Long-Range Forecasting of Iceberg Numbers on the Grand Banks
Ingrid Peterson
0-2 month forecast system of the iceberg population on the Grand
Banks using multiple linear regression
Based on strong relationship between iceberg population and prior sea
ice extent upstream on the Labrador-Newfoundland shelf
Outline of talk
Introduction
Description of data, normalization
Forecast regression model
Results from two-year trial of forecast
Large tabular icebergs and ice islands off East Coast in 2001-2003:
Probable source
Motivation:
Logistics planning for the
offshore oil and gas industry: A
long-range forecast of the
iceberg population would help
offshore operators procure and
commit the appropriate number
of vessels and level of
manpower.
Map of Grand Banks area, showing
well-sites.
Map of iceberg drift pattern
In Baffin Bay and Labrador Sea
(Source: National Ice Center).
In Grand Banks area (Source:
International Ice Patrol).
Previous work
•
Annual iceberg severity predicted based on atmospheric pressure
gradients, sometimes in combination with air temperature or
isobaric height (eg. Smith (1931), Davidson et al., (1986)).
•
January ice extent in Davis Strait is as good or better predictor of
iceberg severity (Marko et al., 1986)
•
Iceberg severity is highly correlated with sea ice extent off
Newfoundland from February to May (r=0.86; Smith, 1931, pg.
185) and in April (Marko et al., 1994).
Data Sources:
Icebergs:
• Iceberg sightings database 1960 - present : International Ice Patrol
– Aircraft
• International Ice Patrol
• Canadian Ice Service (since early 1980’s)
• Provincial Airlines (since 1989)
– Ships
– Lighthouses
– Other (Oil industry, Defense Dept., etc.)
•
Number of icebergs drifting south of 48ºN (annual and monthly) 1900 - present :
International Ice Patrol
Sea ice:
• Digital regional ice charts (weekly) 1969 - present (ArcGIS format): Canadian
Ice Service
• Digitized ice charts (weekly) 1963-1968 (0.5º latitude x 1 º longitude): Canadian
Ice Service
Iceberg sightings
data:
•40-year frequency of
non-zero iceberg
sightings in each
1°latitude by
1°longitude square
(1963-2002)
•data from the
International Ice Patrol
Iceberg Sightings
Database.
Sea ice data:
•35-year frequency of
presence of sea ice in
each 1°latitude by
1°longitude square
(1969-2003)
•from the weekly
digital regional ice
charts produced by the
Canadian Ice Service.
Moderate season length: IIP
(a) Annual number of icebergs
passing south of 48°N
(b) spring sea ice extent east of
Newfoundland
(c) negative mean air temperature at
Cartwright, Labrador (NovemberMay)
(d) March number of icebergs
passing south of 48°N.
The dashed lines represent the
below-normal and above-normal
thresholds computed from the 19631992 data
Annual number of unique
iceberg sightings in each
latitude band.
Comparison of
number of icebergs
south of 48°N from
the I48 data set (red)
and the unique iceberg
sightings at 47-48°N
(blue) from the IIP
sightings database,
scaled to fit the I48
data from 1990-2001.
Marko et al,1994
Smith, 1931
Decadal mean monthly
icebergs south of 48˚N,
sea ice extent, and
negative air
temperature anomalies
at Cartwright,
Labrador.
Monthly histograms
showing the number of
flights at 47-48°N as a
function of the number
of sightings per flight.
Examples of flight
tracks (green lines) and
icebergs sighted
(crosses) from March 924, 1993.
O-2 Month Iceberg Forecast
[z(t+i)]1/2 = c1(t,i)x(t1) + c2(t,i)x(t2) + c3(t,i)y(t3) + c4(t,i) + ε
for i=0,1 and t1 , t2 , t3 ≤ t
z=number of unique icebergs
x=sea ice extent (area from 47-53˚N)
y=offshore distance of ice edge (at 47-49˚N)
Iceberg
Forecast
1-2month
Feb
Mar
Apr
May
Jun
Jul
0-1 month
Feb
Mar
Apr
May
Jun
Jul
t1
t2
Jan
Feb
Feb
Feb
Feb
Jun
Mar
Apr
Apr
-
Jan
Feb
Feb
Feb
Feb
Jun
Feb
Apr
Apr
Jun
-
t3
Mar
Jun
Mar
Apr
Jun
Jun
R
0.85
0.73
0.72
0.82
0.71
0.80
0.85
0.84
0.83
0.82
0.81
0.80
Sea ice extent at the beginning of months t1 and t2,
and offshore distance of ice edge along northern
flank of Grand Banks for t3 are used to forecast the
number of icebergs in the offshore area for the month
given in the first column.
•Observed monthly number of offshore
icebergs for 1963-2004 (grey line)
•Fitted number based on prior sea ice
extent in 1963-2002 for the 1-2 month
(blue line) and 0-1 month (red line)
forecast
•1-2 month (blue dots) and 0-1 month
(red dots) iceberg numbers that were
forecasted for 2003 and 2004.
•Dashed lines represent the below-normal
and above-normal thresholds computed
from the 1963-2001 data.
2003
2004
Correct
Obs
0-1
month
1-2
month
Feb
L
L
M
Mar
M
M
L
Apr
H
H
M
May
H
M
M
Jun
M
M
M
Jul
M
M
M
Feb
L
L
M
Mar
L
L
L
Apr
L
L
L
May
L
L
L
Jun
M
L
L
Jul
M
L
L
9/12
75%
5/12
42%
Categorical comparison (Low, Medium, High) between
observed and predicted iceberg numbers.
(a) Annual number of icebergs
passing south of 48°N
(b) spring sea ice extent east of
Newfoundland
(c) Annual number of monthly 1ºx1º
squares with icebergs sighted (mean
iceberg extent)
The dashed lines represent the
below-normal and above-normal
thresholds computed from the 19631992 data
Annual forecast
Date of
Forecast
issue
Annual Number of
Icebergs South of 48N
(No correction for data
before 1995)
Annual Number of
Icebergs South of 48N
(Correction for data
before 1995)
Annual Total Number of
Monthly Squares
(Mean Monthly Iceberg
Extent)
R
% Correct
(3
categories)
R
% Correct
(3
categories)
R
% Correct
(3
categories)
Jan 1
0.59
54
0.70
67
0.70
64
Feb 1
0.70
54
0.85
59
0.82
71
Mar 1
0.70
54
0.85
59
0.83
71
Apr 1
0.75
66
0.86
76
0.87
80
Critical R (95% confidence level, n=41)=0.31
Critical % Correct (95% confidence level, n=41)=46%
In 2002 and 2003, unusually large number of ice islands and
large tabular icebergs on Grand Banks, up to 20 million tonnes,
with drafts of 65-80m.
Very large tabular iceberg off Newfoundland (47˚N), Apr 2003.
• 480m L x 230m W x 10m H
• Photo courtesy of Pip Rudkin (PAL)
Very large tabular iceberg off Newfoundland (48˚N), May 2002.
•500m L x 300m W x 9m H, 75m draft.
•Photo courtesy of Pip Rudkin (PAL)
Ice Islands
•The term “ice island” was used in 1952 to refer to large ice masses
with lengths as much as 29km, observed in the Arctic Ocean and Arctic
Archipelago between 1946 and 1950 (Koenig et al., 1952) and thought
to have originated from the Ellesmere Island ice shelves (e.g. Ward
Hunt Ice Shelf).
•Ice island (WMO definition): “A large piece of floating ice protruding
about 5 m above sea level, which has broken away from an Arctic ice
shelf. They have a thickness of 30-50 m and an area of from a few
thousand square metres to 500 sq. km or more. They are usually
characterized by a regularly undulating surface giving a ribbed
appearance from the air.”
•Relatively thin tabular icebergs from several glaciers in north
Greenland often resemble ice islands from the Ellesmere Island ice
shelves (Helk and Dunbar (1953)).
Ward Hunt Ice Shelf
Ryder Glacier
Lancaster Sound
Petermann Glacier
Humboldt Glacier
Bylot Is.
GREENLAND
Sources:
(a) the Canadian Ice Service
(b) Stoermer and Rudkin (2002)
(c) the International Ice Patrol
Iceberg Sightings Database
(d) D. Christian of C-CORE.
Baffin Is.
LABRADOR
NFLD
GRAND BANKS
Fi
Map summarizing sightings of ice
islands and large tabular
icebergs in 2001-2003
Ward Hunt Ice
Shelf
Ryder Glacier
Petermann Glacier
Humboldt Glacier
Lancaster
Sound
Bylot
Is.
GREENLAND
Baffin Is.
LABRADOR
NFLD
GRAND
BANKS
Ice island #1 in Navy Board Inlet (74˚N), 23 Oct 2001
•~2.6km L x 2.3km W
•trapped within Navy Board Inlet/ Eclipse Sound from Oct 2001-Aug 2002.
•Photo courtesy Canadian Ice Service
Ward Hunt Ice
Shelf
Ryder Glacier
Petermann Glacier
Humboldt Glacier
Lancaster
Sound
Bylot
Is.
GREENLAND
Baffin Is.
LABRADOR
NFLD
GRAND
BANKS
Ice island #1 one year later off Baffin Island (70˚N) on 26 Oct 2002, with CCG
Icebreaker Louis St. Laurent
• ~2.4km L x 2.0km W x >9m H
Photo courtesy of Dan Crosbie (CIS)
Ward Hunt Ice
Shelf
Ice island
fragment #2
in Baffin Bay
(73˚N), 27
Oct 2001
Ryder Glacier
Petermann Glacier
Humboldt Glacier
Lancaster
Sound
Bylot
Is.
GREENLAND
Baffin Is.
LABRADOR
Believed to
be same ice
island
fragment 6
months later
off Nain
(57˚N), 19
Apr 2002
NFLD
GRAND
BANKS
Photos courtesy of CIS
Debris-covered tabular iceberg in Conception Bay (48˚N), 26 May 2002
•250-300m L x 50-60m W x 6-7m H
•Photo courtesy of Denny Christian (C-CORE).
Debris-covered iceberg in Kane Basin (80˚N)
•90m L x 70m W x 5-18m H
•From geological analysis, Petermann glacier is the most probable source
(Zentilli et al., 2003).
•Photo from Zentilli et al. (2003).
Ward Hunt Ice
Shelf
Petermann Glacier
•
Height of glacier front is 8-10m for eastern margin (1.3km
wide), 4-6m for central region (12km wide) and 2-7m for
western margin (2.2km wide): height for eastern margin is
consistent with height of ice islands/large tabular icebergs
(9-10m)
•
Width of glacier front is about 16km, with a speed of about
0.9km/year
•
Glacier produces an average mass of 530 million tonnes of
calf ice per year, or an area of 14km2 per year
•
Segments up to 12km x 10 km break away at intervals of 510 years or more (Higgins, 1991)
•
Large tabular iceberg with a length of about 750m and a
height of 4-5m was observed east off Newfoundland in 1976
(Robe et al. (1977). Three possible sources: the Ward Hunt
Ice Shelf, the Petermann Glacier, or the Humboldt Glacier.
•
Dunbar (1978): the far most probable source was the
Petermann Glacier, based on surface appearance and
height.
Ryder Glacier
Petermann Glacier
Humboldt Glacier
Lancaster
Sound
Bylot
Is.
GREENLAND
Baffin Is.
LABRADOR
NFLD
GRAND
BANKS
Petermann Glacier looking toward western side, 10 Aug 2003.
• Photo courtesy Dave Forcucci (USCG).
Steep valley walls on eastern side of Petermann Glacier (Hall Land), 10 Aug 2003.
• Photo courtesy Dave Forcucci (USCG).
MODIS Imagery: Petermann Glacier
Large
area
missing
31 Jul 2000
21 Aug 2000
03 Aug 2001
13 Aug 2001
Large
segment
broken
off
MODIS Imagery: Petermann Glacier
13 Aug 2001
07 Sep 2004
Position of calving front in July 2000
and August 2001.
Position of calving front between 1953
and 2001.
Sources: Higgins (1991), Dunbar
(1978), and Zhou and Jezek (2002).
Conclusions
•
A 0-2 month statistical forecast of the iceberg population was developed using regression
analysis and is based primarily on the strong relationship between iceberg numbers on the
Grand Banks and prior sea ice extent on the Northeast Newfoundland shelf. The regression
analysis uses data from the International Ice Patrol Iceberg Sightings Database and digital
ice charts from the Canadian Ice Service since 1963.
•
Iceberg numbers before 1979 were reduced to correct for re-sighted icebergs using a factor
derived from the “Number of Icebergs South of 48°N” data set. A second normalization was
needed for the years prior to the mid-1990’s because since then, the number of icebergs
reported close to 48°N have been high relative to (a) ice extent and air temperature, (b)
iceberg numbers to the south and (c) season length.
•
For 2003 and 2004, a 0-1 month and a 1-2 month iceberg forecast were prepared each
month, and results from the two-year trial of the forecasting system were presented.
•
Uncertainty in the iceberg numbers and their normalization remains a major challenge for
iceberg forecasting. A more robust measure of iceberg severity may be the number of onedegree squares containing icebergs.
•
Regarding the ice islands, the most probable source is the Petermann glacier based on
surface appearance and dimensions. The 2002-2003 event may have been caused by a large
loss of ice in 2000-2001.