JOURNAL
Albedo
Alan
OF GEOPHYSICAL
over the boreal
K. Betts
RESEARCH,
VOL. 102, NO. D24, PAGES 28,901-28,909, DECEMBER
26, 1997
forest
and John H. Ball
Atmospheric Research, Pittsford, Vermont
Abstract. Using the Boreal Ecosystem-Atmosphere
Study mesonetdata, we showthe
annualcycleof albedofor 1994 and 1995 at 10 sites;two over grass,one over an aspen
forest, and an averageof sevenover coniferousforests.Representativedaily average
albedovaluesin summerare 0.2 over grass,0.15 for aspen,and 0.083 for the conifer sites.
In winter the correspondingmean albedofor snow-coveredgrass,aspen,and conifer sites
with snowunder the canopy are 0.75, 0.21, and 0.13. The jack pine siteshave a higher
mean winter albedo of 0.15 than the predominantlysprucesitesfor which mean albedo is
only 0.11. Forest albedo increasesat all sitesin winter (with snowon the ground under
the canopy) as the ratio of diffuse to total solar flux increases.The albedo of the conifer
sitesin winter rarely reaches0.3.
1.
Introduction
A key componentof the surfaceenergybalanceover land is
the albedo c•, defined as the ratio of the reflected solar radia-
tion (S •' ) to the incomingsolarradiation(S $ )
or- (S $ /S I )
(1)
measurements,five of the sites(sites1, 3, 4, 6, and 9), marked
with an asteriskin Table 1, had measurementsof incoming
diffuse radiation and longwaveradiation.
One of the objectivesof BOREAS wasto usemeasurements
to help improve global forecastmodels, and this analysishas
this in mind. Many global forecast models (unlike many climate models,discussed
below) havea singlemodel surfaceand
The surfacealbedo can have a large impact on the net radiation
do not treat albedo well over conifer forests in winter; that is,
and S $ small(as in midwinterat high latitudes),R,, is gen-
deep snowbetweendensecanopiesreflectsrelativelylittle solar radiation, becauseof canopy shading.During the April
1996 field campaignthe daily forecast temperature maxima
over the BOREAS siteswere often low by 10 K in the ECMWF
model (P. Sellers,personalcommunication,1996) simplybecausethe albedo calculatedin the model was as high as 80%.
The effect of canopy shadingon the albedo of forests in
winter and its climatologicalsignificancehas been known for
sometime [McFaddenand Ragotzkie,1967;Federer,1968, 1971;
Robinsonand Kukla, 1984, 1985]. Robinsonand Kukla [1984],
they do not treat the shadingof the snowon the groundby the
forest canopy.In somemodels,forest albedo in winter is too
(2) often as high as 60-80%. The National Center for Environmental Prediction (NCEP) model fixes albedo over snow
(south of 70øN) at 0.6 followingMiyakoda and Sirutis[1986];
whereL $ and L •' are the downwelling
and upwellinglong- while in the European Centre for Medium-Range Weather
wave fluxes at the surface.The net longwaveis typically neg- Forecasts(ECMWF) model, albedo was as high as 0.8 after
ative, becausethe effective thermal radiative temperature of fresh snow [ECMWF, •991], prior to revisionsmade to the
the atmosphereviewed from the surface is cooler than the winter forest albedo in December 1996. In nature, however,
surfacetemperature.The net solarradiation is positiveduring snowfalldoes not stay long on the canopy,particularly if the
the daytime,and unlesscris large (suchasover a snowsurface) solarradiationS $ is large,or if the wind is strong,and the
erally positivein the daytime. PositivedaytimeR,, drivesthe
surface fluxes which in turn are the surface forcing of the
atmosphereboundarylayer [Bettset al., 1996; Eltahir, 1996].
During BOREAS (Boreal Ecosystem-Atmosphere
Study),
10 mesonet
stations were located
in Saskatchewan
and Mani-
toba acrossthe boreal forest [Shewchuk,this issue].Table 1
liststheir locations,elevation, the type of vegetation,and the
station numbers we have assignedto them. Two sites in the
south and southwestwere over grass;one was over a stand of
old aspenin the Prince Albert National Park; and three were
over jack pine sites:two at BOREAS flux tower sites in the
southern and northern study areas (SSA and NSA, respectively) and one at Lynn Lake, the most northerly site. The
remaining four were over mixed spruce and poplar stands
(mostly conifer). Instrumentsmounted on towers extending
abovethe forest measuredatmosphericvariablesand radiation
from aircraft data over southeastern New York State, showed
that
the difference
between
snow-covered
and snow-free
al-
bedowas0.72 for a peatyfield but only 0.1 over a mixedforest.
On a global scale,Robinsonand Kukla [1985], from an analysis
of polar orbitingsatellitedata, showedthat the northern boreal
forestsstand out on a global map as regionsof low maximum
albedo in winter; they quote a mean value of 0.36 for the
fluxes. This note summarizes the albedo measurements
calcumaximum winter albedo of the boreal forest. In a studyusing
lated from upwardand downwardfacingEppley pyranometers
an energy balance climate model, Otterman et al. [1984]
locatedtypicallyabout 5 m abovethe forest canopyfor eight
showedthat the low winter albedo of the high-latitude forests
sitesand at 2 m over grassfor two. In additionto the shortwave
increasethe surfacetemperature at 65øN by 5 K.
Copyright1997 by the American GeophysicalUnion.
There has been considerablework on the developmentof
detailed
snowcover modelsfor climate simulations[e.g.,Loth
Paper number 96JD03876.
0148-0227/97/96J D-03876509.00
et al., 1993], as the issueof the impactof snowcoveron winter
28,901
28,902
BETTS AND BALL: ALBEDO
Table
1.
BOREAS
Mesonet
OVER THE BOREAL
FOREST
Stations
Station
Name
Long, W
Lat, N
Elevation,m
Vegetation
1'
2
3*
4*
5
6*
7
8
97
10
Saskatoon,SK
Meadow Lake, SK
SSA-oldaspen
SSA-OJP
The Pas,MB
Flin Flon, MB
La Ronge, SK
Thompson,MB
NSA-OJP
Lynn Lake, MB
- 106.60
-108.52
- 106.20
- 104.69
-101.05
-101.69
-105.29
- 97.87
-98.62
-101.09
52.16
54.12
53.63
53.92
53.97
54.67
55.13
55.80
55.93
56.89
480
480
587
511
268
305
281
221
282
366
grassland
grassland
aspen
jack pine
spruce/poplar
spruce/poplar
spruce/poplar
spruce/poplar
jack pine
jack pine
SK, Saskatchewan;
OJP, old jack pine; MB, Manitoba;SSA, southernstudyarea;NSA, northernstudy
area.
*Denotesthat incominglong-wave(L $ ) and diffuseradiation(D $ ) were alsomeasuredat these
sites.
?L $ andD $ instruments
wereseparately
locatedat the NSA fen siteat -98.42øW,55.914øN.
climate hasbeen of concernfor sometime [Barnettet al., 1989;
Cohen and Rind, 1991]. A few papers have specificallyaddressedthe impact of the low boreal forest albedo on winter
climate [Thomasand Rowntree,1992;Bonanet al., 1992, 1995].
Thomasand Rowntree[1992] in particularshowthat removing
the boreal forestsand increasingthe surfacealbedo reduced
the surfacetemperature in those regions.Bonan et al. [1992,
1995] showed the impact of the large difference in winter
albedobetweenthe northernforestsand the bare ground(and
tundra) on the net radiationbalanceand surfacetemperature,
using both off-line experimentsand climate model experiments.In the 15-yearclimate runs,the largestdifferenceswere
seen in April, when net radiation over bare snow-covered
groundwasreducedmore than 50 W m-2 at 50øN,and the
correspondingsurfaceair temperatureat 60øNwas 8 K cooler
with climatologicallyprescribedsea surfacetemperaturesand
still colder (-12 K) with an interactiveoceanmodel.
The data we have analyzed are for the first two years of
BOREAS: 1994 and 1995 (measurements
continuedin 1996).
We examinethe mesonetdata as a guide to first-ordermodel
improvements;other more detailed investigationsof snowon
conifercanopiesare in progress[Pomeroyand Dion, 1996].We
discusserrors in winter due to snow on the upward facing
pyranometer,showthe annualcycleof daily averagealbedofor
grass,aspen,and conifer sitesfor 1994 and 1995, present a
towers were carefully sited, but they extended only a few
meters(typically5 m) abovethe canopy,so the effectivefield
of view of the downwardlookingpyranometeris small.To get
someimprovementin representivity,we generateddaily mean
valuesby averagingthe 15-mindata.Daily averagealbedo(&)
was calculatedfrom the daily averagereflected(S •' ) and
incomingsolarradiation(S $ ) for eachsite.
In winter, snowor ice on the upwardfacingpyranometerwill
biasS $ low. This is importantbecausethe forestalbedoin
winter is a key issue,as discussedin section 1. We would like
to know whether snow remaining on the canopy gives the
forest a high albedo on occasions.Three factors might be
important: one is the melting of snow on the canopyby the
incoming solar radiation, the secondis the effect of wind in
shakingthe snowoff the canopy,and the third is a rise of the
air temperatureabovefreezing.The winter air temperaturesat
the BOREAS sites are generallywell below freezing, so we
shallnot discussthis last factor. Unfortunately,the conditions
of low wind and low insolationunderwhichsnowmightremain
on the canopyare alsothe conditionswhensnow(andice) may
remain on the upwardlookingpyranometer.This too givesan
apparently
highalbedo,
bybiasing
• $ low(while• •' isun-
affected),as discussed
in the next section.
In BOREAS, snow depth was measured by an ultrasound
gauge[Shewchuk,
thisissue]in a clearingbetweenthe canopy.
table of mean values for different sites both with and without
We have at each site only a singlemeasurement,so estimation
snow, and show the relation between albedo and incoming of the patchiness
of snowis not possible.Snowdepthsreach
solar and diffuse radiation for selected sites.
half a meter eachwinter, so major snowfallsare readilyseen.
However, drifting of snow may also causechangesin snow
depth, and small snowfallsof a few millimeters,which might
2.
Albedo Measurements
affect albedo of the canopy,are hard to detect in this instru2.1.
Instrumentation
ment record. A separateinstrument,a Belfort precipitation
The BOREAS mesonetsitestypicallysampleddata every5 s gauge [Shewchuk,this issue]containingantifreeze,measures
and the dataloggersgenerateda 15-minaverage.The accuracy precipitation amount and thus gavc an independentestimate
quoted by the manufacturerfor the Eppley pyranometersis of snowfall.
5% [Shewchuk,this issue],which would give an accuracyof
about 1% in albedo for low albedosbelow 20%. However, the
measurementaccuracyalso dependson the alignmentof the
radiometers.The radiometerswere mountedon the top of
tilting towers:theywere servicedregularly,generallyevery3-7
days and aligned usingjigs in a horizontal position,before
being tilted into a vertical position.Although this was done
with care [Shewchuk,this issue],someunknownerrors due to
alignmentmay result.A differentissueis the representivityof
the reflected
radiation
for the forest
as a whole.
The forest
2.2.
Winter
Albedo
Errors
Thereareproblems
withlowandpresumably
incorrect
• {
data in winter at someof the forest sites,almostentirelyin
December and January,when solar radiation levelsare very
low. The probablecauseis snowor ice on the upwardlooking
radiometer. We looked critically at high winter values of albedo and examined
the 15-min
time series of the different
radiation measurements,
particularlythe upwardlookinginstruments,
for consistency.
Five siteshaveonlyS $ andP•4R
_
BETTS AND BALL: ALBEDO
OVER THE BOREAL
(incomingphotosynthetically
activeradiation),while the five
with an asteriskin Table1 havealsoL $ (incominglongwave)
_
1.0
FOREST
28,903
,,,
x
1995
xxx
.
and/5$ (incoming
diffuse
radiation).
_
We rejectedwinterS $ databasedprimarilyon twocriteria:
b$/g$
0.8
Conifer
o
x
Aspen
Rejected
Xx x
x
(3)
because(3) impliesmoresnowon the pyranometermeasuring
ß
0.6
x
• $ thanon thatmeasuring
•1 $ , and
,,
PAR ,•/• ,• > 0.6
ß
(4)
x
becausetypicallyPAR $/S $ • 0.4, since(4) againimplies
moresnowonthe• $ instrument.
Thesefilterstypically
elim-
•i•"•.o•_,•
inated blocksof suspectdata (at primarily,the conifer sites4,
9, and 10 after heavysnowfalls)in midwinter (December and
A ."
oo
•'
..
ß . '.'.
'
January),
whentherecorded
(butbelieved
to betoolow)• $
•< 20 W m-g. We then reexaminedthe dailytime seriesof a
and filtered a few additional isolated days of high albedo,
which narrowlymissedour criteria but were insidesequential
daysof bad data. One further checkwas considered.Snowon
0.0
0
.• I. • I • I • I , I ,
50
100 150 200 250 300
S(Wm'2)
theL$ sensorgivesLy' - L$ • 0, but thiswasrelatively
rare (we calculatedL •' from a downwardlookinginfrared Figure 2. Albedo againstS • for all forest sites in 1995
thermometer).Note that we cannotbe certain that data that (Julian days <110 and >304). Rejected data (crosses)are
passesfilters(3) and (4) in midwinteris correct,astheremight included.
_
be snowevenly distributedon all the sensors.
Figure 1 illustratesthesefiltersby showingthe first 120 days
of 1994, until snowmelt at the SSA old jack pine site 4. Snow again marked with a cross.These include most of the high
depth,shownas a light solidline (right-handscalein centime- valuesin Januaryand an isolatedsnowevent in February.The
ters), risesfrom 150 to 250 mm duringJanuaryand remains highestremainingvalue in January(day 8) doescorrespondto
deep till snowmeltin April. The measuredincomingS $ recent snow,but it passedour simplefilters. On day 33 (Feb(right-handscale,shownwith thickdashes)is verylow in Jan- ruary2), albedo,calculatedfrom the 15-mindata plunges(not
_
uary (•20 W m-g), indicatingeither a very cloudyperiod shown),
as• $ risesduringthedaylight
hours,
butwebelieve
(clearskydailymeanS $ • 60W m g),or baddata,affected that thiswassnowmeltingoff the S $ radiometer,ratherthan
by snowon the S $ radiometer.The valueswe rejectedare off the canopy.
_
marked with a crossand includemost of the Januarydata. The
An independentcheckon the validityof our filtersis based
calculatedalbedo (thick solid on left-hand scale) is low for on the factthat dailyaveragevaluesof S $ are generallyquite
most of Februaryand March (0.15-0.2) exceptfor one day. well correlated(not shown)betweenthe old aspensite 3 and
The valueswe have rejected on the basisof tests 3 and 4 are the old jack pine site 4, which are about 100 km apart (when
the sensorsare snow-free).The fine dotted line (with legend
Alb (4/3)) is an albedo(from days22 to 59) for site4 calculated
Albedo 400
[
Snow -
•
........... AIb (4/3) --
300 •
I
0.4
.go
'..' _
,,
o.2
0.0
I --
Wind Speed
T
'•'
-20
-4O
200
,",'" ,," ,'
30
filters.
•
Alb(4/3)
=••'(4)/•
$(3)
(5)
_Alb(4/3)
islower
than
thealbedos
wehave
rejected
(meaning
S $ at site3 is higher,evenfor two snoweventson day28 and
100
49,andit agrees
reasonably
wellwith&fromsite4 forthose
0
conclude that we were correct in rejecting most of the high
albedo data in January.
The bottom panel showsdaily mean windspeed(on right
dayswherewe havenot rejectedthe site4 • $ data.We
4 • hand
scale),
mostly
<3
ms
•inJanuary,
and
daily
mean
tem2 •
peratures,which are well below zero (dashed on left-hand
scale)in Januaryand February.The meltingof the snowpack
i
inMarch
and
April
ondays
where
mean
temperature
rises
above freezing is clearly visible.
1
0
using
•: •' fromsite4 and• $ fromsite3, whichstarted
collecting
dataonJanuary
20:thedatafromday22passed
our
ß
tt\iI•.
_
60
Day
90
120
2.3. Dependence
of WinterAlbedoon• ,• andWindSpeed
Figure 2 is a scatterplotof the calculatedwinter albedofor
Figure 1. Albedo,snowdepthandS $ (top panel)andwind all the forestsites(for Juliandays<110 and >304 in 1995).
speed,air temperaturefor southernstudyarea old jack pine The crosses,which include all the high albedovalues >0.6, are
site for beginningof 1994.
the data which we rejected, as discussedin section 2.2: they
28,904
BETTS AND BALL: ALBEDO
OVER THE BOREAL FOREST
a better representativelarge-scaleestimate of the fraction of
the incomingsolarenergyreflectedby the primarilyconiferous
boreal forest, we showan averagealbedo calculatedas
0.6
1995
ß
Conifer
o
Aspen
(6)
where the angle bracket denotes an average over the seven
sites4-10. This method of averagingalso reducesthe relative
impactof anyremainingdaysof snow-contaminated
data,since
0.4
theseareall daysof verylow• $ and• q'.
0.2
0.0
,
0
I
,
2
I
•
4
I
•
6
I
8
WindSpeed(ms
-1)
Figure 3. As Figure 2 for albedo againstdaily averagewind
speedbut without rejecteddata.
In Figure 4 the two grasssiteshave the highestalbedo, both
in summer(•0.2) and in winter (•0.7-0.8, wheneverthey are
snowcovered).The most southernsite at Saskatoon(site 1)
has more periods in fall and winter without snow. The aspen
site hasthree albedoregimes:in summerwhen the canopyis in
leaf & • 0.16, in winter & • 0.21, and in springand fall there
are transition periods with the lowest albedo •0.12, when
there is no snow, and the canopy is bare. The conifer siteaveragehasthe lowestalbedoin both summerand winter. The
summer mean is 0.083, while when snow lies on the ground
underthe canopy,{&) • 0.15 typically,with occasionally
higher
values. Both in summer and winter, the conifer sites have a
significantlylower albedo than grass.In summerthe difference
between & = 0.2 for grasslandand 0.083 for conifersis significant in the radiation balance. In winter the corresponding
correspond
to low S $ measurements,
caused,we believe,by albedo difference is still larger: with & as high as 0.7-0.8 over
snow or ice on the upward facing pyranometer. The open snow-coveredgrassand, typically,only 0.15 over the coniferous
circlesare for the aspensite which, with a more open leafless forests.Although the incoming solar energy is small in midcanopyin winter, generallyhashigheralbedos.The dotsare for winterat theselatitudes,
in MarchandApril,• $ hasbecome
all the conifer sites4-10; these have generallylower albedos. significant,while the ground snow cover typically has not
The dayswith 0.4 < & < 0.6, which passedour filters 3 and 4, melted. Consequently,as mentioned above, global forecast
are all in December and Januaryand still havevery low values models that do not predict correctly the low albedo over the
of • • <20 W m-2. We consider
thesealsosuspect,
andasa forest have large errors in surfacetemperature in spring.
result, we can say nothing about the occasionsin midwinter,
Figure 5 showsthe correspondingalbedos for 1995, for
when snowmay stayon the canopybecauseof the low incom- which the year's record is complete. It is similar to 1994, aling solarradiation,becausein just theseconditions,snowalso thoughthere are somedifferencesin winter, presumablyassoappearsto remain often on the upward facingpyranometer. ciated with snowfall events. In midsummer, albedos are lower
Nonetheless
we can draw some conclusions
about the mean
albedo of the conifer forest in winter. Whenever the average
dailyinsolation
• $ > 50 W m 2, thealbedoof theconifer
sitesrarely exceeds0.2.
For the "winter" days shown in Figure 2, we only rejected
2% of the data for the aspensite 3, while for the conifer sites,
we rejected 8%, almostall in December and January.Even in
1.0
[ I
these midwinter months, the mean albedo for the 78% of the
0.8
1994
I-
,I
--
/
I•II
-' Grass
#2
Grass #1
.
,
...........Aspen
unrejectedconifer data was 0.18. Since this midwinter sample
maystillcontainsomebad• • data,it isworthnotingthatin
90% of this sample,the conifer albedowas <0.3. We conclude
that even in midwinter the conifer albedo is low, generally
<0.3.
Figure 3 showsthe correspondingfigure for daily average
albedo against daily average wind speed for the forest sites,
with the rejected data removed.Assuming(which is unlikely,
exceptperhapsfor the aspensite) that the remainingdaysof
higher albedo are correct, we see that all values of & > 0.3
correspond
to dayswith low wind speed<3 m s •, whichis
consistentwith the possibilitythat more snowmay stay on the
canopyin low winds, but clearly, this conclusionis very weak.
2.4.
Seasonal
Variation
of Albedo
for 1994 and 1995
Figure 4 plots daily albedofor grass(sites1, 2), aspen(site
3), and an average albedo for the sevenconifer sites. Gaps
correspondto dayswithout good data. The variation between
the conifer(or mainlyconifer) sites4-10 is small,so to obtain
o0.6, ......
Conifer
<
0.4
i
0.0'
0
60
•
120
180
240
300
360
Day
Figure 4. Daily average albedo for 10 BOREAS mesonet
sitesfor 1994; showingtwo grasssites,the aspen site, and an
average of the sevenconifer sites.
BETTS AND BALL: ALBEDO
I• I
I1'•,
-'Grass#1
i1/I.
-'Grass#2
0.8
11'11U
...........Aspen
......
Conifer
OVER THE BOREAL
FOREST
28,905
calculatedfrom the individual daily albedosto give an indication of their spread. Albedo at the two grasssites increases
from nearly 0.20 in summer to 0.75 with snowon the ground.
For the singleaspensite we showthree albedos:the lowestof
0.116 is with no snow and no leaves; in summer, when the
aspenis in leaf, the mean is 0.156, while when the canopyis
bare and there is snow on the ground, the mean is 0.21. The
next group of three jack pine sites 4, 9, and t0 show some
variation of summeralbedowith the most northern site having
0.6
the lowest albedo:
0.4
the "no snow" mean is 0.086. In winter
this
group has a winter mean of 0.15; a smallerincreasein albedo
than the more open aspensite. The last group of four sites,
describedby Shewchuk[this issue]as spruce/poplarsites,have
a low albedo in summer and the lowest albedos in winter. Their
0.0
0
-"'"180
60
120
240
300
360
Day
Figure 5. As Figure 1 for 1995.
summer average is 0.081, and this increasesonly to 0.108 in
winter. It appears that these towers were sited so that the
downwardradiometerssee mostlyspruceforest. Column 5 in
Table 2 showsthe increasein albedo(as a percent) associated
with snowon the surfacefor each site and vegetationtype. It is
a convenientsummaryof the huge impact of canopyshading
on winter
2.6.
albedo.
Distribution
of Forest
Albedos
in Winter
in 1995 than 1994 at the two grasssites,perhapsbecauseof
lower rainfall in the precedingtwo monthsin 1995.
Figure 6 showsthe normalized frequency distributionsof
daily averagealbedoin winter by forest site. The data are from
1994 and 1995, and we show the total number of daysin 0.02
2.5.
Mean Site Albedos With and Without
Snow
albedo
classesfrom 0.07(_0.0t) to 0.39(+_0.01),normalizedby
Table 2 summarizesthe albedo for the t0 sites, in groups
the
total
number of daysfor whichwe have acceptedthe data
with and without snow. We have excluded days where we
for
that
site
(on the average,there are 330 daysper site). The
rejectedthe S ,• data and alsoa few daysclearlyassociated
with transitions,such as the final melt period in spring.The tail of the distribution for & > 0.4 is representedby a single
albedo averageis here calculated,not as an averageof the point plotted at 0.45, becausewe havelessconfidencein these
high albedovalues,as discussed
earlier (the highest0.45 point
albedos but as
is the most northernjack pine site t0). We have groupedthe
data by forest type. For the aspen site, the distribution is
broadest
with a peak at 0.21. The jack pine sitesin the SSA (4)
where the secondoverbardenotesan averageover all daysin
and
NSA
(9) havepeaksat 0.15,while the mostnorthernjack
the group.This methodof averagingmakesalmostno difference to the summermean, but we think it givesa slightlymore pine site t0, whichhasa low albedoin summeralsohasa lower
representative
andaccuratemeanmeasureof the forestalbedo winter peak. The spruce/poplarsiteshavegenerallythe lowest
in winter, when radiation levelsare low. In winter in particular, albedoin winter, particularlysites5, 7 and 8; their distributions
it givesa low weightingto the lowestS ,• days,whenalbedo peak even in winter in the range (0.09 + 0.0t), an extremely
valuesare often high,but we have the least confidencein the low albedo,consideringthere is snowon the groundunder the
accuracyof the S $ data. Table 2 alsoshowsa countof the canopy.Indeed, for sprucesites7 and 8, almosthalf the days
number of days in each average;and a standarddeviation, with snow under the canopyhave & < 0.t.
_
_
Table
2.
Mean Site Albedos With and Without
Site
1
2
3
4
9
10
5
6
7
8
Type
grass
grass
grassmean
aspen(bare)
in leaf
jack pine
jack pine
jack pine
jack pine mean
spruce/poplar
spruce/poplar
spruce/poplar
spruce/poplar
spruce/poplar
mean
Snow
No Snow(Count)
Snow(Count)
PercentIncrease
0.199_+0.015(408)
0.194+_0.021(409)
0.197
0.116 _+0.011(115)
0.156 _+0.013 (245)
0.090 +_0.005(276)
0.091 _+0.006(304)
0.076_+0.007(227)
0.086
0.088+ 0.008(356)
0.085_+0.990(243)
0.076_+0.004(325)
0.076+ 0.005(231)
0.081
0.72 _+0.10 (182)
0.77 _+0.10(267)
0.75
0.214 _+0.067
51.7%
57.6%
54.7%
9.8%
0.155 _+0.038(264)
0.167 _+0.051(226)
0.129 _+0.011(365)
0.150
0.115_+0.049(348)
0.119 _+0.052(232)
0.100_+0.092(360)
0.907 +__
0.076(253)
0.108
6.5%
7.6%
5.3%
6.5%
2.7%
3.4%
2.4%
2.1%
2.7%
28,906
BETTS AND BALL: ALBEDO OVER THE BOREAL FOREST
clear separation:the grasssites,whenever snowcovered,have
0.45
•
Aspen (#3)
- - - Jack-Pine(•4,9,10)
0.40
......
0.35
a muchhigheralbedoanda dailyR• averaging
50 W m-2 less
thantheforestsites(butthesameaverage
S $ • 155 W m-2
_
Spruce/poplar(#5,6,7,8)
11#4
••., 0.30
0.25
not shown).These observationsare comparableto the model
differencescited by Bonan et al. [1995] for the differencein net
radiation in springover forest and snow-coveredbare ground.
They are also likely to be representativeof the error in R•
introducedinto a globalforecastmodel in March with a model
albedo
of 0.75 instead
of the observed
0.10 to 0.15.
•_ 0.20
•
0.15
3.
0.10
,,,/4•II,,,•
///
0.00
0.1
0.2
0.3
0.4
Albedo (:•0.01)
Figure 6. Normalized frequency distribution of forest site
daily averagedalbedosin winter (from 1994 and 1995data).
Winter
The daily albedo measurementsshow a dependenceon the
ratio of diffuse to incomingtotal solar radiation in both sum_
mer and winter.Incomingdiffuseradiation(D $ ) wasmea0.0
2.7.
Albedo Dependence on Diffuse/Direct Solar
Radiation
Albedo
and Net Radiation
Balance
The impact of the differencebetweenwinter albedo over
grass and forest is large on the net radiation balance R,.
Figure 7 showsdaily averageR, for March 1995 for all 10
BOREASsitesfor all dayswithS $ > 120 W m-g;thatis,we
sured(Table 1) by shadow-bandradiometersat five BOREAS
sites [Shewchuk,this issue]'the grasssite at Saskatoon(see
Table 1), the Old Aspen site, the SSA Old Jack Pine site, Flin
Flon, and in the NSA (at the fen site). For the last two the
diffuse
and solar measurements
were not collocated.
The dif-
fusefluxratio,(D $/• $ ) isanindicator
ofwhether
a daywas
clearandcloud-free
(when•) $/• • --•0.2),orovercast,
when
•) $/• $ mayreachunity.
Thegraphs
of dailyaverage
albedoagainst
(•) $/• $ ) are
interesting.They includeall the data from 1994and 1995which
passedour filters 3 and 4. Figure 8 shows albedo plotted
against
(D $/• $ ) for .thegrass
siteat Saskatoon.
Thedata
are in three groups:the solid circles are bare grass,the open
haveexcluded
overcast
days
(theclearskydailyaverage
• $ at circlesare snow-coveredgrass,and the crossesare transitional
thebeginning
of Marchis =150 W m-g). Thesolidsymbols
are dayswith melting snowor morning frost. Two linear regression
the conifer sites,for which the mean snowdepth is 44 cm, the
crossesare the aspensite which has a similar snowdepth (50
cm), but a more open canopy,and the opencirclesare for the
two sitesover grass,which have a lower snowdepth in March.
At site 2 the mean snow depth is 16 cm, while for site 1, it is
only 3 cm, except for the period of March 11-25, when the
snow melted, and the albedo was correspondinglylow. The
squaresdenoteaveragescorresponding
to the coniferdata and
the clusterof snow-coveredgrassdata with & > 0.5. There is a
lines are plotted for the bare grassand snow.
(7)
The regressioncoefficients.4 and B and their standarddeviation are givenin Table 3 for this and the succeedingfigure.s. In
1.0-0.9-
125 -
ß
.
March 1995
I
Jl
ß Conifer
'. ' .
100 -
+ Aspen
ß
-
el•e
e- +
•o++
75-
0.8-
o Grass
0.70.60.5.
.
ß
.t
o
0.4-
+
x
x
.
50-
. ..•: •...
o
'•
oc•
++
t
25
+
0.3
o d•
o
o
-
o
o.o.
o
-25
0.2
0.4
0.6
Albedo
0.8
1.0
x
ß
', ½'
Transition
Bare Grass
'
o.o
o
' ß
o Snow
0.1-
o øo
0.0
ß
0.2
000•
I
0.2
'
I
0.4
'
I
'
0.6
I
0.8
'
I
1.0
D½/S.J,
Figure 8. Albedo as a function of the ratio of incoming dif-
fuseto solarflux(•) $/• $ ) for thegrasssiteat Saskatoon.
Figure 7. Daily averagenet radiation (R,) againstalbedo Three classesare shown,together with linear regressionlines
for grass,aspen,and conifer sitesfor March 1995.
for bare grassand snow.
BETTS AND BALL: ALBEDO OVER THE BOREAL FOREST
Table 3.
(b
28,907
RegressionCoefficients•1, B for Albedo Dependence in & = /1 + B
$)
No Snow
Site
Type
1
3
3
4
6
9
Snow
A
grass
aspen(bare)
aspen(leaf)
jack pine
spruce/poplar
jack pine
B
0.206 _+0.015
0.122 + 0.010
0.156 _+0.013
0.094 _+0.005
0.093 _+0.007
0.094 _+0.006
-0.018
-0.016
0
-0.009
-0.018
-0.006
A
_+0.003
_+0.004
_+0.004
+ 0.001
+ 0.002
_+0.0015
0.64 + 0.09
0.182 _+0.025
0.16 +_0.02
0.063 +_0.006
0.140 + 0.034
0.119 + 0.036
0.139 _+0.036
0.054 _+0.007
0.023 +_0.009
0.075 + 0.009
summer,
albedo
decreases
slightly
withincreasing
(b ,•/• $ ), with snow on the ground, albedo increaseswith increase difwhile in winter albedo increases,althoughthe scatteris large. fuse flux ratio.
Figure 9 showsthe Old Aspen site. We have partitioned the
Figure 10 showsthe correspondinggraph for the SSA old
days into three groups,which separateclearly on the graphs. jack pine (OJP) site for 1994. The separationis again by day
The lowest group (solid circles) are days in spring and fall into two main classes for this conifer site. Solid circles are for
(days102-140 and 273-304 in 1994 and 122-145 and 278-304 dayswithout snow (days 101 to 304 in 1994 and 122-300 in
in 1995), when there is no snowand the trees are not in leaf. 1995) and open circlesfor the winter, when snowwas on the
These have the lowest albedo • 0.12 and a weak decrease of
ground (there is no diffuse data before Day 50 in 1994). The
albedo
withincreasing
(D $/• $ ). Theregression
coefficientsclear separation is quite sharp: we have also distinguisheda
are againin Table 3. The crossesare when the canopyis in leaf few days in spring (as crosses),when the final thaw of the
(Julian days141 to 272 in 1994 and 146-277 in 1995), with a snow-packoccurs.Without snow,albedois low (•0.09), with a
higher albedo •0.16, and no trend with increasingdiffuse flux slight decreasewith increasingdiffuse flux ratio. With snow
ratio. The open circlesare from the winter (days <102 and under the canopy, an increasingtrend of albedo with diffuse
>304 in 1994 and <122 and >304 in 1995),when there is snow flux ratio can againbe seen,althoughthe scatteris large. There
on the ground(and no leaveson the canopy).We interpret the are no days at this site with & > 0.3. Figure 11 shows the
upper scattered points with • > 0.3 as the occasionswhen correspondingvery similar graph for the OJP site in the NSA.
there may be fresh snowbriefly on the canopy,and the broad The dates with snow lying are different, but the pattern and
band of points with albedo increasingwith diffuse flux ratio regressionlines are the same. We excluded the two isolated
from 0.18 to 0.24 as the generalwinter conditionwith snowon points with albedo closeto 0.5 from the winter regressionline
the ground and none on the canopy.We fitted three regression calculation.For thesejack pine sites,it is fairly easyto detect
lines as shownto the three groupsof data. We excludedall the the presenceof snowunder the canopyfrom the albedo. Note
data above the dotted line shown (a = 0.32) in the linear alsothat the jack pine albedois %0.2 mostof the time in winter
regressionfit to the winter snow-on-groundcase.We see that and only rarely reaches0.3.
0.6
o
Snow
x
ß
Bare
0.5
Leaf
o
-
0.5
-
o
o
o
oø
o
o
o
o
-
o
..............................
;;..................................
o......-• ............
-•..
oo
-
0.2
%0
.oøo• o
•
oo
o
ø o o L•,•A.•--'
0.1
o
-•
0.3-
(J•
O.Ot
I
0.2
•
o
.1•
-
<
0.2--
o
o
o
o
ooo
O
o
øø
x•'
O
x
.....
O .OaOO O
oooo
øø
xu•
o
o
O
øo_m
•Oøo o o • •_•----
--
01
0.0
No Snow
o
o
0.3
Transition
ß
0.4-
o
0.4 _
Snow
x
•
o
'"o c•_8 o
• o o o%•
••
•
"o •o
• x
x x -..
. o
.
-
I
0.4
•
I
i
0.6
I
0.8
I
I
'
1.0
D-,I,/S,/.,
Figure 9. As Figure8 for the southernstudyarea(SSA) old
aspensite.The three classesare bare (no snow,no leaves),in
leaf, andwith snow(and no leaves).The data abovethe dotted
line were excludedfrom the snow regressionline.
0.0
0.0
,
I
0.2
I
I
0.4
,
I
0.•
[
I
0.•
I
I
• .0
D/S{
Figure 10. As Figure 8 for SSA old jack pine site. The three
classesare with and without snow on the ground and the
transition daysduring springthaw.
28,908
BETTS AND BALL: ALBEDO OVER THE BOREAL FOREST
0.5
o
dependenciesshownin Figures 8-12 may have severalexplanations and deservefurther study.
Snow
x Transition
ß
No Snow
0.4
4.
o
o'
0.3
o
o
0.2
..
•"
o
.........
o
ß
x
xo
xo
O. •
•-
•
o
......
•'••'•
o o
o
o
o
o• o
o
(•".½'-"
ß
flux measurements
o
oO•
o•
This paper analyzesthe albedocalculatedfrom the radiation
Oøo
o o o,•
o:
o0
o•' o
o •Ooo
o
ß ß
ß
--
0.0
0.0
,
I
,
0.2
I
0.•
,
I
,
0.•
I
0.8
Conclusions
,
I
1.0
]]. • •.•c
]0 •o• •o•hcm •.dy a•ca (•)
p•c •kc. The two •opmo•
old
at the BOREAS
mesonet
sites. The
mea-
surementsare ongoing,and we have analyzedthe data from
the first two years, 1994 and 1995. We show the annual cycle
for both years,and we have then focusedon the winter season,
becausecurrent global forecastmodels (unlike many climate
models) represent the forest albedo in winter very poorly.
When snowis on the ground,the forecastmodelsat NCEP and
ECMWF too often have albedosas high as 0.6 to 0.8 for the
boreal forest. This has a severe impact on the surface net
radiation budget, especiallyin March and April, before snowmelt. In contrast, the mesonet data showvery low albedosfor
the boreal forest in winter, typicallyin the range 0.1-0.2, and
rarely as high as 0.3.
In summerthe daily averagealbedo of the coniferousforest
is very low: in the range 0.076 to 0.091. The aspensite has a
seasonal variation with a summer value of •0.156, while the
two sitesover grasshave a higher summeralbedo, a little less
than 0.2. The albedo difference between forest and grass,
which is significantin summer is even larger in winter and
Figure 12 showsa corresponding
graphfor the spruce/poplar spring, when there is snow on the ground. The BOREAS
site at Flin Flon. The scatteris a little larger than for the jack mesonetsitesshow that while over snow-coveredgrassalbedo
pine sites,and the separationbetweensummerand winter is •0.75, the mean albedoin winter for the more open aspensite
reducedfor this largelysprucesite, as seenin Tables 2 and 3. is 0.21, for the jack pine sites0.15, and for the mainly spruce
Summer albedo decreaseswith diffuse flux ratio, but the small sitesonly 0.11. There are very few daysin winter when we have
increasein winter shownby the linear regressionfit is arguably gooddata that the forestalbedoreaches0.3. The forestalbedo
may occasionallybe higher on occasionsin midwinter when
insignificant.
There are severalpossibleexplanationsfor the decreasewith snowis on the canopy,and the dailyinsolationS $ •< 50 W
_
it appearsthat thereis often
(D $/• $ ) in albedofor all snow-free
sites(except
aspenin m-2 but undertheseconditions
leaf), and the increasein winterwith snow,whichare shownin
Figures 7-12. Different factorsmay indeed be responsibleat
snowon the upwardfacingpyranometer,and we have rejected
most of these measurements.
The data showthat with snowlyingunder the canopy,forest
is thought to have greater penetration into the canopyand albedo increaseson cloudy days. One possiblereason is that
therefore might be absorbedmore efficiently, thus reducing
albedo. However, the different understoriesmay also be im0.5-portant. At the jack pine sites,the lichen understoryis whiter
o
Snow
and more reflectivewhen dry than wet, so perhapsthe associx Transition
ß
No Snow
ation of drier conditionswith daysof direct sunlightand rainfall and moister conditionswith cloudy daysmay be playing a
0.4role. However, the grass site at Saskatoonshowsa similar
different sites. At the snow-free conifer sites, diffuse radiation
o
decrease
of albedo
with(/5 $/• $ ). It ispossible
thattheleaf
physicaland opticalcharacteristics
changewith direct sunlight
and drier
conditions.
In winter
the increase
of albedo
with
_
o
o
0.:3 ..................................................................................................
o
(D $/• $ ) isseenatallsites(although
marginal
atFlinFlon),
and several explanationsare possible.Snow has a higher albedo to diffuse light, becauseof its different spectralproperties [Wiscombeand Warren,1980; Warren,1982]. In comparison to the direct beam, the spectrumof diffuse radiation is
shifted toward the visible (becauseof preferential absorption
of the near infrared in clouds),where the albedo of snowis
higher. Fresh snowhas a higher albedo than aged snowand is
likely to be associatedpreferentiallywith conditionsof diffuse
rather than direct solar radiation. For the forest sites, the
o
o
o
0.2-
o
o
0
o
0
0
o
o _o
_ ...
G
0
0
0 CD 0 0
•Oo.._o
oo o
-"'•,';:-
.
-
0.0
0.0
0.2
0.4
0.6
0.8
1.0
rather high solar zenith angle in winter makes the canopy
D,,I,/S,J,,
shadingeffect large, as mentioned earlier. However, the diffuse radiation has a smaller effective zenith angle and again Figure 12. As Figure 11 for spruce/poplarsite at Flin Flon.
will penetrate the canopy more effectively and therefore be The points with albedo >0.3 were not included in the snow
reflected more off the underlying snow surface. Clearly, the regressionline calculation.
BETTS AND BALL: ALBEDO
snow has a higher albedo for diffuse radiation, becauseof its
different spectralproperties,anotherfor the forest sitesis that
the canopyshadingis more important for the direct beam in
winter, becausethe solar zenith angle is so high, higher than
the effectivezenith angle for diffuseradiation. In summerthe
data showa small decreaseof albedo at both the grassand the
coniferous sites with increasingratio of diffuse to incoming
solar radiation.
Acknowledgments. This researchwas supportedby NASA GSFC
under
contract
NAS5-32356
and the National
Science
Foundation
under grant ATM95-05018. The mesonetdata were collectedby the
Saskatchewan
Research
Council
under
contract
to NASA.
We
are
grateful to John Pomeroy, Bert Davies, Betsy Middleton, and the
reviewersfor helpful comments.
References
Barnett, T. P., L. Dumenil, U. Schlese, E. Roeckner, and M. Latif, The
effect of Eurasian snowcover on regional and global climate variations, J./ttmos. Sci., 46, 661-685, 1989.
Betts, A. K., J. H. Ball, A. C. M. Beljaars,M. J. Miller, and P. Viterbo,
The land-surface-atmosphere
interaction:A review basedon observational and global modeling perspectives,J. Geophys.Res., 101,
7251-7268, 1996.
Bonan, G. B., D. Pollard, and S. L. Thompson,Effects of boreal forest
vegetationon global climate, Nature, 359, 716-718, 1992.
Bonan, G. B., F. S. Chapin III, and S. L. Thompson,Boreal forest and
tundra ecosystemsas components of the climate system, Clim.
Change,29, 145-167, 1995.
OVER THE BOREAL FOREST
28,909
atmosphericcirculationsin the tropics,J. Geophys.Res., 101, 42554268, 1996.
Federer, C. A., Spatial variation of net radiation, albedo and surface
temperature of forests,J. Appl. Meteor., 7, 789-795, 1968.
Federer, C. A., Solar radiation absorptionby leaflesshardwoodforests,
Agric. Meteorol., 9, 3-20, 1971.
Loth, B., H-F. Graf, and J. M. Oberhuber, Snow cover model for
globalclimate simulations,J. Geophys.Res.,98, 10,451-10,464, 1993.
McFadden, J. D., and R. A. Ragotzkie, Climatologicalsignificanceof
albedo in central Canada, J. Geophys.Res., 72, 1135-1143, 1967.
Miyakoda,K. and J. Sirutis,Manual of the E-physics,1986.(Available
from Geophy. Fluid Dyn. Lab., P.O. Box 308, Princeton Univ.,
Princeton,N.J. 08542).
Robinson, D. A., and G. Kukla, Albedo of a dissipatingsnow cover,
J. Clim. Appl. Meteorol.,23, 1626-1634, 1984.
Robinson,D. A., and G. Kukla, Maximum surfacealbedoof seasonally
snow-coveredlands in the Northern Hemisphere, J. Clim. Appl.
Meteorol., 24, 402-411, 1985.
Shewchuk, S. R., The surface atmospheric sciences mesonet for
BOREAS, J. Geophys.Res., this issue.
Thomas, G., and P. R. Rowntree, The boreal forests and climate,
Q. J. R. Meteorol. Soc., 118, 469-497, 1992.
Otterman, J., M.-D. Chou, and A. Arking, Effectsof nontropicalforest
cover on climate, J. Clim. Appl. Meteorol.,23, 762-767, 1984.
Pomeroy, J. W. and K. Dion, Winter snow radiation extinction and
reflectionfrom a pine canopy:Influenceof interceptedsnow.Paper
presentedat the 53rd Annual Eastern Snow Conference,Williamsburg, Va., May 1996.
Warren, S. G., Optical propertiesof pure snow,Rev. Geophys.,20(1),
67-89, 1982.
Wiscombe,W. J., and S. G. Warren, A model for the spectralalbedo
of snow, I, Pure snow, J. Atmos. Sci., 37, 2712-2733, 1980.
Cohen, J. and D. Rind, The effect of snow cover on climate, J. Clim.,
7, 698-706, 1991.
A. K. Betts and J. H. Ball, AtmosphericResearch,RR 3, Box 3125,
Pittsford,VT 05763 (e-mail: [email protected])
EuropeanCentre for Medium-RangeWeather Forecasts(ECMWF),
ResearchManual 3, ECMWF ForecastModel PhysicalParameterization, 3rd ed., Reading, England, 1991.
Eltahir, E. A. B., Role of vegetation in sustaining the large-scale
(ReceivedJuly 15, 1996;revisedSeptember27, 1996;
acceptedDecember ll, 1996.)