Date post: | 21-Nov-2023 |
Category: |
Documents |
Upload: | independent |
View: | 0 times |
Download: | 0 times |
Hydrochemistry of runoff and subsurface flow withinSahelian microdunes
O. RIBOLZIa , T. BARIAC
b, A. CASENAVEa, J . P. DELHOUME
a, J . DUCLOUXc & V. VALLES
d
aIRD, BP 182, Ouagadougou 01, Burkina Faso, bLaboratoire de Biogeochimie Isotopique, Universite Pierre-et-Marie-Curie, case 120,
4 place Jussieu, 75252 Paris Cedex 05, cLaboratoire Argiles, Sols et Alterations, CNRS, URA 721, 40 avenue du Recteur Pineau, 86022
Poitiers, and dLaboratoire Chimie et Environnement, Universite de Provence, case 29, 3 place Victor Hugo, 13331 Marseille Cedex 3,
France
Summary
The sandy microdune systems of the Sahel are important for biomass production, in that they trap and
store water. We have studied the movement of water over and in a dune and the chemistry of the water to
understand this aspect of the systems. We experimented with simulated rain using a field sprinkling
infiltrometer. We applied demineralized water with a chemical composition similar to that of the natural
rain on a 1-m2 plot. The plot was delimited by a metallic two-level setting: the first enabled us to collect
surface runoff, while the second measured subsurface flow. Water samples were taken at 5- to 10-minute
intervals throughout each simulation for chemical analysis (alkalinity, SO42–, F–, NO3
–, Ca2þ, Mg2þ, Naþ,
Kþ and Si). Mass balances, combined with a simple mixture model involving one tracer (chloride) and two
reservoirs (old and new waters), were calculated. The equilibrating pressures of the CO2 (pCO2) and the
saturation index with respect to specified minerals (e.g. calcite, fluorite, silicates) were also calculated by the
AQUA ion-pair model. The solute concentrations decrease in surface runoff as well as in subsurface water,
except for F– and Si in the subsurface. The pCO2 decreased to a pressure less than the atmospheric pressure.
The difference between measured concentrations and concentrations computed with the mixing model
highlighted interactions between the soil and water. The dissolution of calcite which consumes CO2, and the
cation exchanges, dominated, whereas the dissolutions of fluorite, silicates and gypsum appear secondary.
Reactive mineral stocks were quickly exhausted, especially in the surface flow.
Introduction
In the Sahelian zone of Burkina Faso, overgrazing, extension
of cultivation and increasing dry climate aggravate erosion. In
this context, sandy microdunes are islets of fertility within
degraded landscapes. They are ecological units where infiltra-
tion of water is significant (Casenave & Valentin, 1992) and
are very important for biomass production (Grouzis, 1991).
These sandy aeolian deposits are potential starting points for
the regeneration of eroded surfaces. Preservation of such units
is therefore crucial in this environment.
The aim of the present work is to understand better the
transport of water and solute in the soil of microdunes. Pre-
liminary studies by Biaou et al. (1999) showed the existence of
surface runoff and lateral subsurface fluxes through these
microdunes during storms, and Ribolzi et al. (2000a) showed
that the mixture of rain water (new water) with the water
already present in the soil (old water) affects the solute
concentrations in the flow. However, this process cannot
wholly explain the variations observed. We may wonder
whether the chemical contents are also affected by inter-
actions between soil and solution together with mixture pro-
cesses.
This kind of problem can be resolved by combining the
mass-balance method (e.g. Plummer & Black, 1980) with a
simple mixing model using environmental tracers. Tracers are
currently widely used to identify and occasionally to quantify
hydrogeochemical processes within catchments. The main
results have been obtained by using isotopic tracers of water
(O18 or H2) (e.g. Sklash & Farvolden, 1979; Ribolzi et al.,
1996). Water can also be traced by its physical (e.g.
Nakamura, 1971; Ribolzi et al., 1997) and chemical characteristics
(e.g. Hooper et al., 1990; Ribolzi et al., 2000b). In all cases, a
mixing model is used to separate the contributions of the
different types of water. However, in some cases the mixing
model used to separate the contributions of the different types
Paper given at the Michel Rieu Memorial Colloquium, 8–10 October
2001, in Paris.
Correspondence: O. Ribolzi. E-mail: [email protected]
Received 7 November 2001; revised version accepted 11 October 2002
European Journal of Soil Science, September 2003, 54, 531–542
# 2003 Blackwell Publishing Ltd 531
of water is difficult to apply, either because it fails to
distinguish clearly the different end members or because there
is a significant temporal variation in the rainfall during the
storm event (e.g. Bazemore et al., 1994).
Materials and methods
Site description
The area we studied is in the north of Burkina Faso, some
13 km from Dori (UTM30, WGS84, 80�980470 0E, 15�5009300N).
It is a degraded fan (general mean slope of 1%) crossed and
overgrazed by livestock. The climate is of the Sahelian type,
with a single rainy season (June–September) each year. Aver-
age annual rainfall recorded in Dori is 512mm, with a mean
maximum of 181mm in August. There is a large year-to-year
variation in rainfall (244mm minimum, 784mm maximum).
Mean annual potential evapotranspiration, calculated by the
method of Penman, is about 2396mm.
Most of the soil at this site is Solonetz (in the FAO termi-
nology) and developed from granitic and amphibolitic rocks.
It shows various types of surface characteristics caused by
water (Karambiri et al., 2003) and wind processes. Wind
deposits lead to the formation of sandy microdunes which
are almost exclusively the support of the herbaceous vegeta-
tion (Figure 1). These dunes extend about 10–100m2 in plan
and are up to 0.7m thick.
For our study we chose a dune composed of two main
horizons (Figure 2a,b). The upper 5 cm comprises a recent
deposit of loose sand with numerous macropores formed by
plant roots and soil fauna. The total porosity of this horizon
ranges between 39 and 47%. The horizon below 13 cm has a
laminated structure alternating between continuous sandy
and plasmic layers. Macroporosity is less than in the upper
horizon. It is also more compact with fewer roots. This
second horizon lies over a massive silty-sandy horizon.
Some of the physico-chemical properties of these horizons
are reported in Table 1. The vegetation (30% of the surface)
is dominated by two grasses, Cenchrus biflorus and Schoene-
feldia gracilis.
Experimental method
We used chloride as tracer. To circumvent the problem of
geochemical variation of natural rain, we used simulated
rain. In that way we could ensure a constant geochemical
signature of the rain (e.g. Turton et al., 1995).
A field sprinkling infiltrometer, designed by Asseline &
Valentin (1978), produced rain on a 1-m2 experimental plot
delimited by a metallic two-level setting (Figure 2a,c). The
first level allows surface runoff to be measured and collected
in a tank equipped with a water level recorder (accuracy of
�5%), while the second collects subsurface flow. Subsurface
Sh
rub
(50
cm)
ero
Figure 1 Photograph of the landscape showing a typical sandy microdune (dotted line) and a crusted surface of the erosion type (ero).
532 O. Ribolzi et al.
# 2003 Blackwell Publishing Ltd, European Journal of Soil Science, 54, 531–542
flow was also collected in a tank, but in this case, it was
estimated through capacity discharge (accuracy of �5%).
The hydraulic potential of the soil was measured in the two
sandy horizons with tensiometers. Moisture was measured
before and after each simulation by a direct gravimetric
method from three samples collected within 1m around the
experimental plot.
Four successive simulations were performed (Figures 3 and
4), all with demineralized water with a chemical composition
similar to that of natural rain. Three of the simulations (Sim1,
Sim2 and Sim4) were at constant intensity. The third one
(Sim3) equalled that of a typical annual frequency storm
(Chevallier et al., 1985); the rainfall event was unimodal with
a gentle initial stage, a high-intensity burst, and a tail of
decreasing intensity. Time intervals between successive sprink-
lings Sim1, Sim2, Sim3 and Sim4 are 24, 37 and 6 hours,
respectively.
The drying crust is very fragile. It was partly destroyed at
the periphery of the plot during the installation of the metal-
lic setting. Therefore it was necessary to reconstitute this
ma.
ro.
pl.
18 c
m
ve.
dry
1 m
1 m
Erosion crust
PVC tube
Metallic two-level settingHerbaceous cover
30 cm
Loose sands
Plasmic lamina
Aeoliandeposits
Silty-sandy horizon
Tanks(buried)
(a)
(b) (c)
Figure 2 Diagram of the soil profile and the border setting (a), a photograph of the soil profile showing plant roots (ro.), biological macropores (ma.)
and a plasmic layer (pl.) (b), and a photograph showing the soil surface of the plot (just before Sim2) with tufts of desiccated grass (ve.) and spots of
drying crust (dry) (c).
Table 1 Some physical and chemical properties of the soil near the experimental plot (see Figure 2)
Particle-size distribution Exchangeable cations
Clay Sif Sic Saf Sac OM N Ca2þ Mg2þ Kþ Naþ CEC
Depth /cm /% /g kg�1 pH (H2O) /mmolc kg�1
0–5 3 2 3 67 26 2 0.1 6.8 9.5 4.5 1.7 0.1 18.3
5–13 6 1 3 38 51 2 0.2 6.7 13.5 5.4 1.0 0.2 23.4
13–15 18 3 4 40 33 4 0.2 6.3 26.6 17.4 1.1 0.7 49.8
Clay< 2�m; Sif, fine silt 2–20�m; Sic, coarse silt 20–50�m; Saf, fine sand 50–200�m; Sac, coarse sand 200–2000�m; OM, organic matter; CEC,
cation exchange capacity.
Hydrochemistry of Sahelian microdunes 533
# 2003 Blackwell Publishing Ltd, European Journal of Soil Science, 54, 531–542
drying crust with preliminary simulated rain (Sim1), fol-
lowed by a desiccation phase. Sim1 was made with intense
rainfall (almost 120mmhour�1) to accelerate the reorganiza-
tion of the soil surface (Figure 3). This complete cycle (rain
þ desiccation) totally reconstructed the drying crust, as
shown on the photograph in Figure 2(c), taken 23 hours
after Sim1.
Water sampling and analyses
Samples of water were taken at 5- to 10-minute intervals
throughout each simulation for chemical analysis. A cumu-
lated sample of rain was collected at the end of each applica-
tion in a rain gauge near the plot.
Temperature, electrical conductivity (accuracy of �0.5%
of the measured value) and pH were measured in the field.
After microfiltration at 0.2 �m, samples collected during the
second application (Sim2) were stored in the dark in poly-
ethylene flasks. In the laboratory total calcium, magnesium,
sodium, potassium, silicon and sulphur concentrations were
measured by inductively coupled plasma spectrometry.
Chloride and nitrate ions were analysed by automated
colorimetry, and fluoride was determined by an ion-specific
electrode. All these chemical compounds were determined
by the CIRAD-AMI-AGROMIE laboratory at Montpellier
(France), certified by AFAQ (ISO 9002). Total alkalinity
(Alk) was measured by the method of Gran (1952) by the
IRD laboratory at Ouagadougou (Burkina Faso). Detection
limits and analytical accuracies reported in Table 2 were
determined following the standard French Normative proced-
ure NFX-06-044 of 1984.
Mass-balance and uncertainty calculations
Decomposing a flow from tracer data is equivalent to solving,
at each outlet sampling time, a system of n mixing equations
with n unknown variables (the various contributions to the
flow). Assume that two types of water contribute to flow
0
100
200
300
0 10 20 30 40
Suc
tion
/hP
a
0
20
40
60
0 10 20 30 40 50 60
Time /minutes
Sim2
0
40
80
120
0 10 20 30 40 50 60
0
100
200
300
0 10 20 30 40 50 60
Sim1
0
40
80
120
0 10 20 30 40
Inte
nsity
/mm
hou
r –1
0
20
40
60
0 10 20 30 40
Time /minutes
EC
/µS
cm
–1
Subsurface flow Surface runoff
5 cm deep 15 cm deep
Rain Surface runoffInfiltration Subsurface flow
Figure 3 Variation in the time of intensities,
suctions and electrical conductivity (EC)
during the first two simulations with
demineralized water (Sim1 and Sim2).
534 O. Ribolzi et al.
# 2003 Blackwell Publishing Ltd, European Journal of Soil Science, 54, 531–542
discharge: old water (pre-event water) and new water (event
water). Also assume that in each type of water there is a tracer
of constant concentration but significantly different from that
in the other. Then solving the system of equations leads to the
mixing ratio (Rm):
Rm ¼Ct
c;outlet � Cc;new
Cc;old � Cc;new; ð1Þ
where Ctc,outlet is the concentration of the conservative species
c at the outlet at time t, Cc,new is the concentration of the
conservative species in the new water, and Cc,old is the con-
centration of the conservative species in the old water. The
quantity Rm, which ranges from 0 to 1, gives the fraction of old
water.
Mass-balance calculations quantify the mass of element
and associated mineral or gas phase that has reacted along a
flow path. The mass-balance calculation used along flow
paths A and B relies on Rm (Plummer & Black, 1980).
From the Rm of the successive surface or subsurface samples
collected during the simulated rain, concentrations of
Table 2 Detection limits and analytical inaccuracies determined fol-
lowing the standard procedure NFX-06-044 (in �M)
Analytical uncertainty Detection limit
Chloride � 20%, <30; �3%, >30 3
Sulphate � 10%, <10; �3%, >10 2
Nitrate � 20%, <20; �3%, >20 1
Fluoride � 5%, <50; �1%, >50 1
Alkalinity � 3%, <20; �1%, >20 1
Calcium � 8%, <25; �3%, >25 1
Magnesium � 10%, <40; �1%, >40 2
Sodium � 14%, <40; �3%, >40 9
Potassium � 13%, <30; �3%, >30 10
Silicon � 8%, <50;� 3%, >50 2
0
20
40
60
0 10 20 30 40 50
Time /minutes
EC
/µS
cm
–1
Sim3
0
40
80
120
0 10 20 30 40 50
Inte
nsity
/mm
hou
r–1
Sim4
0
40
80
120
0 10 20 30 40
0
100
200
300
0 10 20 30 400
100
200
300
0 10 20 30 40 50
Suc
tion
/hP
a
0
20
40
60
0 10 20 30 40
Time /minutes
Rain Surface runoffInfiltration Subsurface flow
5 cm deep 15 cm deep
Subsurface flow Surface runoff
Figure 4 Variation in the time of intensities,
suctions and electrical conductivity (EC) during
the last two simulations with demineralized
water (Sim3 and Sim4).
Hydrochemistry of Sahelian microdunes 535
# 2003 Blackwell Publishing Ltd, European Journal of Soil Science, 54, 531–542
alkalinity, SO42–, F–, Ca2þ, Mg2þ, Naþ, Kþ and Si were
estimated by
Eti;outlet ¼ Rm Ci;old þ ð1� RmÞCi;new; ð2Þ
and
�ti ¼ Ct
i;outlet � Eti;outlet; ð3Þ
where Eti,outlet is the predicted concentration of a compound
i at the outlet at time t, based on conservative mixing, and
Cti,outlet is the measured concentration. The mass-balance
model assumed that the differences ð�tiÞ between the observed
and the predicted concentrations, Equation (3), can be attrib-
uted to geochemical reactions. These differences are termed
reaction-derived quantities of a compound. Negative reaction-
derived quantities are taken to represent a loss of element from
solution due to mineral precipitation, ion exchange for the
absorbing ion, or loss of CO2; positive ones are taken to repre-
sent dissolution of minerals, ion exchange for the desorbing
ion, or gain of CO2.
A previous study with 18O as tracer demonstrated that sub-
surface water is exclusively composed of old water at the
beginning of the flow (Ribolzi et al., 2000a). Consequently,
here we supposed that the signature of old water is that of
the first sample collected in the subsurface flow. The composi-
tion of new water is that of the accumulated rain sample
collected in the rain gauge at the end of the simulation.
We adopted the approach proposed by Genereux (1998) to
quantify the uncertainties of Rm and �ti . Each known param-
eter Ctc,outlet, Cc,new, Cc,old, C
ti,outlet, Ci,new and Ci,new of Equa-
tions (1), (2) and (3) was assigned an average error calculated
with analytical inaccuracies (Table 2).
Saturation index of the solutions
Field measurements of pH, temperature, and the concentra-
tions of alkalinity, Cl–, SO42–, F–, Ca2þ, Mg2þ, Naþ, Kþ and
Si were input into the AQUA software ion-pair model. This
model is derived from the GYPSOL program of Valles (1987),
based on the Debye–Huckel law including the deviation func-
tion of Scatchard. This geochemical program accounts for
speciation within the aqueous phase, and evaluated the equi-
librating pressures of the CO2 (pCO2). The pCO2 was estimated
from the pH measured in the field and the carbonate alkalinity
of each solution (Bourrie, 1976).
The program also calculates the saturation index (SM) of the
solutions with respect to specified minerals, M. This index is
defined as
SM ¼ logQ
KTM
� �; ð4Þ
where Q is the ion activity product and KTM is the equilibrium
constant for the mineral dissolution reaction at water tempera-
ture, T. It indicates oversaturation (positive values) and
undersaturation (negative values).
Results and discussion
Hydrodynamic behaviour
Surface runoff occurred after a 5-minute lag in the case of
constant intensity simulations (Sim2 and Sim4), and after
10minutes in the case of the variable intensity simulation
(Sim3). In the first instance (with rain falling at
90mmhour�1), surface runoff increased rapidly during the
first 5–10minutes, then remained constant at around
75mmhour�1 (Figures 3 and 4). Note that the volumetric
water content increases from 10 to 21% (saturation point)
and 15 to 21%, respectively, during simulations 2 and 4.
With regard to Sim3 (daily rainfall of annual frequency), the
lag time was longer on account of the less intense rain at the
beginning of the simulation. Then, surface runoff intensity
changed with changes in the intensity of the rain (Figure 4).
In this case, mean soil moisture increases from 12 to 18%.
To understand hydrodynamic behaviour of our plot, the
chloride ion is useful because it is a conservative tracer that
enters the soil with the rain. In simulation Sim2 the chloride
content of the surface runoff appeared somewhat larger than
that of the rain (Table 3). This discrepancy would have been
much larger if chloride minerals had been present in the soil,
because the solubility of such minerals is important. Therefore,
this discrepancy could be explained better by the mixing of the
rain (new water) with a small proportion of water already
present in the soil before the rain (old water). The concentra-
tion of chloride in this old water is greater than in the new
water because the evaporation phase before Sim2 has
increased the chloride concentration in the soil. Consequently,
we have decided to use chloride as a tracer to calculate Rm
(Figure 5).
Thus, the values of Rm indicate that surface runoff was not
strictly due to excess rain. The occurrence of old water in this
flow can be explained by the surface roughness of the plot. The
surface consists of micromounds made up of wind deposits
trapped in tufts of grass (Figure 2a,b). The latter allow for
infiltration and preferential storage of water close to the sur-
face. During rain, a proportion of this water may flow out at
the basis of the micromound, and thus become part of the
surface runoff.
Subsurface flow represents 0, 7.0, 2.7 and 5.5% of total flow
(surfaceþ subsurface) in simulations 1, 2, 3 and 4, respectively.
This proportion varied during the simulations 2, 3 and 4. In
Sim2 the contribution of subsurface flow was 30% at the early
stage of the simulation, then decreased quickly and became
constant at about 5.5%. In the third simulation (variable rain
intensity), subsurface flow occurred 15minutes after the begin-
ning of the simulation (Figure 4), that is 10minutes after the
complete saturation of the first horizon with water (suction at
5 cm close to 0). When the soil surface was saturated with
water, 36mmhour�1 rain was not intense enough to release
subsurface flow (Figure 4). The latter occurred 5minutes after
rain falling at 120mmhour�1, and persisted for the intensities
536 O. Ribolzi et al.
# 2003 Blackwell Publishing Ltd, European Journal of Soil Science, 54, 531–542
of 90 and 45mmhour�1. Given that subsurface flow stops at
the intensity of 36mmhour�1, it follows that the threshold at
which this phenomenon occurred ranges between 36 and
45mmhour�1.
Surface runoff and subsurface flow started simultaneously
during Sim2 and Sim4 (constant rainfall intensity), that is
5minutes after the beginning of each application (Figures 3
and 4). Subsurface flow increased during the first 10minutes,
and then it stabilized at close to 5mmhour�1. The rapid
response of this flow is typical of a piston effect. This process
has already been explained by using artificial isotopic and
chemical tracers (Ribolzi et al., 2000a): the new water infil-
trates the soil, saturates the upper horizon, and displaces the
old water there, either vertically or laterally. Old and new
waters become intermixed during the simulation: initially the
flow contains almost 100% of old water, but the proportion of
new water keeps increasing.
This hypothesis is confirmed by the observations presented
above. Thus, the chloride content decreased rapidly through-
out the simulation (Table 3). By the end of the simulation, a
small fraction of old water remained in the subsurface flow as
shown by the chloride concentrations in it, which does not
quite reach that of the new water (Table 3). This discrepancy is
probably due to the persistence of a small fraction of immobile
old water in the soil (Angulo-Jaramillo et al., 1996).
Biogeochemical processes
As concluded previously, runoff and subsurface flows both
contain a small proportion of old water varying throughout
simulations. However, this mixing process alone cannot
entirely explain all the chemical variations observed in
those flows. As show by Ribolzi et al. (2000a) the interac-
tions between soil and solution have to be taken into account
too. Therefore we decided to study in detail those interac-
tions using the mass-balance approach of Plummer & Black
(1980).
The waters collected were much less concentrated, their cat-
ionic charge ranging between 22 and 271�M. The subsurface
waters were dominated by Ca2þ and alkalinity with a significant
content in monovalent cations (Kþ and Naþ) (Table 3), whereas
the surface waters were dominated by Kþ and alkalinity.
Table 3 Chemical composition of waters sampled during the second simulation (Sim2)
Time Si Cl– SO42– NO3
– Alkalinity F– Ca2þ Mg2þ Naþ Kþ
Type of water /minutes pH /�M
Demineralized water 0 7.6 39 11 " " 17 " 3 2 10 1
Surface runoff 4.0 6.2 61 29 5 1 127 4 12 11 13 110
9.0 6.4 – 15 3 " 88 2 5 6 15 6
19.9 7.3 56 18 2 " 42 " 5 6 12 4
25.0 7.8 46 16 2 " 50 " 3 4 12 4
34.8 7.2 45 16 2 " 37 " 6 3 12 3
54.1 7.4 48 16 3 " 30 " 6 4 10 2
Subsurface flow 8.7 6.5 95 115 21 6 335 3 109 67 81 14
12.7 6.7 108 89 16 1 326 5 103 62 77 12
22.3 6.9 175 56 8 " 240 10 74 41 75 7
33.9 6.9 165 23 5 " 174 13 48 30 45 5
58.6 6.9 202 35 3 " 143 15 36 24 45 5
", below detection limit; –, not determined.
0
0.1
0.2
0.3
0 10 20 30 40 50 60
Surface runoff
0
0.2
0.4
0.6
0.8
1.0
0 10 20 30 40 50 60
Time /minutes
Subsurface flowMix
ing
ratio
, Rm
Figure 5 Variations in time of mixing ratio, Rm, in surface runoff and
subsurface flow based on chloride concentration. Error bars
determined by the method of Genereux (1998).
Hydrochemistry of Sahelian microdunes 537
# 2003 Blackwell Publishing Ltd, European Journal of Soil Science, 54, 531–542
Thermodynamic calculations show them to be undersatur-
ated with respect to minerals such as calcite, fluorite, gypsum
and amorphous silica (Figure 6), and in equilibrium with regard
to silicates such as potassium feldspar and phyllosilicates.
During the experimentation, the concentrations of solutes
generally decreased, in both the surface and the subsurface
(Table 3). This decrease could be the consequence of progres-
sive dilution of the old water by the new water. In the same
way, the pCO2 decreased steadily, in both the surface and the
subsurface, reaching values less than that of the atmosphere
(Figure 6). Nevertheless, the increased concentrations of silica
and fluoride in the subsurface emphasize the existence of
chemical interactions between the solid phases of soil and its
solution. It is therefore necessary to distinguish between the
mixture and the reactive processes during the transfer of
solutes.
Calculations of reaction-derived quantities, with Equations
(2) and (3) and chloride as tracer, show that solutes have
different behaviours (Figures 7 and 8). At the surface, the
soil releases elements, especially at the beginning of the flow,
then quickly the balance is nil and the transfer of solute
becomes non-reactive. Below the surface, interactions between
soil and water modify the concentrations of the various ele-
ments significantly until the end of the flow. Here the trans-
port is reactive. Fluoride is the only exception (Figure 7). The
calculations show a clear and long-lasting release of this ion by
the soil in subsurface. At the surface, the release was less and
shorter-lived. This mechanism can be the result of the dissolu-
tion of fluorite, already observed in the Sahelian soils
(Barbiero et al., 1995). The undersaturation of the waters in
relation to this mineral confirms this hypothesis (Figure 6).
This dissolution also releases calcium. In the same way, the
release of sulphate may be attributed to the dissolution of
traces of gypsum contained in the aerosols from the northern
Sahara.
At the surface, the soil releases a small quantity of potas-
sium and no sodium, whereas in the subsurface the water is
enriched with these two elements by the soil (Figures 7 and
8). The presence of Naþ and Kþ can hardly be attributed to
the dissolution of the feldspars which would be much too
slow in relation to the duration of the experiment. More-
over, the Kþ was released before the Naþ, and that is not
compatible with the solubilities of the potassic and sodic
feldspars. The ionic exchanges, usually considered to be
rapid, must be taken into consideration, and this implies
the adsorption of divalent ions, especially calcium, in
equivalent quantity.
Nevertheless, the fixation of the calcium resulting in the
ionic exchanges and in the dissolution of fluorite and of
gypsum does not correspond to the balance observed for this
element. The discrepancy can be attributed to the dissolution
of calcite. As it consumes CO2, this dissolution explains the
very small pCO2 in the surface runoff (Figure 6). It also
explains the alkalinity enrichment of the subsurface waters.
Moreover, the dissolution of calcite is possible because of the
undersaturation of the water with respect to the mineral
(Figure 6). Carbonate coats some of the grit (size> 2mm) in
the sandy deposit. Traces of calcite are also present in the soil
matrix, but they are undetectable by X-ray analyses. The
carbonate comes from the nearby erosion surface and in dust
from the Sahara.
–16 –12 –8 –4
2log(F–)
–5.5
–5.0
–4.5
–4.0
log(Ca2+)
+
+ +
+
o o o
o o
+
Fluorite
–4.0 –3.5 –3.0 –2.5
log(pCO2)
8
10
12
14
log(Ca2+) + 2pH + log(H2O)
+ +
+
+
+ +
o o
o o o
Atmosphere Calcite
oSurface runoff Subsurface flow
(b)
(a)
Figure 6 Saturation diagrams with respect to (a) fluorite and (b)
calcite.
538 O. Ribolzi et al.
# 2003 Blackwell Publishing Ltd, European Journal of Soil Science, 54, 531–542
The release of a large quantity of silica suggests the dissolu-
tion of siliceous minerals. It can neither be quartz, because the
waters are oversaturated (Figure 6), nor feldspars for the
kinetic reasons mentioned above. It is probably caused by
the dissolution of phyllosilicates, but the large excess of the
silica balance cannot be justified by it. The dissolution of
amorphous silica must be considered.
The quantification of the interactions between the solutes
and the soil matrix allows us to classify these mechanisms
according to their importance (Table 4). The dissolution of
both amorphous silica and calcite appear to be the major
mechanisms. The ionic exchanges Ca2þ with the monovalent
cations Naþ and Kþ and the dissolution of silicates are of
secondary importance. The dissolution of fluorite, although
clear, is a phenomenon of third importance, as is the dissolu-
tion of sulphate-rich aerosols. In both cases, the quantities of
fluorite and gypsum dissolved are small and quickly
exhausted.
Conclusions
The analyses of alkalinity, Cl–, SO42–, F–, Ca2þ, Mg2þ, Naþ,
Kþ and Si allow us to describe the major hydrogeochemical
processes that occurred inside microdune soils during rain.
Our main results can be summarized in three points.
1 The contribution of subsurface flow to the total water flow
(surface plus subsurface) was of secondary importance (about
5%). However, the proportion of solutes transported by the
subsurface flow was far from negligible (more than 20%).
30
20
10
∆Sili
con
/µM
0
0 10 20 30 40 50 60–10
10
5
0
∆Sul
phat
e /µ
M
0 10 20 30 40 50 60–5
120
80
40∆A
lkal
inity
/µM
0
0 10 20 30 40 50 60–40
6
3
0
∆Flu
orid
e /µ
M
0 10 20 30 40 50 60–3
20
10
0
∆Cal
cium
/µM
–10
0 10 20 30 40 50 60–20
20
0
10
–10
∆Mag
nesi
um /µ
M0 10 20 30 40 50 60
–20
5
10
0
–5
∆Sod
ium
/µM
–10
0 10 20
Time /minutes Time /minutes
30 40 50 60–15
100
40
60
80
20
0
∆Pot
assi
um /µ
M
0 10 20 30 40 50 60
Figure 7 Variations in time of reaction-
derived quantities, �i, Equation (3), of the
major chemical compounds in surface runoff.
Error bars determined using the method of
Genereux (1998).
Hydrochemistry of Sahelian microdunes 539
# 2003 Blackwell Publishing Ltd, European Journal of Soil Science, 54, 531–542
Thus, subsurface water is always more concentrated than
runoff.
2 Both surface and subsurface chemical concentrations
decrease during rainfall simulation.
3 This decreasing trend is explained by soil–solution interac-
tions and by the mixing of old water (pre-event water) with
new water (demineralized event water).
225
150
75
∆Sili
con
/µM
0
0 10 20 30 40 50 60–75
20
10
0
–10
–20
∆Sul
phat
e /µ
M
0 10 20 30 40 50 60
200
100
∆Alk
alin
ity /µ
M
0
0 10 20 30 40 50 60–100
20
15
10
5
0
∆Flu
orid
e /µ
M
0 10 20 30 40 50 60
100
50
∆Cal
cium
/µM
0
0 10 20 30 40 50 60–50
50
100
0
∆Mag
nesi
um /µ
M
0 10 20 30 40 50 60–50
80
40
0
∆Sod
ium
/µM
0 10Time /minutes Time /minutes
20 30 40 50 60–40
75
50
25
–25
–50
0
∆Pot
assi
um /µ
M
0 10 20 30 40 50 60
Figure 8 Variations in time of reaction-derived
quantities, �i, Equation (3), of the major
chemical compounds in subsurface flow.
Error bars determined using the method of
Genereux (1998).
Table 4 Total amount of dissolved minerals (amorphous silica, calcite, fluorite and gypsum) and exchanged cations within surface runoff and
subsurface flow during the second simulation (Sim2)
Amorphous silica Calcite Fluorite Gypsum Exchanged Ca2þ
Type of water /�mol
Surface runoff 729 717 25 516 1691
Subsurface flow 1594 616 83 29 338
540 O. Ribolzi et al.
# 2003 Blackwell Publishing Ltd, European Journal of Soil Science, 54, 531–542
The results, though obtained in controlled experimental
conditions, accord with those of Biaou et al. (1999) for runoff
water at the outlet of a small catchment within the same study
area. The concentrations and tendencies were similar.
The mass-balance method highlighted points related to bio-
geochemical processes.
1 The chemical reactivity of the surface runoff is significantly
different from that of the subsurface flow.
2 Microdune soils appear to be reactive, with kinetic processes
generally rapid, as in the case of ionic exchanges.
3 Dissolution reactions are limited in intensity and in time
because the mineral stocks were exhausted during the simu-
lated rainfall. These reactions probably mobilize either efflor-
escences or less reactive minerals included in impermeable
aggregates.
4 The exchanges between calcium and monovalent cations on
the clay–humus complex are evidently important.
Our results contribute to a better understanding of the
abiotic relationships in pastoral Sahelian environments.
They emphasize particularly the contribution of runoff and
subsurface flow to the lateral redistribution of nutrients
within sandy aeolian deposits. These lateral transfers take
place from the heart of the microdunes, covered by vegetation
and enriched in organic matter, to the recent sandy deposits
around. This process improves the fertility of the peripheral
surfaces of the microdunes, and thus could contribute to the
regeneration of the soil if it were protected from overgra-
zing.
Finally, our data could be used to evaluate the predictive
capability of an existing transport model either using a tracer
in a water flow model or using a combined water flow reactive
transport model.
Acknowledgements
This work was supported by the Unite de Recherche 049
‘ECU’ of the Institut de Recherche pour le Developpement
(IRD), and the INSU Programme National Sol Erosion
(PNSE) project no 99/44. We acknowledge with gratitude the
Institut National de l’Environnement et de la Recherche
Agricole (INERA) of Burkina Faso for providing access to
the site. We also thank the team of the hydrological laboratory
of IRD in Ouagadougou, with a special mention for Barry
Moussa and Tou Boureima for their help in the field. Finally,
we thank the referees and the Editor for their constructive
contributions.
References
Angulo-Jaramillo, R., Gaudet, J.P., Thony, J.L. & Vauclin, M. 1996.
Measurement of hydraulic properties and mobile water content of a
field soil. Soil Science Society of America Journal, 60, 710–715.
Asseline, J. & Valentin, C. 1978. Construction et mise au point d’un
infiltrometre a aspersion. Cahiers de l’ORSTOM, Serie Hydrologie,
15, 321–343.
Barbiero, L., Valles, V. & Regeard, A. 1995. Precipitation de la
fluorine et controle geochimique du calcium dans les sols alcalins
du Niger: consequences pour une estimation quantitative de l’evolu-
tion geochimique des sols. Comptes Rendus de l’Academie des
Sciences de Paris, 321, 1147–1154.
Bazemore, D.E., Eshleman, K.N. & Hollenbeck, K.J. 1994. The role of
soil water in stormflow generation in a forested headwater catch-
ment: synthesis of natural tracer and hydrometric evidence. Journal
of Hydrology, 162, 47–75.
Biaou, A.C., Casenave, A., Delhoume, J.P., Gathelier, R.,
Thiombiano, L., Gineste, P. & Ribolzi, O. 1999. Erosion hydrique
et transferts de solutes en milieu sahelien: etude des processus a
l’echelle d’un petit bassin versant au Nord du Burkina Faso. Sud
Sciences et Techniques, 4, 18–28.
Bourrie, G. 1976. Relation entre le pH, l’alcalinite, le pouvoir tampon
et les equilibres de CO2 dans les eaux naturelles. Science du Sol, 3,
141–159.
Casenave, A. & Valentin, C. 1992. A runoff capability classification
system based on surface features criteria in semi-arid areas of West
Africa. Journal of Hydrology, 130, 231–249.
Chevallier, P., Claude, J., Pouyaud, B. & Bernard, A. 1985. Pluies et
crues au Sahel: Hydrologie de la mare d’Oursi (Burkina Faso)
(1976–1881). Travaux et documents 190, IRD, Paris.
Genereux, D. 1998. Quantifying uncertainty in tracer-based hydro-
graph separations. Water Resources Research, 34, 915–919.
Gran, G. 1952. Determination of the equivalence point in poten-
tiometric titration, Part II. The Analyst, 77, 661–671.
Grouzis, M. 1991. La region de la mare d’Oursi: un milieu sahelien. In:
Un espace Sahelien: La mare d’Oursi (eds J. Claude, M. Grouzis &
P. Milleville), pp. 43–49. IRD, Paris.
Hooper, R.P., Christophersen, N. & Peters, N.E. 1990. Modelling
streamwater chemistry as a mixture of soil water end-members: an
application to the Panola mountain catchment, Georgia, USA.
Journal of Hydrology, 116, 321–343.
Karambiri, H., Ribolzi, O., Delhoume, J.P., Ducloux, J., Coudrain-
Ribstein, A. & Casenave, A. 2003. Importance of soil surface
characteristics on water erosion in a small grazed Sahelian catch-
ment. Hydrological Processes, 17, 1495–1507.
Nakamura, R. 1971. Runoff analysis by electrical conductance of
water. Journal of Hydrology, 14, 197–212.
Plummer, L.N. & Black, W. 1980. The mass balance approach:
application to interpreting the chemical evolution of hydrologic
systems. American Journal of Science, 280, 130–142.
Ribolzi, O., Valles, V. & Bariac, T. 1996. Comparison of hydrograph
deconvolutions using residual alkalinity, chloride and oxygen 18 as
hydrochemical tracers. Water Resources Research, 32, 1051–1059.
Ribolzi, O., Moussa, R., Gaudu, J.C., Valles, V. & Voltz, M. 1997.
Etude des crues de transition entre periode seche et periode humide
par tracage naturel sur un bassin versant mediterraneen cultive.
Comptes Rendus de l’Academie des Sciences de Paris, 324, 985–992.
Ribolzi, O., Auque, L., Bariac, T., Casenave, A., Delhoume, J.P.,
Gathelier, R. & Pot, V. 2000a. Ecoulements hypodermiques et
transferts de solutes dans les placages eoliens du Sahel: etude par
tracage isotopique et chimique sous pluies simulees. Comptes
Rendus de l’Academie des Sciences de Paris, 330, 53–60.
Hydrochemistry of Sahelian microdunes 541
# 2003 Blackwell Publishing Ltd, European Journal of Soil Science, 54, 531–542
Ribolzi, O., Andrieux, P., Valles, V., Bouzigues, R., Bariac, T. & Voltz,
M. 2000b. Contribution of groundwater and overland flows to storm
flow generation in a cultivated Mediterranean catchment: quantifica-
tion by natural chemical tracing. Journal of Hydrology, 233, 241–257.
Sklash, M.G. & Farvolden, R.N. 1979. The role of groundwater in
storm runoff. Journal of Hydrology, 43, 45–65.
Turton, D.J., Barnes, D.R. & Navar, J.D. 1995. Old and new water in
subsurface flow from a forest soil block. Journal of Environmental
Quality, 24, 139–146.
Valles, V. 1987. Modelisation des transferts d’eau et de sels dans un sol
argileux: application au calcul des doses d’irrigation. Memoire des
Sciences Geologiques, 79, 1–148.
542 O. Ribolzi et al.
# 2003 Blackwell Publishing Ltd, European Journal of Soil Science, 54, 531–542