Heat and drought waves in continental midlatitudes
Fabio D’Andrea
With thanks to: Ramdane Alkama, Mara Baudena, Sandrine Bony, Christophe Cassou, Nathalie deNoblet-Ducoudret, Francesco d'Ovidio, Jean-Philippe Duvel, Kerry Emanuel, Masa Kageyama, FrancescoLaio, David Neelin, Amilcare Porporato, Antonello Provenzale, Joe Tribbia, Robert Vautard, Nicolas ViovyPascal Yiou, Matteo Zampieri.
Valsaverenche, June 2008
Outline
Part 1. The “real” world1.1 Climatology of a heatwave. Summer 2003.1.2 Impacts, climate change.1.3 Synoptic aspects, internal dynamics and external forcing.1.4 The importance of soil moisture and vegetation. Feedbacks.
Part 2. An idealized world2.1 A simple model of soil moisture - pbl interaction.2.2 Feedbacks and multiple states2.3 Effect of vegetation dynamics
PART 1
a, JJA temperature anomaly with respect to the1961−90 mean. Colour shading shows temperatureanomaly (°C), bold contours display anomaliesnormalized by the 30-yr standard deviation. b−e,Distribution of Swiss monthly and seasonal summertemperatures for 1864−2003. The fitted gaussiandistribution is indicated in green. The values in thelower left corner of each panel list the standarddeviation and the 2003 anomaly normalized by the1864−2000 standard deviation (T‘’/sigma).
An extreme event : the2003 heatwave in Europe(Schär et al 2004, Nature)
Increased mortality
France
England
Holland
© European Communities, 1995-2006
http://ec.europa.eu/healt/ph_information/dissemination/unexpected/unexpected_1_en.htm
Forest fires in PortugalAugust 3 2003.
Summer 2003 was the worstin 23 years for forest fires.5.6% of forest area was lost.
Decrease of agricultural production
Frenchfries
a, b, Statistical distribution of summertemperatures at a grid point in northernSwitzerland for CTRL and SCEN,respectively. c, Associated temperaturechange (SCEN\u2212CTRL, °C). d,Change in variability expressed asrelative change in standard deviation ofJJA means ((SCEN\u2212CTRL)/CTRL,%).
Schär et al 2004 Nature
Climate ChangeVariability change
Synoptic aspect 2003
Prevailinganticyclonicconditions.
(Black and Sutton 2004)
Anomalous SST
“Horseshoe” pattern
Hot Indian andtropical Atlanticocean
Hot Mediterranean
The effect of global and mediterranean SST (Feudale and Shukla 2007, GRL)
Z500 T
Obs.
Glob.
Med.
Twin GCM simulations forcedwith global and mediterraneanobserved SST.
FIG. 1. (a)–(d) Summer Z500 weather regimes computedover the North Atlantic–European sector from 1950 to 2003.Contour interval is 15 m. (e)–(h) Relative changes (%) in thefrequency of extreme warm days for each individual regime.Color interval is 25% from −100% to 200%, and red abovethat (max equal to 233%). As an example, 100%corresponds here to the multiplication by 2 of the likelihoodfor extreme warm days to occur.
Summer Weather RegimesAnd their associated Temperatureanomalies.
(Cassou et al 2005, J. Clim.)
Percentage of occurrence for each regime for the five warmest summers in France since 1950. The selection of the year issimply based on the five highest anomalous temperatures averaged over the 91 weather stations for JJA (anomalousvalues given in parentheses with the year). Sums for warm regimes (Blocking + Atl.Low) and cold regimes (NAO– +Atl.Ridge) are provided in the fourth and last columns, respectively.
Anticyclonic regimes have a higher frequency of occurrence in the heat-wavesummers.
The problem of predicting a heatwave can therefore be linked to the problem ofpredicting weather regimes frequency.
The main regime dynamics is internal (many authors, e.g. Micheal Ghil)
But also, there are external forcings that influence their:probability. Of these we can mention:
Black and Sutton 2004
1. The effect of SST
2. The effect of tropical forcing
FIG. 2. (a)–(c) Observed OLR anomalies.Anomalies (the base period for climatologyis 1979–95) are computed for the threesummer months taken individually. Greenish(orange) colors correspond to enhancedconvective activity or wetter (drier)conditions. The blue box shows the tropicaldomain where the diabatic heatingperturbations are estimated from OLRanomalies and further imposed in modelexperiments as detailed subsequently.Contour interval is 4 W m−2. (d) Relativechange (%) of occurrence of the fourregimes due to the prescribed tropicalforcing in the atmospheric model. We testedthat the relative changes are not dependenton the choice of the 40-yr period for thecontrol simulation. As a final estimate of thesignificance, the error bars indicate therange of uncertainty due to internalatmospheric variability as given by onestandard deviation of the within-ensemblevariability (Farrara et al. 2000 ).
2. The effect of tropical forcing
Increased convective activity in the tropical Atlanticregion increases the frequency of anticyclonic regimes.
(Cassou et al 2005 J. Clim)
But the change in frequency of the regimes may not betelling the whole story.
n
N
n
n Xfx !=
=1
:!
X 1
!
X 2
!
X 3
!
X 4
The « climate », (i.e. the mean) of a givenvariable can be seen as the result ofsuccession of weather regimes.
If are the composites of the variable, linkedto a given regime, The mean climate can bedecomposed like this:
nX
Example, temperaturein France:
!
ˆ x " x = ( ˆ f i " f i)X ii=1
k
# + ˆ f i(ˆ X i " X i)
i=1
k
#
An anomaly ca be decomposed into change in weather regime frequency:
!
ˆ x = ( ˆ f nn=1
N
" # fn )X n + R
If a change of regime frequency is the only cause of the anomaly, the residual iszero. The complete decomposition is:
Decomposition of an anomaly
Is this decomposition good in the case of the Heatwave?
Figure 1: a) Summertime (JJA)anomaly composites (°C) over thehottest 10 summers in the 1948-2004 period; b) same as a) but forJJA precipitation frequency (%).Stations where the anomaly issignificant at the 90% level aremarked by a black bullet;
We will now check the anomaly decomposition on a 57 year dataset of stationmeasurement in Europe for temperature and precipitation frequency.
We select the 10 hottest years:
Vautard et al 2007. GRL
Hot summers:1950195219591964197619831992199419952003
c) Average anomalies, over the hottest 10 years, of maximal dailytemperatures that would obtain by a sole change in thefrequencies of summertime weather regimes [Cassou et al.,2005], as a function of the corresponding actual averageanomalies. Each point corresponds to a station. See the methodssection for details of calculation; d) Same as c) for precipitationfrequencies.
Reconstructed anomalies ofTmax and precipitationfrequency.
The reconstruction is always toocold and too wet.
Temperature
Precipitation
!
ˆ x =
(ˆ f n
n=
1
N "#
f n)X
n+
R
There is another mechanism that must be responsible for the strong temperatureand precipitation anomalies. A good candidate: SOIL MOISTURE.
Heatwaves have been observed to be preceeded by precipitation deficit (Huangand Van den Dool 1992 J. Clim.
Figure 1: a) Summertime(JJA) anomaly composites(°C) over the hottest 10summers in the 1948-2004period; b) same as a) but forJJA precipitation frequency(%). Stations where theanomaly is significant at the90% level are marked by ablack bullet;
precipitation deficit
Dry soil
Less evapotranspiration
Less latent Cooling.Less clouds
Hot soil
More sensibleHeat flux
Hot PBL
Soil moisture controls the energy balance between the earth surface and theatmosphere by modulating sensible and latent (evaporation) heat fluxes.
Response of the UKMO GCM toprescribed values of soil moisture(Wet – Dry)
Rowntree and Bolton (1983)
TemperatureEvaporation Precipitation
An old example
precipitation deficit
Dry soil
Less evapotranspiration
Less latent Cooling.Less clouds
Hot soil
More sensibleHeat flux
Hot PBL
Soil moisture controls the energy balance between the earth surface and theatmosphere by modulating sensible and latent (evaporation) heat fluxes.
precipitation deficit
Dry soil
Less evapotranspiration
Less latent Cooling.Less clouds
Hot soil
More sensibleHeat flux
Hot PBL
Soil moisture controls the energy balance between the earth surface and theatmosphere by modulating sensible and latent (evaporation) heat fluxes.
The Soil Moisture - Precipitation feedbackHigh amplitude temperature and moisture variationare maintained by a feedback between soil moistureand precipitation (Schär et al 1999).
A more recentexample from thegroup of ChristopheSchär in Zurich
Fischer et al 2007 -GRL
Fischer et al 2005
Regional GCM integrations of the year 2003, with prescribred soil moistureanomalies
Control integrations Perturbed integrations
Fischer et al 2005
Time evolution integrated over France
Mid-way summary
Two ingredients for a good Heatwave:1. A persistent anticyclonic regime2. A dry soil moisture anomaly
Water balance on a layer of soil
!
"Qs
"t= P # E # R # L
In a hot summer day, a maturetree can evaporate 2000 liters ofwater!
…water for one week…
At this point, we really need to study the dynamics of soil water, and its interactionwith the atmosphere. This means studying convection, water fluxes, the physics ofevaporation, the physiology of plants.
To be continued tomorrow…..
PART 2
A simple model of the interactionof the land surface and theatmospheric planetary boundarylayer
1. Box model 1 - the soilmoisture - precipitationfeedback
2. Box model 2 - the effect ofvegetation dynamics
!
"Cpha#$a#t
= Qs +%a%s&TS4 ' "Cpha
#
#t($a +
1
)($a
* '$a )
!
"ha
#qa
#t= E $ "ha
#
#t% ˜ q a + Fq
!
"wCpshs#Ts#t
= Frad $Qs $ LeE $%s&Ts4
!
"#whs$qs$t
= P '%E % L(qs )
Free troposphere
!
"e = "ae
Leqa
c p" a
Box model description (D’Andrea et al 2006 GRL)
Planetary Boundary Layer
Soil
!
"e
#
E P
Qs
1000 m
0.5 m
!
Fq
Leakage
divergence
Evaporation Rate
!
E = 0 forqs " qh
E =qs # qhqw # qh
Ew forqh " qs " qw
E = Ew +qs # qwq$ # qw
EMax # Ew( ) forqw " qs " q$
E = EMax forqs % q$
&
'
( ( (
)
( ( (
!"
#$%
&'=(
)(=
=
=
=
=
)
)
*
d
mm
sm
KgE
q
qqEE
sm
KgE
q
q
q
Max
Sat
aSatMaxMax
w
w
h
2.5106
;
1006.1
EfficiencyPlant Maximum56.0
point Wilting18.0
point cHygroscopi14.0
2
5
2
6
hq wq !
q
MaxE
wE
sq
E
(Laio et al 2001)
Leakage
Sensible heat flux
!
Qs = "CpCD u (Ts #$a )
Precipitation Rate and Runoff (1)
!
P = f1
dt" ˜ q S
!
L(qs)=Ks
ek(q
s"q fc )
"1
ek(1 "q
fc)
"1
(Laio et al 2001)
!
P '= P forP "q0# qsdt
P '=q0# qsdt
forP fq0# qsdt
$
%
& &
'
& &
!
qfc = 0.56
Convection
!
if "e #"e$% 0
!
" ˜ # a =#e $#e
%
1+Le
cp
qRe l&qs
(cooling)
" ˜ q a = qRe l&qs"
˜ # a (drying)
'
(
) )
*
) )
!
"qsat =#qSat#Ta
and qSat = q0e$0.622
Le
c p
1
Ta$
1
273.15
%
& '
(
) *
qRe l =
qa
qSat
Where:
The above formulas for cooling and drying are obtained using the reduction of enthalpy and conserving relative humidity:
humidity) relative ofon conservatiorder (first ~~
variation)(enthalpy ~~
Re aasl
eea
p
ea
qqq
qc
L
!=!
"=!+!#
$%
$$$
!
"e = "ae
Leqa
c p" a
Reminder:
!
"e
!
"e
#
Precipitation Rate and Runoff (2)
!
P = f1
dt" ˜ q a
!
P '= P forP "1# qsdt
P '=1# qsdt
forP >1# qsdt
$
%
& &
'
& &
Not all the water that is liftedin the free troposphere byconvection precipitates.
The fraction of precipitatedwater depends on theintensity of convection.
f is 0.9 for strong convectionand 0.25 for weak convection
!
"ha
# ˜ q a
dt(mm /day)
Equilibrium statesAs a function of initial soil moisture condition
Some parameter values:
!
Frad = 380W
m2
Solar radiation
" = 0.8 Soil emissivity
Fq =1.10#8 Kg
Kgs Lateral flux of humidity
$e
%= 300 &
K Constant equivalent
temperature of the free
atmosphere
Hysteresis cycle obtained by varying Fq
Air temperature
Soil moisture
Dependence on the large scale flux convergence
Stochastically forced integration
Fq is perturbed stochastically every 5days. Values varying between -1 and 3mm/day. Flat distribution
Cool and moistregime
Dry and warmregime
Observed data
D’Odorico and Porporato 2003, PNAS
Free troposphere
Box model description
Planetary Boundary Layer
Soil
!
"e
#
E0 P
Qs
1000 m
0.5 m
!
Fq
Leakage
divergence
b
ET
qsTs
θaqa
Baudena et al., 2007, in preparation
(Baudena et al 2008, WRR subm)
Free troposphere
Box model description
Soil
!
"e
#
E0 P
Qs
1000 m
0.5 m
!
Fq
Leakage
divergence
ET
!
"Cphad#adt
=Qs +$a$s%TS4 & "Cpha
d
dt'#a +
1
((#a* &#a )
!
"ha
dqa
dt= E # "ha
d
dt$ ˜ q a + Fq
!
"wCpshsdTs
dt= (1-#)Frad $Qs $ LeE $%s&Ts
4
!
"#whsdqs
dt= P'$E $ L
!
db
dt= gb(1" b) "µb
Baudena et al., 2007, in preparation Baudena et al.,
(Baudena et al 2008, WRR subm)
Implicit space logistic equation forvegetation dynamics
b fraction of sites occupied byvegetation
!
db
dt= g(s)b(1- b) - µ(s)b
soil moisture
µ local extinction(mortality) rate
g colonizationrate
Baudena et al., Advances in Water Resources, 30 1320-1328 (2007)
Albedo
!
" = b"b
+ (1# b)"0
As in Charney [1975]!
"0
= 0.35
!
"b
= 0.14
EE0 ET
Evapotranspiration
Baudena et al., 2008
!
E = (1" b)E0(qs) + bET(qs)
2 X 2 experiments
• lateral humidity flux Fq:– constant– stochastic (representing synoptic flow
variability)• vegetation:
– “natural” (ability of colonizing new space)– “cultivated” (higher sensibility to drought,
re-planted every season)
Wet/cool
Dry/hot
Wet/cool
Dry/hot
Initi
al c
ondi
tions
of v
eget
atio
n co
ver b
Initial conditions of qs
natural
cultivated
Baudena et al., 2007, in preparation
Costant Fq,
Stochastic Fq, “natural” vegetation
Baudena et al., 2008, WRR. Submitted
Stochastic Fq, “cultivated” vegetation
Baudena et al., 2008, WRR., submitted
Conclusions
Two criteria are necessary for the occurrence of a heat and drought wave:1. The establishment and presistence of an anticyclonic regime.2. A condition of dry soil.When both criteria are met, the soil moisture - precipitation feedback can maintain
large amplitudes of anomaly.
The initial content in soil water (and vegetation state) at the beginning of the summeris a crucial parameter. It is more likely that one summer is either in one or theother state and stay there, rather than a transition occur during the season(counterexample: summer 2006).
Hypothesis: the dependence of precipitation efficiency on convection intensity can beat the base of the soil moisture - precipitation feedback.
“Natural” vegetation increases the probability of the system of being in the wet/coolstate as compared to “cultivated”.
Fig.: The visualization displays TERRA MODIS (MODerate resolution Imaging Spectroradiometer) derived land surface temperaturedata of 1km spatial resolution. The difference in land surface temperature is calculated by subtracting the average of all cloud freedata during 2000, 2001, 2002 and 2004 from the ones in measured in 2003, covering the date range of July 20 - August 20. (Cite thisimage as: Image by Reto St-böckli, Robert Simmon and David Herring, NASA Earth Observatory, based on data from the MODISland team. Correspondance: [email protected]).
Thank you
Northward propagation of drought(Vautard et al 2007, GRL)
Figure: a) Average rainfall frequencyanomaly (% of days) (January–April) forthe 10 years containing the hottestsummers listed in Table 1 of theAuxiliary Material. The color scale is asin Figure 1b. The figures that wouldobtain for the winter only (January-march, not shown) or from January toMay are rather similar to this one; b)Same as a) for averages of the month ofMay only; c) Same as a) for June; d)Difference between the early summer(June and July) maximal hot-summertemperature anomalies when southerlywind occurs at the station and thatobtained for northerly wind. At eachstation, days are classified into 2 classesaccording to the sign of the mean dailysurface meridional wind field. The 58-year average maximal temperature iscalculated and subtracted from the hot-summer average to obtain meananomalies. Then the difference between“southerly” and “northerly” meananomalies is shown in panel d). Thefigure shows that temperature anomalies,in the 10 hottest summers, are higher insoutherly wind conditions than innortherly wind conditions. e) As in d) forrainfall frequency (% of days). Stationswhere the differences are significant atthe 90% level are marked with a blackbullet.
Figure 3: Detrended summertime (JJA) daily maximum temperatureanomalies, averaged over European stations, as a function of year(in black), together with the detrended anomaly of precipitationfrequency averaged in the 42°N-46°N latitude band during precedingwinter and early spring (January to May), in red. Temperatureanomalies are in °C while precipitation frequencies anomalies are in%. In order to assess the sensitivity to the chosen latitude band forprecipitation frequency, 2° northward and southward shifts areapplied and results are also represented. Yellow bars indicate theyears selected in the hottest 10 summer years.
A predictor in themediterranean region?
Spring precipitation at different band oflatitude and summer temperature
Data from weather stations + awater budget model. Wilmottand Matsuura 2001.
Results from 1-d model
Equilibrium
Results from 1-d model
DAY 1
Results from 1-d model
DAY 30
Results from 1-d model
DAY 60
Results from 1-d model
DAY 90
Conclusions
Differential E-P in a meridionally well mixed atmosphere can explain the northwardpropagation of the drought anomaly. (Still very sensitive to the model parameters.It depends strongly on the stability properties of the two regimes.)
Years 1950 1952 1959 1964 1976 1983 1992 1994 1995 2003
Summer T’ 1.33 1.14 1.29 0.52 0.57 0.75 0.66 1.04 0.65 2.44
Summer P’ 9 -23 -20 -18 -55 -54 +16 -29 -30 -58
Summer F’ +1.1 -1.0 -4.3 -2.1 -7.2 -8.1 +0.5 -4.9 -3.7 -9.2
14 d HW Start 02/06 28/06 05/07 14/07 22/06 06/07 29/07 24/07 08/07 01/08
14-day HW T’ 2.91 3.77 2.79 2.65 4.07 2.83 1.93 3.38 2.44 4.19
Test of the drought propagation by mesoscale modeling(credit: Matteo Zampieri, LMD)
MM5 Integration details (credit: Matteo Zampieri, LMD)
Simulations domain: most of Europe (excluding the Northern part of theScandinavian peninsula) and a large part of the Atlantic Ocean.
Forced by the ECMWF (ERA40) or NCAR-NCEP reanalyses with a 6-hour rate.Resolution: 36 km at the centre of the domain. 85x125 grid points and 23 vertical
levels.MM5 3.7.3 model versionReisner microphysics,Kain-Fritsch convection scheme,MRF PBL schemeCCM2 radiation scheme.NOAH land-surface 4-layer scheme.
twin simulation, starting on 1 June, for all the 10 hottest summers, differing onlyby their initial soil moisture south of 46N:
1) 'wet' simulation. Initial volumetric soil moisture content of 30% .2) 'dry' simulation Initial volumetric soil moisture content of 15%.North of 46N initial soil moisture content is taken from the ERA40 ECMWF
reanalysis data, when possible, or the NCAR-NCEP reanalysis data.
twin simulation of MM5, starting on 1 June, forall the 10 hottest summers on record,differing only by their initial soil moisturesouth of 46N:
1) 'wet' simulation. Initial volumetric soilmoisture content of 30% .
2) 'dry' simulation Initial volumetric soilmoisture content of 15%.
dry soil
heat wave
European heatwaves are preceded by drought in the mediterraneanregion: analysis of the physical processes involved by regional modelling.
Propagation of temperature Anomaly (dry-wet runs):
Evolution of incomig solar radiation (left, red line)and of the bowen ratio (right, red line), dry-wetrun.
Temperature
July Dry - Wet July - June, Dry - Wet
(80% run)
Precipitation
July Dry - Wet July - June, Dry - Wet
Soil moisture short wave radiation
Sensible heat flux Latent heat flux
Mid-way summary
Heat propagates northward from the mediterraneanregion:
1. By propagation of drought (over France)2. By propagation of anomalous coudiness (over
Germany)