Fysiikan pivien CRRESDEH posteri

1
Long-term measures of geomagnetic activity and
ring current and their implications on solar
change

Kalevi Mursula Department of Physical
Sciences University of Oulu, Finland e-mail
Kalevi.Mursula_at_oulu.fi http//spaceweb.oulu.fi/ka
levi/publications/ Collaboration Lauri
Holappa, Daniel Martini, Arto Karinen
2
New long-term indices of GA storms
IHV Ah Dst
3
Why new indices of GA ?

• There is NO verifiable, correct centennial index
  of geomagnetic activity among the traditional
  indices !
• aa index is flawed and unverifiable.
• Kp/Ap index exists only since 1932 and does not
  cover the interesting period of rapidly
  increasing geomagnetic activity from 1900 to
  1930s.

4
Why new magnetic storm index ?

• There is NO correct centennial index of
  geomagnetic storms!
• Dst index is flawed and self-inconsistent and
  exists only since 1932.

5
aa index
aa activity antipodal Constructed by Father
Pierre Mayaud Exists since 1868 onwards. Uses K
values of two stations, one in England, one in
Australia, that are in approximate antipodal
location and can give a rough estimate of global
activity. Normalized to correspond to the
activity level at the magnetic latitude of 50.
6
aa index stations
England Greenwich (18681925 correction
factor 1.007) Abinger (19261956
cf0.934) Hartland (1957 cf1.059) Australia
Melbourne (18681919 cf0.967) Toolangi
(19201979 cf1.033) Canberra (1980 cf1.084)
7
Problems in aa index
Jumps in aa value distribution when either
station changed.
8
Problems in aa index
Jumps in aa value distribution when either
station changed.
9
Problems in aa index
Jumps in aa value distribution when either
station changed.
10
Problems in aa index
Jumps in aa value distribution when either
station changed.
11
Problems in aa index
aa North is larger than aa South
12
Problems in aa index
Distribution of different aa values in North and
South is quite different.
13
Ratio of Ap and Ap-normalized aa
Problems in aa index
aa is constructed from observations at 2
antipodal points. Sequence of 3 stations in each
hemisphere. Intercalibration? Long-term
consistency? Known 2 nT scaling error at the end
of 1950s when moving from Abinger to Hartland
(Jarvis et al., 2005 Mursula and Martini, 2007
Lockwood et al., 2008).
ALSO aa depicts too large centennial trend, even
with 2nT correction (reason unknown).
14
Problems archival data
Homogeneity and consistency of archival data are
mandatory requirements for long-term
studies. However Almost all early archives
have problems in this respect ! Sampling changes
in magnetic data Procedure changes and
errors ESK hourly data NGDC aa index
15
Problem due to changing data sampling

• All stations registered hourly samples (one
  sample, e.g., on each sharp hour), not hourly
  means in the early part of 1900s.
• Most stations changed from hourly samples to
  hourly means in 1915.
• Hourly samples include more variability than
  hourly means.
• All indices measure variability and are therfore
  affected by this change in data sampling !
• Without due correction, the level of geomagnetic
  activity in the early part of 1900s will be
  estimated too high and the centennial increase
  remains too low !
• Detailed results show that this is an essential
  correction.

16
Sampling problem CLH/FRD vs. NGK IHV ratio
CLH/FRD / NGK IHV ratio.

• NGK changed its sampling already in 1905.
• We can use NGK to estimate the effect of
  sampling change by calculating ratios of annual
  IHV indices between other stations (here CLH/FRD)
  and NGK.
• The effect of changed sampling is seen as an
  increase of the ratio in 1905 when NGK sampling
  was changed, and as a decrease of the ratio in
  1915 when CLH/FRD sampling was changed.

Decrease
Increase
17
Sampling problem RC calibration factor

• CLH/FRD vs. NGK ratio in 1901-1904 is 0.7433
• CLH/FRD vs. NGK ratio in 1905-1914 is 1.031
• Their ratio 1.031/0.743 1.39 gives an estimate
  of the required sampling calibration factor (to
  be called RC) for NGK IHV index.
• CLH/FRD vs. NGK ratio in 1905-1914 is 1.031
• CLH/FRD vs. NGK ratio since 1915 is 0.7918
• Their ratio 1.031/0.7918 1.30 gives an
  estimate of the required sampling calibration
  factor (to be called RC) for CLH/FRD IHV index.

18
Sampling problem IHV Calibration factors
Table. RC and MH calibration factors for IHV
(raw, and corrected IHV-cor) in 1996 and 2000.
RC and MH calibration factors are quite
consistent for most stations. We use MH factors
because they are more definite. Calibration
factors are indeed quite large, typically 30-40.
They make an essential correction to the
centennial trend based on IHV indices.
19
IHV for CLH/FRD
Difference between aa (lower) and the
recalculated aa
Svalgaard et al. Adv. Space Res., 34, 2004
...any claim based on the rise of aa since 1900
may be inaccurate.
1st International Symposium on Space Climate
20
ESK/NGK IHV ratio

• The ratio of annual IHV averages between ESK and
  NGK depicts a significant RISE from 1931 to 1932.
• This poses a new problem.
• A larger step of about 57 is found in 1932 which
  can not be understood in terms of the SAME
  sampling change.

1st International Symposium on Space Climate
21
ESK yearbook and WDC data
Top ESK H component according to the observatory
yearbook (line with black dots) and according to
the WDC data (line with circles) in January 29,
1911. The WDC data points are shifted by half an
hour forward to demonstrate the fact that they
are two-hour averages of the neighboring yearbook
values. Bottom The hourly ESK H values according
to the WDC data (big open circles), together with
the two-hour running means of the yearbook data
(line with small dots) for a few days in January,
1911.
Coinciding dots and circles prove that the WDC
data are 2-hour running averages of the yearbook
data.
1st International Symposium on Space Climate
22
ESK/NGK IHV correction factors
Annual IHV values for ESK in 1996 and 1999 using
four different assumptions of the treatment of
data. Note 1h IHV values are about 18 larger
than 2 hour running mean 1min IHV values (latter
more dependent!).
BUT 57 increase only explained if WDC data
before 1932 were 2 hour running means of hourly
data!
1st International Symposium on Space Climate
23
ESK/NGK IHV ratio
The ratio of the corrected annual IHV-raw values
between ESK and NGK (thick line with dots). For
comparison the uncorrected ratio is also shown
(thin line with stars). The step from 1931 to
1932 is seen before but not after correction.
1st International Symposium on Space Climate
24
ESK/NGK IHV ratio
The Difference in 1912-2000 between the
reconstructed and original aa index. Top using
uncorrected ESK IHV indices Bottom using
corrected ESK IHV indices.
1st International Symposium on Space Climate
25
Problems in catalogue data
Last jumps in the aa value distribution of NGDC
data is found in 1983, different from 1980 when
stationwas changed (and when ISGI data show
jump).
26
Problems in aa index
Jumps in aa value distribution when either
station changed.
27
Problems with IHV

• Very different definition of geomagnetic activity
  from K method
• - hourly variability measure vs. 3-hourly range
  measure
• - local nightside only vs. all day and night
• Different physics included in IHV vs. K indices
  !
• Extension of K indices by IHV is unfounded !
• Solution Ah index
• (Aamplitude, hbased on hourly data)

Mursula and Martini, GRL, 2007
28
Principles of the Ah index
Based on existing digital data Verifiable as
IHV. Monthly quiet daily curve from five
quietest days. (cf. a similar
treatment for the Dst index). Uses the K-index
method slightly modified Use amplitude, not
quasi-logarithmic K indices. Removes selection
problem between neighboring K-values. Important
esp. for K0 or 1. Removes the fixed K9
selection.
29
SOD IHV vs ak vs Ah
1. Compare the long-term compatibility of the Ah
and IHV indices at different temporal resolution
with the K/ak values at Sodankylä station
(northern Finland). K values measured at SOD
since 1914. ak are the linearized version of
(primary) K values. 2. Study the trends in Ah,
Ap and aa.
30
SOD ak vs. Ah and IHV Part of daily data
Scatterplots of daily averages of SOD Ah vs SOD
ak (cc0.94 upper panel) and SOD IHV vs SOD ak
(cc0.80 lower panel). Only every 20th point is
only included in the plot. Best fit lines ak
0.4257Ah - 0.5417 ak 0.3476IHV 4.792
31
SOD ak vs. Ah and IHV Low end of daily data
The low end of daily values depicts a difference
between Ah and IHV Ah - ak correlation stays
linear until lowest values. BUT IHV - ak
correlation is nonlinear at low values. Only
every 20th point is only included in the plot.
32
SOD ak vs. Ah All 3-hourly data
The ak and Ah indices are 3-hourly indices. The
full set of 3-hourly values is 243374 data
points. Correlation between all 3-hourly ak and
Ah is amazingly good (cc0.873). This is much
better than the correlation between the daily
values of ak and IHV. The best fit line is ak
0.3617Ah 2.178.
33
SOD ak vs. Ah 3-hourly data in each LT sector
The correlation between the eight pairs of
simultaneous 3-hourly ak and Ah values is very
good in each UT (or LT SOD LT UT2.5h) sector,
with correlation coefficients varying from 0.81
to 0.89. Note that in each LT sector the
correlation between ak and Ah is better than
between the daily averages of ak and IHV (the
number of data points is same). The smallest
correlation was found in the third sector
(UT6-9), i.e., in the pre-noon LT sector, the
largest half a day later in the seventh UT sector
(18-21 UT), i.e., in the pre-midnight sector.
34
SOD ak vs. Ah UT variation in Ak and Ah
The average values of the 8 daily ak and Ah
indices in 1914-2000 in the eight three-hourly UT
sectors shows the strong diurnal variation in
geomagnetic activity. The diurnal variation is
very similar in ak and Ah. The correlation
coefficient between the curves is 0.981
(CLgt99.99). Absolute scales are different
because of different normalizations.
35
Correlations of Ah and IHV with Ap
Correlations between the (annual values of)
global Ap index and the local Ah (upper part) and
IHV (lower part) indices for the different
stations. The Ah indices have systematically a
better correlation with Ap than the IHV index.
This is due to the more similar definition
between Ap ad Ah than Ap and IHV.
36
Trends in aa, Ap and global Ah

• aa index depicts an erroneously large centennial
  increase by roughly a factor of two.
• Ah depicts a definite but more moderate increase
  than the aa index.
• Ah index is extremely well correlated with Ap.

Mursula and Martini, GRL, 2007.
37
No step in Ah3 - Ap relation
aa is systematically smaller than Ap until about
1960 and larger thereafter. This error leads to
an excessively large centennial increase in
aa. However, Ah3 (or Ah6) does not depict such
a systematical difference to Ap.
Mursula and Martini, GRL, 2007.
38
Ah indices for 6 stations

• The Ah indices at all 6 stations depict a very
  similar overall evolution.
• Rapid increase to 1960, dropout and weak
  increase thereafter.
• However, the stations depict rather different
  centennial increases.
• The centennial increases are latitudinally
  organized largest at high latitudes, smaller at
  low latitudes and smallest at mid-latitudes.

39
Centennial increases and trends
TABLE The relative increases in the last century
and centennial best-fit trends for the 6
stations, the global Ah3, Ah6 indices, and the aa
index. The stations from the same latitudes
depict the same trends gt consistency Global Ah
indices depict a considerably smaller increase
than the aa index ! Trend in the aa index is
even much larger than the trend of Ah indices at
correponding mid-latitudes!
40
Centennial increase at different latitudes
We find a curious latitudinal variation in the
centennial increase Centennial increase is
largest at high latitudes This is no surprise. It
could result from proximity to auroral oval and
the change of the internal magnetic field. BUT
Centennial increase at low latitudes is larger
than at mid-latitudes! This is surprising.
However, it could also result from the change of
the internal magnetic field. Also, this poses
the question How to define the centennial
increase of GLOBAL geomagnetic activity. It can
not be based only on mid-latitude station data !
41
Dst/Dcx
42
Small problems Abnormal UT variation in 1971
Annually averaged range of the diurnal UT
variation for Dxt in 1932-1956 (blue), Dxt in
1957-2002 (red) Dst in 1957-2002 (black).
Takalo and Mursula. J. Geophys. Res., 106, 2001
Dst exhibits an abnormally large UT variation in
1971 which originates from an erroneous Dst
derivation in that year. Dxt corrects Dst for
this error. NOTE UT variation in Dst is mostly
due to longitudinally sparse network. The problem
is NOT due to exceptional heliospheric
conditions!
43
Small problems Abnormally high level in 1960s
Annual averages of Dxt in 1932-1956 (blue), Dxt
in 1957-2002 (red) Dst in 1957-2002 (black).
Dashed black line depicts the Dst-Dxt
difference 1932-1998.
Karinen and Mursula. Ann. Geophys., 23, 2005
The typical annual Dst-Dxt difference is about
2-3 nT. However, in 1963-1966 the difference is
enhanced. In particular, there are no Dstgt0
years in Dxt, contrary to the official Dst in
1965 !!
44
SPE storm curves yearly Dst
SPE analysis of storms for Dst for each year in
1957-2005. Strongest storms in SC 19 and SC 23.
The erroneous values in mid-1960s show up as
excessively quiet period.
45
Dcx index Corrected Sq treatment
Originally, the linear trend L(t)atb was
removed in order to retain the changing level of
Dst during the recovery phase of magnetic storm.
Slope a Dst change during ring current
recovery. (NOT TO BE CHANGED). Intercept b
seasonally varying quiet time level. (MUST BE
CHANGED). SO If the linear trend is removed
from Sq, also the seasonally varying monthly
level of DH (included in intercept b) is removed.
Since the seasonal variation is removed from
the Sq-matrix, it naturally remains in D(t) and
Dst/Dxt. Our solution (not just remove the
slope b, but) After removing the full trend
(slope and intercept), raise the level back to
the average of the two nights (symmetric
treatment!) ? Dcx index (c for corrected, x
for extended).
Mursula and Karinen, Geophys. Res. Lett., 32,
2005.
46
Monthly superposed epoch storm curves
Average storm level in Dcx is -20.2 nT and in
Dxt -26.2 nT. Average difference between the two
indices is 6 nT during storms, corresponding to
23 average correction. Note also a clear
seasonal variation in the difference between Dcx
and Dxt. Correction indeed varies seasonally!
Monthly superposed epoch storm curves for Dcx
(blue) and Dxt (red) in 1932-2002. Dashed line
shows their difference.
47
Dcx Dt standard deviations without cosine
weighting
Annual standard deviations of Dt values reflect
the strength of the ring current at each of the 4
Dst stations. These depict systematically
different levels! HON depicts the largest
variation, HER the smallest. The variations are
ordered by the magnetic latitude of the station!
48
Dcx Dt standard deviations with cosine weighting
Normalizing the disturbances (Dt values) of each
station by the cosine of magnetic latitude of the
station sets the variations to the same level,
thus correcting the problem. Note This is NOT
the way Dst is normalized now !
49
Dcx Dt annual means No cosine weighting
The Dt annual means of the 4 Dst stations depict
quite different levels. Can this corrected be by
the by normalization with the cosine(latitude) of
the station?
50
Dcx Dt annual means With cosine weighting
NO ! The Dt values of the 4 Dst stations depict
quite different levels even if normalized by
cosine(latitude) of the station. SO Each
station has different weight in Dst/Dxt/Dcx !
51
Superposed epoch storm curves Dcx and Dxt
The Dxt and Dcx SPE curves follow each other
closely in both time intervals. Mean difference
between curves is 5.7 nT in 1932-1956 and 6.1 nT
in 1957-2002. EARLY VS. LATER PERIOD Storm main
phase is less intense and recovery phase longer
in the early period than in the later. Solar or
geomagnetic cause?
Superposed epoch storm curves for the Dst/Dxt
(blue) and Dcx (red) indices in 1932-1956 (left)
and in 1957-2005 (right).
Karinen and Mursula, AG, 23, 2005 Karinen and
Mursula, JGR.,111, 2006
52
SPE storm curves yearly Dcx
SPE analysis of storms for Dcx for each year in
1932-2005. Strongest storms in SC 19 and SC 23. A
few strong storms also in SC17. Long weak times
in 1930s, 1960s and 1990s. (Evidence for a
3-cycle period?)
53
Conclusions
Homogeneity and verifiability are mandatory
requirements that long-term studies set upon
geomagnetic indices. No analogue index is de
facto verifiable. Need fully digital indices
that can be examined in detail. IHV reflects
very different physics than all K indices or
Ah. IHV has a fundamentally nonlinear relation
between SOD ak.
54
Conclusions

• Geomagnetic activity has indeed increased in
  the last 100 years. Earlier claims of no increase
  were due to the accidental choice of a station
  depicting the smallest increase.

• All studied stations depicted the same
  two-step centennial pattern.

• Different centennial increase at different
  latitudes. Largest at high latitudes, smallest as
  mid-latitudes.

• Global centennial increase is smaller than
  depicted by the aa index. In fact, aa index does
  not behave like a mid-latitude station !?

55
Conclusions
Ah is proposed as a fully digital index which
is based on the basic range idea of K indices but
modifies it so that long-term homogeneity is
better guaranteed. A highly correlative, linear
relation between SOD ak and Ah. Ah for all
stations correlate better with Ap than IHV. The
results suggest that Ah rather than IHV should be
used as a long-term (centennial) proxy or
extension of local and global ak/K indices. Ah
verifies the need to correct the aa index. BUT
The centennial trend in aa index is too large
even after 2 nT correction ! (Unknown reason so
far).
56
Conclusions

• The Dst index includes several small accidental
  errors and large systematic errors and
  inconsistencies.
• The small errors (e.g., UT in 1971, Dst level
  in 1960s) are corrected when re-calculating the
  Dst index for full using original data (Dxt
  index).
• The Dst index has a systematic excessive seasonal
  variation which can easily be removed by letting
  the Sq curve include the seasonally varying level
  (Dcx index).
• Correction for some storms can be as high as 44
  nT. The correction affects the classification of
  storms according to their min-Dst value.
• Average correction is 6.0 nT, implying an average
  correction for storms of 23 .
• Correction varies seasonally. The largest
  correction is for Spring equinox (March).

57
Conclusions

• The correction is large enough to seriously
  affect many studies, e.g., earlier estimates of
  the physical causes of semiannual variation based
  on the Dst index.
• Correction improves the correlation between Dcx
  and sunspot activity, as well as with geomagnetic
  activity.
• Storms in the early period (1932-1956) are less
  strong but last longer. This implies that there
  were more of HILDCAA type activity periods,
  probably due to recurrent high-speed streams at
  that time than more recently.
• Dst should be calculated by normalizing each
  station by its own cosine(dipole_latitude), not
  like now in Dst by average of this.
• Each Dst station must be corrected by solar
  illumination effect by dividing by the
  cosine(solar zenith angle).

58
Conclusions

• We have extended the geomagnetic Dst index by 25
  years, i.e., by more than one full solar
  magnetic cycle.
• The non-storm component in the Dst index arises
  from the seasonal variation of the H component at
  Dst stations and from the erroneous treatment of
  Sq elimination.
• The suggested correction removes the non-storm
  component, leaving the Dcx index with a seasonal
  variation which is very similar to that in other
  magnetic indices.
• Average correction is 6.0 nT, implying an average
  correction for storms of 23 .
• Correction varies seasonally. The largest
  correction is for Spring equinox (March).

59
Bottom line
GEOMAGNETIC ACTIVITY IS PERHAPS THE MOST
IMPORTANT AND VERSATILE SINGLE GLOBAL
HELIOSPHERIC VARIABLE !!
60
Bottom line
GEOMAGNETIC ACTIVITY IS PERHAPS THE MOST
IMPORTANT AND VERSATILE SINGLE GLOBAL
HELIOSPHERIC VARIABLE !! ITS CONTINUITY AND
HOMOGENEITY SHOULD BE GUARANTEED FOR DECENNIA
AHEAD!
61
Bottom line
GEOMAGNETIC ACTIVITY IS PERHAPS THE MOST
IMPORTANT AND VERSATILE SINGLE GLOBAL
HELIOSPHERIC VARIABLE !! ITS CONTINUITY AND
HOMOGENEITY SHOULD BE GUARANTEED FOR DECENNIA
AHEAD! ITS AVAILABILITY IN HIGH RESOLUTION IN
DIGITAL FORMAT SHOULD BE MADE POSSIBLE !
62
Bottom line
GEOMAGNETIC ACTIVITY IS PERHAPS THE MOST
IMPORTANT AND VERSATILE SINGLE GLOBAL
HELIOSPHERIC VARIABLE !! ITS CONTINUITY AND
HOMOGENEITY SHOULD BE GUARANTEED FOR DECENNIA
AHEAD! ITS AVAILABILITY IN HIGH RESOLUTION IN
DIGITAL FORMAT SHOULD BE MADE POSSIBLE
! INTERNATIONAL ARCHIVAL DATA CAMPAIGN NEEDED !
63
Thanks