JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 105, NO.

D14, PAGES 18,013-18,025, JULY 27, 2000

Impact of climate variability on the

vegetation water stress
LucaRidolfi,•,2P. D'Odorico,
3 A. Porporato,
•,2andI. Rodriguez-Iturbe
2

Abstract. An important parameterin the assessment of the impact that conditions

of limited soil water availabilityhave on vegetationis the averagelength of the
intervals during the growing season in which soil moisture is below some critical
levels. These levels are related to the plant tolerance to water stress. The
interannualrainfall variability inducesimportant fluctuationson the average
duration and frequencyof the periods of water stresswith important effects on
the spatial and temporal structure of plant ecosystems. It is shown that the
nonlinearitiesembeddedin the dynamicscontrollingthe soil water balancemay
drasticallyenhancethe effectsof the fluctuationspresentin the climatic forcing.
Interannual climate variability leads to strongeryear-to-year changeson the mean
duration and frequencyof periodsof soil water deficit as well as to the emergence
of preferential states in the probability distributions of these two variables. The
sensitivityof thesestatistical propertiesis studiedwith respectto the characteristics
of climate, soil, and vegetation.

1. Introduction through its impact on soil albedo and on the turbulence

in the low atmosphericboundarylayer [e.g., Brubaker
The links between vegetation and soil water content
and Entekhabi, 1996; D'Odorico and Rodriguez-Iturbe,
are numerous and complex, and they are fundamen-
2000]. Thus soilwater dynamicsresultsfrom a number
tal to understand both the behavior of vegetation and
of complex interactions among climate, soil pedology,
the dynamics of soil moisture. Indeed, water is one
and type of vegetation.
of the most important resourcesfor vegetation growth.
The interaction between vegetation and soil moisture
The soil water content affectsdirectly several vital pro-
becomesparticularly evident during the periods of plant
cessesin the plant physiology, such as transpiration,
water stress. These periods play a key role in the life of
photosynthesis and biomassproduction [e.g., Latchet, the plants and occur when the soil water content does
1995;Lainbetset al., 1998],while it indirectly controls not permit the normal courseof the plant physiological
other fundamental aspects such as nutrient adsorption
processesleading to a reduction in their rates. There
[Scholesand Walker,1993;Latchet, 1995]and soiltem- are many symptoms of water stresssuch as the decrease
perature [Lamberset al., 1998]. In situationsleadingto
in turgor and the slowing down of the growth process
water scarcity, soil moisture becomesa key factor in the
ILarchef,1995],but its main macroscopical signatureis
temporal evolution of the vegetation as well as in the
its effect on transpiration. When the plant is under wa-
competitive mechanismsregulating its spatial dynam-
ter stress, its rate of transpiration is reduced, becoming
ics [e.g.,Archer et al., 1988;, Archer, 1994; Rodriguez-
dependenton the availablesoil moisturecontent [Hale
Iturbe et al., 1999a,c]. Vegetation,on the other hand,
and Orchutt, 1987].
strongly affects the evolution of soil moisture through
A hydrological balance of soil water content that ac-
the processof transpiration [i.e., Rodriguez-Iturbeet counts for the influence of climate, soil, and vegetation
al., 1999b].Moreoverit alsoinfluencessoil evaporation is crucial for the quantitative characterization of water
stressin plants and, more generally, to understand how
vegetation dynamics is affected by the occurrenceof pe-
•Dipartimento di Idraulica, Trasporti e Infrastrutture Ci- riods of water scarcity. In a previouspaper [Ridolfi et
vili, Politecnico di Torino, Turin, Italy.
2Departmentof Civil and EnvironmentalEngineeringand
al., 2000] we studied how the duration and frequency
Center for Energy and Environmental Studies, Princeton of water stress depend on climate, soil, and vegetation.
University, Princeton, New Jersey. FollowingRodriguez-Iturbeet al. [1999b],soil moisture
3Department of Civil Engineering, Texas A&M Univer- at a site s was modeled as a random variable whose
sity, College Station. evolution is described by a stochastic differential equa-
tion expressingthe water balance. The rainfall input
Copyright 2000 by the American Geophysical Union.
is random and intermittent, while lossesdue to evapo-
Paper number 2000JD900206. transpiration and leakageare dependenton the state of
0148-0227/ 00/ 2000JD900206$09.00 soil moisture. Rodriguez-Iturbeet al. [1999b]obtained
18,013
 21562202d, 2000, D14, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2000JD900206 by Cochrane Colombia, Wiley Online Library on [01/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
18,014 RIDOLFI ET AL.: CLIMATE VARIABILITY AND VEGETATION

the analytical solution for the probabilistic description in the rainfall regimes among growing seasonsof dif-
of the soil moisture dynamics under steady-state condi- ferent years affect the duration and frequencyof the
tions. vegetation water stress. No considerationis made of
Vegetation water stresscan be characterizedby using the evolutionary dynamics of vegetation in this phase
two different thresholds in soil moisture, whose values of the analysis. This is a necessaryfirst step in the
depend both on the kind of soil and on the type of veg- quantificationof how the interannual variability of cli-
etation. Water stress is assumed to occur when the soil mate impacts the vegetation of water-limited ecosys-
water content is below a threshold value s*, and the ef- tems. Starting from the statistical characteristicsof
fects of the stress are directly related to the time spent rainfall regimesduring the growingseasonfor differ-
in this condition. The other important threshold is the ent stations in the United States, we will focus on the
so-called wilting point s•o, or soil moisture content un- impact of the interannual rainfall fluctuations on the
probabilisticstructure of the seasonalmean values of
der which, if prolongated in time, the vegetation wilts
[Hale and Orchutt, 1987]. In general,given a certain T• and n•. The final objectiveof this line of research
thresholdvalue •c of the soil moisture(with • _< s*), is the quantificationof the impact that natural or man-
two stochastic variables allow the quantification of the induced fluctuations in the rainfall regime have on the
waterstress(seeFigure1): the lengthT• of the time in- temporal and spatial vegetation dynamics,particularly
tervals in which soil moisture is below the threshold and in arid and semiarid ecosystem.
the numbern• of downcrossings
of the thresholdlevel
during a climatic homogeneousperiod, i.e. the growing 2. Modeling Scheme
season.Ridolfi et al., [2000]givethe analyticalexpres-
sions for the expected values of these variables and, in In the absence of strong topographical effects the
particular, show their great sensitivity to the charac- temporal evolution of the relative soil moisture s at
teristics of regional climate. The necessityof consider- site can be described through of the balance equation
ing climatic homogeneousconditions circumscribesthe
validity of the present results to caseswherein no con-
sistent seasonaltrends within the growing seasonsare nZ•ds(t•)
dt
= I(s,t)- E(s)
- L(s)
'
(1)
present. This is not too restrictive, at least for semiarid
ecosystemswith dry and warm winters which preclude
the existenceof any sizable storage of soil moisture at where t is the time, Z• is the soil depth, and n is the
the start of the growingseason[Scholesand Walker, soil porosity. The infiltration rate from rainfall I(s, t)
1993;Scholesand Archer, 1997]. representsthe water input into the system, while the
The presentpaper starts from the resultsof Ridolfi et evapotranspiration
rate E(s) and the leakagerate L(s)
al. [2000]and concentrateson the studyof howchanges representthe lossesfrom the soil. I(s,t) is a random

s(t)

durationof the growing season

o downcrossings of level

Duration
T•
of
the
excursion below 4

i Tseas

Figure 1. Temporal evolutionof soil moistureat a site showingthe two stochasticvariablesT•

and n• related to the crossingof a thresholdvalue •c.
 21562202d, 2000, D14, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2000JD900206 by Cochrane Colombia, Wiley Online Library on [01/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RIDOLFI ET AL' CLIMATE VARIABILITY AND VEGETATION 18,015

variable; consequently, the previous relationship is a monly used expressionK - Kss -c, whoseKs is equal
stochastic differential equation where s itself becomes to the saturated hydraulic conductivity, and the expo-
a random variable. nent c dependson the type of soil [Clapp and Horn-
Formally,equation(1) describes a continuous process, berger,1978;Dingroan,1994]. Sincec is generallyquite
but we interpret the evolution at the daily timescale. high (typical valuesare in the range 10-15), the effect
This scale captures those aspects of the soil moisture of leakageis important only for high values of soil mois-
dynamicswhich are useful to study water stressin veg- ture. Thus for the objectives of this paper, leakagemay
etation. At the same time, the interpretation of the be approximated through a linear relationship between
results at the daily scale helps to maintain the problem (sl,Emax) and (1,Emax + Ks), where s• representsa
analytically tractable, avoiding consideration of those value below which L(s) is negligiblein comparisonto
processes with shorter timescales(e.g., the diurnal cy- Ernax. The different components of the lossesof equa-
cle in temperature). tion (1) are shownin Figure 2. No dynamic interaction
FollowingRodrigucz-Iturb½ ½tal. [19991,rain events with the water table is considered.
are modeled as a sequenceof instantaneous impulses Although quite simple, the previous model captures
having random depths h concentrated at times sepa- the main aspects of the dynamics under study, espe-
rated by random intervals •-. The arrivals of storms are cially those related to the stochastic and intermittent
modeled as a Poissonprocesswith rate /k, namely, the characteristic of the rainfall input as well as the non-
distribution of times •- between storms is linear dependence of infiltration, evapotranspiration,
and leakage on the soil moisture state. For the above
model, Rodriguez-Iturbeet al. [1999b]give the steady
state probability density function of the soil moisture
as well as its moments. From these results Ridolfi et al.
while the depth h is assumedto be exponentially dis-
[2000]derivethe expectedvalue of the duration of the
tributed,
_

stressperiodsduring a seasonT½and the meanseasonal

numberof stress
periods
•f. Theexpected
valueff of
0(h)--eI -•a
; , (3) the duration of an excursion below a given threshold (
of soil moisture is given by
with "7equal to the averageprecipitation in a rainy day.
The relative soil moisture s is limited to the inter-
val [0,1], and the infiltration I(s,t) in equation (1)
is assumed identical to the rainfall input if the soil 1 1•(••)-(1+•)
r/r

does not reach saturation. Surface runoff takes place

whenever precipitation exceeds the available soil vol-
umeandsaturationoccurs.Thusequation(1) describes
x F 1+ -F l+-,-
r/ r , (4)
a bounded shot noiseprocesswhere infiltration depends where
on the state of soil moisture, and runoff only takes place
through the saturation-from-below mechanism studied
in detail by Dunne [1978].Lossesfrom canopyintercep-
tion may also be incorporated by considering a thresh-
rl- s*nZ• r = nZ•'
old of rainfall depth A below which the precipitation
does not effectively reach the ground. In this case the and F[a] and F[a,x] stand for the gammafunctionand
the incomplete gamma function, respectively.
rateofstormarrivals
ismodified
to ;ke-zx/v[Rodriguez-
Iturbe et al., 1999b].
The evapotranspirationrate E(s) is assumedto de-
pend linearly on the soil water content in the interval
Losses(s)
[0,s*], for s > s*, E(s) -- Ernax. The value of s* de-
pends on the field capacity of the soil and on the type
of vegetation, while the value Em•x involves effects of
both climate (temperature, solar radiation, air mois-
ture, wind velocity) and vegetation. Notice that in this
model, lossesfrom evapotranspiration continue also be-
low wilting point' the linear decreasefrom s,o to zero
soil moisture stands for the evaporation lossesthat take
place up to the hygroscopicpoint, which can be consid-
ered zero for modeling purposes. s* s• 1.0 s
The rate of leakageL(s) is equal to the vertical un-
saturatedhydraulic conductivityK(s). It strongly de- Figure 2. Lossesfrom evapotranspiration
and leakage
pends on soil moisture and is modeled through the com- as function of the soil moisture state.
 21562202d, 2000, D14, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2000JD900206 by Cochrane Colombia, Wiley Online Library on [01/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
18,016 RIDOLFI ET AL.- CLIMATE VARIABILITY AND VEGETATION

The previousrelationshipappliesto a hypothetical

very longperiodof homogenous climaticconditions.In
reality, we have to considerthe limit imposedby the_

durationof the season,Tseas.When T½is greaterthan

Tsea•,the expectedvalue of the stressduration coin- _

cides
withtheseason.
Thusthemeanduration
T•, con- o c5 c5 c5 c5 c5 c5 o
strained to the length of the season,is
("',1 ("',1 ("',1 ("',1 ("',1

• -- min[Tseas,
•½]. (5)
The mean seasonalnumberof downcrossings
is [Ri-
dolfi et al., 2000]

Tseasc•/•/ -F (6)

with

C- tl( keX
R[ktl
( (ex-• - eAsl/•ls*d-•-Z*
)
+ - + ,
where
- o. o o o o. • •
d o, d d o,' o, o,' o,'
R- t/s* X •- x- +-
••OoOoO.

ns*(1- s•) •s• s*

oO.o . .o--
rk rls* r c5 o, cSC:?,cSO, d c5

k k ' rk

k k ' rk

The previousexpressions wereobtainedanalyzingthe

crossingpropertiesof the soilmoisturedynamicswith
the frameworkdevelopedby Masoliver [19871for pro-
cessesdriven by shot noise. The methodologyis based
on the reconstruction of the temporal evolution of the
trajectories.The detailsof the derivation,alongwith
other stochastic variables useful to quantify the vegeta-
tion waterstress,are givenby Ridolfiet al. [2000].The
dependence of all thesevariableson climate,soil, and
vegetationis alsodiscussed
there.
 21562202d, 2000, D14, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2000JD900206 by Cochrane Colombia, Wiley Online Library on [01/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RIDOLFI ET AL' CLIMATE VARIABILITY AND VEGETATION 18,017

f[•.] NEW BRUNSWICK (NJ) f[7] NEW BRUNSWICK (NJ)

0.4
0.35 0.4

0.3
0.3
O.25
0.2
0.2
0.15
0.1 0.1
0.05
, !

0.210.250.290.330.370.410.450.490.53 7.4 8.8 10.2 11.6 13. 14.4 15.8 17.2

f[•.] INDEPENDENCE (KS) f[7] INDEPENDENCE

(KS)
0.4
0.35 0.4

0.3
0.3
0.25
0.2
0.2
0.15
0.1 0.1
0.05
, I
0.225 0.27 0.315 0.36 0.405 0.45 0.495
.1
8 10 12 14 16 18 20

f[•] SAVANNAH (GA) f[7] SAVANNAH (GA)

0.4
0.35 0.4

0.3
0.3
O.25
0.2
0.2
0.15
0.1
0.1
0.05
!

0.26 0.3 0.34 0.38 0.42 0.46 0.5 0.54

, 8.2 10.4 12.6 14.8 17.
[ ,,
19.2 21.4

Figure 3. Histogramsof relativefrequencyfor the rate of arrival of storms• and the average
storm depth -/estimated from daily data of precipitationfor someof the stationsin Table 1.

3. Interannual Seasonal Rainfall tion, and thus the corresponding sample sizes of ,• and
Variability 7, oscillates between 49 and 95 years. The limited pe-
riods of observation do not allow precise quantitative
Interannual seasonalrainfall variability is character- analyses;however, they are acceptable for the purposes
ized here through the fluctuations detected on the rate of this section, where the main objective is to obtain
of storm occurencesA and the average depth of rain- a realistic range of values for ,• and • as well as for
fall events7- Daily rainfall data for the period May their interannual fluctuations. In relation to vegetation
through Septemberwere analyzedat eight stations lo- changes,it has been well documented[Archer et al.,
cated in different regions of the United States. This 1988] the impact that climate and, more specifically,
analysis is complementary to the more extensive one of rainfall fluctuations have on timescales from a decade
D'Odorico et al. [2000],whichwas mainly centeredin to a century.
Texas. The chosenperiod broadly comprisesthe grow- In general, the identification of the growing seasonis
ing seasonand the assumption is made that the rainfall a complex problem that depends on the type of vege-
regime is more or lesshomogeneousthroughout the sea- tation, the climate, and the indicator chosen to single
son. Although this assumptionis an obvious oversim- out the growingperiod ILarchef, 1995]. However,the
plification of reality, we are more interested here in the
precise duration of the growing seasonis not a funda-
interannual fluctuationsobservedfor the chosenperiod mental aspect in the present study, since it does not
rather than in the variability inside the period. Table 1
greatly affect the objectives mentioned above. Thus we
showsthe period of observationfor each station and, have chosenthe same period of 153 days as a common
in parentheses,the maximum number of missingdays season during which the scarcity of water may play a
for a growing season.The number of years of observa- key role in plant growth for all the stations.
 21562202d, 2000, D14, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2000JD900206 by Cochrane Colombia, Wiley Online Library on [01/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
18,018 RIDOLFI ET AL.- CLIMATE VARIABILITY AND VEGETATION

The daily rainfall data at each station allowsthe es- ficient of variation for "y considerably larger than that
timation of a value of • and a value of "yfor each year for • in the caseof many stations in Texas. In all cases,
of record. They represent the frequencyof occurrence two-parametergamma distributions provide a good fit
of rainy days and the averagerainfall in a rainy day, re- to the histograms.
spectively.A significantvariability from a year to year
is often present, and the fluctuations may be as high 4. Duration and Frequency of Water
as 30 ø70of the mean of the series. No significant cross Stress Periods
correlation between • and "y seem to exist in the time
series. This was confirmed by extensiveautocorrelation We focus now on the impact of the interannual rain-
and cross-correlationanalyses similar to those carried fallvariability
onthemeanduration
T• andonthemean
out by D'Odorico et al. [2000]. Table 1 givesthe es- frequency•g of periodsof water stress.The statistical
timated lag-1 autocorrelationcoefficientas well as the nature of X and "y confers a probabilistic structure to
_

lag-zerocrosscorrelationbetween"yand •. T• and•g whichin turndepends

onthevegetation
and
Figure 3 showssomeexamplesof the histogramsob- soil properties existing at the site.
tained for • and "y,while Table 1 gives the mean, vari-
ance, and coefficientof variation estimated for both pa- 4.1. Climate Variability and Duration of Water
Stress
rameters at each station. As expected, there is a wide
variation among the values of the statistical parameters The objective here is the evaluation of the mean and
_

characterizing the interannual fluctuations of "7 and • the variance

of T• whenthe parameters
X and-yare
at the different stations. Some common features may considered random variables which change from year
be noticed: the histograms are always unimodal and to year accordinglyto their respectivegammadistribu-
the coefficients of variation of • and "7 are similar in tions. We will first consider the case where only one of
thesetwo parameterspresentsrandom interannualfluc-
value, although the relative variability of • appears to
be somewhat weaker than that of "y. This is not a gen-tuations, while the other remains fixed. This will allow
eral result, sinceD'Odorico et al. [2000]founda coef- us to study the individual impacts of the fluctuations

(7)- days

......... 7(mm/storm)
30 40
•seas
10 20
' ' 7(mm/storm)

20

40

f7(7)
60

Figure 4. Derivationof the probabilitydistributic;n

of the seasonal
meandurationof periods
withwaterstress
• based
onitsfunctional
dependence
ontheseasonal
meanprecipitation
ona
rainy day •. The distributionof "yis assumedto be a two-parametergamma, and the valuesused
in this figuresare 5:12 mm/d, CV["/]-0.2, /k:0.3 storm/d, CV[/k]-0, nZ•-300 mm, Emax:4
mm/d). The thresholdlevel• usedin this exampleis onewhichapproximates
the start of water
stressin many semiarid regions, •:s*=0.35.
 21562202d, 2000, D14, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2000JD900206 by Cochrane Colombia, Wiley Online Library on [01/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RIDOLFI ET AL' CLIMATE VARIABILITY AND VEGETATION 18,019

of A and 7 on the statistical characteristicsof the mean P.•,•

(Tseas)
seasonalduration of water stressperiods. The analyt-
ical framework is identical in both cases, and thus we
will describehere the situation where ' fluctuates from 1 •..:••
0.8
%k
'". (a)
CV[71=0.20
year to year, but A remains fixed. _

0.6 "• ........CV[7]=0.25

Themeanof therandom
variable
T• is givenin this 0.4
•...•
":.'.,
--- CV[7]=0.30
case by
0.2

/z½•
-J0
f½•
(•)•d• + p½•
(Tseas)Tseas. "-"-'
..... - (mm/storm)
10 15 207
The first term of the right-hand side representsthe con-
P.•,,.
(Tseas)
tribution
_
to /z2•fromthecontinuous
partofthepdf 1 • Emax=3mm/d
of T•. The second
termis the contribution
fromthe ........Emax=4mm/d ..........
0.8 ___Emax=5mm/d
...' "'""'"" ..........
_

discrete
probability
correspondingto T• - Tseas.
The
random
variable
• isfunctionally
related
to7 through 0.6
equation (4) where all other parameters,includingA,
are consideredas fixed. For notational purposes,equa-
_
0.4

tion(4) will bewrittenasT• - g('), whichis shown

in Figure 4. One observesthat for very low values of
_
0.2 (b)
7, T• - Tsea•.For7 largerthana value")/seas
there- nZr (cm)
lationship decaysmonotonically reflecting the fact that 30 40 50

the average duration of an excursion of soil moisture

Figure 5. Impact of climate, soil, and vegetation
_

belowthelevel•, T•, becomes

smaller
whenthemean
rain depth 7 becomes larger. The density component characteristics
ontheatomofprobability,
p•
when only 7 is randomly changing from year to year
_

f2•(T•)maythenbewritten
as (•:s*:0.35, Tseas--150d). (a) A:0.2, CV[A]=0,
nZr=300 mm, Em•x=4 mm/d; (b) •=10 mm/storm,
- f(7)
_

•s(•)' (8) cv[]=o.25, - 0.2,

d-y

where the numerator is the pdf of 7 and the negative

sign accountsfor the fact that g(-) is a monotonically where
thelimitscorrespond
_
tothecases
T• - 0, _

decreasing
function.
Thederivation
of f2•(T•)usingandT• - Tse•,- - %ea•-Similarly,
_
the variance
is
obtained as
the functional
dependence
of T• on 7 givenby equa-
tion (4), and a two-parametergamma distribution for
7 is schematicallyshownin Figure 4 for someparticular
parametervalues.One observes that the part of
f•oo
fv(f)g2(f)d7
+/•}•(1
seas -p½•(Tseas
corresponding
to the interval [0,%e•] yields an atom
- 2t• f-•(7)S(7)d7 (10)
ofprobability
at • - T•e•s.Thisatommaybemore seas

or less pronounced depending on the characteristicsof

climate, soil, and vegetation. In many cases(e.g., wet
q-p•'•
(Tseas)(Tseas
-//•,•)2.
climatesor low soil moisturethresholds•) the atom of
probability at T•e• is negligible, but in arid or semiarid
Theatomofprobability,
P•(Tsea•),
in equations
(9)
and (10) may be estimatedby
climates, it may be quite important for the character-
ization of the periods when plants are under stress. It
-- '7seas
_

isimportant
to clarifythatthebimodal
character
ofT• P,•(Tseas) f.•(7)dT. J0
(11)
results solely from the finite duration of the seasonbe-
_

ingconsidered. Thusall values

of T• _>Tseas
arethen Equations(9) and (10) allow the numericalevaluation _

concentrated at of the coefficient

of variation of T• for anyparticular
Substitutingequation(8) into equation(7), one ob- parameter values. Identical relationships to equations
tains
(9) and (10) are valid for the casewhen A is considered
a random variable from year to year, and 7 is assumed
constantthroughout time. Both equationsare still valid _

withthesubstitution
_
of f,x(A)forfv(-) andT• - r(A)
ß• oo d• forT• - g(7).
Figures 5a and 6a show the influenceof climate vari-
seas
fw(•)g(•)d•
+ p• (Tseas)Vseas
, (9) ability
ontheatomofprobability
PTi(Tseas).
Thede-
 21562202d, 2000, D14, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2000JD900206 by Cochrane Colombia, Wiley Online Library on [01/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
18,020 RIDOLFI ET AL.- CLIMATE VARIABILITY AND VEGETATION

P•,..
(Tseas) practically always above 1 and that they reach values
1 in the 2-4 range. This implies a strong amplificationof
0.8 (a)
CV[k]=0.20
climate fluctuations on the fluctuations
mean duration of periods with water stress. The cru-
of the seasonal

0.6 cial role of soil depth and of evapotranspiration emerges

"• ........CV[,t]=0.25
• ..... CV[,t]=0.30 again in Figures 7a and 7b and 8a and 8b. Figures 7c
0.4 and 8c show the influence of the threshold • on the
0.2
relativefluctuations
of T• withrespect
to thoseof the
climate. It is interestingto notice the pronouncedam-
..'....'•..•..
•..-.,
•.(storms/d) plificationof the climatefluctuationson the fluctuations
0.2 0.3 0.4

P•,..
O'seas) cv['H/cvrr]
1 • Emax=3rnrn/d 5
..... Emax=4mm/d ..--"" (a)
0.8 __ Emax=5mm/d""" -.............

0.6 ,'' o o
4• Emax=3
mm/d
........ Emax=4mm/d

0.4
2 "'""'"'.......
O.2 (b) 1

' ' ' nZr (cm)

30 40 50 cv[•,]
0.2 0.3 0.4

Figure 6. Impact of climate, soil, and vegetation cv["H/cvw]

4
characteristics
ontheatomofprobability,
P½•(Tseas),
when only h is randomly changing from year to year 3.5 • (b)
(•c=s*=0.35, Tseas=150d). (a) •=10 mm/storm, _
3 nZr=20
cm
CV["/]=0, nZr=300 mm, Em•x=4 mm/d; (b) A=0.2, 2.5
........nZr=30
cm
0.2, cv[l:0. 2

1.5

1
pendence is highly nonlinear with the strongest sensi-
0.5
tivity
ofp½•
(T•e•s)
withrespect
to• (or•) intherange .... cv['r]
of values where those parameters are most commonly 0.2 0.3 0.4
found in nature. It is also apparent that the sensitivity
to • (or•) islarger
thanthatto err2 (or cr•). Finally,
4
theimpact
onp½•
(Tse•s)
whenvarying
• isquite
similar _

3.5 (C)
to the one obtained when varying A.
Figures
5band6bshow
theimpact
onpf•(Tse•s)
of
the active soil depth and the evapotranspiration. Both 2.5
2
/ '....... • se=s
,
variables strongly affect the atom of probability which 1.5
may change drastically with relatively small changesin 1
n•7r or Emax. Since especially in arid and semiarid cli- 0.5
_

mates
p• (Tseas)
--!
affects
allthestatistical
characteristics 10 15 20 2'53,
(m
m/storm)
ofT•, thesensitivity
shown
in theaboveanalyses
makes
evident the crucial role of evapotranspiration and soil
depth in modulating the impact of climate fluctuations. Figure 7. Impact of climate, soil, and vegetationchar-
acteristics on the coefficient of variation of the seasonal
Figures 7 and 8 summarize the dependence of the
mean duration of periods with water stress. Here only
coefficient
ofvariation
oftheduration,
CV[•], oncli- 3' is randomlychangingfrom year to year (Tse•s=150
mate, soil, and vegetation. The coefficientof variation d)' (a) ½=10 mm/storm,A=0.2 storm/d, CV[X]:0,
is normalized•ith respectto that of 3', or A, sothat the nZr-300 mm, • - s*=0.35; (b) •=10 mm/storm, _

figures show how much the nonlinear dynamics ampli- A:0.2 storm/d, CV[X]-0, Em•x:4 mm/d, • = _

fies the interannual variability of climate. One observes s*:0.35; (c) CV[3']:0.25, A:0.2 storm/d, CV[X]=0,
_ _

that the ratiosCV[T•]/CV["/]andCV[T•]/CV[A]are nZ•-300 mm, Em•x=4mm/d, s*=0.35, &o=0.15.

 21562202d, 2000, D14, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2000JD900206 by Cochrane Colombia, Wiley Online Library on [01/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RIDOLFI ET AL.' CLIMATE VARIABILITY AND VEGETATION 18,021

ofT• fora particular

range
ofvalues of be a veryimportantfactorfor the valueof fi•, especially
ofif. Thisrange
valuesis itself dependenton soil and vegetation charac- in regionswith relativelycoldandhumidwinters,where
teristics. Moreover, the impact of the interannual cli- transient conditionsmay last during a considerablepart
matic fluctuations is highly sensitiveto the threshold of the growing season.
level being considered(e.g., soil moisture at which wa- Figures11 and 12 showthe influenceof climaticvari-
ter stressstarts s*, or wilting point ability, both in '7 and in ,X, and the role of soil and
Monte Carlo simulations were performed to investi- vegetation on f(fi•). Threequalitativelydifferentcases
gate the effect of the combinedvariability of the fre- may be distinguishedin the results:
quency of rainy events ,X and their mean depth '7. 1. The temporal evolution of soil moisture remains
Fifty thousand pairs of random samplesfor often above(e.g., wet climate)or below(e.g., dry cli-
yield stable and smooth probability distributionsfor mate)the threshold,andthusthe crossings duringthe
the seasonalmean duration of periods with water stress.
Figure 9 showstypical probability distributions corre-
spondingto incipient stresscondition, ( - s*, for dif-
ferent climate, soil, and vegetation characteristics.The
previouscommentsmade when consideringseparately
the impacts of '7 and ,Xare all confirmed. The influence 4•
3.5 (a)
Emax=3mm/d
of the climate is alwaysstrong: both the mean and the 3 • ........
Emax=4
mm/d
variance of the climatic parameters have a great impact 2.5 Emax=5
mm/d
2............. --•
_

onthe probability
distribution
of T•. Veryimportant
are also the roles of the active soil depth and the evap- 1.5
otranspiration. 1
Figure 10 is similar to Figure 9, but the threshold 0.5 -- -'"
level is now the permanent wilting point, &o. As ex-
cv[x]
pected, the atom of probability at Tseasis strongly re- 0.2 0.3 0.4
duced, and in many conditions, it is practically absent.
Aside from this, the distributions confirm the same con- cvr;/cvm
4
clusionsobtained for the casewhen ( = s*. Again, one
notesthe strong influenceof the climate and the funda- 3.5 (b)
mental roles of the soil and vegetation characteristics.
--!
3• ......... nZr=20
nZr=30 cm
cm
Since
thedistribution
of T• [Ridolfi
et al., 2000]does
not dependon the initial soil moisturestate (e.g., soil
moisture content at the start of the growing season),
2.5 •=40
1.5 ...............
cm
the results obtained here are also valid, regardlessof
the initial soil moisture value. 0.5

, cv[x]
0.2 0.3 0.4
4.2. Climate Variability and the Frequency of
Water Stress
6
Equation(6) givesthe analyticalexpressionfor the (c)
expectednumberof downcrossings of an arbitrary level 5 •:=s*
by the soilmoistureprocessasfunctionof climate,soil, 4 ........ =
and vegetationcharacteristics.Its dependenceon cli-
3
mate parametersis not monotonicand exhibits a pro-
nounced maximum. Such non monotonic character, 2

quite different from that shownin Figure 4 for the 1

caseofT•, makes
intractable
thederivation
of analyt- _

ical results for the dependenceof the mean number of 0.2 0.3 0'.4
x(storms/d)
downcrossing on the interannualclimatic fluctuations.
Thus this section is based on Monte Carlo simulations Figure 8. Impactof climate,soil,andvegetationchar-
with '7 and ,Xbeing sampledfrom their respectivetwo- acteristics on the coefficient of variation of the seasonal
parametergammadistributions.The pdf of fi• is then mean duration of periodswith water stress.Here only
,X is randomlychoosingfrom year to year (Tsea•-150
computeddirectlyfromequation(6). Noticethat differ-
_

d). (a) X=0.2storm/d,•:10 _mm/storm,CV[']=0,

entlythanforthecaseofT•, theresults forfi• depend nZr=300 mm, •=s*=0.35; (b) ,X=0.2storm/d, •=10
on the existenceof a steady-statecondition[Ridolfi et mm/d, CV[h']:0, Em•x:4 mm/d, •=s*=0.35; (c)
al., 2000]. Thus the valueof the initial soil moisture CV[/k]=0.25,•=10 mm/storms,CV[h']=0, nZr=300
content(e.g., at the start of the growingseason)may mm, Em•x=4mm/d, s*=0.35, &o=0.15.
 21562202d, 2000, D14, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2000JD900206 by Cochrane Colombia, Wiley Online Library on [01/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
!8,022 RIDOLFI ET AL.' CLIMATE VARIABILITY AND VEGETATION

0.08
f(•,w)
0.25
f(Tseas)=0.34
0.07 (a) (a)
0.2
0.06

0.05 wet climate

0.15 wet climate
...... dry climate
ß

0.04

0.03 0.1

0.02
0.05
0.01
.:'. .......... ". •"."•": ................. r......... , ....
20 40 60 80 100 120 140 20 40 60 80 1 O0 120 140

f(•';.) f(•'sw)
0.15
0.15
(b)
(b)

0.1 0.1
CV[y]=CV[?,]=0.2
CV[y]=CV[A]=0.2
........ CV[yl=CV[?,]=0.3
........ CV[7]=CV[X]=0.3
0.05 0.05

20 40 60 80 100 120 140 20 40 60 80 100 120 140

f(•',•.) f(•;w)
0.15

(c) 0.14
(c)
0.12

0.1 0.1
Emax
=3 mm/d
0.08 Emax=3mm/d
........ Emax
=5 mm/d
........ Emax=5mm/d
0.06
0.05
0.04

0.02

l; ......... ' ...... ,.......... •...I. ............................

20 40 60 80 1O0 120 140 20 40 60 80 1 O0 120 140

0.15
f(•'sw)
0.12
(d) (d)
0.1

0.1 0.08
nZr=20 cm
nZr=20 cm
........ nZr=50 cm 0.06
........ nZr=50 cm
0.05 0.04

0.02

20 40 60 80 100 120 140 20 40 60 80 100 120 140

Figure 9. Impact of climate, soil, and veg- Figure 10. Impact of climate,soil, and vegetation
etation characteristics
on the probabilitydistribu- characteristics
on the probabilitydistributionof the
tion of the seasonalmean durationof periodswith seasonal durationof periodswith waterstressfor the
water stressfor the caseof •=s*=0.35 (s•-0.8, caseof •=sw=0.15(s•-0.8, K•=90 cm/d, Tse•=150
K•=90 cm/d, Tseas-150d). (a) Wet climate' _d).(a) Wet climate'•:12 mm/storm,CV[71:0.25,
½=12mm/storm,CV['7]-0.25,X=0.35,CV[A]=0.25; A=0.35,CV[A?0.25;dry climate:•=10 mm/storm,
dry climate:•-10 mm/storm,CV[7]=0.25,•=0.25, CV['7]=0.25,A:0.25, CV[A]:0.25 (nZr:30_0mm,
CV[A]=0.25(nZr:300 mm, Emax=4mm/d); (b) •:12 Em•x=4 mm/d); (b) •:10 mm/storm, A:0.25,
mm/storm,
•=0.3, nZ•=300mm,Em•x=4mm/d;(c) nZ•=300mm,Em•x:4mm/d; (c) •:10 mm/storm,
•=12 mm/storm,
CV[-7]:0.25,
X=0.35,CV[X]=0.25,CV[-7]:0.25,
•=0.25,CV[X]:0.25,
nZ•=300mm;(d)
nZ•-300 mm; (d) •=12 mm/storm, CV[-7]:0.25, •:10 mm/storm,CV[-7]=0.25,A:0.25, CV[A] =0.25,
X=0.35,CV[A]=0.25,Em•x:4mm/d. Emax=4
mm/d.
 21562202d, 2000, D14, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2000JD900206 by Cochrane Colombia, Wiley Online Library on [01/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RIDOLFI ET AL.- CLIMATE VARIABILITY AND VEGETATION 18,023

f(•s.) f(•s.)
1 1

(a) (b)
0.8 0.8

: wet climate CV['),l=CV[,t.]=0.3

0.6
'. ,
........
dry climate 0.6 ........ CV[.),l=CV[,t.]=0.2

0.4 '• 0.4

: :,

0.2 0.2

ns, ' ns,

2 4 6 8 10 2 4 6 8 10

f(•.,.) f(•s,)
1 1

(c) (d)
0.8 0.8

Emax=3 mm/d nZr=20 cm

0.6 0.6
........ Emax=5mm/d ........ nZr=50 cm

0.4 0.4

0.2 0.2

' ' ' ' ' ' ' ' ns, ' ' ' ' ' ' ' ; ............ ns,
2 4 6 8 10 2 4 6 8 10

Figure 11. Impact of climate, soil, and vegetationcharacteristicson the probability distribution
of the seasonalfrequencyof periodswith water stressfor the caseof •c=s*:0.35 (s1=0.8, Ks=90
cm/d,Tseas-150
d). (a) Wetclimate:•:12 m•n/storm,
CV[7]:0.25,•:0.35, CV[A]:0.25;
dry
climate:•=10 mm/storm,CV[7]=0.25,A=0.20,CV[A]=0.25(r•Z?=300mm, Emax-4 mm/d);
(b) •=12 mm/storm,X=0.3, nZ?=300mm, Em•x:4 mm/d; (c) •=12 mm/storm,CV[7]=0.25, _

•:0.35, CV[A]:0.25,r•Z•=300 mm; (d) •:12 mm/storm,CV[?]-0.25, A:0.35, CV[A]:0.25,

Emax:4 mm/d.

growing season are rare. In this case the distribution 5. Conclusions

has a strong mode at • - 0 with a sharp decreasefor
The objective analysis of water stressrequires the
•>0.
2. The temporal evolution of soil moisture involves quantificationof the duration and frequencyof the pe-
frequent crossings of the threshold. Then the mode riods with water deficit. The analytical expressionsde-
shifts toward valuesof • different from zero and a scribingthe mean seasonalcharacteristicsof thesevari-
number of stress periods are expected during the sea- ables exhibit a complex and strongly nonlinear depen-
son (e.g., see Figures lib and 12b for the case when denceon the climate propertiesas well as on the soil
and vegetationcharacteristics.Sincethe seasonalcli-
3. Most commonly, the pdf is bimodal with one mode matic characteristics fluctuate from year to year, it is
at zero (e.g., lack of crossingduring the season)and important to study the impact of suchfluctuationson
another at a value which depends on the climate, soil, the hydrologicdescriptionof the periodswhere plants
and vegetation characteristics. Two qualitatively dif- undergo water stress.
ferent preferential states coexist during the evolution It was found that the seasonalfrequency of daily rain-
of soil moisture and the mean of the distribution has fall occurrences and the mean precipitation on a rainy
little meaning in describing the impact of interannual day exhibit important interannual fluctuationswhich
climatic fluctuations. have a strong impact on the temporal evolutionof soil
The importance of the rainfall regime is clearly shown moisture,particularly on the frequencyof level crossings
in Figures 11a and lib and 12a and 12b, while Figures and the length of excursionsbelow thresholdlevels.
11c and lid and 12c and 12d highlight the crucial role 1. In the presenceof interannual climatic fluctuations
played by the soil and the vegetation. there are numerous combinations of climate, soil, and
 21562202d, 2000, D14, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2000JD900206 by Cochrane Colombia, Wiley Online Library on [01/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
18,024 RIDOLFI ET AL.' CLIMATE VARIABILITY AND VEGETATION

f(•l.w) f(•:,w)
1 1

(a) (b)
0.8 0.8

wet climate CV[7]=CV[•]=0.3

0.6 0.6
........ dry climate ........ CV[7]=CV[•]=0.2

0.4 0.4

0.2 0.2

, , , , , , , i , , , , , , , , , , , ' n,w
2 4 6 8 10 2 4 6 8 10

f(•sw) f(•sw)
1 1

(c) (d)
0.8 0.8

Emax=3mm/d
0.6 0.6l nZr=20
cm
........ Emax=5mm/d

0.4

0.2
0.4
,. .."'"'..
0.2
.........
_

ß ' ' nsw nsw

2 4 6 8 10 2 4 6 8 10

Figure 12. Impact of climate,soil, and vegetationcharacteristics

on the probabilitydistribution
of the seasonal
frequencyof periodswith water stressfor the caseof •=sw=0.15 (s1:0.8, Ks:90
cm/d, Tseas-150
d). (a) Wet climate:•=12 mm/storm,CV['7]:0.25,A=0.35,CV[A]:0.25;dry
climate:•:10 mm/storm,CV[q•]:0.25,
_
•:0.25, CV[A]-0.25(nZ•=300mm,Emax:4mm/d);
(b) •=10 mm/storm,A=0.25,nZ•=300 mm,Emax=4mm/d; (c) •-10 mm/storm,CV[q•]=0.25, _

•=0.25, CV[A]=0.25,nZ•-300 mm; (d) •=10 mm/storm,CV[q•]-0.25,A=0.25,CV[A]=0.25,

Emax-4 mm/d.

vegetation characteristics that lead to bimodal proba- an additional crucial impact on its control of the per-
bility distributions, both for the duration and for the manentwilting point and the moisturecontentat which
frequency of water stress periods. Thus it is common water stress starts. These threshold levels are central
to find the existenceof two preferential states for the hy- to the temporal and spatial vegetation dynamics,and
drologic variables related to soil moisture which better the statistical structure of the water stress periods is
characterize water stress in vegetation. These prefer- very sensitiveto their valueswhen subjectto climatic
ential states are generally very different from the mean fluctuations.
value, and interannual rainfall fluctuations tend to force The above conclusions are considered important to
the permanency of the seasonal mean duration of water understand the evolutionary dynamics of vegetation
stressperiods and their frequency of occurrencearound and the hydrologiccontrolsacting upon this dynamics
these states. on the presenceof interannual climatic fluctuations.
2. The coefficient of variation for both the seasonal
_

meanduration
of periods
withwaterstress
T• andthe Acknowledgments. This researchhas been supported
mean seasonalfrequencyof theseperiods• is almost by grants from NASA (grants NAGW-4171 and NAGW-
always greater than that of the climate fluctuations. 4766) and NSF (grant EAR-996180).
Thus the soil moisture evolutionary dynamics amplifies
thevariability
ofT• and• withrespect
to thatexisting References
in the interannual fluctuations of the rainfall regime.
3. Both the active soil depth and the evapotranspira- Archer, R. A., Woody plant encroachmentinto southwestern
tion have fundamental roles in modulating the impact of grasslandand savannas: Rates, patterns and proximate
causes,in EcologicalImplications of LivestockHerbivory
climatic variability on the soil moisture evolution. Rel- in the West, edited by M. Vavra, W. A. Laycock,and R.
atively small changeson these parameters may lead to D. Pieper, pp.13-68, Soc. Range Manage., Denver, Colo.,
1994.
significant
differences
onT• and•. Vegetation
alsohas
 21562202d, 2000, D14, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2000JD900206 by Cochrane Colombia, Wiley Online Library on [01/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RIDOLFI ET AL.: CLIMATE VARIABILITY AND VEGETATION 18,025

Archer, S., C. Scifres,C. R. Bassham,and R. Maggio, Auto- Iturbe, Duration and frequency of water stress in vegeta-
genic successionin a subtropical savanna: Conversion of tion: An analytical model, Water Resour. Res., in press,
grasslandto thorn woodland, Ecol. Monogr., 58, 90-102, 2OOO.
1988. Rodriguez-Iturbe, I., P. D'Odorico, A. Porporato, and L.
Brubaker, K. L., and D. Entekhabi, Analysis of feedback Ridolfi, Tree-grass coexistence in savannas: the role of
mechanisms in land-atmosphere interaction, Water Re- spatial dynamics and climate fluctuations, Geophys. Res.
sour. Res., 32(5), 1343-1357, 1996. Lett., 26, 247-250, 1999a.
Clapp, R. B., and G. M. Hornberger, Empirical equations Rodriguez-Iturbe, I., A. Porporato, L. Ridolfi, V. Isham,
for some hydraulic properties, Water Resour. Res., 1•, and D. Cox, Probabilistic modelling of water balance at a
601-604, 1978. point: The role of climate, soil and vegetation, Proc. R.
Dingroan, S. L., Physical Hydrology,Prentice Hall, Engle- Soc. London, Set. A, •55, 3789-3805, 1999b.
wood Cliffs. N.J., 1994. Rodriguez-Iturbe, I., P. D'Odorico, A. Potpotato, and L.
D'Odorico, P., and I. Rodriguez-Iturbe, Space-time self- Ridolfi, On the spatial and temporal links between veg-
organization of mesoscalerainfall and soil moisture, Adv. etation, climate, and soil moisture, Water Resour. Res.,
Water Res., 23(4), 349-358, 2000. 35 (12), 3709-3722, 1999c.
D'Odorico, P., L. Ridolfi, A. Porporato, and I. Rodriguez- Scholes, R. J., and S. R. Archer, Tree-grass coexistence in
Iturbe, Preferential states of seasonal soil moisture: The savannas. Ann. Rev. Ecol. $yst., 28, 517-544, 1997.
impact of climate fluctuations, Water Resour. Res., in Scholes,R. J., and B. H. Walker, An African Savanna, Cam-
press, 2000. bridge Univ. Press, New York, 1993.
Dunne, T., Field studies of hillslope flow processes,in Hills-
lope Hydrology,edited by M. J. Kirkby, pp.227-293, John
Wiley, New York, 1978. P. D'Odorico, Department of Civil Engineering, Texas
Hale, M. G., and D. M. Orchutt, The Physiologyof Plants A&M University, College Station, TX 77843.
Under Stress, John Wiley, New York, 1987. A. Porporato and L. Ridolfi, Dipartimento di Idraulica,
Trasporti e Infrastrutture Civili, Politecnico di Torino,
Lambers, H., S. F. Chapin, and T. L. Pons, Plant Physio-
logical Ecology,Springer-Verlag, New York, 1998. Corso Duca degli Abruzzi, 24, 10129 Torino, Italy.
Larcher, W., PhysiologicalPlant Ecology,Springer-Verlag, I. Rodriguez-Iturbe, Department of Civil and Environ-
New York, 1995. mental Engineering, E-220 E-Quad., Princeton University,
Masoliver, J., First-passage times for non-Markovian pro- Princeton, NJ 08544. ([email protected])
cesses:Shot noise,Phys. Rev., Set. A, 35(9), 3918-3928,
1987. (ReceivedJuly 16, 1999; revised January 27, 2000;
Ridolfi, L., P. D'Odorico, A. Porporato, and I. Rodriguez- acceptedMarch 22, 2000.)