behaviour of the electron density and
temperature at Millstone Hill during the equinox
transition study September 1984, J. Geophys.
Res., 94, 16.969-16.975, 1989. Richards, P.G.,
Torr, D.G., Buonsanto, M.J., and Sipler, D.,
Ionospheric effects of the March 1990 magnetic
storm comparison of theory and measurements, J.
Geophys. Res., 99, 23,359-23,365, 1994.
Shubin,V.N. and Anakuliev, S.K., Ionospheric
storm negative phase model at middle latitudes,
Geomagn. and Aeronom., 35, 363-369, 1995 (English
translation). Tsagouri, I. and Belehaki, A., A
new empirical model of middle latitude
ionospheric response for space weather
applications, submitted to Adv. Space Research
(2005). Wintoft, P. and Cander, L.R.,
Twenty-four hour predictions of foF2 using time
delay neural networks, Radio Sci., 35, 395-408,
2000. Wrenn, G.L., Time-weighted accumulations
ap(t) and Kp(t), J. Geophys. Res., 92,
10125-10129, 1987. Wrenn, G.L., Rodger, A.S., and
Rishbeth, H., Geomagnetic Storms in the Antartic
F-region. I. Diurnal and seasonal patterns for
main phase effects, J. Atmos. Terr. Phys., 49,
901-913, 1987. Wu, J. and Wilkinson, P.J.,
Time-weighted magnetic indices as predictors of
ionospheric behaviour, J. Atmos. Terr. Phys.,
57, 1763-1770, 1995. Zevakina, R.A., Zhulina,
E.M., Nosova, G.N., and Sergeenko, N.P.,
Short-term prediction manual, Materials of the
World Data Centre B, Moscow, pp.71, 1990 (in
Russian). Zevakina, R.A. and Kiseleva, M.V.
F2-region parameter variations during positive
disturbances related to phenomena in the
magnetosphere and interplanetary medium. In The
diagnostics and modelling of the ionospheric
disturbances, Nauka, Moscow, 151-167, 1978 (in
Russian).