Download Contrasting responses of mean and extreme snowfall to climate

Response of snowfall extremes to climate
change: theory and simulations
Paul O’Gorman, MIT
Workshop on Water in the Climate System, 2014
Simulations of climate warming: declines in
annual-mean snowfall in many regions
709
710
711
712
713
714
715
716
Figure 2. Multi-model ensemble trends in snowfall (2006 to 2100, units cm year-1 decade-1).
Annual (A), SON (B), DJF (C), and MAM (D). Contours of 2-m temperature at intervals of 10
ºC are also shown from the multi-model ensemble for the period 1986-2005. Hatching denotes
regions of statistically significant trends (p <= 0.01).
CMIP5 multimodel mean trends in
snowfall depth
(2006-2100 based on rcp4.5)
Contours of 2m temperature shown
in degrees Celsius for 1986-2005
35
Krasting et al, J. Climate, 2013
What about snowfall extremes?
(heavy daily snowfall events)
•
Important because of disruption of transportation (roads, air, rail),
business, schools
•
May not respond to climate change like mean snowfall
e.g., heavy snowfall events in both anomalously cold and warm years (Kunkel et al, 2013;
Changnon et al 2006)
1928
Regional studies of observed snowfall extremes:
Decadal variability but inconsistent long-term trends
JOURNAL OF CLIMATE
VOLUME 14
FIG. 6. Nationally averaged 20-yr return values (relative to the values for 1971–90) of annual maximum daily
precipitation, rainfall, and snowfall. The 20-yr return values are first estimated using 20-yr running windows for every
station, and then normalized by the values estimated for the period 1971–90. Values are plotted in the center of the
20-yr window.
Zhang et al, J. Climate, 2001 (Canadian observations)
ally fewer events after the mid-1970s during both spring
ing. The observed 18C warming trend over Canada
Effect of climate change on daily snowfall
extremes in global simulations
•
High percentiles of daily snowfall in liquid water equivalent
•
CMIP5 (use 20 models) under RCP8.5
•
Compare warm climate (2081-2100) to control climate (1981-2000)
Ratio
0.5
Analyze according Extremes
to climatological temperature in
Mean
control climate
0
Ratio (warm/control)
1.5
1
0.5
0
−20
−15
−10
−5
0
5
10
Climatological temperature (°C) in control climate
Grid boxes and days binned by climatological monthly
surface air temperature in control climate
99.99th
99.99th percentile
(mm day−1)
40
30
20
Observations
Simulations (control)
10
Simulated snowfall
extremes compare
well with estimates
from observations
0
99.9th percentile
(mm day−1)
40
30
99.9th
20
10
0
99th percentile
(mm day−1)
40
99th
30
20
Northern hemisphere land only
10
Observational estimates based on daily
precipitation rates (GPCP 1DD), snowfall
fraction (Feiccabrino et al 2013), and daily
surface temperatures (NCEP2)
0
Mean
(mm day−1)
Mean
1
0.5
0
−20
−15
−10
−5
0
5
Climatological temperature (°C) in control climate
10
Ratio
0.5
Extremes
Response of mean
snowfall to climate warming:
Mean
ratio
of warm over control-climate values
0
Ratio (warm/control)
1.5
1
0.5
0
−20
−15
−10
−5
0
5
10
Climatological temperature (°C) in control climate
(Northern Hemisphere land below 500m)
Rati
0.5
Extremes
Mean
Weaker response of daily snowfall extremes as
compared to mean snowfall
0
Ratio (warm/control)
(warm/control)
Ratio
(b)
(a) 1.5
1.5
11
Increasing
Percentile
0.5
0.5
99.99th
Extremes
Theory
Mean
Theory Simple
00
/control)
(b)
−20
1.5
1
−15
−10
−5
0
5
10
Climatological temperature (°C) in control climate
99th, 99.9th and 99.99th percentiles of daily snowfall
Features of the response of snowfall extremes that
would like to understand:
•
Climatological temperature at which snowfall extremes response
goes from positive to negative
•
Weaker fractional changes at higher percentiles
Simple theory (based on known physics/
observations) for the response of snowfall
extremes to changes in climate
ds on the vertical temperature
ple theory is next developed that accounts for the main features of the resp
Theory assumptions 1:
22
s
to
climate
change.
Surface
precipitation
on the
Relate
daily
snowfall
rate
(s)
to
(p)vertica
d to surface air temperature precipitation
.type dependsrate
and surface air temperature (T)
e lower troposphere23 , but to first order it may be related to surface air t
pitation rate p by s = f (T )p,
nowfall rate s in the theory is related to the daily precipitation rate p b
the daily
air
temperature,
andfalls
f (T ) is the fraction of precipita
Snowfall
fraction:
tion
ofsurface
precipitation
that
100
Snowfall fraction (%)
Control
snowfall fraction) 80at a given temperature T . The
snowfall fraction in th
fall fraction in the simulations
Warm
60
p decline near freezing in the multimodel median (Fig. 2) and in most of
40
.)S2).
need not occur precisely at a surface t
andThis
in rain-snow
most oftransition
the individual
20
0
−9
−6
−3
0
3
6
9
Daily surface air temperature (°C)
Figure 2 do not immediately
e hydrometeors
change phase as they cross the mel
Fig. 2. Importantly for the theory, and as expected given mod
Theory
2:
ow
is almost
exactly
sametheory
in the control
and wa
ilytransition
precipitation
rateassumptions
inthethe
is assume
Relate precipitation rate (p) to temperature (T)
and simulated snowfall in Fig. 2. Importantly for the theory, and as expected gi
ure
according
to
p
=
e
p̂,
where
=
0.0
in lapse
rates,
the rain-snow
transition
almostaexactly
the dependence
same in the contro
n rate
in the
theory
is assumed
toishave
simple
on
T
p
(Fig. (T
2). Tc )
=e
p̂,
and simulated snowfall1in Fig. 2. Importantly for the the
where
= 0.06 C
is a representative ther
in lapse
rates,
rain-snow
transition
is almost
ase ofTheextratropical
extremes
wi
daily precipitation
rateprecipitation
in thethe
theory
is assumed
to have
a simpleexactly
depen
opical
extremes
respect= to
surface
(T with
Tc ) p̂, where
1 is a tempera
(Fig.
2).evariable
temperature
according
to p =
0.06
C velocity;
representa
Normalized
precipitation
behaves like
upward
• precipitation
follows gamma distribution on wet days
9 rate
daily
precipitation
inwith
the theory
rate of increase
of extratropical
precipitation
extremes
respectistoassumed
surface
specific
humidity
(9). The
The
normalized
precipitation
variabl
aturation specific humidity . The normalize
(T Tc ) p̂, where
temperature
according
to
p The
= e normalized
= 0.
Temperature
is normally
distributed
and(9).
independent
of
to •increases
in saturation
specific
humidity
precipitation
variable closely related to upward motion in the atmosphere; it
rate
of increase
ofrelated
extratropical
precipitation
thought of as a dynamic
variable
closely
to upward
motion inextremes
the atmosw
amic
variable
closely
related
to
upward
mo
bution
on
wet
days
with
scale
parameter
1/
and
shape
param
to increases
in saturation
humidity1/(9).and
The
n
to follow a gamma distribution
on wet
days withspecific
scale parameter
shap
With these assumptions, the qth percentile of snowfall, sq , is exceeded if
1 normalized precipitation rate p̂
here.) The
and) the
s that p̂ > hsq where
h(Ttemperature
) = e T fT (T
. Assuming sq is non-zero, t
Integral expression for
eded may be written as
T
th
p̂
e
f (T ) >of
sq ,snowfall
q percentile
(sq)
With these assumptions, the qth percentile of snowfall, sq , i
T
1
which requires thatZp̂ >
hs
where
h(T
)
=
e
f
(T
)
. Assuming sq is non-z
q
Z
1
1
q
w k k
1
=
dT
dp̂
p̂
that sq is exceeded
may
100
1 be written
hsq as (k)
(T T )2
1
1
e p̂ p̂pe T f (Te ) >2sq2, .
2⇡
Z 1
Z 1
1 T )2
k h(T ) = e T f (T ) (T
which the
requires
thatofp̂ wet
> h(T
)sis
. Ass
where
days
q fraction
w and
1
q where
k 1
p̂
p
1
=
dT
dp̂
p̂
e
e 22 .
100
(k)
2⇡
1
hsq
otic methods are
next
to evaluate
thewritten
double
bility
that used
sq is exceeded
may be
as integral in (S3) i
→ Evaluate using asymptotic methods
for large sq
Z
Z
1
1
k
q
w
of large
s
.
The
integral
in
p̂
is
first
evaluated
using
a
standard
1
q
1 used =
dT the double
dp̂
p̂kasympto
e inp̂ (p
Asymptotic
methods are next
to evaluate
integral
100
(k)
1
h(T )sq
S1
plete
gamma
function,
snowfall limit of large sq . The integral in p̂ is first evaluated using a standard asy
Z
methods
are next used to evaluate the double
S1
1 Asymptotic
for the incomplete gamma
function,
⇥
⇤
k 1
t
k 1
z
1
dt t
e
=z
e
1 + O(z ) ,
Asymptotics gives expression for snowfall extremes
of snowfall sqthat
is given
by
involves optimal temperature Tm
m
( sq hm )
3
2
k
e
sq hm
=
w
1
q
100
asymptotically for large sq , and where
(k)
s
hm
e
00
hm
(T
Tm )
2 2
2
,
is the gamma function, h
is the second derivative of h at Tm . For a change in mean tempe
igible changes in all other parameters, the change in snowfall extrem
expression
f(T) exp( T)
Temperature dependence of snowfall reaches a
maximum at Tm (roughly -2°C)
−6
−4
−2
0
2
Surface air temperature (°C)
Competition between increasing precipitation and
decreasing snowfall fraction with increasing temperature
✓
◆s
✓
◆
ble
in all gives
other
in
snowfall
ext
srearranging
Ts
T parameters,
k the
2 (Tchange
s
2
terms
q changes
q
Tm )
rearranging
terms
gives
w
h
=
T
+
T
+
log
1 + changes
.
(S
m
c
k
sq hmSimple
result
if
only
mean
temperature
2
2
2
2
sq ew ✓ = sq hm (Tq◆
2T2m )
s
e
, sq
q
✓
◆
00
✓
◆
e ◆2
,(k) 2 h✓m
(1)
1
q
T
T
k
s
00
q
100 k
ression
T100 T(k)
T
2
s
h
q
m
+is beingTctaken,
Tc ++(S11) log
log1 +1that
+ sq./sq. ! 0,(S11)
(S11)
T
+
of
s
!
1
implies
and
the sec
2
q
2 ssq
✓
◆
2
s
s
2
sq q
sq q
q
t hand side of (S11) may be neglected.
(TheTsecond term on the right h
T
oruldlarge
s
,
and
where
is
the
gamma
function,
h
qthe
m is h eval
s
=
T
+
T
,
where
is
gamma
function,
h
is
h
evaluated
q
m
m
not
betaken,
neglected
if implies
either
sq that
!s /s
sqq/s
or
sq /s
!
1,
but
neithe
2 hthat
q and
1isisbeing
being
taken,(S11)
(S11)
implies
s
!
0,
the
second
1
!
0,
and
the
second
q
2
q
q
m
with
(S11)
if k < 2.)
The
final
result
is that
eofof(S11)
(S11)may
maybe
beneglected.
neglected.
(The
second
term
on the
econsistent
(The
second
term
on
the
rightright
handhand
Trivative
change
inmmean
temperature
of ◆
T andtemperature of T
h
at
T
.
For
a
change
in
mean
m . For aof
✓
sqsq!! sq sor
1,1,
but but
neither
of of
eneglected
neglectedififeither
either
sqq/s!
neither
q orsq /s
q !
upplementary
Information.
T
T
th
with
final
result
is that
+
T
,
with(S11)
(S11)ififksk
<2.)
2.)The
The
final
result
iscsnowfall
that
Change
in
qT
percentile
of
with climate change (S
q<=
2
ers,other
the change
in snowfall
extremes,
2 inssnowfall
q , is givenextremes, s , is
all
parameters,
the
change
q
✓✓
◆◆
TT ChangeTinTmean temperature
me as
main
According to (S12),
the cha
ssq(2)
= in the
T T+body
(S12)
cTc,paper.
+2 ofTthe
,
(S12)
q =
22
o
the
approximation
(2),
changes
in
snowfall
extremes
transit
Variance
of
daily
temperature
2
es
is
independent
of
q,
,
f
,
and
k,
although
>
0
is
required
for
✓
◆
T
✓
◆
lt
(S8)
to
be
valid.
If
it
is
found
that
sq to
< (S12),
sq when
applying (S
inin
main
body
of
the
paper.
According
the
change
Tthe
+
T
,
(2)
Optimal
temperature
m
T
T
the
main
body
of
the
paper.
According
to0,(S12),
the
change
2
point
(S3)
is
invalid
because
it
assumes
s
>
and
we
must
instead
npendent
temperature
in
the
control
climate
of
T
T
/2.
Changes
i
s
+
T
and
k,
although
is
required
for
the
q = of 2q, , f , T
m , > q0 m
pendent of q,hm , fInverse
, andprecipitation
k,
although
scale > 0 is required for the
2
be valid. If it is found that sq < sq when applying (S12),
be valid.
If
it is found
that
sq < sq when applying
(S12),
2
tesimal,
the
slightly
simpler
result
) is invalid
because it assumes
sq >as0,measured
and we must
set mak
sely
on temperature
variability
by instead
, which
Ratio
0.5
Extremes
Theory
matches simulations
Mean
(and dynamic changes don’t matter very much)
0
Ratio (warm/control)
(b)
1.5
1
0.5
99.99th
Theory
Theory Simple
0
−20 −15 −10
−5
0
5
10
Climatological temperature (°C) in control climate
Shading shows interquartile range of model ratios
le changes in all other parameters, the change in snowfall extr
Changes in snowfall extremes don’t depend on
percentile q!
ession
✓
◆
T
T
sq =
T+
Tm ,
2 h
2
m
upplementary Information.
Example: same change for 99th percentile
as 99.99th percentile
of the percentile considered, such that the fractiona
o the approximation (2), changes in snowfall extremes transiti
percentiles ( sq /sq ! 0 as sq ! 1). This is the
temperature in the control climate of Tm
T /2. Changes in
dependencies of precipitation extremes and the rain-
Rati
0.5
Extremes
Mean
Fractional changes in snowfall extremes tend to zero
for high percentiles
0
Ratio (warm/control)
(warm/control)
Ratio
(b)
(a) 1.5
1.5
11
Increasing
Percentile
0.5
0.5
99.99th
Extremes
Theory
Mean
Theory Simple
00
/control)
(b)
−20
1.5
1
−15
−10
−5
0
5
10
Climatological temperature (°C) in control climate
99th to 99.99th percentiles of daily snowfall
Joint PDF
Intuition: Probability of optimal temperature (Tm) for
snowfall extremes does change as climate warms
← Snow
Rain →
Joint PDF
Control
Warm
T m 0◦ C
(b)
Tc
(b)
Joint PDF
(a)
T
Also mean snowfall decrease substantially
0
Control
Warm
s
Results in changes
Tm in
0◦ Csnowfall percentiles
T
(b)
Joint PDF
Control
Warm
0
sq sq
p̂/hm
Results in changes
Tm in
0◦ Csnowfall percentiles
T
(b)
Joint PDF
Control
Warm
0
sq sq
p̂/hm
...but fractional change in sq is fairly small and is
similar for all high percentiles
Ratio (warm/control)
(a)
1.5
1
Increasing
Percentile
0.5
Extremes
Mean
0
Ratio (warm/control)
(b)
1.5
Similar results above
500m elevation
(but models have
issues there)
1
0.5
99.99th
Theory
Theory Simple
0
−20 −15 −10
−5
0
5
10
Climatological temperature (°C) in control climate
cf. Kapnick & Delworth, 2013
mptotically for large sq , and where
Conclusions
is the gamma function,
he second derivative of h at Tm . For a change in mean temp
• Simulations: Smaller fractional changes in snowfall extremes
le changes
all other
parameters,
the change in snowfall extr
than ininmean
snowfall
in many cases
• Simple asymptotic theory: captures main features of response
ession
✓
◆
T
T
sq =
T
+
T
,
m
2 h
2
m
pplementary
Information.
Implications:
detection and perception of climate change,
•
changes in snowfall extremes still likely to have impacts
o the approximation (2), changes in snowfall extremes transiti
By contrast:
Probability of snowfall and mean snowfall decrease
substantially
Snow
←←
Snow
Rain
Rain
→→
Joint PDF
Joint
PDF
(a)
(a)
Tm 0◦ C
Big decrease in area under curve to left of
rain-snow transition temperature
(b)
(b)
Control
TT
19504
AND BROWN: SPRING SNOW COVER EXTENT REDUCTIONS
L19504
RapidlyDERKSEN
changing
snow cover in Northern Hemisphere
L19504
DERKSEN AND BROWN: SPRING SNOW COVER EXTENT REDUCTIONS
L19504
igure 2. Time
series
Northern
June snow
(NOAA
snow
chart
seaice
iceextent
extent
(NASA
TEAM)
Figure
2. of
Time
series ofHemisphere
Northern Hemisphere
June cover
snow cover
(NOAA
snow
chartCDR)
CDR)and
and sea
(NASA
TEAM)
or 1979–2012for(1979–2011
for sea ice).
denotes
5-yr running
mean.
1979–2012 (1979–2011
for Thick
sea ice).line
Thick
line denotes
5-yr running
mean.
andtoBrown,
2012
declines
are being
drivenbylargely
by pervasive
warming
rcp8.5
[Riahi
2011]Derksen
(2006 to
create
a GRL,
150-year
CE declines SCE
are being
driven
largely
pervasive
warming
rcp8.5
[Riahi
et et
al.,al.,2011]
(2006
to2100)
2100)
to
create
a 150-year
pan-Arctic temperatures (as described in Screen et al. [2012]), time series of model simulated and projected snow cover.
an-Arctic temperatures
(as described in Screen et al. [2012]), timeOne
series
of model simulated and projected snow cover.
independent of these low-frequency climate variables.
model run (typically the first member) was selected from
Probability of snowfall given that precipitation
occurs
3-hourly synoptic weather reports over land aggregated globally
Dai, GRL, 2008
Conditional probability of snowfall (%)
Rain-snow transition in climate models (CMIP5)
versus observations
100
Multimodel Median
Observations
80
60
40
20
0
−9
−6
−3
0
3
Surface air temperature (°C)
6
9
- 3-hourly observed (Dai 2008): 2 curves depending on whether mixed counted as snow
- Daily accumulations in multimodel median (black) and individual models (gray); snowfall
Conditional
probability of snowfall over land as in Fig. 2 but showing
taken to occur if precipitation is 50% solid
2.
the
l median only in the control climate (the black solid line) and showing the prob
r individual models in the control climate (gray lines).
Intuition: snowfall extremes occur when
temperatures close to freezing
Temperature (°C)
(otherwise too cold to snow heavily)
10
0
−10
−20
Snow Extremes (Control)
Snow Extremes (Warm)
Mean (Control)
Mean (Warm)
−20
−15
−10
−5
0
5
10
Climatological temperature (°C) in control climate
Note snowfall extremes remain at roughly same temperature (with same
humidities) as climate changes - unlike rainfall extremes
Quiz: World record daily snowfall
Where?
How much (inches of depth)?
World record (probably): 75.8 inches in 24 hours
MONTHLY WEATHER REVIEW
38
FEBBUABY
1963
RECORD SNOWFALL OF APRIL 14-15,1921, AT SILVER LAKE, COLORADO
J. L. H. PAULHUS
U. S. Weather Bureau, Washlngton. D. C.
[Manuscript received January27, 19531
ABSTRACT
A snowfall of 87 inches in 27% hours on April 14-15, 1921, was reported at Silver Lake, Colo. This snowfall, if
INTRODUCTION
Although the meteorologist and hydrologist are generally
interested in the water equivalentof a snowfall rather than
in the snow depth, there are many, including the general
public, highway and street maintenance engineers, etc.,
who are very much interested in the depth. Consequently,
Changes in precipitation extremes:
some aspects are well understood!
•
Theory
•
Climate models
•
Observed trends
matological mean temperatures; for example, at latitud
here precipitation is related to poleward movement of
Simple scaling captures changes in simulated
asses, they may be tied more closely to mean temperatu
precipitation
extremes
Theory:
factors
controlling
intensity of precipitation
rther equatorward, and mean temperatures
change differen
extremes
different latitudes in global
warming simulations (25).
Taking into account these factors, we can express the intens
Saturation specific humidity
precipitation extremes at a given latitude as
Precipitation rate
$ ! %
dq s
Pe # " ! e
dp
.
[
" *,T e
Vertical velocity
ere, Pe is a high percentile
of precipitation, !e the correspon
Moist adiabatic integral o
g upward vertical velocity, {!} is a mass-weighted
derivative
e troposphere, and the moist-adiabatic derivative of saturat
O’Gorman
and Schneider, PNAS,
2009 temp
ecific humidity is evaluated at the
conditional
mean
O’Gorman
and
Schneider,
2009
O’Gorman
and
Schneider,J. Climate,
PNAS, 2009
occurs.
A large-scale
av
ure Te when extreme precipitationO’Gorman
and Schneider,
J. Climate, 2009
e over precipitation systems is implied, so that !e is a
Growth of precipitation extremes
with warming in simulations with climate models
Fig. 1. The 99.9th percentile of daily precipitation (millimeters per day) for
CMIP3 multi-model mean 99.9th percentile of daily precipitation
the periods 1981–2000 (blue) and 2081–2100 (red) in the SRES A1B scenario
(20c3m
A1B)
and
observations
(GPCP)Project
(multimodel
median),toand
based
oncurrent
Global Precipitation
Climatology
(GPCP) data for the period 1997–2006 (black). Model
scatter
(shading)
the
O’Gorman
and
Schneider, for
PNAS,
2009
period 1981–2000 is shown using the interquartile range (50% of models lie
Fig. 2.
(blue), zo
water va
extreme
itation e
Problem:
Response
of tropical
precipitation
NATURE
GEOSCIENCE
DOI: 10.1038/NGEO1568
extremes varies widely between models
Climate change (% K¬1)
20
Inferred
10
Observed
0
0
20
40
Variability (% K¬1)
60
BCCR BCM2.0
CGCM3.1 T47
CGCM3.1 T63
CNRM¬CM3
CSIRO¬Mk3.0
CSIRO¬Mk3.5
GFDL¬CM2.0
GFDL¬CM2.1
FGOALS¬g1.0
ECHAM4/INGV
ECHAM5/MPI
INM¬CM3.0
IPSL¬CM4
MIROC3.2¬med
MIROC3.2¬hi
MRI¬CGCM2.32
NCAR¬PCM1
NCAR¬CCSM3.0
Figure 2 | Sensitivities (% K 1 ) of the 99.9th percentile of precipitation
Problem:
Response
of tropical
precipitation
NATURE
GEOSCIENCE
DOI: 10.1038/NGEO1568
extremes varies widely between models
Climate change (% K¬1)
20
Use observed
variability to constrain:
10±4%/K
Inferred
10
Observed
0
0
BCCR BCM2.0
CGCM3.1 T47
CGCM3.1 T63
CNRM¬CM3
CSIRO¬Mk3.0
CSIRO¬Mk3.5
GFDL¬CM2.0
GFDL¬CM2.1
FGOALS¬g1.0
ECHAM4/INGV
ECHAM5/MPI
INM¬CM3.0
(O’Gorman,
Nature
IPSL¬CM4
MIROC3.2¬med
MIROC3.2¬hi
MRI¬CGCM2.32
NCAR¬PCM1
NCAR¬CCSM3.0
20
40
Variability (% K¬1)
60
Geo. 2012)
(see also Muller, O’Gorman, Back 2011)
Figure 2 | Sensitivities (% K 1 ) of the 99.9th percentile of precipitation
Observed trends in annual-maximum
precipitation over land
FIG. 12.
Variation
estimated
sensitivity
of annual
maximum
to a 1 K
Sensitivity
(%)in
of the
annual
maximum
precipitation
per kelvin
warmingprecipitation
of global nearincreasesurface
in global
mean temperature
byrecords
latitude.
(top)
Thewith
number
stations
within
each 658
temperature
(1900-2009;
> 30
years),
lightof
blue
shading
indicating
latitudethe
band.
(middle)
The fraction
of stations exhibiting significant positive association, with
upper
97.5% confidence
bound
light blue shading indicating the upper 97.5% confidence bound and dark blue shading indicating the median of the confidence interval. (bottom) Sensitivity (%) of annual maximum
precipitation per kelvin warming of global near-surface temperature, with light blue shading
Westra et al, J. Climate, 2013
indicating the upper 97.5% confidence bound.
Intuition: snowfall extremes occur when
temperatures close to freezing
Temperature (°C)
(otherwise too cold to snow heavily)
10
0
−10
−20
Snow Extremes (Control)
Snow Extremes (Warm)
Mean (Control)
Mean (Warm)
−20
−15
−10
−5
0
5
10
Climatological temperature (°C) in control climate