Early Alert of Solar Radiation Hazard
Paul Evenson (firstname.lastname@example.org), T. Kuwabara, J. W. Bieber, J. Clem, R. Pyle
Bartol Research Institute and Department of Physics and Astronomy, University of Delaware
Abstract We describe our automated Ground Level Algorithm for issuing an alarm We define three levels of alarm (Watch, Warning,
Enhancement (GLE) Alarm that is now operational in beta testing. and Alert) based on the number of stations that record a
Using our neutron monitor stations at Fort Smith, Inuvik, McMurdo, significant intensity increase.
Nain, Newark, Peawanuck, South Pole, and Thule, automated e-mail A threshold level Ith is set for the cosmic ray intensity
alarms can be sent to subscribers within minutes of large GLE onsets. increase, and an alarm is generated when the number of
In brief, the system generates a WATCH when one station observes an stations that exceed the threshold level is
increase of 4% or more in a three minute moving average, a 1 - Watch, 2 - Warning, 3 –Alert.
WARNING when two stations observe such an increase, and an Baseline to calculate the intensity increase is defined
ALERT when three or more stations record such an increase. (A higher by two parameters b and 0. Then, by using a trailing
threshold for percent increase may be imposed for some stations with moving average value for the current count rate, the
lower count rates.) A scientific discussion of the alarm system can be intensity at time t = is calculated each minute from
found in the article "Development of a Ground Level Enhancement Condition for issuing three levels of alarm. Intensity
the observed count rate N() averaged over the
Alarm System Based upon Neutron Monitors", T. Kuwabara, J. W. increase recorded at three stations during a schematic
preceding c minutes, shown as left equation.
Bieber, J. Clem, P. Evenson, and R. Pyle, Space Weather, 4, S10001, GLE are illustrated.
To reduce fluctuations and accurately detect the GLE,
doi: 10.1029/2006SW000223, 2006. A web page specifically devoted all parameters (threshold and baseline) were optimized
1 1 0
I N t N t
to the GLE alarm, including graphical displays is available at:
http://www.bartol.udel.edu/~takao/neutronm/glealarm/index.html as by backtesting against past neutron monitor data as
c t c b t 0 b c=3 min, b =75min, 0=10min, Ith = 4%.
South Pole neutron monitor is supported by NSF award ANT-0838839 At this condition, the false alarm rate for Watch,
and McMurdo is supported by NSF award ANT-0739620 c : average time 0 : time interval between the Warning, and Alert was ~40/yr, less than 1/yr, and 0/yr
b : duration of the baseline and current time respectively, during the 4.4-year period of our
base-line Ith : threshold level backtesting study,
GLE event on January 20, 2005
Alarm issued for January 20, 2005 event On Jan 20, 2005 event, alarm times for
Warning and Alert are 6:49 and 6:50
respectively; a separate Watch alarm was not
generated. The SEC alert was issued at 7:02 from
>10 MeV data, and 7:04 from >100 MeV data.
Proton flux observed in the >10 MeV channel
exceeds 100 pfu at 6:55. This flux level is defined
as a “Moderate Storm” (S2 on NOAA Space
Weather Scale for Solar Radiation Storms) that
has the possibility to cause single-event upsets
aboard satellites, and our system can produce
alarm before this.
Space Weather Message Code: ALTPX2
Serial Number: 27
Issue Time: 2005 Jan 20 0702 UTC
GOES proton flux and cosmic ray intensity increase. In upper ALERT: Proton Event 10MeV Integral Flux exceeded 100pfu
Begin Time: 2005 Jan 20 0701 UTC
panel, black arrows indicate the onset time of SEP event(>100 NOAA Scale: S2 - Moderate
MeV), and the start time of the Moderate storm. Vertical lines Space Weather Message Code: ALTPC0
drawn in lower panels also show the onset time of GLE. Serial Number: 17
Colored arrows indicate the time when each alarm is issued Issue Time: 2005 Jan 20 0704 UTC
(proton monitor) or generated (neutron monitor). ALERT: Proton Event 100MeV Integral Flux exceeded 1pfu
Begin Time: 2005 Jan 20 0701 UTC
Comparison between the alert times from our system
and alarm issue times from proton data at SEC/NOAA for nine GLE events.
GLE event on January 20, 2005. Upper panel shows the low energy
proton integral flux recorded by the GOES-11 satellite (solid line >10MeV,
dotted line >100MeV). Lower panel shows neutron rates detected in
several neutron monitors. These data are normalized to average counting
rate at Inuvik (05:30-06:30).
The GLE (Ground Level Enhancement) of January 20, 2005 was the largest
in half a century. The onset of the intensity increase in neutron monitors is
earlier than that of the low energy proton flux, and time to reach maximum
intensity is shorter for neutron monitors.
The GLE particles have large mean free paths and travel almost at the
speed of light, and can be detected with high accuracy by detectors with
large volume such as ground-based neutron monitors. A GLE alarm can
provide a very useful early warning of an impending solar radiation storm.
We developed a system that watches (in real time) for count rate increases in
our data, and gives an alarm when a GLE is detected. We compare the alarm Number of Minutes by which each alert
Number of Minutes by which GLE Alert precedes
time produced in our system with that of the system operated by precedes Moderate Storm (>100pfu in >10
earliest SEC Proton Alert
SEC/NOAA by using the GOES proton data. MeV channel)
Eight neutron monitors are used in this work. The station name, detector Summary
type, average cosmic ray background count rate (in 2005),
We have developed a real-time GLE detection system using eight high-latitude neutron monitors
geographical latitude, longitude, and altitude are listed.
• GLE alarms are produced at three levels (Watch, Warning, and Alert) corresponding to the number of
Station Type Count/hour Lat., Long., Altitude stations that exceed the 4% intensity threshold
• Intensity increase are calculated from a 3-minute moving average counting rate relative to a 75-minute
Inuvik, Canada 18NM64 6.6×105 68.4 N, 133.7 W, 21 m baseline extending from 85 minutes to 10 minutes before the current time
Fort Smith, Canada 18NM64 7.4×105 60.0 N, 111.9 W, 203 m • During the 4.4-year period of our backtesting study, the false alarm rate for Watch, Warning, and Alert
Peawanuck, Canada 18NM64 7.3×105 55.0 N, 85.4 W, 52 m was ~40/yr, less than 1/yr, and 0/yr respectively
Nain, Canada 18NM64 7.3×105 56.5 N, 61.7 W, 46 m Alert times decided from this algorithm in the past 9 GLE events were compared with the earliest alert issued
Thule, Greenland 18NM64 8.0×105 76.5 N, 68.7 W, 44 m by SEC/NOAA based upon GOES (100 MeV or 10 MeV protons) data
McMurdo, Antarctica 18NM64 9.4×105 77.9 S, 166.6 E, 48 m • Alert times produced by our system are ~10-30 minutes earlier than alert issue times from SEC/NOAA
• Alert times are also substantially earlier (around 60 minutes) than the time when dangerous amounts of
South Pole, Antarctica 3NM64 10.3×105 90.0 S, 0.0 E, 2820 m low energy particles reach the satellite (S2 storm level)
South Pole Bares* 6NM64 3.2×105 90.0 S, 0.0 E, 2820 m
*Bare : Neutron counters without the usual lead shielding that respond to a slightly lower These results suggest that our system can provide valuable added minutes of advance warning for radiation
energy primary cosmic ray than the standard monitor. events of concern for satellites, astronauts, and air crews.