Early Warning Systems Defined - Reducing Disaster: Early Warning Systems for Climate Change - page 95

Catalysts/Patterns : There is a need for a defi ned 'triggering' mechanism or regular
pattern for sending out information. A trigger could be anything from a quantitative
indicator to an anecdotal comment. A regular pattern needs to be frequent enough to
keep people engaged and familiar with the messages, but not so frequent so that
people get irritated or ignore the large volume of messages. The information sent
out is not necessarily only about a specifi c vulnerability or hazard. It could also be
a reassuring message that 'no hazard is imminent' or 'vulnerability has been
reduced'. 'All clear' or 'improvements have happened' messages are indeed impor-
tant components of EWS and they, too, need defi ned triggering mechanisms or
regular patterns. For a hazard example, when Mount Pinatubo in the Philippines
was ramping up to a major eruption in 1991, different warning alert levels were
developed and issued. After a higher alert level was decided, the EWS forbade the
alert level to drop lower until a mandatory waiting period had elapsed of 72 h (from
Alert Level 3 to Alert Level 2) or 1 week (from Alert Level 4 to Alert Level 3)
(Punongbayan et al. 1996 ; this alert level system has now been entirely revised).
Since explosive volcanoes can often quieten down for a short time before a massive
explosion, that delay helped to avoid complacency.
One challenge which every EWS needs to address explicitly is how to defi ne
success. From a hazard perspective, so-called near misses (such as when a warning
was not issued, but it was nearly needed) and false alarms (such as when a warning
was issued, but it was apparently not needed) should be defi ned for the EWS and
described as part of the EWS' performance metrics. Yet it is not clear that either near
misses or false alarms indicate failure (Handmer 2000 ).
Barnes et al. ( 2007 ) argue that, for US tornadoes, the way in which 'false alarms'
are measured and recorded does not do justice to forecasting accuracy and ability,
thereby obscuring instances where people did need to take action even without a
tornado touchdown nearby. They also suggest that so-called false alarms for torna-
does in the USA do not make people less likely to respond to tornado warnings in
the future; that is, their evidence is that the 'Crying Wolf' syndrome is not usually a
concern. In contrast, Simmons and Sutter ( 2009 ) found that a higher, recent false
alarm rate for tornadoes in the USA signifi cantly increases casualties from torna-
does in that area, suggesting that so-called false alarms for tornadoes do make peo-
ple less inclined to react to subsequent warnings. The discrepancy could be a result
of people's expectations in terms of an EWS' structure and function not being com-
municated properly, leading to expectations which cannot be fully met. As such, no
specifi c or universal answers can be given regarding what 'failure' means for
an EWS.
Similarly, metrics for 'success' can be defi ned only for each specifi c context.
During the 2004 Indian Ocean tsunami, the Pacifi c Tsunami Warning Center fi lled
its mandate admirably and without fl aw, issuing international warnings for a major
tsunami within minutes of the earthquake and using all available channels to dis-
seminate the message. Yet places such as Thailand and Sri Lanka did not evacuate
coastal areas despite having hours of warning lead time before the tsunami struck.