Professional Documents
Culture Documents
www.elsevier.com/locate/advwatres
Universita di Padova, Dept. IMAGe, via Loredan 20, I-35131 Padova, Italy
Received 17 June 2004; received in revised form 14 February 2005; accepted 22 February 2005
Available online 10 May 2005
Abstract
On hillsides of semiarid catchments regular bands of vegetation have been observed to form under low rainfall conditions. Many
authors have observed the existence of a slope gradient threshold below which no banded patterns were observed; this increases with
the mean annual rainfall. A simple model for soil moisture balance and vegetation growth, explicitly accounting for basic soil physics, is demonstrated and discussed. The inuence of relevant soil characteristics on vegetation patterns and their patchiness is
addressed by linear stability analysis. The results conrm the experimental evidence of a threshold slope gradient depending on
the mean annual net water supply, and demonstrate the inuence of the soil properties (saturated conductivity and capillary rise)
on the stability condition and on the threshold slope. Nevertheless, the concluding remark concerns the oversimplied models for
vegetation patterns, such as the one discussed here, that require properly dened eective soil parameters in order to become predictive tools.
2005 Elsevier Ltd. All rights reserved.
Keywords: Ecohydrology; Semiarid catchments; Vegetation banding; Stability analysis; Patterns initiation
1. Introduction
Regular bands of vegetation have been observed to
form on hillsides of semiarid catchments under low rainfall conditions. Numerous experimental studies have
been conducted in dierent parts of the world, mainly
in Africa and in Australia, in order to collect data and
classify patterns of vegetation organized in a two-phase
mosaic as well as to dene their connection with geomorphic and climatic conditions. Valentin et al. [24]
present a summary of the experimental evidence collected in the last 30 years circa. There is a continuous
interest in trying to understand and model vegetation
patterns formation and evolution. Lefever and Lejeune
[8] explain the formation of tiger bush through a mechanism of resources competition. Gilad et al. [5] attribute
*
0309-1708/$ - see front matter 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.advwatres.2005.02.009
957
2. Theory
Turing [23] provided a hypothesis to explain the generation of patterns through mechanisms of reaction and
diusion. By superimposing an initial pattern of given
wave-number upon the steady state homogeneous solution and allowing reaction and diusion to act over
time, the initial disturbance is expected to either grow
or decline. Obviously, the patterns which grow faster
in time will be most visible. The mathematical framework for this stability analysis is the Fourier analysis.
The relations between the attributes of the unstable patterns (growth rate and characteristic length) and those
of the environment (mean rainfall, hill slope and soil
properties) are investigated below.
The characteristic length of the bands of vegetation
considered here is much larger in the direction perpendicular to the slope than in the parallel one. The soil
moisture follows a spatial distribution similar to that
of the vegetation. Furthermore, the water content variability in the direction perpendicular to the soil surface
may be neglected since the soil depth that inuences vegetation growth is shallow as compared to the surface
extension of the domain (a similar hypothesis is commonly adopted in remote sensing studies, see e.g. [20]).
Thus, based on the above simplied hypothesis, the
theoretical analysis of the soil moisture and of the vegetation balance equation is limited here to the one-dimensional case only. On at ground the extension of the
model and the related results to the two-dimensional
case is straightforward, by replacing the squared mode
considered here with the summation of the squared
unstable modes in two directions.
2.1. Linear stability analysis of soil moisture
and biomass balances
The model of Klausmeier [7] is a pair of partial dierential equations for soil moisture (W) and plant biomass
(N) that are dened on an innite one dimensional domain indexed by X as a function of time T.
oW
oW
A LW RWN 2 V
oT
oX
2
oN
o
N
RJWN 2 MN D 2
oT
oX
958
plant dispersal, and RWN2 is the plants uptake. According to Klausmeier [7] plants take up water at rate
RG(W)F(N)Nwhere G(W) is the response of plants to
water and F(N) is a function that describes how plants
increase inltration. Klausmeier [7] linearizes the problem assuming G(W) = W and F(N) = N, and arguing
that the results are not sensitive to the exact form of
these functions. The lateral loss V oW
, that is responsible
oX
for the instability of the model, will be disclosed in the
next section.
Equation (1) can be reduced in dimensionless form
assuming T = L1t, X = D1/2L1/2x, V = (LD)1/2v, M =
Lm, W = Wsw, N = WsJn and A = LWsa, and obtaining:
ow
ow
a w rwn2 v
ot
ox
2
on
o
n
rwn2 mn 2
ot
ox
where r W 2s J 2 RL1 .
The spatially uniform steady state solution (w0, n0) of
the coupled soil moisture and vegetation problem (1) is
the solution of
lw; n a w rwn2 0
3
hw; n rwn2 mn 0
p
1
namely: w0 0.5a a2 4m2 r1 and n0 mw1
0 r .
Adding a small perturbation to the steady state
(w0, n0),
w w0
u
4
n n0
in the form
X
ck ekt U k x;
u
hn
v 0
0 0
;V
;D
0 0
0 1
ln
and lw, hw, hn and ln are the derivative of l and h with respect to w or n, evaluated in (w0, n0). Substituting (5) into
(6), the following expression for the eigenvalue k may be
found:
k A jkV k 2 D;
Kh
i
8
dx dx
dx
where W 6 0 is the negative pressure head in the unsaturated soil, h is the soil moisture content, K(h) is a mean
value of the soil moisture dependent conductivity in the
dz
unsaturated root zone i dx
is the slope gradient (the
vertical coordinate z is taken as positive upward). Set
h = hs eaW and K = Ks eaW = h Ks/hs [4], where a is
the inverse of the mean capillary rise and depends on
the soil characteristics [11,12]. a represents the ratio between the gravity and the capillary forces. Substituting
K = K(W) and h = h(W) into (8) yields to
dq
d2 h
dh
Dw 2 V w ;
dx
dx
dx
10
959
Fig. 1. Growth coecient Rek as a function of wavelength k. Depending on A and i no wavelength may lead to instability (Rek < 0), or a continuous
range of unstable modes may do this. At the critical stability condition only one k has Rek = 0; the other modes are stable. In the case shown here, the
parameters that characterize the marginal stability condition are A = 374 mm, and i = 0.02. The other relevant parameters are M = 1.8 y1,
J = 0.003 kg m2 mm1 y1; R = 100 m4 kg2 y1, D = 1 m2 y1, L = 4 y1, hs = 0.4, H = 0.1 m, Ks = 3000 m y1 and a = 100 m1.
960
continuous range of unstable modes may arise if stimulated. In the next section, the focus is on the pair Ai
leading to the marginal stability condition for dierent
biomass and soil parameters.
3. Results
A sensitivity analysis of the critical stability condition
(12) for soil parameters was developed. All variables
were set according to literature data or reasonable
assumptions.
Rainfall in semi-arid regions is between 50 and 750 mm
y1 [24]. A is set in this range. Klausmeier [7] assumes for
grass M = 1.8 y1; J = 0.003 kg m2 mm1 y1; R =
100 m4 kg2 y1 (obtained from the expression for equilibrium plant biomass); D = 1 m2 y1, and L = 4 y1.
Thus, m = 0.45 and r = 0.36. The maximum stored soil
moisture Ws hsH depends on the active soil depth H.
If hs = 0.4, and H = 0.1 m, then Ws = 40 mm. In the depth
averaged water balance only the transverse conductivity
comes into play. It was set Ks = 3 102 and 3
103 m y1 (for more clayey and more sandy soils respectively). Rather low wettability of the bare zones and lateral diusion due to capillary forces are expected to
support the so called runo runon mechanism of water
redistribution, which is responsible for the formation of
vegetation patterns. Thus, very high values of the ratio between gravity and capillary forces a have been considered:
a = 10, 100, 1000 m1. Klausmeier [7] does not take into
account soil properties and assumes v = 182.5 and d = 0.
Here, for i = 4%, L = 4 y1, and a = 100 m1,
1=2
v iK s h1
16, 160 and d = Ks(hsaD)1 = 7.5,
s LD
75 depending on Ks (case 1). Setting L = 2 y1, v = 22,
220 and d = 11, 110 (case 2). Assuming that a = 1000,
Fig. 2. Case 1. Grass critical stability condition for dierent slope gradients i and net inow A. ( ) Experimental data taken from literature. Open
symbols: analytical solutions of the marginal stability condition problem Rek = 0 (12). Water loss due to evaporation and leakage per unit soil
moisture L = 4 y1 and coecient of plant uptake R = 100 m4 kg2 year 1. Soil parameters: saturated conductivity Ks = 3000 m y1 (left) and
Ks = 300 m y1 (right); a = 1000 m1 (n); a = 100 m1 ( ) a = 10 m1 (h).
i = 0 in the lower range of precipitation (between a minimum Amin and the intersection of the marginal stability
condition with the axis i = 0). In this case, instability is
impelled by capillary diusion only (d is maximum when
a is minimum). In general, the inuence of Ks, a and i on
the critical stability condition is more evident for large
Ks (left side of Fig. 2).
In Fig. 2 the analytical model results are plotted with
the critical stability conditions observed by dierent
authors at dierent sites in Africa and Australia (bold
circles in Fig. 2). All the experimental data are referred
by Valentin et al. [24] in a review paper. Unfortunately
the sparse details concerning the soil characteristics reported in the literature references do not allow us to
accurately parameterize our model. At any rate, the
hypothesis of homogeneous soil adopted here for the
sake of simplicity sounds rather unrealistic, and thus it
does not justify a very accurate description of the subsoil, other than in terms of homogeneous, eective
parameters. The eective parameters are meant here to
be those values that somehow match the experimental
evidence and thus make the model soil behavior similar
to the observed one. Indeed, only the processes in action
and the relevant parameters are discussed, whereas the
exact comparison of the experimental data with the theoretical prediction is retained to be a complementary
target of this study. Further, the data shown are taken
from diverse experimental sites, requiring, in principle,
separate interpolations. Many of the experimental data
shown in Fig. 2 match poorly with the theoretical prediction, since the experimental precipitation rate A is
below the theoretical minimum meaningful precipitation: Amin = 2 mr1/2 (see the constraint for w0 and n0
to be real). In order to overcome this limitation, a second case was tackled. R and L have been adjusted maintaining the same ratio LR1 in order to leave the plant
961
Fig. 3. Case 2. Grass critical stability condition for dierent slope gradients i and net inow A. ( ) Experimental data taken from literature. Open
symbols: analytical solutions of the marginal stability condition problem Rek = 0 (12). L = 2 y1; R = 50 m4 kg2 year1. Ks = 3000 m y1 (left) and
Ks = 300 m y1 (right); a = 1000 m1 (n); a = 100 m1 ( ) a = 10 m1 (h).
962
Fig. 4. Case 1. Grass critical wave length for dierent slope gradients i and net inow A. L = 4 y1 and R = 100 m4 kg2 year1. Soil parameters:
saturated conductivity Ks = 3000 m y1 (left) and Ks = 300 m y1 (right); a = 1000 m1 (n); a = 100 m1 ( ) a = 10 m1 (h).
Fig. 5. Case 2. Grass critical wave length for dierent slope gradients i and net inow A. L = 2 y1; R = 50 m4 kg2 year1. Soil parameters:
saturated conductivity Ks = 3000 m y1 (left) and Ks = 300 m y1 (right); a = 1000 m1 (n); a = 100 m1 ( ) a = 10 m1 (h).
ties: the soil saturated conductivity and the ratio between the gravity and the capillary forces. The model
is based on the Klausmeiers [7]. All terms and parameters were set accordingly, except those related to subsurface ow. The model results were compared with
literature data. In order to t the experimental data,
the soil parameters had to be stretched beyond the usual
range. The result suggests that the subsurface processes
that have a relevance for banded vegetation formation
may not have been captured and other relevant processes must be taken into account. At least three terms
within the soil moisture balance should be parameterized with care: net vertical inow, leaching, and lateral
destabilizing ow. These depend on the upper and lower
boundary conditions that do not come into play within
the horizontal soil moisture balance. In this case, the
eective parameters are the upscale keys to the scale of
the soil moisture balances relevant processes taking
place at smaller scales.
References
[1] Ares J, Del Valle H, Bisigato A. Detection of process-related
changes in plant patterns at extended spatial scales during early
dryland desertication. Glob Change Biol 2003;9:164359.
[2] Fernando TM, Cortina J. Spatial patterns of surface soil
properties and vegetation in a Mediterranean semi-arid steppe.
Plant Soil 2002;241:27991.
[3] Galle S, Ehrmann M, Peugeot C. Water balance in a banded
vegetation pattern. A case study of tiger bush in western Niger.
Catena 1999;37:197216.
[4] Gardner WR. Some steady state solutions of the unsaturated soil
moisture ow equation with application to evaporation from a
water table. Soil Sci 1958;4(85):22832.
[5] Gilad E, von Hardenberg J, Provenzale A, Shachak M, Meron E.
Ecosystem engineers: From pattern formation to habitat creation.
Phys Rev Lett 2004;93:098105.
[6] HilleRisLambers R, Rietkerk M, van den Bosch F, Prins HHT, de
Kroon H. Vegetation patterns formation in semi-arid grazing
systems. Ecology 2001;82(1):5061.
963