Aug 23, 2001 - Results show the bias in altitude from multiple .... the potential bias effect from seasonal and interannual ...... Data tape documentation.
POPULAR
SUMMARY
Atmospheric Expected
Ashwin
Multiple Errors
in Antarctic
Mahesh,
Code
James
D. Spinhime,
David
P. Duda,
Edwin
from
location, atmosphere path.
scatter
This adds
such scattering
cloud
size
of the
fraction
bias
even
when
result
from
blowing
atmospheric altimetry
photons, spent
scattering
snow
near
launched
deviating
to quantify
variations
to significant
year-to-year
can
be
limited
reduce
the
number
for the GLAS
height,
that with careful
observations
similar
discusses selection
can be met.
in
the
to-and-fro
of the height biases
cloud
properties
of measurements
paper
the direct estimate
variations
that
laser
from
to a minimum.
in these
to
it is likely
mission
them
layers
the expected
in cloud
variations
This
a biased
of the
of
of the satellite's
Cloud
from
important changes
time
knowledge
them
to keep
An
elevation
travel
elevation.
Year-to-year
the surface.
The
precise
and produces
by
in 2002.
and monitor
surface
of analysis
and demonstrates
specifications
with
It is important
are present,
Sensing
Antarctica.
of
is affected
greatly
no clouds
along
thereby
methods
lead
be
and
in travel,
satellite.
measurements will
Greenland
measurements
depth.
could
but this
Also,
laser
the
will
is to measure
and back,
and to develop
elevation
conditions
over
laser
and optical
cover
Surface
the
and Remote
System
laser system
make
below
magnitude
particle
Altimeter
to the time
of the ice sheet
of
Measurements
in Geoscience
to the earth to
Part Ih Analysis
of Wisconsin-Madison
Trans.
ice sheets
is used
Altimetry.
University
of this space-borne
the satellite
on GLAS
912, GSFC/NASA
Laser
the high-latitude
Altitude
University
to IEEE
Geoscience
objective
The
Code
W. Eloranta,
Effects
912, GSFC/GEST
Hampton
For submission
The
Scattering
effective
as well
in the altitude under that
as in bias.
cloud-free can
be used.
scattering
effects
the likely
biases
and analysis
will from
of data,
Atmospheric Part II: Analysis
Multiple
Scattering
Effects
of Expected
Errors
in Antarctic
Ashwin Goddard
Earth
Sciences
NASA
Goddard
Space
Flight
University
submitted
Trans.
Ashwin
author
address
Mahesh
Mail Code
912
NASA Goddard Space Greenbelt MD 20771
Flight
Email:
@gsfc.nasa.gov
ashwin.mahesh
Center
Greenbelt
MD 20771
Center,
Greenbelt
MD 20771
Hampton
VA 23681
W. Eloranta
in Geoscience August
Corresponding
Center,
of Wisconsin-Madison,
to IEEE
Measurements
P. Duda
University,
Edwin
Altitude
D. Spinhirne
David Hampton
Altimetry.
Mahesh
and Technology
James
on GLAS
2001
Madison
W153706
and Remote
Sensing
ABSTRACT
The altimetry ing from
bias in GLAS
atmospheric
properties
over
multiple
the Antarctic
are presented.
Results
nificant
source
error
free observations, used
here,
affected
in these
in cloud
in the altitude
measurements
snow, also a source altimetry
correction.
to minimize
properties
from
height,
Altimeter
is studied Estimates
selective
analysis
bias. space,
Although over
can be met.
scattering
delay.
With
The
could
may often
careful
in the bias
would clouds
magnitude
size and optical
observations
these
changes
depth
include
selection
depth.
biases
or cloudmethod is
Interannual year-to-
in surface
near-surface
and analysis
be a sig-
of the bias
lead to significant reduce
resultof cloud
such as by the Gaussian
errors.
fraction
cloud-free
Antarctica
pulse
particle
cover
knowledge
in clouds
use of low optical
altitude
effective
to current
and interannual
of the return
the surface
or other laser altimeters
in relationship
from multiple
The
cloud
System)
of seasonal
as well as in cloud
of scattering-induced
specifications
Laser
the bias in altitude
as well as improved
are necessary
year variations vation
show
scattering Plateau.
without
by variations
variations
(Geoscience
ele-
blowing
of data, laser
1. Introduction
With increasing sound
concern
monitoring
also grown.
programs
in the expectation
decades,
A major
especially
aspect
of the Earth's potentially notably
changes
change
of polar
parison
developed lite radar
ments
errors
new methods
on existing
Earth's
polar ice sheets
GLAS
is a laser-based
2002 ments, return
and Antarctica
from
surface
as part of the Earth the mean pulse.
elevation
To permit
temperature
changes
changes
and moni-
in the mass balance
Global
to other
warming
climatic
inter-annual
could
changes,
changes
of continent-wide
methods,
in the
elevation
the ice-sheet
of the ice sheets drawbacks;
radar
have ever been attempted. and will record
which
rely on a com-
and total ice loss - that are each subject mass
balance
by airborne measurements
into snow, and relatively
observations,
of
will be observed
is to measure
regions.
by typical
to measure
measurements
focus
[2].
is limited
each contain
namely
the first estimates
methods
change
(GLAS)
to measure
accumulation
and radar penetration
over Greenland
an improvement
- total snow
stage has
a particular
surface
in turn leading
proposes
ice sheets
mass balance
altimetry
System
in the polar
of these ice sheets,
more accurate
using repeated [4]. These
are concentrated
GLAS
at an early
have been climate
and implement
[1].
Altimeter
and will provide
of ice-sheet
Recently,
slope
which
to develop
changes
of significant
in the high latitudes,
and Greenland
of two large numbers
large errors.
change
climate
signs of global
Antarctic
Laser
in sea level.
ice sheets,
in the Antarctic
The determination
surface
of climate
the mass balance
a possible
thickness
Geoscience
the need
the polar ice shelves
by measurements
in the coastal
large ice sheets
alter
around
surface
large-scale
that the early
bolstered
goal of the orbital
tor a particular
of the earth's potential
ice zones
This view has been
in recent
warming to detect
The marginal
such programs, here.
over
temporal
been
[3] and satel-
are sensitive
few airborne GLAS
lidar
have
to
to
iidar measure-
measurements
will mark
changes
in the thickness
scheduled
to be launched
of the
space.
altimeter
Observing
and atmospheric
System
of the laser's the determination
(EOS)
surface
profiler
program.
For the surface
spot will be estimated
of mass
balance
changes,
altimetry
from a centroid individual
ice-sheet
in
measureof the altitude
measurements mustbemadewith uncertaintiessmallerthan 10cm. A numberof factorsaffect the accuracyof the altitudemeasurement, including surfaceslope,atmosphericpropagationand signalnoise. A cross-overtechniquethataveragestheelevationdifferencesat selectedpointson the icesheetsis designedto reduceerrorsin orderto measureregionaliceelevationchangesto an accuracyof 1.5cm per year,the statedgoal of theice-sheetaltimetry [5].
In PartI of this paper([6] hereafterreferredto asDSE),the authorspresentedcalculationsof path delaysby cloud andaerosolscatteringfrom ananalytic double-scattering modelandMonte Carlo simulationsof lidar surfacereturns. Both methodsdemonstratedthatmultiple scatteringby optically thin polarcloudscould seriouslybiasthealtituderangingof GLAS. Forexample,if the surfaceheightweremeasuredfrom thecentroidof thereturnpulse,a thin arctic stratuscloud with an optical depthof 0.5,a meanparticle radiusof 6 microns, anda thicknessof 3 km would produce ato-and-fropathdelayof 30cm, correspondingto analtitudebiasof 15cm.
Sincethe effectof multiple scatteringis to alwaysintroducea delay,themeanheightchangewill beaffectedby the changesin cloud andaerosollayersin the atmosphere.If not corrected,seasonalandannualvariationin cloud propertiescould significantlyaffectthe determinationof changesin the surfaceheight. Surfacealtimetryis a primaryobjectiveof the GLAS mission,and multiple-scatteringinduceddelayin theobservationshasthe potentialto seriouslyunderminethis goal. Using currentknowledgeof Antarctic cloudproperties,we studytheimpactof multiplescatteringon the determinationof surfacealtitude,aswell ason theinterannualvariability of it. The useof the atmosphericlidar signalsof GLAS to eliminatesucherrorswill alsobediscussed.
2. Variability
The purpose ranging
of ice-sheet
bias would
the critical of cloud
of Cloud
altitude
on the Antarctic
observations
not be a problem
issue is to determine properties.
ple scattering-induced height.
Properties
The variability
DSE found delays,
is to measure
if polar cloud
the potential that several including
Plateau
properties
bias effect cloud
cloud
in each is now considered
depth,
in turn.
changes
were constant
from seasonal
parameters
optical
temporal
over time.
and interannual
can affect cloud
in ice thickness.
particle
the magnitude
Therefore, variability of multi-
size, and mean
cloud
A
2.1. Cloud
Occurrence
Due to the harsh conditions of polar cloud surveys
properties
[7]. Despite
rent knowledge observations
of polar
- which
weather
records
percent)
observations
of clouds
report
cent of the observations. and the coastal
entirely
skies usually
Cloud
occurrence
clouds
over
generally
cycle winter.
regardless
1. The values
fewer
stations
annual
most of the Arctic
higher
during
Greenland occurs
of the time, while of clouds,
Mean
Ocean
to over
80 percent
Antarctic wintertime
ranges
even
stations, cloud
stable
from values
- from with val-
55 and 70
values
however, occurrence
during clouds over
which
in between
10 percent
surface
40 and 60 per-
inversions
as diamond
et aI. find that in winter,
from 65 percent in the Siberian the summer,
the South
cloud
of under
dust.
observations occurrence
of the observations
Arctic.
A similar
and lower
are seen far more
in the inte-
Observations
temperature
Arctic
the
in the Arctic
[8].
known
report
located
sky observations
50 to 70 percent. range
stations
seen
ice cryslals,
Hahn
of observations
with higher
occurrence
by them
(between
the few stations
less than
of near-surface
when
cloud
the
by observ-
70 and 80 percent,
amounts
typically
of clear
is usually
over the poles.
the summer,
over Antarctica,
At coastal
high plateau.
seasonally
between
are from coastal
frequency
of Antarctica
of Arctic
of the sky filled
and smaller
over Antarctica
lead to the formation
varies
values
annual
surface
over
made
cur-
Canada.
occurrences
The mean
variations
Mean
Spitsbergen,
70 to 80 percent
spatial
and
from
summarize
of sky conditions
are typically
and northern
can be made
observations
of the fraction
clear skies are rare at the high latitudes;
clear
western
of clouds
of clouds
between
rior of the continent
Ocean
the world.
than 80% in the area around
Most surface
surface
indicate
cloud
over Greenland
et al., for example,
The observations
routine
Greenland
characteristics
of Hahn
observations
Even the most comprehensive
data from the ice sheets
of polar cloud
around
in Figure
over western
presence
from
the presence
cloud
The surveys
the globe.
stations
Hahn et al. are shown ues greater
across
are summarized
ers at various
estimates
cloudiness.
as well as their remoteness,
than elsewhere.
only sparse
this, some
of clouds
They
and Antarctic,
have been far fewer
such as Hahn et al. contain
Antarctica
poles.
in the Arctic
values
frequently
Pole ranges
from
of is over
seasonal during
the
than over
the
30 to 40 per-
cent,while in the summer,cloudsareobservedin 45 to 70 percentof the observations.On the otherhand,observationsbetween1971and1980atthe coastalSyowaStation(69 S,39 E) showa maximumin the latesummer,whennearly80 percentof observationsareof clouds,andminima in the earlysummer(53%)andwinter (60%) [9]. Hahnet
al. also determined
IAV was defined
the interannual
as the standard
deviation
1991. The interannual
variation
the Arctic
near 5 percent,
(DJF)
was usually
observations.
percent
north
ations
ranged
The most (2001,
in press,
Pole station
throughout
made
at three
much
strained
height
ground
based
obtained
angles
face-based
inversion
kilometers.
The higher
from
1982-
for land stations
in
variations
show
IAV values
the summertime
from 20
the standard
at the Antarctic
coastal
devi-
stations
and
an annual
plateau
spectral
cycle
observations
of cloud
was not specifically had a limited
et al found
with Hahn
is that of Mahesh of clouds
at South
particle
effective
base heights,
designed
et al.
to quantify
cloud
occur-
field of view, and observations
clouds
were
the year to consistently in approximately
et al.'s multi-year
record
43% of their
average.
depth
plateau,
absence
bases
surface-based
of topographical
method
layer,
Ocean
During
longwave
Mahesh
consistent
and optical
The cloud
for the period
occurrence
- 45 °, 60 °, and 75 °- throughout
direction.
roughly
of the CO2-slicing
over the poles.
for December-January-February
over the Antarctic
their spectrometer
by the relative
observations.
of clouds
the year;
of the Antarctic
cloud
over the Arctic
This study
observations,
2.2. Cloud
version
From
zenith
occurrence
10 percent
over Spitsbergen.
occurrences
near 5 percent.
study
in the same viewing
spectral
and was near
thicknesses.
viewing
cloud (JJA)
The interannual
in 1992, the authors
rence
Over
year-long [ 10,11]).
radii and optical
clouds
to 2 percent
Pole are generally
recent
in seasonal
observations
from 2 to 5 percent.
at the South
(IAV) in cloud
in June-July-August
Wintertime
of Alaska
variation
have
to determine a bimodal
and a seasonally
clouds,
visual
estimates
reference the base
points. height
distribution,
of cloud Mahesh
of clouds,
secondary
i.e. most of the clouds
with bases
are con-
et al. use a modified from longwave
with the primary
dependent
height
maximum
maximum between
spectral in the sur-
2.0 and 2.5
in the 2.0 - 2.5 km range
have
smalleroptical depths(lessthan 1),whereascloudswith basesnearthe surfaceareoftenthicker, althoughmanyof thesetoo haveoptical depthsof lessthan2.
Icecrystalprecipitationcanhavea wide rangeof optical depths,but it is commonlymuchthicker duringthe winter.Wilson et between
2.7 and
springtime, optical cally
10.7, although
the observed
depths much
2.3. Cloud particle
crystal
Curry
measured
to 1.9. Mahesh
Nearly
depths
[12]. In the Arctic
et al.'s
findings
over the Antarctic
of the continent. smaller
optical
of cloud
plateau
are opti-
95% of the clouds
at South
than 5.
size
The multiple-scattering the clouds.
from 0.015
depths
of ice crystal
as 21 have been
held view that clouds
at the coasts
seen to have optical
observations
as large
ranged
the generally
than those
were
wintertime
thicknesses
thicknesses
confirmed
thinner
Pole station
al. report
induced
et al. report
distributions
path delays
that the most comprehensive
show modal
gm [8]. Summertime
Arctic
will also depend
radii between
stratus,
on the microphysical
measurements
of wintertime
10 to 80 I.tm, and an average
on the other
hand,
have much
properties
smaller
Arctic
effective mean
of
radius
radii,
ice
of 40
ranging
from 2 to 7 microns.
In the Antarctic,
Smiley
itation
during
observed
smaller
et al. reported the wintertime
than 50 microns
particles
were
could
not reported.
time from
radiometric
composed
of small particle
cloud
particle
winter
sizes using
effective
determined
particle
rately
and only a lower
here indicated
replicator,
are optically
clouds [14].
Lubin
limit to those particle
sizes
particles in winter
Mahesh
could
A particular
ranged
thin and
between
retrieved
summer
a median
than 25 microns
mostly
the winter-
and Harper
that the mean
is given.
crystals
during
[15].
precip-
and smaller
most
1992 data, and obtained larger
However,
clouds
and estimated
radii of particles
ice crystal
of Antarctic
of 4 to 16 microns radiances,
[13].
on their particle
properties
radii from their
the effective
that cloud
of clear-sky
Pole are 12.3 and 5.6 ktm, respectively
effective
study,
observed
infrared
sizes
50 and 200 microns
and estimated
sizes on the order AVHRR
common
measured
cloud
measurements,
of 15 Iam; in their determined,
inferred
radii over the South
cloud
are between
not be reliably
Stone
profile
that the most
and et al.
particle
size
not be accu-
seasonal
pattern
10 and 20 _m,
whereasin summerlargerparticles,with effectiveradii largerthan25_tm,weredominant[10,11].
3. Results
3.1. Altitude
bias
The observations would
summarized
lead to seasonal
bias for a particular climatological
The mean
in Section
and interannual
period,
frequency
altitude
2 indicate variation
the Monte of various
cloud
bias for a given _ B=
Carlo
period
]_i]_jZk
some
in the altitude
path delay
b('l:i,hj,rk)
thickness
is the computed
"1:,cloud
tion as a function fraction,
is defined .b(x
of cloud
or interannual
(year)
to the next:
In this paper, ties reported ('1:), cloud
by Mahesh
made
altitude
from the ground
well as the entire over
using
depths,
Altitude
in 1992. these
be obtained.
the Antarctic
Plateau
altitude by the
hj, rk) cloud
biases
Carlo
equation
and mean
coverage
of cloud
frac-
cloud
heights,
can be examined
calculations
using
observations, from infrared
(1), an altitude
altitude
optical
cover and r k
for sea-
the mean bias from one season
of cloud
radii (r) derived
on cloud
F is the overall
thus obtained
the frequency
Following
cloud.
between
based
is the cloud
hj is the distribution
from Monte
particle
properties,
year can
• F.r('ri,
as the difference
include
effective
the mean
that
hj, rk)
size r, and FT('l:i,hj,rk)
bias estimates
et al.; these
properties
from DSE can be weighted
F.r('l:i,
for each transmissive
sizes.
bias. To estimate
bias for each transmissive
optical
computed
(h) and cloud
for each measurement
conditions
particle
variation,
we obtain
heights
of cloud
cloud
as:
i, hj, rk)-
particle
same variables
1:i is the distribution
sonal
altitude
h, and cloud
of those
is the distribution
ments
height
results
in polar
types.
( 1 - F) + ]_i]_j]_k where
variability
biases
Due to interannual
variability,
that are not identical
with those
optical spectral
properdepths measure-
bias can be computed
over different GLAS from
cloud
seasons
will record
as
cloud
1992; nevertheless
thesedatarepresentthe bestcombinationof severalcloud propertiesrelevantto multiple-scattering induceddelayfrom a singleobservationprogram;also,atthis time Maheshet remain
the only available
Not all clouds the GLAS
will contribute
Carlo
elevation
of the GLAS
not included calculations
dataset
mission,
in any altimetry by DSE show
estimates.
that when
transmissivity
et al. at South
Pole suggest
that this upper
ments
the year.
made during
section.
old of cloud
Figure
optical
2 shows
conditions
depth
bias in Figure
of the measurements, clear-sky further
However,
2a is primarily
expected
it was assumed
in GLAS
that the scattering-induced
by Mahesh
only those
measure-
later in a later
the lower
measurements
thresh-
using
sky
for all observations
(2a)
with little or no scatter-
of clear-sky, scattering
limit
from consideration.
peak of observations
whose
depth
75% of the clouds
as shown
are plotted
due to observations
are from clouds
obtained
by using
of the observations
biases
The large
depths
in such an approach
in 1992; histograms
(2b).
the optical
thin clouds,
to
lidar, the Monte
the use of nearly
optically
According
of less than 0.25
of the GLAS
biases
by
which
effect
comprise
is minimal.
57% For the
bias is zero, this is explored
in a later section.
cloud
obtain
the altitude
and optical cal cloud
Also,
alone
the remainder
observations,
Using
Mahesh
cases
in such cases.
is considered
altitude
numbers
altitude
by the interferometer
as well as for the cloudy ing-induced
greater
will not be penetrated
transmissivity
scattering
limit still permits
altimetry.
the scattering-induced
recorded
forward
clouds
will be made
For the geometry
or through
accurate
eliminates
thick
the plateau.
is as large as 2. The optical
conditions
more
over
with a two-way
Also, one might minimize
cloud-free
This will permit
bias; optically
clouds
to the above
during
properties
measurements
corresponding
observed
of cloud
to the altitude
lidar, and no surface
the specifications would
year-long
al.'s findings
properties
depth. thickness
obtained
by Mahesh
bias that would Consistent
in a few mostly
from radiosonde
calculations
were performed
of cloud
height,
data taken
during
particle
to radius
the year,
a typi-
used in the modeling.
only a lower
winter
Carlo
from each combination
with indications
of 1 km was
et al. determined
result,
et al., Monte
cases
bound
of thick
in particle clouds
radius
only a lower
in a number
of summer-time
limit to the optical
depth
cases. was
determined.The MonteCarlo calculationsusedto obtainthe valuesin Figure2 wererun only for thoseobservationsof clouds(approximatelythree-fourthsof the total numberof cloudsobserved) in which bothparticleradiusandopticaldepthwereknown,i.e.if only a lower limit to eitherparticle sizeor optical depthis availablethosecloudsareomittedin Figure2. Theseomittedvalues, however,areshownin Figure3, andarespecificallyindicatedasthosewith only a lower limit to optical depth(diamonds),thosewith only a lower limit to particle size(opencircles)andthose with only a lower limit to bothparticleradii andopticaldepths(filled circles)known. In these specialcases,it mustbeassumedthatthe altitudebiascorrespondingto scattering-induced delay is at leastaslargeasindicatedin Figure3. The medianvalueof the altitudebiasfor the entire year,from only the cloudobservations,is 10.8cm, andthe meanis 16.2cm. For a givenvalueof the opticaldepth,thebiasin altitudewill changedueto variationsin both particlesizeandin theheightof the cloudabovethe surface. Low cloudsscatterphotonswhich, despitethescatteredpath,still remainwithin the field of view of the instrument. Scatteringby higherclouds,which aremorecommonin the non-wintermonths(October-March),tendsto removethe scatteredpathlengthsfrom the field of view,therebybiasingthe altitude less. With increasingparticleradius,however,a cloud of a given opticaldepthwill biasthe altitude increasingly, until a limiting particlesizeis reachedat which valuethe variationin the forwardscattering peakis small. The winter altitudebiasesin Table3 (discussedin a latersection)aresmallerthan non-wintervalues;this suggeststhatthe effectof particlesizesin the non-wintermonthsis more significantthanthe fact thatin winter,cloudsoccurnearerthe surface.
3.2.
Variability
If the altitude
in altitude
bias.
bias were invariant
surements
as a result
of multiple
determine
interannual
changes
of clouds bias varies
which
cause
as well.
cloud
occurrence
sitive
to changes
delay
from one year to another, scattering
could
in elevation, by multiple
The interannual
as well as the fractional in the specific
be neglected,
could
scattering
variability cloud
microphysical
errors
introduced
since
still be fulfilled.
the objective, However,
are not constant
in bias can result cover.
More
and radiative
into altimetry
mea-
namely
since
to
the properties
from one year to the next, from changes
significantly, properties
in frequencies
the bias values
of clouds
the of
are sen-
from one year to
the next.
We examine
the interannual
stand the impact
of these
from the spectral such clouds. ability
occurrence
reports
to the next,
the averages
considers
whereas
et al. to obtain
in the estimated reported
bias using two different
In the first method,
of Mahesh
by Hahn
that would
changes
approaches
we use cloud
altitude
biases
to under-
properties
that would
obtained
result
from
biases
is then obtained
using
the inter-annual
et al.
In the second
approach,
we obtain
from
during
1992-94,
of the fraction
bias changes
approach
in altitude
variables.
measurements
and the interannual former
different
The variability
in cloud
tine synoptic
variability
of the sky covered
result
that result
from
from
the latter deals with having
variations
having
more
by clouds in the cloud
more (or fewer)
vari-
fractions.
clouds
(or less) of the sky covered
rou-
The
from
one year
by clouds
when
they are present.
3.2.1
Variability
To estimate
from
the uncertainty
consider
the average
plateau.
The average
Hahn et al. is about variation
spectral
clouds
in altimetry
interannual 5 percent,
particle
sizes,
If, on the other
If clouds
during
biases
a given
will be a corresponding record
of variations
must assume bounds
in Table
properties
average
thickness
1 will increase
optical
in bias.
°
other
delay
(or reduced)
We can assume,
10
To assess
the variation
the Antarctic Pole from
the impact
differ
from
a different
from those
seen in
correspondingly.
than those as well.
seen during There
Absent cloud
(or fewer)
seen in the 1992 dataset,
bias using data years
of this
in their optical
year the additional
or decrease
thickness
over
at the South
than those
from one year to another.
of the increased
variability
altitude
during
in the scattering-induced
properties
the winter. know
to the next we may
occurrence
occurrence
If in any given
properties
year are of different
in optical
of the interannual
in the annual
if the cloud
change
the optical
heights.
in their average
computed
cloud
during
observations
in cloud
we must additionally
and the clouds different
hand,
in summer
while it is 11 percent
measurements,
seen are negligibly
1992 the average
one year of GLAS of variability
variation
then we may well see no change year.
from
as well as the extremes
on altimetry
thicknesses,
observations
is, however,
this information,
occurrence,
for instance,
1992, there
to obtain
that any increases
no we the (or
decreases)
in cloud
obtaining clouds
occurrence
the minimum
relate
only to optically
and maximum
variability
with the most and least impact
on altitude
variability
in cloud
occurrence
(5% in summer,
thin (or alternately,
of the bias. biases
thick)
clouds,
By thus removing
(or adding)
from the 1992 data along
11% in winter),
we can obtain
thereby the
with the known
new annual
average
bias values.
The altitude
biases
are tabulated seasonal
in Table
and annual
of thin clouds, variability
average
sion's
specified
A second
altitude
biases
are viewed
to be entirely
in the altitude
are computed
expected,
the seasonal
numbers
to another being
suggest
produces
comparable
able determination of the various
up to three
that the variation
variation
in the altitude
to or greater of altitude
changes
considered
from
in Tables
The interannual
reported
interannual
of the average
for Table optical numbers
the GLAS
occurrence
mission
in the mis-
(1.5 cm).
variability
values,
in
also from Hahn
1, in this case too, the additional depth
regimes,
are shown
and the annual
in Table
in bias are now even mission
2. As one
more different
specification.
and optical
bias that is significant.
than the GLAS changes
these
in cloud
bias.
than the GLAS
of variability
or four times
altitude
of the
the addition
is larger
values
values
Conversely,
of the table)
of the extreme
and annual
the values
the change
instead
again;
increases
seen in 1992
is high;
the maximum
27% in winter)
depths
occurrence
years
As was done in obtaining
from the 1992 numbers,
These
in summer,
using
the average
cloud
bias between
the optical
of thin clouds.
reduces
1992 (last column
from
of thick clouds
the removal
clouds,
can also be made
(13%
clouds
the addition
or reduced
limit for the relative
et al.'s measurements.
would
of thick
bias from
such differences
as does
such increased
altitude
occurrence
average
biases,
or the removal
calculation
(or fewer)
by considering
1. As is expected,
seen from
annual
cloud
obtained
thickness
The values
specification,
from one year
of such variability,
will clearly
impede
the reli-
one year to the next.
Indeed,
the most advantageous
1 and 2 still produces
interannual
bias variations
sizes instead
of or in addition
of 1
to 1.5 cm.
Similar optical
assessments depth
changes.
can also be made with changes The results
in Table
in particle
1 and 2 implicitly
11
assume
that whereas
optical
to depths
from oneyearto anotheraredifferent,the particlesizesandoptical depthsarecomparable betweenthe two years.Thepotentialimpactof changesin thosecharacteristicscannotbe overlooked. However,our intentionhereis to suggestthatvariability in cloud occurrencecanmanifest itself in significantvariationsin the altitudebiasof GLAS measurements from oneyearto another.Withoutquantifyingthe potentialimpactonaltitudebiasfrom everyconceivablechange in cloud characteristics,we haveattemptedto definesomerangeof valuesto suchvariability. This effort showsthatvariationin the altitudebiascouldbe of the samemagnitudeasor larger thanthe accuracyrequirementspecifiedfor the GLAS missionitself. The determinationof surface altitudes,alreadyuncertaindueto thepresenceof clouds,mustadditionallybereconciled with year-to-yearchangesin the uncertaintyin suchmeasurements.
3.2.2.
Variability
In Section
3.2.1,
from
we obtained
cloud
occurrence
result
from variations
absence
synoptic
of clouds,
as likely
in the cloud
the cloud
rence,
The routine
by surface depths,
surface
the ones of Mahesh uniform
topography
heights.
However,
in the altitude
typical
the mere
in the previous
from one year to the next.
basis,
that would
presence
- it is the portion induced
delay
or
of the sky results
sub-section,
cloud
no variability
not
but just
An alternate
bias is to use fractional and to assume
in
approach
cover
informa-
in cloud
occur-
sizes.
and synoptic
South
reports
For this reason,
the multi-year at South
12
visual
cover
heights
knowledge
observations
observations
data, in contrast
of reference
the accurate
the visual
Pole.
cloud
the advantage
this precludes
measurements,
conditions
that contain
and without
Pole station;
the spectral
sky-cover
occurrence
information
we examined
in the altitude
et al., are made visually
unlike
cover
on a regular
observations
around
the variability
The multiple-scattering
which
cloud
bias due to variation
we determine
cloud
additional
by clouds.
variability
observers
Whereas
contains
occurrence,
or particle
line of sight.
that describes
fraction
in fractional
the interannual
optical
particular
in cloud
from changes
tion reported
variability
fractions.
that is filled
variation
to obtaining
the interannual
from one year to the next; in this section
from each observation only from
reports
in the
of cloud
are not limited
provide
to
a useful
to a
dataset
From the WMO
synoptic
as the variability of the years data
in those
1992-94.
values
values,
at South Pole station,
were obtained
The average
was 42% during
ity in these
data taken
the winter
cloud
(as measured
can be computed.
in an average
These
The variations
interannual
numbers
case we have and thick tively,
are considered
3.3.
Methods
The results
tering
from
of cloud-scattering optically
clouds,
of interannual of the plateau
that seasonal/interannual distribution,
across
clouds
at which
extremes
clouds
variabil-
in the winter
obtained in cloud
to variation
in Section fractions
in the interan-
the rest of the year; this results 0.8 cm.
of all optical
depths.
delay
is expected
of all optical
biases
3.2.1 ; this is expected,
the ranging
the altimetry available
since in this
Very (optically)
is least and largest
to be larger
thin
respec-
than when
variations
thicknesses.
cause
using
errors
altimetry
eliminating all clouds,
will be used as a "stand-alone"
properties.
then we could
that the altimetry which
measurements
on cloud
were known,
are chosen
effects
thin clouds,
estimates interior
observation
of observations
and 0.85 cm during
the 1992
bias
are small enough
effects
subsets
across
so far assume
under
correspond
we saw in section
at these
with no information
of the clouds cal depths
of changes
to reduce
fractions
months
around
(11.5%)
from changes
in the bias of approximately
the extremes
to be manifest
cover
as well
Inter-annual
larger
altitude
cover,
and non-winter
months.
is slightly
result
cloud
year period
the other
average
that would
months,
the variability
represent
presented
measurement,
the winter
than the values
distributed
and the average
in cloud
variability
are lower
clouds
or decreases
for the winter
deviation)
the seasonal
of fractional
for the three
and 52% during
Using
increases
nual bias of 0.75 cm during
fraction
by the standard
than in the rest of the year (8.1%). 3.1, the corresponding
separately
cover
months,
values
select
expected
biases,
highly however,
However, those
scattering because
layers.
variability.
Also,
it is the only way of relating
variability
we saw (in the previous
Antarctic
of thin clouds section)
does not match
the opti-
when
our discussions
of climatologies
for
they are the only available
or the Arctic.
that it can provide
13
First,
depths
Multiple-scat-
be smaller
We began
in the absence
data is still useful.
when
are small.
will understandably
cover
from the coastal
instances
from them
the total cloud
to those
if the optical
cloud
statistics
Although
the variability useful
boundaries
from
the
it is possible of the total cloud to variability.
Wenow turn our attentionto subsetsof observations
that include
only optically
cloud-free
and aerosol
profiling
provide
conditions.
The use of data from the cloud
the necessary
obtained
information
from the green
both above
and below
and a substantial
Using
optical
optical
depth
by using
channel
3 shows
ent thresholds few clouds
GLAS
method
set of observations, the seasonal
depth
depths
for such analysis enough
bias could
can
can be
is detectable
is between
to permit
then be reduced
observations.
to analyze
- 0.1, 0.5, at South
Additionally,
the lidar surface
from all the data.
optical
algorithm
I and 2,
such a determina-
by setting
biases
returns.
Both
approach
for example, calculations to Table
and one-third
threshold
to detect
could
a cloud be reduced
approaches
variability
eliminates
a significant
3. The median during
threshold
are
fraction
delays altitude
the other months;
14
from
the bias
shown
bias (0.1)
as small
as 1.5 cm a
cloud-free
observa-
is to use a more fit method
delay.
values
in paper
1
Table 4
this table is readily
with this fit is nearly
the mean
sophisticated
described
from this method;
bias obtained
very
the altitude
threshold
of the scattering-induced
obtained
Since
bias.
altitudes
The Gaussian
differ-
than 2, the bias
is lowered,
or nearly
thresholds.
several
different
in ice thickness
in altitude
using
larger
at the lowest
changes
depth
all observations.
depths
bias to such cloud-free
measurements.
smaller
depth
the bias in estimated
of scattering-induced
from
optical
bias obtained
secular
the GLAS
obtained
set to 2 is not significantly
as the optical
the interannuai
to limiting
have
lower optical
biases
with the numbers
10-15%)
of altitude
using
of the altitude
in altitude
requirements
removes
to analyze
comparable
values
the computation
tions also largely
An alternate
depth
However,
The
the GLAS
Limiting
values
Pole (approximately
correspondingly.
can be selected
1.0 and 2.0 - along
obtained
approach
subsets
and annual
with the cloud
winter,
optical
thin so that a lidar signal
are transmissive
the altitude
for acceptable
obtained
shows
optical
Cloud
on GLAS
below.
Table
(DSE),
are sufficiently
clouds
so determined,
threshold
the entire
year.
data.
channel
or
depth.
depths
From
drops
such selective
The limiting
of Antarctic
a more sophisticated
discussed
if clouds
them [16].
fraction
tion of layer optical
to obtain
thin clouds
40% smaller
are reduced
by even
in
greater
amounts.
than
1.5 cm.
3.4.
Observations
As discussed vations
At very low optical
under
above,
blowing
clouds
nation,
conditions.
of relatively
Throughout are prevalent teau creates made
namely
much
blowing
by the surface
winds,
weather
spectral snow
measurements
year is less
if we selectively
It will be especially
observations
channel
which
on GLAS
downslope
The settling
which
can disturb at South
[17].
exclude
advantageous,
are made
will permit
observation.
obser-
under
known
such a determi-
There
is, however,
Blowing
used in Mahesh
surface
winds
known
as katabatic
of cold air at the higher
elevations
loose
Visual
and recent
Pole station snow
et al. suggest
snow.
indicate
is typically
blowing
depth
of the pla-
observations
snow
not very optically
that an optical
winds
conditions thick,
in
and
of 0.1 is as thick
as the
may be.
The concern
for GLAS,
surface.
Blowing
special
operational
near-surface theless
times
snow
that even if GLAS
is not just the optical
typically
extends
altimetry
physical
a typical
value
and optical
is limited the altitude
(100 microns) thicknesses
data at 50-m
layer
is close
of the GLAS
by such scattering
depth
from the surface
GLAS
a scattering
the footprint
the 532 nm channel,
Using
to process
When
within
caused
however,
mode
layers.
remain
travel
using
over the entire
snow.
observers
up to a third of all observations
depths.
to those
plateau,
of the year.
surface
values
is not used as a stand-alone
of the Antarctic
during these
optical
The use of the atmospheric
concern,
much
large
of altitude
so that the 1064 nm channel
an additional
bias averaged
bias can be held to small
in fact, to limit the determination cloud-free
the altitude
snow conditions
the altitude
that include
depths,
measurement.
to nearly
or entirely
obtained
for the blowing
snow
15
photons
As a result,
in altimetry cloud-free
from them
particle
radius,
50-300
is needed
to the surface
included
for the snow
up to the lowest resolution
becomes
values
of the snow, but its proximity
to detect
these
thin
by it never-
the delay
in their
conditions
This means as determined
be in error.
and several Carlo
and a
scattered
calculations.
might
layer, Monte
meters,
to the
combinations
calculations
of
were per-
formedasbeforeto obtainanestimateof the altitudebiasdueto blowing snow. Figure4 shows the altitudebiasdueto blowing snowfor two differentopticaldepths(filled circlesandsquares)at severaldifferentphysicalthicknessvaluesfor the snowlayer. Also shownarethe lower biasestimatesobtainedwhenthe calculationsarerepeatedwith the Gaussianfit (correspondingopencirclesandsquares)describedin DSE. A blowing snowlayer50-100m thick with anoptical depth between0.05and0.1 will biasthealtitudesderivedby between2 and4 cm approximately;this biascanbeconsiderablyreduced(to between0.5 and1.0cm)by the useof the Gaussianfit methodto determinethe centroidof the returnpulse.
4. Summary
and
Atmospheric
multiple
ments
of surface
other similar cloud
Conclusions
interannual
altitude
space
properties
scattering
as envisioned
missions.
variations.
Using
of such clouds
variability
in altitude
measurements
visual
observers,
in altitude larger itself
However, which
altimetry
are known
Laser
of polar cloud
altitude
study
Altimeter
System
cloud
variation
(GLAS)
seasonal
properties,
The
likely
in the relevant
that the atmospheric
measureor
that most of the
have significant
is quantified.
from year-to-year
laser
indicates
of Antarctic
measurements
suggest
for precision
observations
lidar measurements completed
calculations
from observations
of the mean Antarctic
variability
in cloud
made
at the South
summer
and winter
occurrence
year-to-year
by clouds
than the accuracies is substantial;
a recently
derived
the likely
introduced
error source
and the poten-
inter-annual cloud
scattering
effects
properon
are not insignificant.
properties
the interannual
spaceborne
These
data from DSE, estimates From
a survey
on GLAS
a large
for the Geoscience
bias that will result
ties is also determined.
Using cloud
Also,
that will affect
tial impact
GLAS
is potentially
variation
in the altitude
in the path of the lidar pulse
specified
for the mission.
and a uniform
altitude
measurements
can be confined
to be under
and cloud
cloud-free
bias cannot
or optically
16
estimated
by surface
to be significant,
interannual
variability
conditions;
made
The bias and is often
in the bias
out of observations
observations
thin-cloud
bias were computed.
bias was also obtained.
be subtracted
to those
altitude
fraction
appears
Further,
Pole as well as the path delay
made.
from the satellite
this reduces
the alti-
tudebiasesa greatdeal. To overcomethe limitationsin altimetrymeasurements causedby the biasresultingfrom scatteringwithin cloud layers,ice sheetelevationsshouldthusbe determined only from cloud-freeobservations.This canbe achievedusingthe atmosphericchannelat 532 nm for cloud-detection,alongsidethe 1064nm channel'saltimetrycapability. The useof improvedwaveformanalysistechniques,moresophisticatedthanmerelyacceptedthe centroidof returnpulses,canfurtherreducethe biases. Evenwith the selectiveuseof clear-skyconditionsfor altimetrycalculations,near-surfaceblowing snowwhich occursfrequentlywill remainunaccountedfor. The proximity of the snowto the surfacemakesthis scatteringlayermorepotent(perunit optical depth)thanclouds,sincescattered,delayedphotonsremainwithin the field of view of theinstrument. An altitudebiasof 1-3 cmfrom the snowlayeraloneis likely. However,aswith clouds,the useof improvedmethodsto analyzethereturnpulsewill help in substantiallyreducingthe biasunderblowing snowconditions.
The upcomingGLAS mission,by monitoringice-sheetaltitudechangesover Antarcticaandelsewhere,is expectedto provideinformationon whetherglobalwarmingis affectinga sensitiveand importantpart of theplanet. Potentialmelting of high-latitudeice sheetsfrom warmingwill likely leadto significantrisesin sealevel,andconsequentlyto catastrophicoutcomesalongcoastlinesaroundtheworld andin manyislandnations. This papersuggeststhatthe measurement accuraciesnecessaryto permitthe requiredmonitoringareachievableunderconditionsof thin or no cloud cover. Carefulselectionof datafrom which GLAS altimetrymeasurements aremadeis thereforenecessaryto ensurethatrangingdelaydueto scatteringis accountedor correctedfor.
A factor
that has not been included
bias. The results scattering cult.
of DSE suggest
on the path delay,
In addition, These
factors
other
forms
of return
els.
Further
study
that sloped
and make
other factors
ined.
in this analysis
pulse
analysis
is necessary
surfaces
noise
and surface
to determine
how signal
17
noise,
pulse
roughness
of Gaussian to reduce
slope
the effects
of the return
the effectiveness may be required
of surface
may obscure
the determination
such as signal
may also reduce
is the effect
fitting
altimetry rough,
on the altitude of cloud
centroid have
multiple more diffi-
not been exam-
on the path delay, biases
sloped
to acceptable
surfaces
and
and lev-
advanced
waveform
analysis
of the return
pulse
may affect
the multiple
scattering-induced
alti-
tude bias.
Acknowledgments.
Part of this work 33015
and NAG5-7522.
Department National
was supported
of Energy
by USRA
The total cloud through
its Carbon
contract cover
NAS5-32484,
database
Dioxide
Laboratory.
18
of Hahn
Information
and by NASA
contracts
et al. was provided Analysis
Center
NAG5-
by the U.S.
at Oak Ridge
REFERENCES
[1] Vaughan,
D. G., and C. S. M. Doake,
on the Antarctic
[2] S. Cohen,
J. Degnan,
ing system,
[3] W. Krabill,
[4] H. J. Zwally,
IEEE
Trans.
ice sheet:
[5] H. J. Zwally,
B. Schutz,
D. Hancock, Thomas, mitted
measured
Geoscience
W. Abdalati,
Laser
J. D. Spinhirne,
GLAS
altimetry.
Remote
Sensing,
S. Warren,
face observations,
K. Kuivinen,
and E. W. Eloranta,
1982-1991:
data package
sis Center,
Oak Ridge
laser
Greenland
22, 2341--2344,
1589,
thick1995.
Growth
of the
1989.
A. Brenner,
K. Quinn,,
1987.
ice sheet
and J. G. Marsh,
1587--
altimetry/rang-
pp. 581--592,
Res. Lett.,
C. Bentley,
J. Bufton,
S. Palm,
Ice, Atmosphere,
Atmospheric
of single
92-101,
J. Dezio,
J. Spinhirne,
Ocean,
and R.
and Land.
Sub-
Data
Lab,
[8] J. A. Curry,
W. B. Rossow,
D. Randall,
radiation
characteristics,"
J. Clim.,
pulse
multiple
bias,"
IEEE
scattering Trans.
effects
on
Geoscience
2001.
"Climatological
NDP-026A, National
246,
of Polar
of ice shelves
2001.
Part I: Calculations
Numerical
Geophys.
B. Minster,
Measurements
and J. London,
GE-25,
and S. Manizade,
J. Abshire,
T. Herring,
vol. 39, pp.
The geoscience
Sensing,
Science,
and retreat
1996.
R. A. Bindschadler,
Measurement,
to the J. Geodynamics,
[6] D. P. Duda,
Remote
J. A. Major,
warming
and J. Abshire,
by laser altimetry.
D. Harding,
ICESat's
atmospheric
379, 328-331,
J. Garvin,
K. Jezek,
A. C. Brenner,
Greenland
Nature,
J. Bufton,
R. Thomas,
ness changes
[7] C. Hahn,
Peninsula.
Recent
data for clouds
tape documentation Available
over
the globe
for the total cloud
from Carbon
Dioxide
from
sur-
edition,"
Information
Analy-
1995.
and J. L. Schramm, vol. 9, pp. 1731-1764,
19
"Overview 1996.
of Arctic
cloud
and
[9] Japan
Meteorological
Rec.,
Agency,
"Climatological
vol. 73, pp. 273-303,
[10] A. Mahesh,
V. R Walden,
over the Antarctic
[I 1] A. Mahesh,
Plateau.
Applied
crystal
[13] V. N. Smiley,
[15] D. Lubin, infrared
System
[17] V. R Walden,
L. Smith
at South
winter
Res.,
of cloud
of Applied
properties
Meteorology,
and K. Stammes,
retrieval
Sizes",
Pole during
clear-sky
properties Journal
of lower-tropospheric
and J. A. Warburton,
clouds
radiative
derived
of
ice
1993.
"Lidar
determinations
precipitation,"
from radiometric
properties
E. J. Welton,
Algorithm
912, Goddard
South
of cloud
vol. 6, pp. 1467-1472,
vol. 9, pp. 3405-3418,
(GLAS),
sensing
and Particle
vol. 98, pp. 12,961-12,971,
"Cloud
and S. G. Warren,
skies over
Antarct.
of
J. Appl.
1980.
of austral
J. Climate,
able from Code
cloudy
layers
W. Hart, D. L. Hlavka,
Altimeter
J. Climate,
B. M. Morley,
and D. A. Harper,
[16] S. E Palm,
Depths
"Satellite
J. Geophys.
data."
Optical
and T. P. Ackerman,
ice crystal
South Pole,"
sensing
remote
2001.
in the polar regions,"
"Properties
Antarctica,"
Journal
"Ground-based
vol. 40, pp. 1279-1294,
vol. 19, pp. 1074-1090,
[14] R. S. Stone,
remote
Base Heights.",
Part II: Cloud
B.M.Whitcomb,
atmospheric Meteor.,
Part I: Cloud
Plateau.
J. A. Curry,
clouds
"Ground-based
and S. Warren,
Meteorology,
[12] L. D. Wilson,
station,
2001.
V. P. Walden,
over the Antarctic
at Syowa
1981.
and S. Warren,
vol. 40, pp. 1265-1278,
data
Space
Flight
"A spectral
Pole," Eds.,
Center,
over the South
A. Deepak
2O
Pole from
AVHRR
1996.
Basis
Document,
Greenbelt
downward
IRS '96: Current
at the
1993.
and J. D. Spinhirne,
Theoretical
profiles
longwave Problems
Publishing,
"Geoscience Version
Laser 4.1."
Avail-
MD 20771,2001.
climatology in Atmospheric
ttampton,
for clear
and
Radiation,
VA, pp: 54-57,
1996.
W.
List of
Figures
Figure
1. Mean
Figure
2. Histogram
using cloud
annual
(a) includes
delay
under
obtained
observations
separately.
sky observatrions
by Monte
depth
and particle
delay
is calculable,
only a lower
Figure
in DSE.
induced
Carlo
calculations.
radius
are known.
either
because
limit to the particle
depths, The filled
to be zero.
value
obtained
by Monte
panel,
induced
Carlo
during
conditions;
In the lower
of the scattering
[6].
made
as well as cloudy
using
altitude
The pluses In other
effective
calculations,
1992.
The upper
scattering-induced only the cloudy
cases
bias from only the cloudy-
are at an optical obtained
from the Gaussian
radius
all observations
(+) represent
cases,
bias from
both the centroid
circles
bias from
only a lower
altitude
of 0.05; each of these were ues obtained
errors
from
measurements
of both clear-sky is assumed
derived
is 10.8 cm.
4. Scattering-induced
ent optical
altimetry
The median
3. Multiple-scattering
obtained
over the Arctic
from interferometer
clear sky conditions
are considered
Figure
occurrence
of scattering-induced
properties
panel
cloud
data when
only a lower
blowing
of the return depth
(open
snow. pulse,
depth circles),
Results
fit at the two optical
respectively.
21
of the return depths,
both cloud
is known
optical
squares
pulse.
are shown
induced
or both (filled
are shown
1992,
(diamonds),
or
circles).
for two differ-
as well as the Gaussian
of 0.1 and the filled
from the centroid
during
limit to the scattering
limit to the optical is known
of clouds
fit discussed
at an optical
The corresponding
as open circles
depth val-
and squares
Table1: Altitude biasvalues,andchangesin thosevaluesfrom 1992annualandseasonal averages,assumingthat average 11% in the winter)
is contained
More
clouds
year-to-year entirely
than
variation
in either
in 1992
in cloud
thick Fewer
occurrence
('r>2) clouds clouds
than
(5% in the summer,
or in thin (z