NASA Technical Reports Server (NTRS) 19970006984: Ice Processes and Growth History on Arctic and Sub-Arctic Lakes Using ERS-1 SAR Data

14 Pages·1.3 MB·English

by  NASA Technical Reports Server (NTRS)

/
NASA-CR-202626 /J
Polar Record 31 (177): 115-128 (1995). Printed in Great Britain. J_'f._ _ _¢'Y& 115
Ice processes and growth history on Arctic and sub-Arctic
lakes using ERS-1 SAR data
K. Morris, M.O. Jeffries, and W.F. Weeks
Geophysical Institute, University of Alaska, PO Box 757320, Fairbanks, AK 99775-7320, USA
Received June 1994
ABSTRACT. A survey of ice growth and decay processes on aselection of shallow and deep sub-Arctic and Arctic lakes
was conducted using radiometrically calibrated ERS- 1SAR images. Time series of radar backscatter data were compiled
for selected sites on the lakes during the period of ice cover (September to June) for the years 1991-92 and 1992-93.
A variety of lake-ice processes could be observed, and significant changes inbackscatter occurred from thetime of initial
ice formation in autunm until the onset of the spring thaw. Backscatter also varied according to the location and depth
of the lakes. The spatial and temporal changes in backscatter were most constant and predictable at the shallow lakes
on the North Slope of Alaska. As a consequence, they represent the most promising sites for long-term monitoring and
the detection of changes related to global warming and its effects on the polar regions.
Contents the result of studies of ice on shallow lakes on the North
Introduction 115 Slope of Alaska. Since the early t970s, they have been
Study lakes and methods 116 studied occasionally, and primarily in late winter, using
Background on lake-ice growth, structure, airborne X- and L-band real aperture radars in combina-
and backscatter 116 tion with surface observations and measurements (Sellmann
Shallow Arctic lakes 117 and others 1975a, 1975b; Elachi and others 1976; Weeks
Shallow sub-Arctic lakes 120 and others 1977, 1978, 1981). Only one time series of
Deep lakes 121 radar images of lake-ice development in this region was
Discussion 124 compiled (Mellor 1982). This early research identified
Conclusions !26 some of the factors that affected backscatter and showed
Acknowledgements 127 that it was possible to differentiate between floating ice
References 127
and ice that was frozen to the lake bottom, and thus make
Introduction inferences about water depth. Unfortunately, those studies
relied on uncalibrated photographic products and the analy-
Lakes represent an important natural resource in the Arc-
sis was qualitative. Recently, radiometrically calibrated,
tic. Many communities depend on lakes that do not freeze
geolocaled European Remote Sensing Satellite- 1(ERS- 1)
completely to the bottom for their fresh-water supply
synthetic aperture radar (SAR) data (C-band), available in
during the winter. These lakes also provide a habitat in
digital fi)rm from the Alaska SAR Facility, have been
winter for animal life. Consequently, the ability tomonitor
combined with field observations and modelling in more
the freeze-up and break-up dates of lakes -- that is, the
duration of the ice-covered season -- and to determine quantitative studies of backscatter variations inrelation to
which lakes do not freeze to the bottom inthe winter isvery ice-growth processes and structure throughout the ice-
growth season on shallow North Slope lakes (Jeff ties and
important for wildlife and water resources management.
others t993, 1994; Wakayabashi and others 1993a, 1993b).
In recent decades, some areas of the Arctic have experi-
enced a climate wanning (for example, Chapman and Because spaceborne SARs such as ERS- 1can obtain
Walsh 1993), and numerical climate models suggest that images of the polar regions regardless of light and cloud
the effects of global warming will be experienced first and conditions, and because the data are available asgeolocated
will beamplified inthe polar regions (for example, Manabe and radiometrically calibrated digital products, they offer
and others 1992). There isevidence that the duration of the previously unavailable opportunities for quantitative stud-
lake-ice cover isclosely related toclimate variability, and ies oflake-ice growth processes and history throughout the
that changes in the length of the ice-covered season could winter. In order to realise the full potential of SAR for
have significant consequences for lake hydrology and studies of high-latitude lake ice in relation to climate
biology (Palecki and others 1985; Palecki and Barry 1986; change, and wildlife and water resources management, it
Schindler and others 1990). Since the timing of lake is important to understand the lake-ice growth processes
freeze-up and break-up is primarily a function of air and history that are being observed by these instruments.
temperature variations, the availability of a long-term This paper presents further results from the study of the
record of these lake-ice benchmarks could contribute to shallow, Alaskan North Slope lakes, and broadens the
the identification of the onset and subsequent effects of scope oflake-ice investigations using SAR toinclude deep
climate change at high latitudes. and shallow Arctic and sub-Arctic lakes. Lake-ice
Remote sensing provides the most practical method for backscatter variations observed inERS-1 images and their
long-term continuous monitoring and study of polar lakes. relationship to fitctors such as ice growth, thickening and
Much of the present understanding of high-latitude lake- freezing to the bottom, the development of air inclusions
ice properties and processes by remote-sensing methods is inthe ice, ice defl_rmation, and tile efl_'cts of preciffitalion
116 MORRISJ,EFFRIEASN, DWEEKS
considered, as fieldwork would be
B_rrowLakes ',__.' ';.l=._
performed on these lakes (Jeffries
(.ttt,K(.H! ,.. '_ _ ..... Bt-AL.'FORT SI'5'_ _ t :
and others 1994). The latitude and
• 'B' Lakes "-* _. j4:$+, _ _" 1. l [ longitude of each site in an ERS-I
SAR image were identified and the
closest obtainable co-ordinates in all
of the other inmges were sampled.
"-_. ALASKA wl.,,ke_{ \ _-j;> .Lake ]
Since the standard processed data are
geometrically corrected with the
reslituted state vectors of the orbit,
Kwiguk _ ) l i' "_h I which introduces an error into the
..._.',]l [ Tazlinalmke [". CANADA _.. %;.j registration of the extracted coordi-
nates, itwas sometimes necessary to
_ ,4 _¢,'_'*_ I "" -'-, _ " - /
visually select the appropriate pixels
.... } t.akesf" _'¢ - ,:r,, /"ih _,. ' -J" ! for sampling. Because many of the
:;"_ _" ?_ .I,'l.r or.._+,_.ss'A'-_'y._;;'. . !
lakes are quite small or have distinc-
tive shorelines, this did not pose a
serious problem. The o'_value repre-
sents the mean ofa nine-pixel square
that has the site location as its center.
Fig. 1.Map showing the location of individual lakesor groups of lakes investi-
The backscatter of undeformed ice
gated inthe study.
adjacent to deformation features in
and air temperature changes on the snow and ice surface the early stage of ice development was also determined at
are discussed. some lakes. When possible, that is, when sufficient data
were available, time series for consecutive seasons (1991-
Study lakes and methods
92 and 1992-93) were compiled and compared. Data
The locations of the lakes chosen for this study are shown
availability was primarily dependent onwhether SAR data
in Figure 1. The lakes include:
acquisition had been scheduled by the European Space
1. small shallow (many <2 m deept Arctic lakes
Agency.
located on the North Slope ofAlaska (Barrow lakes
Weather records lk_rthe nearest meteorological station
at 71°13'5"N, 156°39'W; 'B' lakes at 70°22'5"N,
were obtained. In some cases, the records of several
156°30'W, and 'C' lakes at68°45'39"N, 156°15'W)
and in the interior of Alaska near Fort Yukon stations were consulted for a single group of lakes. This
(66°45'N, 145°30'W); was necessary because the records for the closest station
2. two groups of sub-Arctic lakes (Kwiguk lakes in were incomplete or because there isno single station close
the Yukon delta at 62°40'N, 163°20'W, and the to the desired lake location and its weather conditions had
Naknek lakes near Bristol Bay at 58°37'N, 157°W); to be inferred from the records of several stations over a
and broad region. The weather records were used to investi-
3. two deep-water lakes, one sub-Arctic (Tazlina Lake gate whether particular backscatter changes correlated
at 61°52'N, 146°45'W) and one Arctic (Great Bear with meteorological changes.
Lake at 66°5'N, 119°W).
A total of 233 ASF Standard low-resolution ERS-1 Background on lake-lee growth, structure, and
baekscatter
SAR images obtained from the Alaska SAR Facility were
used for this study. These images cover a ground area of Much of the present knowledge of lake-ice growth, struc-
100 km by 100 km and have aspatial resolution of 240 m ture, and backscatter is based on investigations of ice on
and a pixel size of 100m. Acquisition times for the images the shallow lakes of the North Slope of Alaska. This
cluster around midday and midnight local time. For the section summarizes the present state of knowledge and
quantitative analysis of backscatter variability within a provides a basis fl)rthe interpretation of the SAR images
single scene and between multiple scenes, these data were of the lakes that were investigated in this study.
converted from DN values toradiometrically calibrated o° Lake ice iscomprised primarily ofcongelation ice that
values using MacSigma-0 public-domain software devel- isthe product of downward growth of ice crystals into the
oped at the Jet Propulsion Laboratory. Backscatter (o"_] water as heat isconducted through the ice from the growth
time series from September to June were created for interface to the atmosphere. This ice can be either clear or
selected sites on the lakes. The lakes to be studied were contain bubbles or both. Typically, the ice on the shallow
selected from an ERS- 1SAR scene of the area acquired in lakes of the Alaskan North Slope has a bi-layer character,
late winter 1992. Lakes that are inclose proximity toeach with a layer of clear ice, sometimes several decimeters in
other and that have interesting and differing backscatter thickness, overlying a layer of ice containing tubular
returns were chosen. In the case of the Barrow lakes, easy bubbles (Weeks and others, 1977, 1978, 1981; Mcllor
accessibility to the lakes from the town-site was also 1982; Jcffries and others 1993, 1994). The tubular bubbles
ICE PROCESSES AND GROWTH HISTORY 117
() I I I I 1 I I I I I I I I I I I I I
1991 1992 1993
-5 o b° -5
i_ _ _" °°° 7_
E-
o h q,. n /
Z -10 -I0
i !
z_ T_ "._// e -15 7Z
Z
,-.]
-20 I' P d -20
SITE I GO
• SITE 5 ,--]
,-<
-25 v' SITE 6 -25 -'--
< • SITE 7 et'__
t.r, SITE I1
-30 I I I ! I ! I I I I I I I I I I I I I I -3O
270 3(X)330 360 31) 6(1 90 120 150 180 210 240 270 300 330 360 30 60 90 120 150
JULIAN DAYS
Fig. 2. Time series of backscatter intensity (o°)for selected Barrow lakes from September 1991 to June 1993.
Several recurring features are identified: (a) the onset of freezing at about Julian day 270; (b) the prolonged
maxima from floating ice from mid-winter to spring; (c) the grounding of ice on the bottom of lakes; (d) the
prolonged minima from grounded ice until spring; and (e) the onset of melting in May.
are oriented with their hmg axes parallel to the growth ice on backscatter from the North Slope lake ice is un-
direction. The clear ice represents the earlier stages of ice known. On the other band, Hall and others (1994) sug-
growth when all gases are rejected back into the water gested that it may be an important source of backscatter
during freezing, and the appearance of the tubular bubbles from alpine lakes.
probably represents the time when these shallow reser-
Shallow Arctic lakes
voirs become supersaturated with gas (Jeffries and others
1994). Backscatter time series were compiled for the period
September 1991 to June 1993 for the North Slope lakes
As long as the ice remains afloat, there will be specular
(Barrow, 'B', and 'C'), and for the period September 1991
reflection of the radar signal off the basal ice-water inter-
to May 1992 for the Fort Yukon lakes. The Barrow lakes
face because of the strong dielectric contrast between the
data are representative of all of these datasets; thus, the
two media (Weeks and others 1978; Leconte and Klassen
following discussion will focus primarily on the Barrow
1991). When there are few or no inclusions present, this
lakes with reference to the three other shallow Arctic lake
specular reflection leads to most of the signal being re-
data sets as necessary. The time series for the Barrow lakes
flected away from the radar and the ice appears relatively
is illustrated in Figure 2. A number of annually recurring
dark in a radar image. Ice that is frozen to the bottom of a
features arc:
lake also appears dark in SAR images because there is a
low dielectric contrast between the ice and lake sediments 1. a steady increase in backscatter from early autumn
to early January;
and the signal is absorbed into the lake bottom; conse-
2. a backscatter maximum that is attained at many
quently, there is little or no return to the radar (Weeks and
lakes in early January and is maintained until April/
others 1977, 1978, 1981 ;Mellor 1982; Leconte and Klassen May;
1991; Jeffries and others 1993, 1994; Wakabayashi and 3. a backscatter decrease at a number of lakes at
others 1993b). Ithas been noted that, since tubular bubbles different times during the winter, with minimum
occur in both floating ice and grounded ice on the shallow backscatter levels maintained thereafter until April/
North Slope lakes, the tubular bubbles must act as forward May;
scatterers producing a strong return to the radar from 4. rapid backscatter changes in late April/early May;
and
floating ice and bright signatures in SAR images (Weeks
and others 1977, 1978, 1981; Mellor 1982; Jeffries and 5. high backscatter at all lakes during the short sum-
mer.
others 1993, 1994).
These features also occur in the time series for 'B'
After initial freeze-up, alayer of snow accumulates on
lakes, 'C' lakes, and Fort Yukon lakes.
the ice surface. The base of the snow can be flooded by
water reaching the surface through cracks formed by Initial ice formation and early stages of ice growth
thermal stresses or the weight of snow on the ice surface. During the short summer and immediately prior to freeze-
Freezing of the wet snow forms snow-ice that contains up, radar returns from the Barrow lakes are high, probably
many spherical bubbles. Jeffries and others (1994) sug- due to strong backscatter from the typically wind-rough-
gested that the magnitude of the effect of the bubbly snow- cncd watcr surfaccs. The onset of freczing in mid- to late
118 MORRIS, JEFFRIES, AND WEEKS
ens on all the lakes (Fig. 2).
I)EFORMATI()N F[iAI'I RE NtlMBIR
[ 2 "_ 4 5 l_ 7 H U IO [[ A visual representation of
{I the increase in backscatter
l I I I l I I I I I I .n
• 29September 1991j > during the early stages of ice
growth at Barrow is shown in
-4 Figure 3. In late September,
the lakes are the clearly vis-
ible dark areas, which con-
2C} trast with the lighter tones of
the tundra (Fig. 3a). As time
25
progresses, backscatter from
the lakes increases and the
I I I I I I I I I Ii ;= contrast between the lakes
8October 1991
f. "z-:' and the tundra decreases (Figs
3b and 3c). By late Novem-
-4-4 ber, the lakes are not readily
distinguishable from the sur-
7
--, rounding land (Fig. 3d). AI-
zm though the lakes typically
have a dark surface during
-'_2S
the early stages of ice growth,
I IIIIIIII1 there are localized areas of
tl
. I Iltllllll 26October1991 =e. strong returns, which are vis-
g.
5 ible as narrow, bright, and
generally linear features in
.-,- the SAR images (Figs 3a and
_ _ i °
15 = 3b). These are probably de-
formation features on the ice
surface. They might be cracks
related to thermal stresses
25
generated as the thin ice cover
IIIIIIIIli {11 cools, or they might be small
tI
IIIIIIIII 22November 1991 ridging or rafting features that
h. S z develop as the wind fractures
>_ and displaces the initial, thin
IH -_4-4 ice cover. In either case, the
_: strong backscatter from these
IS
= features is probably due to
2N _. their rough, multi-faceted,
o Deformation featturt.
angular surfaces.
• Non-deformed ice
Mean and standard de-
=c
I I I I I I I 1 I 1 viation O"°values of I1 se-
2 _ _ 5 (_ 7 h L1 Ill I lected deformation features
DI-FC)RM,VII()N F:ti,\IL RE NI 7,,IBI R
and the adjacent undeformed
Fig. 3. Time sequence of SAR image sub-scenes of the tundra and lakes near Barrow ice are shown inFigures 3e to
and the differences in mean backscatter intensity between the deformation features 3h. Soon after the initial ice
visible on the lake-ice surfaces and the surrounding undisturbed floating ice from 29 formation in September, the
September to 22 November 1991. The numbers in(a) identify the bright deformation backscatter contrast between
features for which backscatter data are shown in(e) through (h). The sub-scenes have
the deformation features and
ground dimensions of 21.3 km by 15.3 km and are taken from ASF Standard low-
resolution images with scene IDnumbers: 8989200 (a); 8438200 (b); 1068200 (c);and the undeformed ice is high
21213200 (d). SAR scenes are copyright ESA. (Fig. 3e). As icegrowth con-
tinues during the autumn, the
September (around Julian day 270) is typified by a sharp contrast decreases until there is only a roughly 2 dB
decrease in backscatter (Fig. 2). This is probably due to difference (Fig. 3h) and the deformation features 'disap-
specular reflection away from the radar off the smooth pear' (Fig. 3d). The final 'disappearance' of the deforma-
upper and lower surfaces of the new, thin, continuous ice tion features in late November might be due to the initial
cover (Jeffries and others 1993, 1994). After initial freeze- development of the arrays of tubular bubbles as these
up, backscatter values rise steadily as the ice cover thick- shallow reservoirs are nolonger able to absorb completely
ICEPROCESSAENSDGROWTHHISTORY 119
thegasebseingexpellefdromthegrowingice.Forward thaw. They suggested that the reversal at these lakes was
scatterinogffthebubblesin,combinatiownithspecularrelated todifferences inice surface properties and surface
reflectioonffthebasaicle-wateinrterfacwe,hichshowas
slope between the floating and grounded ice covers.
highcontrascto,uldincreasbeackscatttoearleveclompa- Changes in the snow cover also must contribute to
rablewith,andsubsequengtrleyatethrant,hebackscattersome of the backscatter change. At temperatures <0°C,
fromthedeformatiofenatures. snow may be considered to be dry and thus transparent to
Periodic backseatter decreases and the backscatter the radar signal since there isno 'free' liquid water present
maximum (Leconte and others 1990). However, once the snow cover
Between late October and late January, backscatter from warms sufficiently for 'free water' to be present, the radar
three of the lakes decreased sharply at different times, signal is absorbed and backscatter is reduced (Stiles and
while backscatter increased and reached a maximum at Ulaby 1980; Rott and others 1992; Hall and others 1994).
other lakes (Fig. 2). The sharp backscatter decrease This should apply whether lake-ice cover is floating or
occurred at the same three lakes at different times each grounded, and thus backscatter should decrease at all
winter -- for example, at site 11, backscatter decreased lakes, notjust from the floating ice. Itshould be noted that,
earlier in autumn 1992 than it did inautumn 1991. Those in both 1992 and 1993, the acquisition time of all the
lakes where backscatter decreased are interpreted as hav- images but one was nominally noon local time. Thus, for
ing frozen completely to the bottom. In any given winter, the spring images acquired inMay and June, amoist snow
cover may be considered an increasingly important factor
the variations in the timing of the backscatter decrease
inthe radar returns. Acontributing factor tothe backscatter
reflect differences in water depth between lakes and thus
increase from the shallow lakes might be related to lake
the time required for the growing ice to reach the lake
vegetation. Hall (in press) has suggested that tundra
bottom. The differences in timing between winters prob-
vegetation protruding through a shallow autumn snow
ably reflect differences in the timing of initial ice forma-
cover causes a roughening of the snow surface relative to
tion inthe autumn and subsequent ice-growth rates, which
the C-band wavelength resulting in an increase in
are primarily afunction of air temperature and snow-cover
backscatter, especially when the vegetation is wet. In a
depth. A number of Barrow lakes that froze to the bottom
similar fashion, vegetation protruding through adiminish-
also had identifiable deformation features early inthe ice-
ing spring snow and/or ice cover on a shallow lake may
growth season, and these features remained visible after
cause astrong backscatter regardless of the state of the ice
the ice grounded. The fact that these surface features
surface.
remain visible after the ice has grounded, unlike those on
Spatial and year-to-year backscatter variability on the
floating ice that are obscured as the ice thickens but
North Slope
remains afloat, provides an indication of the magnitude of
The backscatter time series for the Alaskan North Slope
the signal loss in the lake bed and the relative strength of
lakes are quite similar from year to year, both in terms of
the surface scattering from the deformation features.
the spatial and temporal variability and the magnitude of
Those lakes where backscatter remained at a constant
the values (Fig. 4). Figure 4a shows two backscatter
high level after early January are interpreted as having an
records for aBarrow lake where the ice remained afloat all
ice cover that remained afloat all winter. Jeffries and
winter in 1991-92 and 1992-93. There is very little
others (1994) have drawn attention to the fact that
difference between the curves, except for an apparent
backscatter from the floating ice covers reaches a maxi-
missing backscatter minimum during freeze-up in 1992,
mum and saturates inJanuary, before the ice cover and the
which was due to data being unavailable between early
layer of ice containing tubular bubbles have reached their
September and early October. The backscatter records for
maximum thickness. It has been suggested that either the
one of the 'B' lakes that grounded in 1991-92 and 1992-
layer of ice with forward-scattering tubular bubbles need
93 are similar, except for the greater length of time it took
only be a few centimeters thick to cause such strong
for backscatter to change from high to low when the ice
backscatter, and/or there isalimit tohow much backscatter
grounded in 1993 (Fig. 4b). Prior togrounding, maximum
the tubular bubbles can cause (Jeffries and others 1994). backscatter was close to that maintained at the Barrow
The spring thaw lakes all winter (Fig. 4a). The timing of initial ice forma-
In May (Julian days 120-150) 1992 and 1993, unusual tion at the Barrow and 'B' lakes is coincident and clearly
backscatter variations occurred (Fig. 2) at the Barrow identified by the backscatter minimum in late September
lakes. At those lakes where backscatter was consistently (Figs 4a and 4b). The backscatter record for a 'C' lake that
high from January onwards, backscatter decreased sharply was frozen to the bottom for most of the winter shows a
to values similar to those from lakes where the ice was negligible year-to-year difference between the magnitude
aground for much of the winter. At those lakes where of the backscatter values (Fig. 4c). These o'_values are
backscatter was consistently low, backscatter increased similar to those from grounded ice at the 'B' lakes (Fig. 4b)
sharply to values similar to those from ice that had been and Barrow lakes (Fig. 2). These backscatter records for
afloat all winter. Jeffries and others (1994) first reported the Barrow, 'B', and 'C' lakes also show the onset of the
this unusual backscatter reversal, which occurs at the onset spring thaw, which occurs at almost the same time each
of air temperatures >0°C and during the subsequent spring year atthe same lake, and almost concurrently inany given
120 MORRISJ,EFFRIEASN, DWEEKS
backscatter differences is that the ice at
JUI.IAN DAYS site 5 was frozen to the lake bottom in
2_4 275 _IMI 7125 t511 2S &} 75 RiO 125 1511
-7- (t "I ..N..Il'liSII9..1..t92t I..... I.... I'" .... I.... I.... I .... t.... I.... Ii'''li." "5 _ lfarotezenDecteomtbheer,botatonmd theiniFceebarutasriyte. 3Twhaes
Y- '_ ice at the other lakes remained afloat all
[ I • • o o _ °• • ° _ i°e •
I i. _ winter.
7_ Oi i *o_1 . B 0 _5_ Beginning in early March,
15 backscatter from the floating ice de-
b<- 2{1 o 2.> creased, while that from the grounded
G
ice increased, converging at about -11
-25 -25 g dB in mid-March (Fig. 5). This back-
.SEn'. iOCT. NOV. I)1!{'. JAN tl;B. MAR. APR. MAY JUI
i
tl1 "111,,,i t, ,,I .,,I .... I1, tt.,l ii ill •,I .... I.... I.... I,,," tl) scatter change occurred when mean daily
II 'I .... i.... i.... i.... i'' .......'._. '..."._' '' d! temperatures were above freezing for 13
SI11i(,(qI/92)
• Sllli 60_2/9 I) o 5 of 18days from 7to 24 March. Conse-
quently, there was probably an increase
m \ 0
Z I0 g i• • /,/ -I0 in the free-water content of the snow,
z o o o o __. with resultant absorption of the radar
_ -f5 , @ II o oi o: e--,'% ;o. signal. The large drop in o'_values at all
o
_ -211 c_ 2O sites on 24 March may be the result of
I .-a
the cumulative effect of those above-
i i 2S
SEPT. (X'T, NOV. [)I';C. JAN. I'liB. I MAR. APR. MAY [JUI freezing temperatures combined with
30 .i .... t .... I.... i .... I.. .,.i, ,iiI.... I.... I.... '1 .... li, , precipitation on 23 and 24 March (total
D
.', "i .S..I.TE 7i(.9..1.f_)2) i .... 1. ;, ,i ' ' .... l.... I ....l I .... l''" _I .... I''12'. 17mm water equivalent). The o°values
-5 • S[11]7lq2.,aal1 - 5 _, rose rapidly in the following few days,
_._ reaching levels similar tothose immedi-
OI. IO >"
-I(I " _ ately prior to the precipitous drop of 24
0
Z t" _ _'2 March. Mean daily temperatures below
15 o 0 C_ ' (I •( 0,. .o " o % ogq -5°C which persisted until the end of
"< 21} 2. =_ the month, probably froze the wet snow
i '_ -e4_ cover, and reduced the free-water con-
,;EFF. OCT. NOV. _I)EC. JAN. I.I!1,',. iMAR. APR. MAY JUb
_= tent, thereby reducing absorption of the
,I .... I.... i.... i.... 1.. ,., I.... ii i1. I.... i....
251 275 Illl _25 _%(1 2_ 5_t 75 [INI 125 I_t radar signal and increasing backscatter.
JtlLIAN DAYS The relative lack of data for April
and May makes itdifficult to make any
Fig.4. Backscatter time series for the 1991-92 and 1992-93 iceseasons for
detailed assessment ofthe processes that
selected sites onshallow Arctic lakes: a.afloating Barrow lake; b.agrounded
'B' lake; and c. agrounded 'C' take. might have been occurring onthe lakes.
Itappears that byearly May the iceon all
year at all the lakes. The decrease in backscatter that the lakes had melted. On 7May, backscatter from Kvichak
occurs at sites where the ice is floating (Fig. 4a) and the Bay at the northern end of Bristol Bay was very strong due
increase that occurs where it isgrounded (Figs 4b and 4c) to wind roughening the water surface. Wind-roughened
are evidence of the onset of the spring thaw. water probably also accounts for the high radar returns
Shallow sub-Arctic lakes from the lakes at the end of the time series (Fig. 5).
The Naknek lakes are located adjacent to Bristol Bay in o 0
southwest Alaska, and the Kwiguk lakes are located inthe _
Yukon River delta (Fig. 1). Each set of lakes issituated in _ -5 -5
a sub-Arctic marine environment. A backscatter time
series for January to June 1992 was compiled for the _ -10 I0
Naknek lakes (a complete winter record could not be
Z
compiled because no data were available for autumn ,,,-15 -15
1991). A similar backscatter time series was compiled for
the Kwiguk lakes as well as a complete backscatter time _ -2o t l s'r'''] _ i " -20
EI st,,;5
series for winter 1992-93. _,
FI " sill s
v
Naknek lakes < -25 -25
Hi SITI 9
The backscatter record l_)rthe Naknek lakes was similar to ,',".E,.,7.",.,, ,MA",..
3O 3O
that tbr the Barrow lakes, that is,there was astrong contrast
25 50 75 IIX) 125 150
in backscatter between site 5and the other lakes from late
JULIAN DAYS
December to late February, and at site 3 backscatter
changed from strong to weak inearly February (Fig. 5). As Fig.5.Backscatter timeseries from December 1991 toMay
1992 for sites on the shallow sub-Arctic Naknek lakes.
with the North Slope lakes, the interpretation of the
ICE PROCESSES AND GROWTH HISTORY 121
strongly enhanced backscatter (M_itzler
o and Schanda 1984). During May 1992,
._ backscatter decreased atmost sites prob-
ably indicating the onset of the spring
_.__ thaw. Only site 7had strong backscatter
at this time, perhaps due to vegetation
15 protruding through the ice cover.
_' The contrast in timing of the
-20
._ backscatter maximum, and the apparent
-<
-2__ absence of grounding in 1993 at the
_o Kwiguk lakes might be a reflection of
(} "' .... ' .... '"'i_9'i ....1993l .... l .... l ' ' I" I ' ' ' rI .... l'" '- 0 winter temperature differences. In win-
: I b'i ter 1992-93, temperatures often reached
:.- 5 i -5 '_ almost O°C and the weather was not as
v I #---T ' ,.
z -_o_ persistently cold as it had been the pre-
:_ I(1 v6i:! •ova'i _."A,, I-- "_ U vious winter. Consequently, ice-growth
.".:.e./' ,• SSIIT'II'IE2 A I", " rl , d5 7_ rates might not have been as rapid and
,;_i-A_ zx
-2._ the ice did not grow as thick as itdid the
•"' mITESIT4E53SII"E i; [ t' \ ,,I ,'/,,"
20 I • SITE 6 ! previous winter, and thus the ice at most
I ," SITE7 ' I #
,_<Q. 25 I I ! -25 - sites did not ground on the bottom. A
" ,X'T t NOel'. DI;XT.,!qN.,l._7_y_'.._,_...i ._,_.R..,._.._y.i!L:- -3o colder winter in 1991-92 is also sug-
-3(1 I.... I .... ia
275 3011 325 351) 25 50 75 I(XI 125 150 gested by the 1992 backscatter record,
JUI.IAN DAYS
which indicates that the onset of the
spring thaw occurred later in 1992 than
Fig.6. Backscatter time series fortheshallow sub-Arctic Kwiguk lakes from:
it did in 1993; thus, the ice cover had a
a. January to June 1992; and b.September 1992 toJune 1993.
longer duration. It should be noted that
Kwiguk lakes the backscatter values from most of the grounded sites in
The 1992-93 baekscatter time series (Fig. 6b) for the 1992 are neither as low as those at site 7(Fig. 6a) nor those
Kwiguk lakes shares some features with time series for the from grounded ice at the shallow Arctic lakes (Figs 2and
Barrow lakes. From an initial backscatter minimum asso- 4).
ciated with freeze-up in late September, backscatter rose
Deep lakes
steadily for the remainder of the winter until a maximum
Tazlina Lake
was reached in March. During March, there was a short
period of fairly constant maximum backscatter values Tazlina Lake is a sub-Arctic lake 22 km long located in
before they decreased in April at the onset of the spring south-central Alaska on the north side of the Chugach
thaw. None of the lakes appear to have frozen to the Mountains (Fig. 1). Although Tazlina Glacier islocated at
bottom in winter 1992-93. the south end of the lake, the two are not indirect contact.
The winter 1992 backscatter time series (Fig. 6a) was The depth of the lake isunknown, but, inview of its glacial
quite different from that observed at the same time the origin, it is assumed to be much deeper than the shallow
following year (Fig. 6b). The consistently low backscatter Arctic and sub-Arctic lakes discussed above. Backscatter
from site 7 from early January 1992 onwards (Fig. 6a) was measured at 14locations along the length of the lake
suggests that itwas grounded while the ice atthe other sites (Fig. 8d) and backscatter time series were compiled for
was afloat. However, itshould benoted that the backscatter each location during winters 1991-92 and 1992-93. None
values for site 7 were unusually low; only the initial ice of the 14 sites is believed to have grounded at any time
cover onthe shallow, Arctic lakes had asimilar value (Figs during either winter because of the lake's depth.
2 and 4) and it was not sustained through a period of A selection of the 14 backscatter time series is shown
months, as was the case at site 7. At those sites where the in Figure 7. Both the 1991-92 and 1992-93 records have
ice apparently remained afloat, maximum backscatter was features similar tothose observed at the shallow Arctic and
reached on different dates inJanuary, unlike the following sub-Arctic lakes. However, there are noticeable differ-
year when maximum backscatter occurred in March. Af- ences between the two Tazlina Lake records. In each time
termaximum backscatter was reached atthe Kwiguk lakes series, there isabackscatter minimum inlate December or
inJanuary !992, itdecreased tominimum levels until mid- early January, probably evidence of final freeze-up and the
March. This might be aperiod when the ice was grounded establishment ofastable icecover. Thereafter, backscatter
on the bottom. During April 1992, amoderate backscatter values increase steadily as the ice thickens. However,
increase occurred (Fig. 6a) when airtemperatures atnearby whereas a backscatter maximum was attained in late
Emmonak fluctuated about 0°C. This could have caused February 1992, in 1993 it occurred two months later, in
thawing and refreezing of the snow cover, creating ice April. One exception tothe steady increase in backscatter
layers, which are known to cause specular reflection and in 1992-93 was site 3, which had strong backscatter
122 MORRISJ,EFFRIEASN, DWEEKS
JUI.IAN DAYS throughout the winter. It is located at the
northern tip of the lake, where the strong
325 350 25 _) 75 I(X) 125 150
I) t 9"l''9'9.'.z.................... backscatter is evident as a bright, roughly
'1'''' | I I I I _1 1 a,.'"
>- -5 _7.•,: SSIITTEE 23 i• i -5 tpriraonbgaubllayr aaprleuag ooff diceeform(Feigd. 8ic).e thTahtiswaiss
[-,
: SITE 7 I I_ blown into this region from further south
-I(I _:"_V::w_ -tO ;i.,.
Sll'l._ 12 onthe lake early inthe ice-growth season,
[--
Z :l sm_13 _ .I..'W. -/-•, before a stable ice cover was established.
-i5 g,/o
,,,,, Similar features have been observed on
kid [] ° l
Canadian lakes and rivers (Leconte and
._ -20 • i -20
Klassen 1991)• During April 1993,
i
_ 25 ! backscatter values decreased sharply, prob-
< •
i , -25 ably as a result of the onset of the spring
-30 ,,,J,AN. I ..].. FEB I. ,i,, MARI.,,,, AIPR.......... ,MAY JUN,, 30 thaw. The timing of the onset of the spring
0 0 thaw is not readily apparent in the 1992
record.
-- 5 -5
7_ Pronounced spatial backscatter varia-
[...,
tions occurred on the ice cover at Tazlina
Z -ill -I(I 2.-
Lake. In December 1992, apart from the
z -is -15 strong backscatter from the deformed ice
Z--I at the northern end of the lake, the ice had
rr_
_ 211 Z afairly uniform dark tone with a network
_o
.--I ofbrighter, narrow, linear features criss-
-<
-25 -25 crossing the lake (Fig. 8a). On 24 January
.<
1993, this network had become more
-30 prominent (Fig. 8b) and by 9February, it
325 350 25 54) 75 100 125 150
had expanded across much of the southern
JUIJAN DAYS
two-thirds of the lake (Fig. 8c). These
Fig. 7. Backscatter time series forselected sites onTazlina Lakefrom: a. bright features are interpreted as icedefor-
December 1991 to June 1992; and b. November 1992 toJune 1993. mation features, which cause strong
Fig.8. Time sequence ofSAR image sub-scenes of Tazlina Lake from 20 December 1992 to 28 February 1993. A
network of deformation features isclearly visible on 20 December (a). These features appear to expand spatially
and increase indensity (b and c) until 28 February, when the lake takes on a near-uniform tone (d). The sampling
sites marked in (d) were used to compile the backscatter intensity time series in Figure 7b. The sub-scenes have
ground dimensions of 15.7 km by 35.4 km and are taken from ASF Standard low-resolution images with scene ID
numbers 78092200 (a); 60029200 (b); 60126200 (c); and 60228200 (d). SAR scenes are copyright ESA.
ICE PROCESSES AND GROWTH HISTORY 123
Fig. 9. Time sequence ofSAR image sub-scenes of Tazlina Lakefrom 9February to 28 March 1992. Anetwork of
deformation features isfaintly visible on 9 February (a). Two zones ofhigh backscatter are evident at the north and
south ends ofthe lake (zones Aand B),withazone of lowerbackscatter inbetween (zone C). Zone Bbecomes very
bright on 27 February (b). Onemonth later thelake has taken on amottled appearance (c). The sampling sites seen
in (c) were used to compile the backscatter intensity time series in Figure 7a. The sub-scenes have ground
dimensions of 15.7 km by 35.4 km and arefrom ASF Standard low-resolution images with scene IDs:65470200 (a);
65501200 (b); and 13250200 (c). SAR scenes are copyright ESA.
backscatter, surrounded by undeformed ice with lower have changed the nature of the snow cover sufficiently to
backscatter. The increased density of the deformation cause higher and near-uniform backscatter from the lake
network may be due, inpart, to katabatic winds that sweep on 28 February.
down the glacier and are funnelled between the mountains The spatial variability of backscatter in winter 1991-
on either side of the lake, thus exerting strong wind stress 92 differed sharply from that in 1992-93. One significant
on the ice. This hypothesis issupported bythe fact that the difference was that any deformation features that existed
density of the deformation is greatest close to the glacier in 1992 (Fig. 9a) were not as well defined as in 1993 (Fig.
and decreases at the north end of the lake, where katabatic 8c). This might be an indication that the ice was less
wind stress will be lower and the effects on the ice cover deformed in 1992 than in 1993, or it might indicate that
will be reduced. Regardless of location on the lake, the other factors, such as a modified snow cover, played a
deformation features all but 'disappeared' by28 February stronger role in affecting backscatter in 1992. In winter
(Fig. 8d). The 'disappearance' of deformation features at 1992, there was strong zonation of backscatter across the
the Barrow lakes was attributed tothe initial development lake at any given time; zones A, B, and C had different
of the arrays of tubular bubbles in those shallow lakes. backscatter characteristics that gave the lake a non-uni-
However, it is unlikely that such dense arrays of tubular form appearance (Fig. 9a). The contrast between the
bubbles develop at Tazlina Lake, if they develop at all, strong backscatter from zone Bat the south end of the lake
because the lake is deep and the gases expelled during ice compared to the lower backscatter from zone C in the
growth are absorbed bythis large reservoir. An alternative central part of the lake was particularly striking (Fig. 9b).
explanation for the 'disappearance' of deformation fea- Similar backscatter variability has been observed at alpine
tures isachange inair temperatures. At both Gulkana and lakes inMontana, where ithas been attributed to localized
Glenallen weather stations, 50km cast-northeast ofTazlina flooding of the snow cover and snow-ice formation (Hall
Lake, daily maximum temperatures >0°C occurred on 26- and othcrs 1994). Jcffrics and others (1994) have drawn
28 February. Similar temperatures at Tazlina Lake could attention to thc fact that flooding of the basc of the snow
124 MORRISJ,EFFRIEASN, DWEEKS
Fig. 10. The extensive net-
work of deformation features
in McTavish Arm, the north-
eastern section of Great Bear
Lake, is clearly visible in this
ERS-1 SAR data acquired on
1January 1992 (ASF image
ID14507200). Thesub-scene
hasground dimensions of71.6
km by71.3 km. This network
becomes established in De-
cember and persists until the
spring melt. The numbered
locations are the sampling
sites used to created the
backscatter time series inFig-
ure 11. SAR scene is copy-
right ESA.
coveriscommoonnAntarcticseaice,andthatstrong ment of astable icecover, backscatter rises steadily and in
backscathtearsbeenreportefdromfloodeidceduetothe early January reaches maximum values, which are main-
strongdielectriccontrasbtetweetnheuppelrayeorfdry rained for the rest of the winter (Fig. 11). The time series
snowandtheunderlyinsglush(Lytleandother1s990)A. does not show the onset of the spring thaw because data
similarmechanismmighctaussetrongbackscattferorm were not available for McTavish Arm after early May. By
floodedlakeice.Oncetheslushfreezess,trongback- 27 May, deformation features were barely visible in SAR
scattemr ightcontinubeecausoefthebubblynatureof images ofthe lake-ice cover immediately west of McTavish
snow-ice.Regardlesosf thespecificcauseosfthe Arm, suggesting that the spring thaw had begun.
backscattvearriabilityinFigure9,itseemlsikelythat The SAR images used to investigate lake-ice processes
backscattinereachzonewascontrollebdydifferenftac- onGreat Bear Lake showed that astable icecover was not
torso,therwisthee lake might be expected to have amore established until late December. In subsequent images, the
uniform appearance, such as that observed in 1993 (Fig. position of the delbrmation features remained the same, an
8d). indication of the stability of the ice cover once it had been
established. Prior to late December, ice was forming on
Great Bear Lake
the lake, but itwas subject towind stress and displacement.
Great Bear Lake is a deep Arctic lake located in the
The network of deformation features rellects this initial
extreme northwest of the Canadian Northwest Territories
instability. The presence of ice of different ages and
(Fig. 1). Backscatter time series were compiled for 13sites
thicknesses, resulting from the convergence and diver-
inMcTavish Arm, located inthe northeast region of Great
gence early in the ice-growth season, probably accounts
Bear Lake. The SAR image of McTavish Arm (Fig. 10)
for the variability in backscatter from the darker-toned ice
shows a dense network of deformation features. The
adjacent to the deformation features.
strong backscatter from these features is exemplified by
Discussion
sites i and 11, located along the northern and southern
shores of the arm, respectively. Backscalter from the ice Mean o+ values li>r grounded ice and floating ice were
adjacent to the deformation features has a much darker calculated forallcategories of lakes (Fig. 12). The shallow
appearance, with variable grey tones. A selection of the Arctic lakes have fairly consistent mean backscatter val-
backscatter time series (Fig. 11)shows apattern similar to ues for both l]oating and grounded ice, from site tosite and
that observed at the other lakes, that is,after the establish- year to year, with a mean o° value of-6.4 +_0.5 dB lor all