NASA Technical Reports Server (NTRS) 19980023498: Climate Sensitivity Studies of the Greenland Ice Sheet Using Satellite AVHRR, SMMR, SSM/I and in Situ Data

20 Pages·1.3 MB·English

by  NASA Technical Reports Server (NTRS)

NASA-CR-Z05169
Meteorol. Atmos. Phys. 51,239 258{19931
Meteorology
and Atmos pheric
Physics
t Springer-Verlag 1993
Printed in Austria
(_ : " /_,P' UC r:
551.583:551:579.3(988)
Cooperative Institute for Research in Environmental Sciences, University ofColorado, Boulder, Colorado, U.S.A.
Climate Sensitivity Studies of the Greenland Ice Sheet Using Satellite
AVHRR, SMMR, SSM/I and In Situ Data
K. Steffen, W. Abdalati, and J. Stroeve
With 16 Figures
Received January 25, 1993
Revised February 8, 1993
Summary the hemispheric circulation through its topography
and strong temperature gradient towards the
The feasibility of usingsatellite data forclimate research over
northwestern Atlantic. The present ice sheet has
the Greenland ice sheet is discussed. In particular, we
a surface area of 1.75-106 km 2 and a volume of
demonstrate the usefulness ofAdvanced Very High Resolution
Radiometer (AVHRR) Local Area Coverage [LAC) and 2.65" 106 km 3. These correspond to 11 and 8_o of
Global Area Coverage (GAC) data for narrow-band albedo the global glacier surface area and the ice volume,
retrieval. Our study supports the use of lower resolution
respectively (Ohmura et al., 1991). The ice volume
AVHRR (GAC) data for process studies over most of the
is also equivalent to 6.7 m of water relative to the
Greenland icesheet. Based on LAC data time series analysis,
present global sea level. The ice sheet has a typical
wecan resolve relative albedo changes on the order of 2 5%.
In addition, we examine Scanning Multichannel Microwave shield shape climbing relatively steeply from sea-
Radiometer (SMMR)and Special Sensor Microwave Imager level to 1000m and then gently to the highest
(SSM/I) passive microwave data for snow typing and other point of 3210m a.s.l, at 72.3'_N and 38.0°W,
signals ofclimatological significance. Based on relationships where the thickness of ice is estimated at about
between insitu measurements and horizontally polarized 19
3000 m. The average thickness of the ice sheet is
and 37GHz observations, wet snow regions are identified.
1515m.
The wet snow regions increase in aerial percentage from 9°0
of the total ice surface in June to a maximum of 26% in The Greenland ice sheet is a remnant of the
August 1990. Furthermore, the relationship between bright- Pleistocene ice sheet. During the last ice age the
ness temperatures and accumulation rates inthe northeastern volume was estimated at 2.92 to 5.59"106km 3
part of Greenland is described. We found a consistent
(Ohmura et al., 1991), suggesting that the surface
increase inaccumulation rate for the northeastern part ofthe
ice sheet from 1981 to 1986. altitude was higher in the past. Presently the
equilibrium line of the ice sheet lies at 1800 m a.s.l.
1. Introduction at the southern end, and it descends to 750 m a.s.1.
in the northern Greenland (Ambach and Kuhn,
1.1 Greenland Ice Sheet - Overview
1985). The accumulation and ablation area occupy
The Greenland ice sheet is one of the major ice 82 and 18%, respectively. Within the accumulation
sheets in the world. It is the largest one in the area the dry snow line is found at 3100m in the
Northern Hemisphere and is generally thought to south descending to 1650 m in the north, with the
have survived throughout the Quaternary. The surface area of the dry snow zone being approxi-
ice sheet presently plays an important role in mately 50,q_ of the entire ice sheet.
240 K. Steffenet al.
1.2 Greenland Ice Sheet - Climatology permanent cyclones, the Baffin Bay low to the
west and the larger Icelandic low to the southeast.
Snow and ice are key variables in the global The ice sheet is located under the weak saddle
climate system. They can influence the global heat between the two depressions. The summer circu-
budget through their regional feedback mechanism lation is dominated by the pressure ridge extend-
in the varying exchange of heat, moisture and ing from the northeast towards the center of
momentum between the surface and the atmosphere the ice sheet. Changes in the surface radiation
(Dickinson et al., 1987). Therefore, accurate infor- balance and/or in the cloud amount over the ice
mation on cryospheric variations (e.g. large scale sheet due to global climate change may alter the
surface albedo increases or decreases due to change overall pressure distribution and consequently
in dry snow area extent) are essential for the the precipitation. It is beyond the scope of this
prediction of future climate change. Information paper to derive precipitation amounts, but the
on the response of the ice sheet to climate warming aerial extent of dry snow area and its seasonal and
is of crucial importance for reliable projection of annual variations as derived from satellite data
future sea level. As the Greenland ice sheet is will help to understand the present situation.
located in a relatively warm climate, it is particu-
larly vulnerable to climate change. Present com-
1.3 Greenland Ice Sheet - Research Camp
puter model simulations of the greenhouse scenario
predict a regional cooling for the Greenland ice Scientists from the Swiss Federal Institute of
sheet, contrary to the general warming trend Technology (ETH) in Ziirich established a perma-
over the Arctic sea ice areas (Manabe et al., 1992). nent research camp at the equilibrium line of the
Trends for the annual temperature computed Greenland ice sheet near Jakobshavn (69° 34' N,
from land surface stations for the period 1961 to 49° 17' W)in spring of 1990 (Fig. 1).Climatological
1990 show a slight cooling for the southern half measurements were carried out to study the energy
of Greenland (Chapman and Walsh, 1993). How- exchange and the mass balance above, at, and
ever, our present knowledge of the mass and below the ice surface during April-September
energy exchange, and its sensitivity to environ- of 1990 and 1991. Furthermore, the structure of
mental conditions, is insufficient to make reliable the atmospheric boundary layer within the kata-
quantitative estimates of future changes associ- batic wind regime on the ice sheet was recorded
ated with the greenhouse warming. An observed with daily radiosonde profile measurements. Pre-
0.23 m/year thickening of the Greenland ice sheet liminary results from the two field seasons are
south of 72°N was reported by Zwally (1989) summarized in progress reports (Ohmura et al.,
based on satellite radar altimeter measurements. 1991, 1992). The research station has been trans-
The northern part of the ice sheet was not covered ferred in summer 1992 to the University of Colo-
by satellite altimeter measurements and at this rado at Boulder to continue the field experiments
point we don't know if the Greenland ice sheet for three additional seasons.
north of 72°N is growing or shrinking. Micro-climatological measurements at a single
Annual total precipitation and annual accumu- point are necessary and essential for the study of
lation on the Greenland ice sheet were evaluated the magnitude of the different energy fluxes such
by Ohmura and Reeh (1991). The analysis isbased as the turbulent fluxes and radiation balance at
on accumulation measurements of 251 pits and the ice sheet surface. However, a single point
cores obtained from the upper accumulation zone measurement isnot sufficient to discuss the clima-
and precipitation measurements made at 35 tology of the entire Greenland ice sheet. Therefore,
meteorological stations in the coastal region. The we propose the use of multispectral satellite data
mean annual precipitation for all of Greenland to derive certain surface parameters which are
was found to be 340 mm water equivalent, with an relevant to the surface climatology. The purpose
increasing gradient from north to south. The of this paper is to demonstrate the feasibility of
amount of precipitation is regulated primarily by satellite remote sensing as a technique for climate
atmospheric conditions, such as stability, water- studies over large homogeneous snow and ice
vapor content and circulation. The winter covered areas. The ultimate goal of the project is
circulation is strongly dominated by two semi- the statistical analysis of the regional and seasonal
Climate Sensitivity Studies of the Greenland Ice Sheet 241
,- y, 85ON
85oN
/; Nord
ETH/CU
Station
Fig. 1. Overview of the Greenland ice sheet.
The outline of the ice sheet is shown with a
dotted line, and the land is depicted with a
solid line. The location of the joint research
camp of the Swiss Federal Institute of Tech-
nology and the University of Colorado at
Boulder is shown at the west side of the ice
sheet near the coastal village of Jakobshavn.
The labels A through K give the locations
of the SMMR brightness temperature time
series as given in Fig. 14 and 15
variations of surface parameters. This statistical homogeneous surface, is an ideal test area for the
analysis over a period of ten years is needed to development of remote sensing techniques in
improve the present knowledge and understanding climate research. Unlike sea ice and land areas,
of the Greenland ice sheet surface climatology. the three major surface types of the ice sheet
(glacier ice, wet snow, and dry snow) extend over
2. Method several tens of kilometers at a minimum, which
makes the application of low resolution satellite
2.1 Multispectral Satellite Data
data feasible for surface type classification.
Past and ongoing research have shown that several The types of satellite imagery used for this study
polar surface properties can be mapped through cover a range of spatial resolutions and wave-
a combination of multispectral satellite and aircraft lengths. For the visible and thermal infrared
data (Steffen et al., 1993; Haefliger et al., 1993; wavelengths we have used Advanced Very High
Williams et al., 1991). These properties include: Resolution Radiometer (AVHRR) Local Area
spectrally integrated surface albedo, surface tem- Coverage (LAC: 1.1km at swath nadir) and Global
perature, shortwave and longwave radiation Area Coverage (GAC: 4km at swath nadir) data.
balance, and snow properties. All of these param- For the passive microwave region, Nimbus-7
eters can be linked to the surface energy balance. Scanning multifrequency microwave radiometer
The Greenland ice sheet, due to its large size and (SMMR) data (10/25/80-8/20/87), and DMSP
242 K. Steffen et al.
1
Satellite Input Data: _
p.u,,o ._,o,,.,,o m (
Algorithm
SMMR &SSM/I m |
-Radiance2sChannels I_ Development
- Resolution 50 km (19GHz) m "_ I -o,_elolM_, I I
-WoSt .o. I I
-o,_G=ac__ce I I
-FloodeGd_ae_r ?.........I........
.iiii!iiiiiii(o_p.t)_iiiiiiiiiiiiiiiiiiiiii!iiiiiiiiiiiii_ili
iiii_ilE_,r0mii_:!i!iii_ii!ii!_iiiiiii_i!i
i!iii!!i i!!iiiii! i!ili iiiiiiiiiiiiii!iiiiii iiiiii!'iiiiiii iiiiiii i!iiii !
Meteorological Data:
RNeemhCamp
- Radiosonde Profde T,H,V (25 kin)
-Air & Sudace Temperature
- Incoming Shortwave Radialk:_
- Inco_ng I_ongwave Rad_atK)n
- Reflected Shortwave Radiation
- Longwave Outgoing Radiation
-Emissivity at37 GHz
-Dislrib_dion of ,Snow Deplh
[.,sfL,,J=A,_o [I
ii_iii_iii__iliiii_iiii!:iiiii_iiiii!iiiiiiiiiiiiiiiiii_i!iil
Satellite Input Data:
VII & TIR
NOAA-AVHRR
Radiances 5 Channels
DMSP OLS Observables
--2RReaRsdeoiaslnoucltuextsi)onn6s200C1hkmainnn,e4lskin A-SlgurofarciethmTemp/erature 1
-Snow Types -.
- Radiarces 5 Channels -Onset ofMelt
-2 P,eso_utio_ 80m (VIS), -AI3edo
LANDSAT M1S2S0rn ('FIR) --CRiaod¢¢ia1tsive Fluxes
Fig. 2. Processing of multi-spectral satellite data in combination with in situ measurements for the retrieval of the surface
parameters relevant to climate process studies
special sensor microwave imager (SSM/I) data into discrete data sets (level lb data) at the
(7/9/87-9/30/90) were used. They are presently Naval Oceanic Atmospheric Research Laboratory
available for both polar regions, including the (NOARL). AVHRR LAC and GAC channel 1
Greenland ice sheet, on CD-ROM from the (0.58 /am-0.68gLm) and 2 (0.725 /_m-l.10 #m)
National Snow and Ice Data Center (NSIDC), images have been calibrated over the entire Green-
Boulder, Colorado. land ice sheet using pre-launch calibration coeffi-
In order to retrieve surface parameters relevant cients to obtain percent albedo values. The cali-
for climate process studies from satellite imagery, bration technique follows the description given in
a number of sensitivity analyses and intercom- the NOAA users guide (Kidwell, 1991) and in-
parisons with ground measurements are needed. cludes the non-linear corrections. In order to
These are described below (Fig. 2). correct for the geometric distortions caused by the
curvature of the earth and the scanning geometry
the images have been geo-registered and mapped
2.2 Processino of AVHRR Data
to a polar stereographic projection using a navi-
The AVHRR data were received as raw data gation code developed by Baldwin et al. (1993).
that have been quality controlled and assembled The accuracy of the navigation routine is at most
ClimatSeensitivSitytudieosftheGreenlanIcdeSheet 243
6 to 8km dueto errorsin thetime tags.The relations which are described below, require that
imageshavebeenfittedtoamapgridto reduce
the data be left in image format. For such applica-
thiserror to lessthan1 pixel.TheCPUtimes
tions an ice mask was developed from a vector file
necessartyocalibrateandnavigatetheLACand of the ice sheet boundaries which was digitized
GACdatafortheentireGreenlandicesheetare
from the Geological Survey of Greenland (GGU)
5minforaGACimage(325×625pixels)and84 geological map (scale= 1:250,000). This mask is
minfora LAC image(1500×2500pixels)ona applied to the images and the result was a set of
IBM6000workstation.OnaSUNII workstation
images of the Greenland ice sheet with no land-
13minforGACand188minforLACareneeded
contaminated pixels.
to processthesatelliteimagery. Other applications such as time series investi-
Ourpresenctomputingfacilitydoesnotpermit gations require that data points from the same
theprocessingofalargenumberofLACscenes location(s) be extracted from all of the images.
dueto extensivecomputingtime necessaryto The first such application is the retrieval of time
calibrateandnavigatetheAVHRRLACimages series of brightness temperatures along a horizontal
for theentireGreenlandicesheet.Thus,it is transect across Greenland. The transect chosen in
desirableto usetheGACimagesfor longtime the first stage of the data analysis is the one that
seriesanalysis.In section3 wepresenta pilot passes through the ETH/CU research station
programto testthefeasibilityofusingAVHRR (Fig. 1). In addition to providing information
GACimagesfor longtermclimatestudiesover about the brightness temperatures along a west/
theGreenlandicesheet.
east profile across the ice sheet, this time series
provides a means of identifying days of bad data.
2.3 Processing of SMMR and SSM/I Data The brightness temperatures between 318-325 ° E
The SMMR isa ten-channel, five frequency linearly longitude are relatively stable with little fluctua-
polarized passive microwave radiometer system. tion. Consequently, days that show very high
The instrument measures surface brightness deviation from the normal (greater than 4 sigma)
temperatures at 6.6, 10.7, 18.0, 21.0 and 37.0 GHz. in this longitude range are flagged as "bad" days,
Vertical and horizontal polarization are provided and are linearly interpolated from the nearest
for each frequency. The SSM/I operates at four "good" days.
frequencies, namely 19.35, 22.24, 37.0, and 85.0 GHz. From the transect, the pixel that corresponds to
Both polarizations are provided for each frequency the station location is extracted, and a time series
except 22.24GHz. For this study we used the 19 of data specifically for the station is produced. In
and 37GHz frequencies which have effective field addition, the horizontally polarized gradient ratio
ofview dimensions of 69 × 43 km and 37 × 28km, is calculated using the following formula:
respectively (along track × cross track). The
GR =(19H- 37H)/(19H + 37H) (1)
SMMR and SSM/I data provided by NSIDC are
gridded in polar stereographic projection in 25 × where 19 and 37 correspond to the channel
25 km grid cells. frequencies in GHz, and H indicates horizontal
The passive microwave data are read from the polarization. Similar time series can be produced
CD-ROM, and all points within the rectangle for any location on the ice sheet, but the climate
whose edges contain the outermost limits of the station is chosen initially so that comparisons to
Greenland boundary are extracted. Then the in situ data can be made.
ocean mask (which is provided on each CD- One final item in the processing of the passive
ROM) is applied. This procedure is followed for microwave data pertains to the relationships
all of the available dates, and it yields a set of between SMMR and SSM/I data. An extended
images 60 x 109 pixels in size for each channel time series is important for the assessment of
and each day of SSM/I and SMMR coverage. A climatological trends, but in order to effectively
total of 1611 images for each SMMR channel and combine the SMMR and SSM/I data sets, a
1186 images for each SSM/I channel have been comprehensive understanding of how the two are
extracted. related must be obtained. Figure 3 shows the
Some applications, such as snow mapping relationship between the 4 common channels on
(described in section 3.2) or SMMR/SSM/I cor- the two instruments for the overlapping days of
244 K. Steffen et al.
260 250
....... SSM/I 19v ....... SSM/I 37v
v __ SMMR 18V __ SMMR 57V
250 240
£ £
&
240 E 230
, ,. -'--'--'----,.... ..... ,- ,.._, .....
c c
_c:,,250 _,220
220 210
10 20 50 40 10 20 30 40
Days Past July 9, 1987 Days Past July 9, 1987
220 220
....... SSM/I 19H ....... SSM/I 57H
_C
__ SMMR 18H __ SMMR 57H
210 2t0
2o o
&
200 zoo
" L/- --*"
_ 190 m 19o
c_ &
180 180 o..........................................
0 10 20 30 40 10 20 50 40
D(_ys Post July 9, 1987 Days Past July 9, 1987
Fig. 3. SMMR and SSM/I brightness temperatures for the overlapping days in 1987 (July 9-August 20) at 72°N and 37°W
longitude
28O I I I I I
__ Tb(SMMR) = 0.80 xTb(SSM/I) + .!,7.0, R = 0.94
Calculated from Greenland Data
260
24O
rv-
220
09
2O0
180
Fig. 4. Relationship between S MMR
160 'J, , , I , . I , , . I . , , I , , , I and SSM/I data over the entire ice
160 180 200 220 240 260 280 sheet for the full period of overlapping
SSMI coverage
operation in the highest area of the ice sheet (72°N These relationships were addressed for Antarctica
and 37° W). Alarge discrepancy (nearly 10degrees by Jezek et al. (1991), but they have not yet been
consistently) is apparent between the 19GHz defined for the Greenland ice sheet. For all of the
SSM/I channels and the 18GHz SMMR channels. overlapping days, July 9, 1987 through August 20,
Similar discrepancies, though not as pronounced, 1987, SSM/I 19GHz vertically polarized bright-
exist for the 37GHz channels. ness temperature measurements are compared to
Climate Sensitivity Studies ofthe Greenland Ice Sheet 245
the corresponding SMMR observations (18 GHz
8O LAC
vertically polarized), and the relationships are
0
assessed. The results, which correspond well to "0
Jezek et al. (1991) results are shown in Fig. 4. The _ 60
0
small discrepancy can most likely be credited to
the fact that Antarctica has much colder regions C0
_ 40
than Greenland. Consequently, the data points t A E;
span a much broader range of temperatures, so o
the behavior of the correlation curve is better L
o 20
i i t I
defined at the low end. Furthermore there is Z
49.8 49.3 48.8 48.3 47.8 47.3
simply more data from Antarctica because of its Longitude (degrees west)
size. The similarity between the two regression
lines, however is encouraging• 8O
__ LAC
Based on this comparison, the SMMR time
O
series can be extended to include the SSM/I data, "ID 70
_0
_D
for a data set that currently spans over 13 years.
O
Similar relationships need to be determined for -O 60
E
the other channels• O C ©
I
5O
8
3. Results L
ZD 40 I I I I
3.I Narrow-Band Albedo of Snow 47.5 44.5 41.5 38.5 35.5 32.5
Longitude (degrees west)
The variation of solar radiation absorbed and
reflected by the earth is a key factor in the Fig. 6. West-east transect plots of narrow-band planetary
understanding of climate change. If the solar albedo (0.58-0.68pm) for AVHRR Local Area Coverage
(LAC: dotted curve) and Global Area Coverage (GAC: line
radiation reflected by the earth were to increase,
curve) data at 69.6_'N of the Greenland ice sheet. Four
a trend toward the extension of the winter season
regions for June 06, 1990 are represented. Region A shows
would result, as can be shown in a simple energy the bare ice, region Bthe wet snow, region C the dry snow,
feedback mechanism (Budyko, 1974). The snow and region D the clouds
1.0-
Very New Snow
NewSnow
0s- OnsetofMelt
Z NewSnow
0,6_ WetSnow
2 i
h
_ 0.4-
0.2-
Fig. 5. Hemispheric spectral albedo for very new
snow (a few hours old), new snow (1 2 days old),
O.O- onset ofmelt, and melting snow surface inthe spectral
3OO 1200 1500 1800 2100 2400 270O range 300-2500nm measured at the ETH/CU re-
Wavelengths (run) search camp in 1991
246 K. Steffenetal.
cover is one of the most reflective naturally Region C: Dry Snow
occurring materials, and therefore, an under-
Over dry snow where the albedo remains relatively
standing of its visible solar reflectance is of great
constant, the agreement between the LAC and the
importance. The spectral albedo of different snow
GAC images are very high, with mean differences
types at the ETH/CU research camp was measured
practically zero. In this region the albedo is very
for the purpose of calibrating and interpreting
homogeneous and thus very little difference between
satellite derived values (Fig. 5). All measurements
the two resolutions is expected (Table 1).
were carried out at solar zenith angles between
50° and 68°.The snow grain diameter varied for
Region D: Clouds
the different measurements: very new snow 0.5 mm;
new snow 0.1-0.2mm; onset of melt and wet Detection of clouds tends to be difficult since
snow 1-5 mm. cloud reflectance is similar, if not identical, to
AVHRR LAC and GAC data were processed as snow reflectance. However, thick overcast regions
described in section 2.2 to study the differences in show noticeable fluctuations in reflectance (5Y/o).
pixel resolution for albedo retrieval over the ice In these regions the agreement between the LAC
sheet. Figure 6 shows transect plots of the planetary and the GAC images decreases slightly (Table 1),
albedo values for both the LAC and GAC channel 1 but due to the large size of the cloud features (4 km
(0.58-0.68/tm) across the ice sheet at 69.6°N. or larger) the GAC data is able to detect the slight
Four different regions can be distinguished: variation in reflectance due to the presence of
clouds as well as the larger decrease in reflectance
Region A: Bare Ice and Slush from the cloud shadows. The ability of the GAC
data to detect the cloud shadows could be a useful
In this region the majority of the snow has melted tool in cloud discrimination over the Greenland
and bare ice remains. Surface lakes tend to form
ice sheet.
on top of the bare ice, and some slush also exists.
Thus, the same information for wet snow, dry
The agreement between the LAC and GAC images
snow, and clouds can be obtained by using either
is low in comparison to the dry snow regions.
high or low resolution data. Using the GAC data
Such a result is expected because the melt lakes
during summer months, when the melt region is
are within the 1km foot-print ofthe LAC data but
relatively large, should be adequate. When.the
the GAC footprint contains information from the
melt season isjust beginning, the lower resolution_
surrounding area as well (Table 1).
data will not be able to accurately detect the
melting since the pixel will be contaminated by
Region B: Wet Snow
dry snow. Higher resolution data such as LAC
In the wet snow region, the agreement between will be needed at the margin of the ice sheet in
the LAC and GAC data is much improved. such cases. In addition, to study the surface lakes
Although, small scale variation is not detected that form during the melting season, high resolution
with the GAC data, the general gradual increase data with pixel size less than 2km (average lake
of albedo is well represented (Table 1). diameter) will also be needed.
To obtain snow surface albedo maps it is
necessary to correct the AVHRR satellite data for
the intervening atmosphere. Such a correction can
Table 1. Statistical Results ofthe ComparisonBetween LAC be accomplished using a linear relationship between
and GAC Datafrom June 06, 1990.Values are given for a clear sky planetary albedo and surface albedo
region 300km2region (Koepke, 1989), or by using a radiative transfer
model, such as LOWTRAN or 5S, to correct for
Region Mean Stddev of Correlation
scattering and absorption. Although presently
difference% difference
there is no consensus among scientists as to which
Bare ice - 2.05 3.06 0.890 atmospheric radiative transfer code is best, most
Wet snow -0.16 0.89 0.970 codes tend to disagree only for large optical
Dry snow -0.04 0.26 0.998 depths of aerosols and large off-nadir angles (60 °
Clouds 0.06 2.21 0.997
or more) (Royer et al., 1988). The 5S code developed
Climate Sensitivity Studies of the Greenland Ice Sheet 247
by Tan re et al. (1990) provides reasonably accurate
Table 2. Input Parameters for the 5S Code Run for the Lz_cation
atmospheric modeling and surface reflectance t_fthe ETH/CU Camp, May 23, 1991
retrieval (Teillet, 1992). The overall scheme involves
correcting the narrow-band planetary albedo as Solar zenith angle 49.0
Sensor zenith angle 29.40 '_
viewed by the sensor to the narrow-band surface
Solar azimuth angle 285.57'
albedo, taking into account illumination, sensor
Sensor azimuth angle 243.69"
view angle, and atmospheric effects. Atmospheric profile (i) H20=0.62gcm 2
To run the code, illumination and observation (ii) 03 = subarctic summer
geometries for each pixel are provided as input. Aerosol model Continental
Aerosol level Visibility = 8km
The water vapor input can be calculated from
Spectral bands NOAA-I 1AVHRR channel 1
the AVHRR channels 4 (10.30 ll.30/tm) and 5
Surface reflectance Dry snow
(11.50-12.50/_m) or it can be provided by radio-
sonde data that has been collected at the ETH/
CU camp during the summers of 1990 and 1991
(Haefliger et al., 1993). Radiosonde data are also
These quantities, which are necessary to calculate
available from the summit of the Greenland
the surface reflectance, are obtained by running
ice sheet and from the coastal station Gothavn the 5S code.
and will serve to derive an "average atmosphere" A three day time series for the AVHRR LAC
for the data available in 1990 and 1991. Further
channel 1was atmospherically corrected using the
parameterization is needed to extrapolate for in situ radiosonde profile data (See Table 2 for 5S
other years. The ozone concentration is taken to
inputs). Figure 7 illustrates the temporal changes
be that of subarctic summer. Aerosol content can
of narrow-band surface albedo for the dates May
vary significantly in Arctic regions (Lindsay and 23, June 01, and June 08, 1991, along a transect
Rothrock, 1993) and can result in large errors in at 69.& N (see also Fig. 1).This time series shows
deriving the surface reflectance. Unfortunately, a continual decrease in albedo over time. On May
optical depths of aerosols over the Greenland ice 23, 1991, the snow cover was dry and the land was
sheet are not known and standard atmosphere also covered by snow with many rocky areas
assumptions must be made. Using the 5S code, it showing through. ByJune 01, melting had occurred
is possible to retrieve surface reflectance by the and the surface reflectance dropped considerably,
following expression: and by June 08 the reflectance has decreased
further by 5'_/o.The ability of the sensor to detect
pi = 100 (Ai + Bi) (2) melt enables us to determine the aerial extent of
(100 + (Ai+ Bi)Si)
the transition regions along the perimeter of the
(pidos)2 (1- 100pa,, ) ice sheet. Figure 8 shows the AVHRR channel 1
Ai - Bi = (3) imagery for the Jakobshavn area on June 8, 1991
TOT.sTvCOS0s' TsZv
(see also Fig. 1 for location). Cloud shadows
where the indexj stands for the AVHRR channel running north-south on the left side of the image
(i = l, 2). (snow covered ice sheet) can be identified. The
snow free areas of the ice sheet along the margin
surface reflectance for channel 1 or 2 in (bare ice) stand out against the higher reflectance
percent of wet snow areas.
pO reflectance at the sensor The narrow-band surface albedo as derived
atmospheric reflectance from AVHRR channel 1and corrected with the 5S
Patm
S= atmospheric spherical albedo radiative transfer model for atmospheric scatter-
Tg total gas transmittance ing, and absorption was compared with in situ
total scattering transmittance in solar measurements at the ETH/CU station. The date
"Cs
direction
used for this intercomparison is 23 May 1991
L, = total scattering transmittance in sensor under clear sky conditions. The in situ hemispheric
direction
spectral albedo, integrated over the wavelength
ds-- solar distance region 580-680nm, is 91.6j.oo/, and the narrow-
0_= solar zenith angle. band albedo derived from AVHRR channel 1 is
248 K. Steffen et al.
1°t°
8O
! 't J :'
I :3 / ;
, :1 J,-;
60 ! :,. 4
o ¢ I.,/
Ii t/
c
b 40 !_
o I
;P
.t
(; /
2orj __ 25 May
.\ ......... 01 June
Fig. 7. Narrow-band surface albedo
_ _ 08 June
(0.58 0.68/_m) time series showing
the melt region at 69.6" N. The time
50 W 47W series is from May 23, June 01 and
Longitude (degrees west) June 08, 1991
Fig. 8. AVHRR channel 1 imagery for
the Jakobshvan area of the Greenland
ice sheet, June 8, 1991. Location and size
of this imagery is given in Fig. 1. Bare
ice, wet snow, dry snow and cloud shadows
can be identified by different gray levels
¸¸ I
DRY SNOW [CLOUD SHADOWS (reflectance)
90.2/oo/. A recent intercomparison with the same channel 1. Furthermore, the comparison suggests
input values but using the LOWTRAN for the that the standard aerosol and ozone concentrations
atmospheric correction gave a narrow-band give reasonable results for this case study. However,
albedo value of 93.3_ (Haefliger et al., 1993). comparisons for different seasons and a larger
This preliminary analysis shows that with in situ number of cases are needed to derive statistically
atmospheric temperature and humidity profile significant values.
data as input variables for the radiative transfer It should be noted that altitude dependence was
model, narrow-band surface albedo can be derived not considered in these computations, as we were
within one and two percent accuracy from AVHRR only interested in relative changes of the albedo.