Based on recent observations of solar activity, Solar Cycle 23
has begun and is expected to reach a maximum smoothed monthly
sunspot number near 160 (between the values of 130 and 190) near
March, 2000 (between June 1999 and January 2001). The corresponding
prediction for 10.7 cm solar flux is approximately 205 solar flux
units (sfu; 10-22 W/m2), between 175 and
This result reaffirms the consensus prediction made in September,
1996, for a solar cycle which is comparable to the last two cycles,
but is unlikely to exceed Cycle 19, the largest cycle on record.
Based on traditional smoothed monthly sunspot numbers and 10.7
cm solar flux, the minimum level of solar activity between Cycles
22 and 23 occurred in May, 1996. However, based on additional
factors including the number of spotless days and the dominance
of new cycle active regions, the Solar Cycle 23 Panel recommends
that October, 1996, be used as the effective onset of Cycle
The date of geomagnetic minimum is not yet certain; therefore
the panel provisionally retained their previous prediction which
calls for intermittent storm activity especially between 1999
and 2005 with storm frequencies and intensities similar to that
experienced during Cycle 22.
In September, 1996, in advance of knowledge of the end of Solar Cycle 22, an international panel of 12 scientists met in Boulder, Colorado to consider 28 collected predictions of the profile and amplitude of Cycle 23 solar and geomagnetic activity. They concluded that a reasonable forecast was a new cycle comparable to old Cycle 22. In particular, the maximum smoothed sunspot count was predicted to be between 130 and 190, with 160 as a representative value (the maximum smoothed Cycle 22 value was 159). This is a large cycle, but is unlikely to exceed Cycle 19, the largest known solar cycle, which reached a smoothed monthly maximum of 201 in March 1958. The panel further estimated that the maximum would occur between January 1999 and June 2001, with March 2000 as a good "middle-ground"; estimate. The wide spread in possible dates for the month of maximum was considered necessary because the date of the onset of Cycle 23 was unknown. This prediction was disseminated on the NOAA Space Environment Center Internet home page in October, 1996, and a summary was published [Joselyn et al. 1997].
In the year since that meeting, there has been a slow but steady
increase in sunspot numbers and monthly 10.7 cm flux, and a rapid
upturn in activity beginning in late August, 1997. The minimum
in smoothed monthly sunspot numbers (weighted 13-month running
average) is May, 1996, with a value of 8.0. The panel reconvened
to review their solar activity prediction in light of these new
data as well as several new forecasts recently published.
The most problematic point discussed was the time of the cycle
minimum. While the traditional numerical prescription as well
as other measures of solar irradiance and activity agree that
May, 1996, was the minimum smoothed month of the cycle, Karen
Harvey noted that there are several factors that argue that this
date is misleading as a fiducial for Cycle 23 onset. In particular,
no new cycle spots were observed before this month - a situation
never before recorded. The date of minimum is expected to represent
that time when new cycle activity becomes dominant; that is, the
new cycle should have been in progress as the old cycle declined,
the minimum then marking the crossover. But for Cycle 23, new
cycle regions did not outnumber old cycle regions until December,
1996. The resurgence of activity in the months following May is
due to old cycle regions (low latitude spot groups with the appropriate
magnetic polarity of old cycle regions). Another factor that indicates
May is a misleading date is that the maximum number of spotless
days occurred in September and October, 1996. Finally, the Panel
noted that the epoch tables specifying cycle maxima and minima
included in Waldmeier  are not consistent with a strict
determination based on smoothed sunspot number.
The Panel then reviewed the published definition of sunspot extrema [McKinnon 1987] which includes 5 criteria:
When observations permit, a date selected as either a cycle minimum or maximum is based in part on an average of the times when extremes are reached (1) in the monthly mean sunspot number, (2) in the smoothed monthly mean sunspot number, and (3) in the monthly mean number of spot groups alone. Two more measures are used at time of sunspot minimum: (4) the number of spotless days and (5) the frequency of occurrence of "old" and "new" cycle spot groups.
Because the smallest monthly mean sunspot numbers were achieved
in September (1.6) and October (0.9), 1996, and the most spotless
days occurred in October, the Panel agreed that October, 1996,
was the effective onset of Cycle 23.
There are significant implications for this choice of Cycle 23
onset, especially for statistical prediction techniques that key
on the date of minimum such as the McNish-Lincoln method. Since
this method is inherently short-range, McNish and Lincoln 
recommended that it not be used for making projections longer
than one year in advance. Modifications of this method, such as
that developed by the NASA Marshall Space Flight Center (MSFC)
[Niehuss et al. 1996], can be used with 13-month smoothed data
to provide estimates over an entire current solar cycle. Jerry
Owens showed results for this method provisionally initialized
using the May 1996 minimum, which are also reported in the September
1997 MSFC Solar Memo. These results indicate a mean (50th
percentile) maximum 10-cm solar flux of 149 sfu (sunspot number
of about 103) and a 95th percentile value of 224 sfu (sunspot
number of about 180). While the 50th percentile value is less
than the Panel's expected lower bound of 175 sfu, MSFC
routinely recommends use of the 95th percentile values for spacecraft
and space vehicle applications, within the range of 175 to 235
agreed by the Panel. He also showed results demonstrating that
use of a later date to initialize the calculation results in a
profile with faster rise time and higher peak value.
By means of a similar modified McNish-Lincoln technique, Alan
Thomson used the method of Kerridge et al. , based on Holland
and Vaughan , to calculate a mean (50th percentile) maximum
10.7 cm flux of 175.5 sfu (± 45 for upper and lower 95th
percentile values) in March 2001 with an earlier peak of 173.9
in December 1999. The latter calculation uses only cycles 18-22,
whereas the MSFC calculation uses solar cycles 1-22.
Richard Thompson presented an analysis of historical data that
indicates that the early rise of a solar cycle is a poor indicator
of the eventual amplitude of that cycle. This is an important
limitation in the use of projection techniques such as McNish-Lincoln
in determining the amplitude of the cycle.
Werner Mende continued his work with spectral filtering, which
accurately predicted the date of smoothed minimum approximately
2 years in advance. His research isolated a band of filters of
approximately 4 years in length; these filters remove multiple
maxima but do not alter cycle length. He scaled available cycles
to unit amplitude, and found that the inflection point (where
the derivative of the slope changes sign) on the increasing flank
of the cycle profile was the most repeatable characteristic from
one cycle to the next. His estimate for the date of the peak of
the new cycle is the sum of 2 years from the numerical minimum
to the inflection point, and then 1.9 years more to maximum. He
expects a duration of 5.4 years from maximum to the end of the
cycle. Using his filtered date of minimum, the inflection point
is in mid-1998 and the resulting maximum date and amplitude are
in close agreement with the Panel consensus.
The Panel also considered 10 new and two updated predictions of
Solar Cycle 23 behavior. All of the new submissions used techniques
which were already represented among the predictions that were
evaluated in 1996. Most of these newly furnished forecasts fell
within the Panel's range of maximum values. Those that
predicted an average or low cycle were discounted for the same
reasons that such predictions were discounted previously: they
each relied on the entire record of cycle activity. The Panel
believes that recent high activity cycles and the precursor techniques,
successful in predicting previous cycles, should be more heavily
weighted. Ultimately, the Panel found no compelling reason
to alter the previous estimate of the amplitude or shape of Cycle
However, because the onset of Cycle 23 is now established, the
Panel agreed that the range of possible dates of maximum could
be narrowed. Based on an historical range of rise times for cycles
of the expected amplitude, the month of smoothed maximum should
lie between June 1999 and January 2001, with March 2000 as a likely
mid-range estimate. Figure 1 illustrates the mid-range prediction
for 10.7 cm flux solar initialized to October, 1996, along with
the data observed through September, 1997 (smoothed Cycle 23 data
are shown, except for the last 6 months which are the monthly
mean values). The monthly mean values for Cycle 22, matched so
that Cycle 22 minimum, September 1986, is aligned with October,
1996, are shown for comparison.
Figure 1: Actual smoothed or monthly mean values (last 6 months)
of 10.7 cm solar flux and the predicted profile of 10.7 cm solar
flux for Cycle 23, along with the observed values for Cycle 22
aligned so the predicted cycle and Cycle 22 begin in October,
The record of geomagnetic aa indices, beginning in 1868,
shows that the minimum in geomagnetic activity typically lags
the minimum in solar activity. That behavior appears to be continuing
for this cycle. The annual average aa for 1996 was 18.6;
thus far for 1997 (January through August), the annual average
aa is 15.9. Until the date and level of geomagnetic minimum
is firmly established, the Panel predictions made in 1996 for
Cycle 23 geomagnetic activity will be retained. These are repeated
in Table 1. The precursor approach is based on the same data and
analysis used by Richard Thompson for his solar cycle precursor
prediction, and the climatological approach is based on the median
behavior of the activity observed in the record of odd-numbered
cycles of geomagnetic aa index [Joselyn et al. 1997].
Because of the relevance of geomagnetic activity for upper atmospheric
density calculations, NASA MSFC predicts the geomagnetic Ap
index with the same modified McNish-Lincoln algorithm used for
the solar flux prediction; 13-month smoothed Ap data
are substituted for the solar flux data [Niehuss et al. 1996].
That calculation, initialized in November, 1996, and reported
in the September 1997 MSFC Solar Memo, predicts a maximum
smoothed monthly Ap of 24.9 (95th percentile, approximately
equivalent to an aa of 31) in August of 2004. This value
is consistent with the upper envelope of the Panel prediction
made one year ago. The newest British Geological Survey McNish-Lincoln
result using the aa index, also based on a preliminary
initialization date of November 1996, is for a 50th percentile
maximum aa of 28.2 nT in mid-2002. This is slightly less
than the aa of 29.2 nT used to estimate the climatological
number of occurrences shown in Table 1.
The Panel continues to recommend that more research is necessary
in the area of long-term geomagnetic activity predictions.
Holland, R.L. and W.W. Vaughan, Lagrangian Least-Squares Prediction of Solar Flux (F10.7), J. Geophys. Res., 89, 11-16, 1984.
Joselyn, J.A., J.B. Anderson, H. Coffey, K. Harvey, D. Hathaway, G. Heckman, E. Hildner, W. Mende, K. Schatten, R. Thompson, A.W.P. Thomson, and O.R. White, Panel achieves consensus prediction of Solar Cycle 23, EOS, Trans. Amer. Geophys. Union, 78, pages 205, 211-212, 1997.
Kerridge, D.J., V. Carlaw, and D. Beamish, A Review of Methods for Solar and Geomagnetic Activity Forecasting for Application in Space Missions Planning, BGS Technical Report, WM/89/14C (62 pp.) 1989.
McKinnon, J.A., Sunspot Numbers: 1610-1985, Report UAG-95, World Data Center A for Solar-Terrestrial Physics, NOAA, Boulder, CO, January 1987.
McNish, A.G., and J.V. Lincoln, Prediction of Sunspot Numbers, Eos, Trans. Amer. Geophys. Union, 30, p. 673, 1949.
Niehuss, K.O., H.C. Euler, Jr., and W.W. Vaughan, Statistical Technique for Intermediate and Long-Range Estimation of 13-Month Smoothed Solar Flux and Geomagnetic Index, NASA TM-4759, Washington, DC, September, 1996.
Waldmeier, M., The Sunspot Activity in the Years 1610-1960,
Zurich Schultess and Company, Switzerland, 1961.
The NOAA Space Environment Center (SEC) convened the Solar Cycle
23 Project Panel with the support of the NASA Office of Space
Science and the respective institutions of the panel members.
The Panel Members are Dr. Jo Ann Joselyn (Chair, NOAA Space Environment
Center, Boulder, CO); Helen Coffey (NOAA National Geophysical
Data Center, Boulder, CO); Dr. Karen Harvey (Solar Physics Research
Corp, Tucson AZ); Dr. David Hathaway (NASA Marshall Space Flight
Center, Huntsville, AL); Gary Heckman (NOAA Space Environment
Center, Boulder, CO); Dr. Ernie Hildner (NOAA Space Environment
Center, Boulder, CO); Dr. Werner Mende (Institute of Meteorology,
Free University of Berlin, GERMANY); Dr. Jerry Owens (NASA Marshall
Space Flight Center, Huntsville, AL); Dr. Kenneth Schatten (NSF/Upper
Atmospheric Division, Arlington, VA); Dr. Richard Thompson (IPS
Radio and Space Services, NSW, AUSTRALIA); Dr. Alan W.P. Thomson
(British Geological Survey, Edinburgh, Scotland, UK); Dr. Oran
R. White (NCAR High Altitude Observatory, Boulder, CO).