NOTE: Charges only apply to receiving a copy by mail
Free Copy: TR 93-03, Water Equivalent vs. Rain Gauge Measurements From the March 1993 Blizzard (PDF File)
Use Adobe Acrobat
PDF Reader
by Neal Lott, Physical Scientist
August 1993
On March 12-14, 1993, the eastern seaboard of the U. S. was struck by
what is now referred to as: 1) The Storm of the Century, 2) The Blizzard of
93, or 3) The Big One! One of the interesting and more overlooked aspects of
the storm was the discrepancy between liquid water measurements by the rain
gauge and water equivalent 'core samples' of the snow/ice on the ground. In
examining these data, I found that many stations appeared to either have a
problem with 'undercatch' of snowfall by the rain gauges, or a systematic
problem with water equivalent measurements. This report will attempt to show
that: 1) The most likely scenario is a problem with 'undercatch' of
snowfall, and 2) the careful measurement of water equivalent is an important
element for hydrological interests and the climatic records.
Water equivalent measurements are taken by extracting core samples, or
slices, of the snow/ice on the ground, and then melting the sample to
calculate the water equivalent of the snow/ice. The measurement can be done
carefully with a rain gauge (e.g., an 8-inch sample of 16 inches on the
ground melts to 1 inch of water, indicating an 8-1 ratio and 2 inches of
water equivalent). It is very important that the sample be representative of
the full profile of the snow/ice depth. Care must be taken not to compress
or compact the snow for the sample, since the density of snow in the sample
should be the same as that on the ground.
The problem with 'undercatch' of snowfall by rain gauges has been
documented in previous studies (Larson and Peck, 1974; Peck, 1972). These
studies have established empirical relationships between gauge measurements,
water equivalent measurements of snow on the ground, and wind speeds. From
these relationships and field testing of the ensuing equations, unshielded
gauges were shown to 'undercatch' precipitation by 70% or more (40% or more
for shielded gauges) during snowfall events with wind speeds of 20 MPH or
higher. Suffice it to say that this storm presented an excellent opportunity
to observe this problem.
The accompanying table presents statistics on the measurements made
during the storm at 40 stations. By carefully studying the table, several
aspects become readily apparent. These aspects are discussed below.
The snowfall storm totals (INCR1) were quite substantial, with totals of
1 to 2 feet being common throughout much of the eastern U.S. These totals
are based on snow depth reports before, during, and after the storm, in order
to better correlate them with water equivalent 'core samples' taken at about
the same times. Generally, the greatest depths were observed on the 14th.
The water equivalent storm totals (INCR2) were also unusually high for a
snow event in this part of the country, with amounts of 2 to 4 inches being
rather common. If these totals are accurate, then the 'true' amounts of
liquid deposited by the storm should be equal to or greater than those shown
in the INCR2 column. In other words, the 'true' liquid total from a storm is
always equal to or greater than the final water equivalent measurement. The
observer's measurement of the greatest water equivalent amount (WTEQ2) does
not account for melting of snow/ice near a ground surface with above freezing
temperatures at the beginning of the storm. Since this storm was a late
season event, most of the affected areas did have above freezing ground
temperatures at storm onset, and, in some cases, several hours of snow fell
before actual accumulation began. Then, as heavy snow fell and accumulated,
some melting from underneath occurred throughout the storm at many locations.
This affect can be seen in looking at the synoptic-hour observations
(every three hours) of snow depth, which show a decrease in snow depth for
some stations during the hours immediately after the storm--with temperatures
still well below freezing--an indication of melting from underneath. Also,
the true amount of liquid would tend to be higher than 'INCR2' since most
stations reported some additional snowfall after their report of the maximum
water equivalent amount (WTEQ2).
'RATIO1' is a measure of the apparent undercatch by the gauges (rain
gauge vs water equivalent), although the factors mentioned above have to be
considered. For the 28 stations reporting 'SNOW' as the predominant
precipitation type, 4 are within a range of .01 to .40, 15 are within .41 to
.80, 8 are within .81 to 1.20, and 1 is over 1.20. In light of previous
studies mentioned above ('undercatch' precipitation by 70% or more...), these
figures are not surprising since wind speeds generally averaged 10-30 MPH at
most stations during the heavy snowfall. Also, they show more than half (15)
of the values to be within the second of these four ranges.
Some of the snow/water ratios (RATIO2) seem quite low (i.e., high water
content). However, given that this was a very deep 'spring-type' storm which
drew in large amounts of Atlantic and Gulf of Mexico moisture, it is likely
that the snow had a higher liquid content than would be expected in a
'normal' winter storm. Unfortunately, I am not aware of any studies which
have attempted to correlate the severity and type of storm system with the
resulting liquid water content of snowfall. Such a study is not within the
scope of this paper.
Although there were a wide range of temperatures among stations during
the heavy snowfall, the general range was 25-34 degrees Fahrenheit at storm
onset, falling to 10-25 degrees Fahrenheit when heavy snowfall ended. Dew
point depressions averaged about 2 degrees Fahrenheit throughout the storm.
The dew points are probably a better indicator of the liquid content of
snowfall in a storm system with large amounts of moisture entrained, since
they give us a 'reading' of atmospheric moisture content. Historically, dry
bulb temperatures at the surface have been used as a 'rough' indicator of
liquid content, with readings of around 30 degrees Fahrenheit indicative of
'wet snow.' However, a 30-degree dry bulb with a 25-degree dew point would
tend to indicate less liquid content than a 30-degree dry bulb with a
28-degree dew point.
In looking at this aspect further, I chose 9 stations along the
Appalachian chain where geography and the synoptic situation would be similar
for this event. The table below presents each station's mean dew point
temperature during the time of moderate to heavy snow (i.e., when most of the
snow fell) as compared to the snow/water ratio (RATIO2).
STATION MEAN DEW PT RATIO2
Asheville NC 30 4.2
Hickory NC 29 6.0
Charleston WV 24 6.8
Roanoke VA 27 7.6
Huntington WV 23 7.9
Beckley WV 20 8.1
Elkins WV 18 9.5
Pittsburgh PA 20 11.9
Binghamton NY 14 14.0
One would expect that as the dew points fell, the snow/water ratios would
increase. The table shows this to generally be true. To investigate this
further, the correlation coefficient can be calculated which will estimate
the affect of the mean dew point on the snow/water ratio. A 'perfect'
correlation here would be -1.00 since a decrease in the mean dew point yields
an increase in the snow/water ratio. The actual correlation coefficient for
these values is -.90, and 80.3% of the variation in snow/water ratio is
accounted for by a linear relationship with the mean dew point. Also, by
restricting the list to Asheville, NC plus the four West Virginia stations
(five stations with very similar terrain influences), the coefficient
'improves' to -.98, and 96.1% of the variation in snow/water ratio is then
accounted for by the relationship with mean dew point. Although this cannot
be said to validate the water equivalent measurements for these stations, it
certainly adds more credibility to the values.
Other factors that may come into play in a storm such as this would be
the proximity of the station to large bodies of water, and the location of
the station to the leeward/windward side of mountain ranges. However, in my
evaluation of the data (for all 40 stations), I did not find any direct
correlations with these factors. As for the distribution of 'RATIO2' values
for the 28 'SNOW' stations, 3 are within the .1 to 4.0 range, 13 are within
4.1 to 8.0, 10 are within 8.1 to 12.0, and 2 are over 12.0. Therefore,
nearly half (13) have ratios of from 4.1 to 8.0. This is certainly a
significant departure from the 10.0 ratio that is commonly used when an
actual measurement is not taken.
The inconsistency of reports between stations is especially apparent for
'RATIO1' and 'RATIO2.' As the distributions shown above indicate, the data
do show some clustering of values within certain ranges. However, there is
considerable variability in the data and spatial (i.e., geographic)
clustering cannot be shown. A few cases of large differences over short
distances (e.g., Rochester vs Syracuse, NY) may indicate a problem in the
validity of a few of the values. Some of these variations can probably be
explained by geographic/terrain differences; snow melt during the storm at
some locations (melting underneath mentioned above); and possibly some subtle
differences in observing practices between stations.
Other important considerations are the wind speeds associated with the
storm and the existence (or not) of windshields for each of the gauges. Some
of the National Weather Service (NWS) stations in the Northeast are equipped
with windshields, while most in the Southeast are not. Wind speeds generally
averaged 10-30 MPH for most stations during the moderate to heavy snowfall,
but were quite gusty. Estimating an average wind speed for the storm for
each station is of questionable value due to this gustiness, and since the
wind's affect on the gauges would depend on the gauge exposure. However,
data for mean wind speeds and the existence of windshields are included in
the data table (see WND and SHLD columns) to provide some indication of how
these data correlate with the 'undercatch' of the gauges (RATIO1). Following
are correlation coefficients calculated from these data by correlating
'RATIO1' with 'WND':
19 'SNOW' stations without windshields = -.39
9 'SNOW' stations with windshields = -.33
Here, a 'perfect' correlation would be -1.00 since 'RATIO1' should
decrease as 'WND' increases. These poor correlations may be due not only to
the factors mentioned above, but also to variations in snow density (i.e.,
snow weight, which is directly proportional to water content and inversely
proportional to 'RATIO2'). Heavier, wetter snow would tend to be less
affected by the wind than drier snow. However, I am not aware of any
previous research of this effect. Correlations can be calculated using
groups of stations with similar values for 'RATIO2' and/or similar
geographical influences, but I found most combinations to only yield
coefficients of between -.30 and -.60. It is my opinion that these low
correlations are the result of the uncertainties cited above as well as
possible inconsistencies in reporting practices among the various stations.
TABLE
MARCH 12-14, 1993 "STORM OF THE CENTURY"
All stations which reported at least 1 inch of water equivalent
of snow/ice on ground at some point during storm:
STATION DPTH1 DPTH2 INCR1 WTEQ1 WTEQ2 INCR2 DPTH3 GAUGE PRECIP RATIO1 RATIO2 WND SHLD
Albany NY 4 28 24 1.2 3.2 2.0 28 1.83 SNOW .91 12.0 20 YES
Allentown PA 0 18 18 0 1.7 1.7 16 2.01 SNOW 1.18 9.4 23 YES
Asheville NC 0 18 18 0 3.8 3.8 16 1.85 SNOW .49 4.2 15 NO
Beckley WV 0 30 30 0 3.7 3.7 30 2.00 SNOW .54 8.1 18 YES
Binghamton NY 13 35 22 3.2 4.7 1.5 34 1.00 SNOW .67 14.0 23 YES
Bristol TN 0 13 13 0 1.1 1.1 12 1.35 SNOW 1.23 10.9 15 NO
Buffalo NY 4 16 12 .3 2.0 1.7 15 .82 SNOW .48 6.5 26 NO
Burlington VT 13 31 18 2.6 3.7 1.1 25 .59 SNOW .54 10.9 18 NO
Caribou ME 20 36 16 6.5 11.8 5.3 36 1.05 SNOW .20 3.0 21 YES
Charleston WV 0 19 19 0 1.9 1.9 13 1.16 SNOW .61 6.8 13 NO
Chattanooga TN 0 20 20 0 1.8 1.8 18 1.44 SNOW .80 10.0 18 NO
Cleveland OH 3 11 8 .7 1.7 1.0 7 .55 SNOW .55 4.0 25 NO
Concord NH 6 23 17 2.2 4.5 2.3 17 .75 SNOW .33 4.8 17 NO
Elkins WV 1 19 18 0 1.9 1.9 19 1.22 SNOW .64 9.5 13 NO
Erie PA 4 17 13 .3 1.3 1.0 16 .70 SNOW .70 12.0 31 NO
Hickory NC 0 10 10 0 1.5 1.5 9 1.76 SNOW 1.17 6.0 17 NO
Huntington WV 0 22 22 0 1.9 1.9 15 1.08 SNOW .57 7.9 16 YES
Jackson KY 0 20 20 0 3.1 3.1 20 .47 SNOW .15 6.5 16 YES
Knoxville TN 0 15 15 0 1.6 1.6 12 1.49 SNOW .93 7.5 12 NO
Mansfield OH 2 9 7 .3 1.3 1.0 9 .51 SNOW .51 7.0 25 NO
Pittsburgh PA 0 25 25 0 2.1 2.1 25 1.12 SNOW .53 11.9 22 NO
Portland ME 17 34 17 4.4 6.1 1.7 34 1.58 SNOW .93 10.0 24 NO
Roanoke VA 0 16 16 0 1.7 1.7 13 1.97 SNOW 1.16 7.6 18 NO
Rochester NY 7 25 18 1.7 8.1 6.4 25 1.09 SNOW .17 2.8 32 NO
Syracuse NY 5 37 32 1.6 3.5 1.9 34 2.03 SNOW 1.07 15.3 20 YES
Wilkes-Barre PA 1 21 20 0 1.4 1.4 12 1.24 SNOW .89 7.9 18 NO
Williamsport PA 1 15 14 0 1.8 1.8 13 1.10 SNOW .61 6.7 15 NO
Worcester MA 9 26 17 3.2 6.1 2.9 26 1.23 SNOW .42 5.9 24 YES
Baltimore MD 0 9 9 0 2.3 2.3 9 2.48 SNOW, IP 1.08 3.9 25 NO
Hartford CT 0 16 16 0 2.0 2.0 16 2.05 SNOW, IP 1.03 8.0 18 NO
Philadelphia PA 0 12 12 0 1.5 1.5 11 1.80 SNOW, IP 1.20 7.3 31 NO
Washington-Dulles 0 13 13 0 1.8 1.8 13 1.55 SNOW, IP .86 7.2 19 NO
Boston MA 1 12 11 0 5.0 5.0 9 1.95 MIXED .39 1.6 44 NO
Bridgeport CT 0 10 10 0 2.2 2.2 9 2.64 MIXED 1.20 4.1 34 NO
JFK Apt NY 0 9 9 0 2.1 2.1 8 2.39 MIXED 1.14 3.8 28 NO
LaGuardia Apt NY 0 9 9 0 3.7 3.7 8 2.49 MIXED .67 2.2 35 NO
Newark NJ 0 13 13 0 2.7 2.7 13 2.81 MIXED 1.04 4.8 23 YES
Providence RI 0 6 6 0 2.0 2.0 4 2.58 MIXED 1.29 2.0 23 NO
Washington-Natl. 0 6 6 0 1.3 1.3 5 2.31 MIXED 1.78 3.8 20 YES
Wilmington DE 0 10 10 0 2.4 2.4 9 2.33 MIXED .97 3.7 21 NO
KEY:
DPTH1 = Snow depth in inches before the storm (on March 11).
DPTH2 = Greatest snow depth in inches during storm.
INCR1 = Snowfall storm total in inches (DPTH2 - DPTH1) as calculated by
subtracting the depth before the storm from the greatest depth
reported. The actual snowfall totals may be slightly higher.
WTEQ1 = Water equivalent of snow/ice on ground before storm (on March 11)
in inches and tenths.
WTEQ2 = Greatest water equivalent of snow/ice on ground during storm
in inches and tenths.
INCR2 = Water equivalent storm total (WTEQ2 - WTEQ1) in inches and tenths.
DPTH3 = Snow depth in inches at time of WTEQ2 report.
GAUGE = Liquid 'catch' by rain gauge (storm total) in inches and hundredths.
PRECIP = Predominant precipitation type during storm:
IP = ice pellets
MIXED = snow, ice pellets, and rain
RATIO1 = GAUGE / INCR2. This is the amount of precipitation caught
by the gauge as a proportion of the WTEQ storm total. In
theory, this value should always equal or exceed 1.00
since melting from underneath the snow cover (due to above
freezing ground temperature) and precipitation after the
WTEQ2 report are not accounted for in the INCR2 column.
In effect, this is an estimate of the gauge 'undercatch.'
RATIO2 = (DPTH3 - DPTH1) / INCR2. This is a measure of the water
content of snow/ice from the storm by calculating the ratio
of snow/ice accumulation to water equivalent. DPTH3 minus
DPTH1 is the storm's snow/ice total at the time of the water
equivalent measurement used for calculating INCR2.
In effect, this is an indication of the average weight of the
snow and/or ice from the storm. Over the years, stations not
taking water equivalent measurements have often assumed a
value of 10.0 for this calculation.
WND = The average wind speed in MPH during the storm for the period when
moderate or heavy snow and/or ice pellets were reported.
SHLD = The existence (YES or NO) of a windshield for the rain gauge.
Of the 28 stations which reported 'SNOW' as the predominant precipitation
type, 23 (82%) show gauge 'catch' lower than the water equivalent storm total
(see RATIO1 column). However, of the 12 which reported 'SNOW, IP' or 'MIXED'
as the main type, only 4 (33%) show gauge 'catch' lower than the water
equivalent storm total. Therefore, those stations which received less
snowfall and more ice pellets/rain showed a much lower tendency for
'undercatch.' Also, 15 of the 28 'SNOW' stations (54%) have a RATIO1 value
of from .41 to .80, with a mean for all 28 stations of .68. These statistics
indicate significant 'undercatch' of snowfall by the gauges.
The mean for 'RATIO2' for 'SNOW' stations is 8.2. This seems to indicate
a rather wet snow as compared to the typical 10-1 ratio that we're accustomed
to using. (Perhaps it's time to reevaluate this 'typical' ratio, since
snowfall often has a somewhat lower or higher ratio than 10-1.) These
statistics also point to a need for further study of the methods used in
measuring water equivalent of snow/ice on the ground, as the variability of
the data indicate possible problems with a few of the values.
However, this should not diminish the importance of water equivalent
measurements/data for climatic records and for hydrological interests (river
forecasting, etc.). In fact, considerable flooding occurred in parts of the
eastern U.S. shortly after the storm mainly due to snow melt. In 'extreme
events' of this nature, it would be wise for hydrologists and climatologists
to take note of how the water equivalent reports compare with the rain gauge
reports. This is especially true for a month such as March 1993 when this
event contributed greatly to the month's precipitation total (based on rain
gauge measurements), but where the 'official' totals for the month probably
fell significantly short (20% or more in some cases) of the actual liquid
amounts received. In summary--the water equivalent reports are very
important for the climatic records--not only to get a true picture of the
liquid amounts received, but also to provide a baseline for studying the
problem of 'undercatch' by gauges.
As to what can be gained from all of this, I suggest the following:
a. Although water equivalent has not been one of the more used/studied
meteorological elements, this storm is a prime example of the usefulness of
these data when measured correctly. I would encourage future observing
practice standards (i.e., the FMH) to emphasize the measurement of 'core
samples' (even in the age of automation). The measurements are especially
important to hydrological interests and climatological records, as discussed
above.
b. Recent studies have indicated that the Canadian 'Nipher' shield may
be the best available due to its structure allowing for better 'catch' of
snowfall. The use of this shield could be implemented on a 'test' basis at
several stations with frequent snowfall. Optimally, a three-way test could
be conducted with the Nipher shielded gauge vs the standard NWS shield vs an
unshielded gauge. Of course, the gauge type would have to be the same in
each instance. These data could be compiled with follow-on recommendations
for precipitation measurements.
c. Some equations have been developed (Larson and Peck, 1974) for
estimating the 'true' liquid amount during snowfall events. These equations
use (as input) the rain gauge measurement and estimated average wind speeds.
One equation is used for shielded gauges and another equation for unshielded
gauges. This storm would be an excellent case study for the application of
these equations. Such a follow-up study could add to the table shown in this
report by calculating:
1) Estimated 'true' liquid amounts using the equations mentioned
above.
2) Comparison of these estimates to the water equivalent storm
totals (INCR2).
This report has shown that the March 1993 "Storm of the Century"
presented an excellent opportunity to study the problem of 'undercatch' of
snowfall by rain gauges. Also, it has shown that the accurate measurement of
water equivalent is important for both hydrological interests and for the
climatic records. In fact, these measurements provide one of the bases for
studying the afformentioned 'undercatch' problem. For further information
about this storm, you may contact the National Climatic Data Center (phone
704-271-4800, fax 704-271-4876, internet orders@ncdc.noaa.gov) in Asheville,
NC. We have a complete report about the storm, along with several digital
datasets of observations taken during the "Storm of the Century."
L. Larson and Peck, E.L., 1974: Accuracy of Precipitation Measurements for
Hydrologic Modeling. "Water Resources Research," 857-862.
Peck, E.L., 1972: Snow Measurement Predicament. "Water Resources Research,"
244-248.
 Top of Page
Please see the NCDC Customer Support page if you have questions or comments.
|