tal ice control structure in 1981. Since that time,
Perception stage bias
there have been no major ice jam floods.
One bias that must be considered for ice events
reported by all sources is the perception stage.
Perception stage, as defined by Gerard and
Spatial bias
A spatial bias toward ice events at USGS gaged
Karpuk (1979), is the minimum stage that must
sites also results from reliance on USGS records,
be attained before a particular source will perceive
so that ice events at sites without gages are likely
and report an event. Different sources will have
to be underrepresented or missed entirely. For ex-
different perception stages for the same event, and
ample, we know from anecdotal evidence that the
the reliability of each source must be assessed
Waits River near Bradford, Vermont, experiences
when comparing stages reported by different
both freezeup and breakup ice jams almost an-
sources. The effect of this bias is that ice events
nually, yet the database contains only four entries
that are not damaging or that have stages less than
for this ungaged site. The information sources are
the perception stage are often unreported.
one newspaper story each in 1964, 1976, and 1992,
For example, ice events that flood a residential
and brief reports by two CRREL personnel that
area may be reported, but similar ice events that
mention ice events in 1973, 1976, and 1992. It is
do not cause flooding may not be reported. Or,
probable that many more ice events could be
events that took place before development may
identified through a search of local newspapers
simply not have been observed by anyone and
and historical records.
therefore have gone unreported. It is also possible
CRREL's proximity to many ice jam sites in
that ice events that did not cause an annual peak
New England has introduced another spatial bias
stage will not be reported in the USGS series on
into the database. The database is more likely to
contain information for ungaged sites, and from
concerned with peak events. In other words, if an
sources other than the USGS, for locations in the
ice event that caused a high winter stage is fol-
New England states than for other states. Like-
lowed later in the year by a higher open-water
wise, ice jam problem areas that have been the
flood stage, usually only the open-water stage is
focus of detailed study by CRREL or another gov-
reported.
ernment agency, such as the lower Platte River
Perception stage can vary; for the USGS, gages
Basin in Nebraska or the Salmon River near
might have a precise perception stage at which
Salmon, Idaho, will have more complete informa-
water goes overbank. Sometimes perception stage
tion than a gaged site that has not been the focus
can be quite high. For example, when contacted
of such a study.
by telephone about ice jam flooding at Akiak in
1983, Hugh Olsen reported "It's not really a prob-
lem (but) I'm standing in about four inches talk-
Single source bias
The database's heavy reliance on the USGS
ing to you" (Campbell 1983).
sources at this time also introduces a source bias.
Because of the perception stage bias, the data-
For example, ice events that took place at gaged
base should be used with care when developing
sites, but outside the period of record, will be
ice jam stage-frequency curves. Even unspectacu-
underreported. For instance, the USGS gage on
lar ice events can be quite important.
Collyer Brook at Gray, Maine, operated during
water years 1965 through 1982. In 11 of these 18
Potential bias must be considered
years, the maximum annual gage height resulted
All of these examples underscore the need to
from an ice jam at or downstream from the gage.
consider potential biases when evaluating infor-
The annual USGS Water-Data Reports (e.g., USGS
mation in the database. If data for a particular site
1981) is the only information source listed. No
appear to be biased, additional sources should be
other sources were identified, with the result that
examined to determine whether the apparent bias
there are no ice event entries for years outside the
is real or a reflection of actual conditions. As a
gage's period of record, although it is very likely
general rule, bias effects can be minimized if the
that there were many such events.
database is augmented by other sources, such as
In a case such as this, additional sources (e.g.,
old newspapers and other historical documents,
local newspapers) must be reviewed to obtain any
photographs, field notes made by USGS person-
ice information for years outside the period of
nel during normal winter discharge measure-
record, as well as for more complete information
ments, and field observations by others, such as
on the events listed in the database.
Corps or emergency response personnel.
5
Previous Page
