M
AGRICULTURAL
ELSEVIER
Agricultural and Forest Meteorology 83 (1997) 49-74
AND
FOREST
METEOROLOGY
An approach to couple vegetation functioning and
soil-vegetation-atmosphere-transfer models for
semiarid grasslands during the HAPEX-Sahel
D. Lo Seen a'*, Á.[Chehbouni b, E. Njoku a, S. Saatchi a,
E. Mougin ', B.,Monteny
15f-W.
Jet Propulsion Laborarory, 4800, Oak Groue Driue, Pasadena, CA 91109-8099,US4
ORSTOM- Laboratoire d'hydrologie, 911, avenue Agropolis, BP 5054,34032 Montpellier Cedex, France
CESBIO/ CNES, 18, auenue Edouard Belin, 3105 Toulouse Cedex, France
a
Received 18 September 1995; revised 7 March 1996; accepted 27 March 1996
Abstract
This paper presents a model which has been developed to simulate the major land surface
processes occurring in and and semiarid grasslands. The model is composed of a hydrological
submodel which describes the water and energy budgets, and a vegetation growth submodel which
groups the processes associated with biomass production. Emphasis has been placed on developing a realistic representation of the interaction between these subprocesses taking account of the
different time scales involved. The hydrological submodel couples the energy balance of the
soil/canopy with the soil moisture and thermal dynamics. It interacts with the vegetation growth
submodel by exchanging information needed to account for the influence of plant water status and
canopy temperature on photosynthesis, and the influence of the vegetation canopy on the boundary
layer within which transport processes are taking place. The model has been tested with
meteorological, biomass and energy flux measurements made on a grassland site during the
HAPEX-Sahel experiment, Niger, in 1992. Model simulations of biomass over the growing
season are all found to be within a 15% error margin allowed on biomass measurements. Hourly
values of net radiation, as well as latent and sensible heat fluxes, are simulated with an RMSE of
less than 50 W m-2. Given the relative simplicity of the model and the long period of
uninterrupted simulation, these results are considered satisfactory. Overall, the results show that
* Corresponding author. Present address: CIRAD-Forêt, CIRAD/Maison de la Télédétection, 500 Rue J.F.
Breton, 34093 Montpellier Cedex 5,France.
0168-1923/97/$17.00 Copyright O 1997 Elsevier Science B.V. All rights reserved.
PII SO 168-1 923(96)02350-7
50
D.Lo Seen er al./Agricultural ancl Forest Meteorology 83 (1997)49-74
the model behaves consistently at different stages of vegetation growth, and satisfactorily
reproduces the interdependence of vegetation growth with the physical processes giving rise to the
water and energy balances.
1. Introduction
Interconnected land surface processes control the energy, water and carbon transfers
between soil, vegetation and the atmosphere. Several studies have put emphasis upon the
necessity to correctly describe the land surface processes while investigating the
interactions between land surface and regional climate (Wilson et al., 1987; Pielke and
Avissar, 1990; Raupach, 1991). Soil-vegetation-atmosphere-transfer (SVAT) models
have therefore been developed to simulate these mass and energy transfers for different
land surfaces and vegetation types (e.g. forests, grasslands). But more than anywhere
else, the need for an accurate description of the processes is crucial in arid and semiarid
regions where neither the soil nor the vegetation dominates the exchange of heat and
water with the atmosphere. Moreover, the land surface itself may change considerably
during a year, like for example in the Sahel where the predominance of annual grasses
makes the surface vary from bare soil during the dry season, to a sometimes luxuriant
vegetated surface during the rainy season.
The model presented in this paper seeks to describe the major land surface processes
occurring in.semiarid grasslands as typically found in the Sahel. Unlike other SVAT
models which normally consider vegetation to be invariant in time, the model couples
vegetation growth with the heat and moisture dynamics in the soil and the sparse
canopy. From the point of view of the ecosystem modeling of primary productivity, the
model uses an improved description of the water and energy budgets. The model
simulates Sahelian grassland vegetation growth, evapotranspiration, sensible heat flux
and the evolution of surface temperature and soil moisture during the growing season. It
can be considered as a compromise between the need for a simplified representation of
the processes, while retaining the main mechanisms which ensure a realistic simulation
of these processes. After description of the land surface process model, the paper deals
with the application of the model to a grassland site of the HAPEX-Sahel experiment,
Niger, (Goutorbe et al., 1993; Prince et al., 1995) for which biomass, vegetation height
and energy fluxes measurements have been made, and for which meteorological and
precipitation data necessary to run the model have been recorded during the growing
season of 1992.
2. Model description
The model can be schematically presented as two interacting submodels (see Fig. 1).
The first submodel groups the hydrological processes such as soil water and thermal
dynamics, coupled with the water and energy balances in a sparse canopy. The second
submodel groups the main processes related to vegetation growth, such as photosynthesis, respiration and biomass production. The hydrological submodel runs with an hour
SURFACE PROCESSES MODEL
I
Meteo. data
(air temp., humidify,
wind speed, in. solar
radiation)
Rainfalldata
I
Soil fiermal and
hydraulic properties
soil texture
(%clay, %sand)
l
%C3/C4
I
-
e
OUTPUT
.
.Water and heat
fluxes
.Soil-moisture
.Temperature of
-ground
-canopy
.Aerodynamic
and radiative
temperatures
. Biomass
.Net primary
productivity
.Photosynthesis
.Respiration
I
Fig. I. Diagram showing the interaction between the vegetation growth submodel and the water and energy balance submodel, by the exchange of the necessary
variables. The major input data needed, as well as the main processes and surface variables simulated, are also shown.
D.Lo Seen et al./Agricultural and Foresr Meteorology 83 (1997)49-74
52
time step, whereas the vegetation growth submodel has a daily time step. The interaction
between the two submodels, and the manner in which the difference in time steps is
dealt with, are described later in this section. The notation used in this paper is listed in
Appendix A.
2.1. Energy balance equations
The Sahelian grassland considered here, like most arid and semiarid vegetation, is
typically sparse vegetation in which sensible and latent heat originate both from the
canopy and the soil in comparable amounts. This energy partitioning gradually changes
throughout the growing season as the vegetation grows, along with the recharge and
depletion of soil moisture in the system following rain events. The model uses the
one-dimensional two-layer approach proposed by Shuttleworth and Wallace (1985),
which considers the soil and the vegetation as separate sources (or sinks) of latent and
sensible heat. The approach has already been evaluated for sparse crops (e.g. Ham and
Heilman, 1991) and sparse natural vegetation (e.g. Nichols, 1992). As the herbaceous
vegetation present is exclusively composed of annuals, a model simulation carried out
for the growing season should also account for the transition between bare soil (before
the first rains, usuaIly in June) and fully grown vegetation (occurring around September).
The general two-layer model is briefly presented here, knowing that it can be simplified
to represent both the case of bare soil and that of closed canopy.
The net radiation available above the canopy is partitioned between the soil and the
canopy, such that:
R,
=R,,
(1)
+ R,,c
An energy budget written separately for the soil and the canopy gives:
R,,, - AE, - H, - G = O
(2)
- AE, - H , = O
(3)
and
R,,
The model assumes the existence of a mean air-flow at a ‘source level’ within the
canopy. Energy exchanges are then considered to occur between the soil surface, the
canopy, the within-canopy source level and an above-canopy reference level. Fig. 2
shows the network of resistances across which differences in vapor pressure and
temperature give the fluxes of latent (Eqs. (4)-(6))and sensible heat (Eqs. (714911, by
analogy to Ohm’s law.
AE=-
PCp ( e o -
Y
raa
ea)
(4)
b
s
k?
9
%
n
7
\
----
e,
Mean canopy flow
source height
>
2-.
E
c
E"
P
k21
2
2
T2
w2
3
8a
F
B
Fig. 2. Diagram showing the network of resisfances (from Shultleworlh and Wallace, 1985) between the soil surface, the within-canopy source level and the
above-canopy reference level, across which differences in vapor pressure and temperature give latent and sensible heat fluxes. The surface and root layers are also
shown with their corresponding temperature and moislure confent.
8
2
u
2
L
A
I
\o
u
A
u
W
l
D.Lo Seen et aL/Agricubural
54
and Forest Meteorology 83 (1997)49-74
Substituting Eqs. (4)-(6) in the conservation of latent heat of the canopy air flow gives
and similarly, Eqs. (7)-(9) with the conservation of sensible heat give
To=
+ raaracT, + racrasT,
rac ras + ‘,aras f rac raa
raarasT,
-
The way the foliage prevents the radiation from reaching the ground, is expressed by
the quantity wf (the shielding factor used by Deardroff, 1978; Taconet et a1.,-1986; Ben
Mehrez, 1990). Assuming no heat storage in the canopy, radiative budgets written for
shortwave and longwave radiation can be arranged such that R , , and Rn,s are
expressed in terms of incoming atmospheric and shortwave radiation, the shielding
factor, and the temperature, emissivity and albedo of the canopy and the soil.
D.Lo Seen et al./Agricultural and Forest Meteorology 83 (1997)49-74
55
The incoming shortwave radiation is usually measured whereas the atmospheric radiation, when not measured, can be estimated by (Brutsaert, 1975):
Thus, if the different resistances are known, the latent, sensible and net radiation heat
fluxes of the soil and the canopy can all be written in terms of q, and the measured
Ta and e,.
r,
2.2. Ground temperature and soil moisture time-dependent equations
The evolution of ground temperature and soil moisture are described following the
force-restore method proposed in Deardroff (1978). The method was originally proposed
by Bhumralkar (1975) for the computation of ground surface temperature, where the
forcing by the ground heat flux is modified by the restore term which depends on the
deep soil temperature T2 (Eq. (15) and Eq. (16)). The soil moisture is treated similarly
(Eq. (17) and Eq. (18)) as proposed by Deardroff (1978) and modified by Noilhan and
Planton (1989).
8%
Cl
= -(P
-E,) -
aw,
Pwd,
P-E,-E,
at
Pw
at
-=
7‘ws - we,)
c
2
when O < w, 5 wfc
when O < w2 5 wfc
d2
2.3. Determination of T, and T,
The system is solved for the two unknowns q, and T,, by first writing an energy
budget separately for the canopy (Eq. (3)) and the soil (Eq. (2)), then by forwarding the
soil heat flux G obtained in terms of q, and T, (by replacing Eq. (61, Eq. (9) and Eq.
(13) into Eq. (211,as the forcing term in the soil surface temperature time dependent
equation (Eq. (15)). Integration in time is done using the Crank-Nicolson method after
linearization of the non-linear terms (Eq. (19) and Eq. (20)).
[ q 1 + ~ q 4
=
[qq4 + 4[qq3[qr+~f
-
\
(19)
Once T, and T, are obtained for a given time step, the soil moisture can also be
,
. .
,
56
D.Lo Seen et al./Agricultural und Forest Meteorology 83 (1997)49-74
calculated using Eq. (17) and Eq. (18). The two component temperatures then allow the
calculation of the aerodynamic temperature To using Eq. (111, as well as the total
outgoing longwave radiation, which in turn is used to estimate the radiative temperature
Tr by considering an average surface emissivity of 0.97.
2.4. Resistances
The Shuttleworth and Wallace (1985) approach adopted here involves the use of five
resistances. For the present study, except when more appropriate formulations are
available in the literature, the resistance formulations used generally follow those found
in Shuttleworth and Gumey (1990). Each resistance formulation is briefly described in
the following:
2.4.1. Resistance rac
Assuming the eddy diffusivity and wind speed attenuation are the same inside the
canopy, and that energy is exchanged by molecular diffusion through a laminar layer
around the leaves, the bulk boundary layer resistance to heat and water vapor in the
canopy, is computed according to Choudhury and Monteith (1988).
1
(21)
2.4.2. Resistance ros
The aerodynamic resistance between ground surface and within canopy source height
is estimated using the approach proposed in Shuttleworth and Gurney (1990), where the
eddy diffusion coefficient, which is assumed to decrease exponentially in the canopy, is
integrated between height = O and height = z, d. The dependence of this resistance on
plant density is accounted for by relating the roughness length of the canopy z, and the
zero plane displacement height d to the leaf area index (Eq. (421, Eq. (431, in
Shuttleworth and Gumey, 1990).
+
r, =
where,
and
h
D.Lo Seen et al./Agricultural
and Forest Meteorology 83 (1997)49-74
57
2.4.3. Resistance raa
The aerodynamic resistance between within canopy source height and above canopy
reference height is estimated with the method proposed by Mahrt and Ek (1984) which
takes into account the influence of atmospheric stability.
1
raa= Cq ua
The exchange coefficient Cq depends on atmospheric stability which can be expressed in
terms of a Richardson number Ri (unstable case: Ri < O, stable case: Ri > O):
For the unstable case, Cq is given by:
r
12
with
1
while for the stable case, C, is given by:
2
k
cq=[ln[zrd-:+zo)
1
(1
+ 15Ri)(l +5Ri)'12
As the Richardson number depends on the ground surface temperature
namic resistance needs to be computed using an iterative procedure.
q, the aerody-
2.4.4. Resistance rsc
The bulk stomatal resistance of the canopy is expressed as the proáuct of a minimum
and different factorstwhich are always 2 1, and which vary in
stomatal resistance rSmin
time. As in Noilhan and Planton (1989), the factors considered here are a solar radiation
factor f i , a water stress factor f2, a factor related to the pressure deficit of the
atmosphere f 3 , and an air temperature dependence factor f4.
rsc =
zfl()f2(
R
w 2 ) f:
3 ( e * ( T ) - ea)f4(k
- Ta)
(30)
58
D.Lo Seen er al./Agricultural and Forest Meteorology 83 (1997) 49-74
with
I
fZ(W2)
=
’
1
when
( Wfc - ww,)
( wz - wwp)
when
wfc z w2 z wwp
when
wz < wwp
II
%
wz
Wfc
(32)
(33)
and
ar,- T,)
‘
1
=
1 - O.O016( Tc - T,)’
(34)
2.4.5. Resistance rss
Evaporation from a soil surface is the result of complex processes occumng inside a
porous medium and at its interface with air. Numerous empirical formulations based on
in situ data have been proposed to relate the soil surface resistance to the near surface
soil moisture content. (See Mahfouf and Noilhan, 1991, for a comparative study of
different formulations.) It is clear that the relationships proposed depend on the soil
textures used in the different studies, as well as on the thickness of the soil surface over
which the moisture content is averaged (Kondo et al., 1990). The formulation used in the
present study assumes a simple linear relationship between r,, and the surface soil
moisture as in Camillo and Gumey (1986):
r,, = 414O( wsat- w,) - 805
(35)
2.5. Vegetation growth model
The model used to simulate herbaceous vegetation growth in the Sahelian site under
study is taken from Mougin et al. (1995). The total standing aboveground biomass
present in a given day during the growing season is the sum of green biomass BG and
dead biomass B,, both expressed in kilogram dry matter per hectare (kg DM ha-’). The
amount of green biomass changes in time according to the balance between gross
photosynthesis P g , the incremental term and respiration R, and senescence S, the
D. Lo Seen et al./Agricultural and Forest Meteorology 83 (1997)49-74
59
decremental terms (Eq. (36)). Similarly, dead biomass increases with the senescing of
green material, and decreases due to litter production L (Eq.(37)).
At the beginning of the season, when the surface soil moisture is above wilting point
for 5 consecutive days, vegetation growth is made to start with an initial green biomass
BG(0).Thenceforth, the values of BG and B,, are calculated with a daily time step using
Eq. (36) and Eq. (37). Gross photosynthesis Pg is given by the product of intercepted
photosynthetically active radiation (PAR), a conversion factor cg which can be considered as a growth efficiency in optimal temperature conditions and in the absence of
water limitation, and two terms which account for the influence of water availability and
temperature on vegetation growth.
Pg = P A R . & , . & J - ( F , ) . f ( T , )
(38)
with
1 .64rS,,
f(FP)=
1.64rsmin(1
and
f(
c)
=1
+
+ r,
(39)
[
- 0.0389(Topt-
+ r,
c)
Total respiration R , is written as the sum of photorespiration R , , and dark respiration
which can be further identified as being composed of construction respiration R, and
maintenance respiration R,. Photorespiration is estimated as a constant fraction p r of
gross photosynthesis for C3 grasses, and is neglected for C4. Construction respiration is
proportional to gross photosynthesis, and maintenance respiration, to green biomass.
R , =R ,
R,
+ R , i-R ,
= p r . Pg
(41)
(42)
R c = ( l - Y G ) ( l -pr)P,
(43)
R , = mYG BG
(44)
Throughout the growing season before pea$ biomass, senescence is roughly estimated
except when
as a constant fraction of the green biomass (a constant senescence rate SI,
the vegetation suffers a severe stress, in which case the senescence rate follows a
Q10-type relationship with plant water potential F,. At seed maturation, the senescence
rate is made to increase drastically to simulate the fapid drying of the vegetation after
peak biomass.
,
D.Lo Seen et al./Agricultural and Forest Meteorology 83 (1997)49-74
60
The standing dry vegetation eventually reaches the litter pool under the mechanical
effects of rain, wind, animals, etc. The modeling of litter production and decomposition,
which would have been useful in the study of N cycling, is not included here. A constant
value is used for litter production in order to simulate the decrease of total standing
biomass at the beginning of the dry season. The growth and distribution of the root
system is also not modeled implicitly; it is assumed that the rooting system is well
developed enough not to limit the extraction of the water needed by the vegetation at
any time during the growing season.
2.6. Coupling the vegetation growth submodel with the water and energy balance
submodel
Closely related to the amount of latent heat coming from the canopy (transpiration) is
the amount of CO, absorbed by the vegetation in an inverse pathway of the water flow
through the stomates. In general, and especially in the arid and semiarid grasslands,
vegetation growth depends on the water status of the plants. On the other hand, the
water and energy balance depends on how the vegetation influences the boundary layer
in which the transport processes are taking place. Here, the coupling of the vegetation
growth model with the water and energy balance model is performed simply by
exchanging the necessary variables between the models (see Fig. 1). As the vegetation
growth model runs with a daily time step and the water and energy balance model with
an hourly time step, the LAI and h needed for the calculation of the resistances every
hour are considered constant throughout a given day. Conversely, and Fp used in the
vegetation growth model are computed from hourly values; the daily canopy temperaa,
ture is taken as the average of hourly temperatures, whereas the daily equivalent
vegetation water potential is iteratively estimated from the cumulated daily transpiration,
as being the potential necessary to extract from the soil (with soil moisture W)enough
water to balance daily transpiration (Camillo and Schmugge, 1983):
TC
"
with
,
,
i,'
3. Application of the model
The one-dimensional model presented above has been tested using meteorological
data and flux measurements acquired during the HAPEX-Sahel experiment, Niger, in
1992 (Prince et al., 1995). The objective was to determine whether the vegetation
D. Lo Seen et al./Agricultural and Forest Meteorology 83 (1997)49-74
b
61
growth and the water and energy balance submodels interact satisfactorily throughout
the growing season, and whether simulations of energy fluxes as well as biomass
compare well with measurements. This would then constitute a partial validation of the
model, allowing a certain degree of confidence in the simulations obtained for other
important variables which could not be monitored during the experiment. The energy
fluxes were estimated using the energy budget/Bowen ratio method based on air
temperature and vapor pressure measurements made at two different heights. A complete
description of the instrumental setup and the data acquisition procedure can be found in
Monteny et al. (1996). The fluxes and meteorological data (air temperature, air
humidity, wind speed and incoming radiation) used are hourly values computed from
20-min averages of acquisitions made every 10 s. Rainfall data used are daily totals
obtained with the nearest rain gauge of the EPSAT network (Lebel et al., 1992). As
hourly values are needed in the model, the daily total is arbitrarily distributed aver 3 h
during late evening. The consequence of this bias on the simulations is not known but is
suspected to be important only when rain intensity is high; significant run-off may then
occur and is not accounted for in the equations, as it may happen that soil saturation is
not reached when the same amount of precipitation is distributed over a longer period of
time.
3.1. The site during the growing season
Simulation is made to start before the first rains (beginning of June) and runs
uninterrupted throughout the growing season until the vegetation has dried out (November). The rainy period itself lasted around 3 months and for that particular site it brought
a total of about 418 mm. Vegetative growth occurred after DOY 220, and peak biomass
was reached around DOY 280. During that period, rainfall was well distributed leaving
no significant interval of prolonged drought (Fig. 3(a)). However, periods of more than 5
consecutive days without significant rainfall were not uncommon, such that situations of
water stress as well as water abundance have both been encountered during the season.
Aboveground biomass measurements made regularly (approximately every 10 days)
during the growing season showed that +e vegetation growth stages were not critically
affected by any severe climatic event. In Fig. 3(b) simulated biomass is compared to
measured biomass. The average amount of C3 vegetation with respect to that of C4 for
the grassland site was estimated to be of the order of 35%. This value was used in the
simulations. The simulation results for biomass are all found to be well within a(n
optimistic) 15% error allowed on the biomas,s measurements and are therefore quite
satisfactory.
Daily transpiration is obtained by summing hourly values of latent heat flux from the
canopy (AE,) converted into millimeters per day. Fig. 3(c) shows how daily transpiration evolves over the season. It roughly follows the amount of green vegetation, and as
expected from the formulations of resistances usèd, is further modulated both by
climatic factors and the availability of water from the soil reserve. By the end of the
season, transpiration reaches high values of more than 1.5 mm day-' for fully grown
vegetation.
D.Lo Seen et al./Agricultural and Forest Meteorology 83 (1997)49-74
62
0
f
2
m
2ooo:
__
Green
: _.___
Dead
15001
x
Y
i 1000;
E
üi
500
-
o:.
150
200
250
DOY
300
350
150
I
,
,
f
200
,,-,,-.,-.;\,
250
DOY
,
300
,
,
,
,
:
350
Fig. 3. (a) Rainfall measured at a grassland site in the Central Est Supersite during the HAPEX-Sahel
experiment, Niger, in 1992. (b) Simulated live and dead biomass, compared to biomass measured about every
10 days during the 1992 growing season (crosses represent biomass measurement&$)
Simulated daily
banspiration throughout the season.
t
3.2. Canopy structural variables
In the vegetation growth submodel, the main variables which express the change in
the amount of vegetation present at the site throughout the season are the green biomass
B,, and the standing dead biomass B,. These do not convey much information about
the structure of the canopy, and the variables used in the resistance formulations which
describe the canopy are only the L A I and h, the height of the vegetation. As the way a
given amount of biomass is displayed in a plant is essentially species dependent, it is not
readily acceptable to derive LAI and h from biomass directly. However, it can be
assumed that for a given site containing a variety of species of herbaceous vegetation, an
average value of Lkl and h can still be estimated from biomass. Here, an estimate of
M is obtained from green biomass by using an average value of the specific leaf area
SLAG equal to 70 cm2 g-' DM (0.0007 ha kg-' DM), whereas the empirical
relationship used to estimate h is obtained by fitting a second order polynomial to
measurements of vegetation height and biomass made during the season.
D.Lo Seen et al./Agricultural and Forest Meteorology 83 (1997)49-74
150
200
250
DOY
300
350
150
200
250
DOY
63
300
350
Fig. 4. Canopy structural variables derived from simulated green biomass: (a) leaf area index; (b) vegetation
height (crosses represent measurements).
The empirical relationships used to estimate LAI and h are shown in Eq. (47) and Eq.
(481,
LAI= SLAG.BG
+
h = ao a,BG
(47)
+ a,B;
(48)
with ao= 5.0, a , = 0.072, a2 = -0.000024, h in cm and BG in kg DM ha-'.The
evolution of LAI and h during the season are shown in Fig. 4(a) and (b). Eq. (46)is
found to underestimate the vegetation height for low values of biomass. The bias could
have been corrected by weighting the data points in favor of the low values, but this
would also suggest an abnormally high value of the height at vegetation emergence.
Also, when the height is computed using the polynomial, it is arbitrarily prevented from
decreasing from one day to another, even when green biomass actually decreases.
3.3. Comparison of energy fluxes simulations with measurements
The observation period during which energy fluxes were continuously measured
started on DOY 202 (July 20, 1992) and lasted for more than 2 months, covering stages
of vegetation installation, vegetative growth and fully grown vegetation. One week in
the middle of the growing season is chosen to show more closely the different input data
used by the model and the main variables and processes simulated, as compared to
measurements. Fig. 5 shows the meteorological data used by the model during that
particular week which can be considered a typical week of the rainy season, with 2 or 3
rainy days and enough soil moisture in the soot layer to sustain vegetation growth. The
biomass present was about 350 kg DM ha-', the height of the vegetation, about 25 cm
and the LAI was estimated to be of the orcler of 0.25 (see Fig. 4). Days 242 and 243
received, respectively, 14 and 15.4 mm of rainfall. DOY 243 was also generally cloudy
during the day as shown in Fig. 5(c). The main vqriables simulated by the model on an
hourly basis, namely the temperatures and soil moistures, are shown in Fig. 6 . The
difference between soil temperature and canopy temperature is quite important during
the day, and is found to be of the order of up to more than 10°C, as shown in Fig. 6(a).
The canopy temperature, however, is not greater than air temperature by more than a
.r.3
.
. .
I
.
. ,,
.""
~
.'
,
. . .
....
... .
.,.,..I..
... .
242
244
246
DOY
248
250
1200
-
2 1000
h
-e
v
800
C
.U
5
-=.
244
246
240
250
246
248
250
DOY
40
O
\
z
--
242
L1
600
30
L
a
E 20
400
c
.L
O
200
O
242
244
246
DOY
240
250
10
242
244
DOY
Fig. 5. Hydrological submodel main input parameters used for 1 week in the middle of the growing season (DOY 242-249): (a) relative humidity of the air; (b) wind
speed; (cl incoming shortwave radiation; (d) air temperature.
I
.
...."..
TC--
Ts-
Tr-
T2-."
'
"
"
~
To-"
'
'
'
(b)
50
Y
v
h
h
e
a
e
u
2 40
-e
v
40
a
c
30
E
f
30
I-
20
20
3
242
244
246
248
250
DOY
0.40
us'
'
'
'
'
I
242
244
0
246
248
1
250
DOY
GIZ-."
'
.
"
'
'
'
(.>
]
E
--.-.-._.-._._.
242
244
246
248
250
DOY
Fig. 6. Simulated temperatures and soil moistures for 1 week in the middle of the growing season (DOY 242-249): (a) soil surface temperature q, canopy
temperature T, and the temperature of the deeper soil layer T,;(b) radiative temperature T, and aerodynamic temperature To; (c) surface soil moisture w, and root
layer soil moisture w2.
LE (simulated-
measured 0)
600
> o o 400
6
<
'
H'
measured O)
(simulated'
'
'
'
I
"
'
I
"
'
400
<
h
h
N
N
200
200
3
3
v
v
O
O
t
-200
1
i
242
244
246
DOY
Rn fsimulated-
800
248
250
measured 0)
600
;
; 400
E
.c
.-?
,a 0.1
C
e
+
\
z
200
O
y-------
-2oot.
242
.
.
I
244
. . .
I
246
DOY
.
,
.
I
248
. . . 1
250
0.0
242
244
246
DOY
248
250
Fig. 7. Simulatea heat fluxes compared to fluxes measured with the Bowen ratio method, for 1 week in the middle of the growing season (DOY 242-249): (a) latent
heat flux; (b) sensible heat flux; (c) net radiation flux; (d) hourly simulated transpiration.
__
L
D. Lo Seen et al./Agricultural and Forest Meteorology 83 (1997) 49-74
67
few degrees, thus avoiding the higher temperature ranges which may become inhibitory
for some vital plant processes. The deep ground temperature, as expected, remains fairly
stable (about 29°C) over such a short time period. Fig. 6(b) shows the difference
between radiative surface temperature T, and aerodynamic temperature To, and gives an
indication of the overestimation which would be obtained when expressing sensible heat
flux in terms of T, instead of To over sparsely vegetated areas (Chehbouni et al., 1996).
The evolution of soil moisture in the surface layer and the root layer are shown in Fig.
6(c). The rain events of DOY 242 and 243 have brought the surface soil moisture up to
saturation level during the night, and by noon the following day, the soil surface has
already dried completely. For the non-rainy days, moisture from the root layer diffuses
upwards during the night such that the surface soil moisture nearly reaches the root layer
soil moisture in the morning. The soil surface then quickly dries up during the first hours
of daylight, and around 10:00-11:00, the soil is completely dry again. The deeper root
layer shows much less moisture variation and acts mainly as the storage which provides
moisture to the vegetation in replacement of water lost by transpiration. The soil
moisture conditions of the surface indicates that during the day.the latent heat flux is
mainly composed of transpiration except for the first hours of the day when evaporation
from the soil is also important. This can be seen in Fig. 7(a) and (d), where the
differences between the simulated latent heat flux and transpiration occur mainly in the
'
"
600
'
'
'
'
'
'
'
. . '... Y
400-
.-z
N
E
-I
u
w
c
'
'
200
/
-
O
-z
,
,,
-200
-200
z
.
rmse 47.82 W/m2
o
200
400
LE measured (W/m2)
,
-200.
-200
'
/
3
u
-
0-
I
. ,. /
400-
\
200
'
/
n
/
f
YI
w
.
(b)
\
3
u
P-
" " ' " *
600
~
/
n
7
'
"
: (a>
.
600
o-200
,,
/
.
-200
.
.
.
rmse 47.27 W/m2
I
o
H
.
.
.
,
.
'
I
'
200
400
measured (W/m2)
'
'
600
rmse 48.56 W/m2
I
O
,
.
.
#
200
.
.
.
*
.
'
400
S
.
600
.
.
800
Rn measured (W/m2)
Fig. 8. Scattergrams showing comparison of measured and simulated daytime (8:00-18:00) hourly heat fluxes
during the observation period: (a) latent heat flux; (b) sensible heat flux; (c) net radiation flux.
D. Lo Seen et al./Agriculrural and Forest Meteorology 83 (1997) 49-74
68
morning. Fig. 7(d) is also to be paired with Fig. 3(c), to illustrate how the transpiration
observed during that week compares with the evolution of daily transpiration over the
season. Fig. 7(a)-(c) shows the comparison between simulated and measured energy
fluxes. A fairly good agreement is obtained in general, but differences of more than 50
W m-' are quite common, especially for latent and sensible heat fluxes.
In order to assess more objectively the performance of the model over the season,
measured and simulated hourly energy fluxes are compared for the whole observation
period. Flux values derived from measured Bowen ratios which are either missing or
inexplicably too high or too low (> 600 W m-' or < O W m-' for day-time latent and
sensible heat flux values) have been replaced by time interpolated values, and when
several erroneous hourly fluxes are obtained during 1 same day, all the flux data for that
day are rejected. Out of 70 days of observation, only 5 days of data have been removed
in that way, suggesting that the error screening used in the present study has not been
very severe. Fig. 8 shows the scattergrams obtained for (a) latent heat, (b) sensible heat
and (c) net radiation with day-time (8:OO to 18:OO) values only (when night-time values
are included, better RMSE are artificially obtained due to the high percentage of
near-zero values). The RMSE computed for the three cases give values which are all of
the same order of magnitude (about 50 W m-'1. These values are satisfactory even
though they are slightly higher than the standard error tolerance normally associated
6001
"
'
a
"
.
t
. . .
/
.
-200
.
t
'
'
"
I
"
o
t
,
,
.
1
.
' 1
.
.
200
400
LE measured (W/m2)
'
'
'
'
'
" '
" '
"
600
" '
6001
'
'
'
'
"
,
,
,
rmse 23.22 W/m2
-200
800
'
'
-200
-200
. .
.
I
:
"
I
"
'
I
rmse 25.68 W/m2
I
o
H
,
.
I
200
400
measured (W/m2)
,
,
'
600
'I
/
(cl
/
,
600-
E
\
5
400-
U
.c
200-
YI
C
=
0/
-200
-200
,
. . .
29.83 W/m2
rmse
,
O
.
.
.
t
.
.
.
I
.
.
.
I
200
400
600
Rn measured (W/m2)
..
800
Fig. 9. Scattergrams showing comparison of measured and simulated daily averages of daytime (8:00-18:00)
heat fluxes during the observation period: (a) latent heat flux; (b) sensible heat flux; (c) net radiation flux.
D.Lo Seen ef al./Agricultural andForest Meteorology 83 (1997)49-74
69
with energy fluxes measured using the Bowen ratio method. It has been shown that
HAPEX-Sahel hourly sensible heat flux and latent heat flux measurements have
confidence limits of +15% and +20%, respectively (Lloyd et al., 1996). When
measured and simulated daily average fluxes are plotted (Fig. 9) better RMSE values are
obtained, but it becomes also more evident that the model somewhat underestimates net
radiation and sensible heat in favor of latent heat flux (in both cases, a line fitted to the
data points lies about 35 W m-2 below the 1:l line). This tendency could be minimized
or reversed by an appropriate calibration of resistances like rss and r,,, as well as
relationships between soil thermal characteristics (heat capacity and thermal conductivity) and soil moisture. Unfortunately, the measurements necessary to carry out such
calibrations have not been made at the time of the experiment. However, even with a
better calibrated model, it is not certain that the accuracy of the simulations can be
significantly increased. Given the relative simplicity of the 1-D model and the long
period of uninterrupted simulation which characterizes this experiment, the model can be
considered able to satisfactorily and realistically simulate the main processes occurring
in a semiarid grassland.
4. Discussion and perspective
A model describing the major land surface processes occurring in semiarid grasslands
has been presented in this paper. The model couples the water and energy balance of a
sparsely vegetated surface with the vegetation growth and soil moisture and thermal
dynamics. This coupling ensured a consistency between root uptake, transpiration and
photosynthesis, as well as between 'rapid' (temperature, moisture content) and 'gradual'
(amount of vegetation) changes in the state of the surface. Model simulations of biomass
over the growing season are found to be within a 15% error margin allowed on biomass
measurements. Hourly values of net radiation, as well as latent and sensible heat fluxes
are simulated with an RMSE of less than 50 W m-2. Given the relative simplicity of the
model and the long period of uninterrupted simulation, these results are considered
satisfactory. However, additional field measurements would be beneficial to calibrate
resistance formulations, and relationships between soil thermal properties and soil
moisture, in order to correct a slight underestimation ( 35 W m-2) of sensible heat
flux in favor of latent heat flux.
The evolution of the state of the surface during the growing season is described in the
model using variables that include soil moisture, leaf area index, vegetation height and
temperature of the soil and canopy, which' are variables commonly used as inputs in
radiative transfer modeling. This would enable the land surface process model to be used
with radiative transfer models to study the sensitivity of remote sensing observations to
varying surface conditions and associated processes. Ongoing research based on this
modeling approach is directed towards the incorporation of remote sensing data in
controling simulations of land surface processes. Through such studies, the surface
variables and parameters by which the land surface process model may be controlled
using remote sensing data can be objectively identified.
-
70
D. Lo Seen er al./Agriculrural and Forest Meteorology 83 (1997)49-74
Acknowledgements
This work was carried out in part at the Jet Propulsion Laboratory, California
Institute of Technology, under contract with the National Aeronautics and Space
Administration. Field data supplied by contributors to the HAPEX-Sahel experiment
database are gratefully acknowledged. DL was at JPL as a National Research Council
Resident Research Associate.
Appendix A. Notation
A S . Variables
Aboveground standing herbaceous biomass, green and dry (kg DM
ha-')
Coefficient used in the surface soil moisture time dependent equation
Cl* c
2
~~~;
=
(C,(dimensionless) = 0.082 ( w , , , / ~ , ) ~ ~C,(dimensionless)
3.9w2/(w,,, - w; O.OOl), value estimated for sand, as in Noilhan
and Planton, 1989)
AE, AE,, AE, Latent heat flux above the canopy, from the canopy and from the
ground (W m-2)
G
Ground heat flux (W m-2)
H , H,, H,
Sensible heat flux above the canopy, from the canopy and from the
ground (W m-')
L
Litter production (kg DM ha-' day- '1
LAI
Leaf area index (m2 m-2)
P
Mass of precipitation reaching the soil surface per unit area and unit
time (kg m-' s-I)
PAR
Photosynthetically active radiation (W m-')
Gross photosynthesis (kg DM ha-' day- ')
pg
R,, R , , , R , , Net radiation flux into the canopy ground, into the canopy and into
the ground (W m-2)
R,, R,, R,, R , Construction and maintenance respiration, photorespiration and total
respiration (kg DM ha-' day-')
Longwave radiation flux at the top of the canopy (W m-2)
Ri,
Shortwave radiation flux at the top of the canopy (W m-2)
R,1
S
Senescence (kg DM ha- day- I)
Air temperature at above canopy reference height (K)
r,
T
Temperature
of the canopy (K)
E
Daily
average
temperature of the canopy (K)
T,
Temperature
of
the ground surface (K)
T,
Air temperature at within canopy source height (K)
T2 .;'
Mean temperature of soil layer of depth d, (K)
Daily
transpiration (kg m-2 day- I)
Tì-*
Soil
temperature
at time step t (K)
T,', T,'
d
Zero plane displacement height for canopy (m)
BCì* B D
+
I
+
'
r,
..
D. Lo Seen et al./Agricultural and Foresr Meteorology 83 (1997)49-74
71
Vapor pressure at above canopy reference height (mbar)
Vapor pressure at within canopy source height (mbar)
Saturated vapor pressure at temperature T (mbar)
Vegetation height (m)
Richardson's number (dimensionless)
Aerodynamic resistance between within canopy source heig..t and
above canopy reference height (s m-I)
Bulk boundary layer resistance of the canopy (s m-'>
Aerodynamic resistance between ground surface and within canopy
source height (s m-')
Bulk stomatal resistance of the canopy (s m- I)
Surface resistance of the ground (s m-I)
Wind speed at reference level zref (m s-I)
Wind speed at top of the canopy (m sFriction velocity (m s-')
Characteristic leaf width (m)
Volumetric soil moisture content of ground surface and mean soil
moisture content of root layer (dimensionless)
Surface soil moisture when gravity balances capillarity forces; it is
equivalent to T2 in the restore t e m of the soil moisture time dependent
equation; it depends on soil type, wsat and w 2 (dimensionless)
Roughness length of canopy (m)
Roughness length of substrate (m)
Daily average soil water content (dimensionless)
Bulk canopy water potential and soil water potential in the root zone
(bar)
Daily equivalent bulk canopy water potential (bar)
Efficiency of interception of PAR by the canopy, estimated from B,
as in Mougin et al. (1994): ì?,(dimensionless)= 0.2 Log$
BG/180)
Ground surface albedo, estimated from soil moisture content as in Ben
Mehrez (1990): a,(dimensionless) = 0.28 - 0.21( ws/wsat>
Shortwave radiation shielding factor of ground by the canopy, estimated from LAI as in Kanemasu et al. (1977): uf(dimensionless) = 1
- exp( - 0.4 X LAI)
Heat capacity of soil, estimated from soil moisture content as in
DeVries (1963): p,cs (J K-' m-3) = 4.18 lo6 (0.3 w), where
w = w, or w2
Thermal conductivity o$ soil, estimated from soil moisture content as
in DeVries (1963) and Ben Mehrez (1990): A, (W m-' K-') = 0.06
O . ~ ( W ) ' /where
~,
w = ws or w 2
'
+
+
+
A.2. Constants
Specific heat of air at constant pressure (- 1012 J kg-' K- '1
Constants used in the ground temperature time-dependent equation
(cI= 27r1l2,c2= 27r, dimensionless)
cP
C I ? C2
8
D.Lo Seen et al./AgriculturaI and Forest Meteorology 83 (1997)49-74
Acceleration due to gravity (9.81 m s-’1
von Kármán’s constant (0.4, dimensionless)
Psychrometric ‘constant’ ( 0.66 mbar K- )
Latent heat of vaporization (- 2.45 lo6 J kg- ‘1
Density of air (- 1.2 kg m-3>
Density of water ( 1000 kg m-3)
Stefan-Boltzmann constant (5.6705 lo-* W m-’ K-4>
l-day period (86 400 s)
-
’
-
o-
A.3. Parameters
m
Pr
Exponent in power function relating soil ‘matric’ potential to soil
moisture content (3.79, derived from soil texture (clay = 6%, sand =
85%), Cosby et al., 1984)
Initial aboveground biomass (30 kg DM ha-’)
Percentage by mass of C3 species present (35%)
Normalization depth for surface layer (10 cm), thickness of root layer
(60 cm)
Maintenance coefficient (0.02 day-’ )
Fraction of gross photosynthesis used as photorespiration (0.4 dayfor C3 species, negligible for C4 species)
Minimum and maximum stomatal resistance (100, 5000 s m-’>
Resistance of the mesophyll (183 s m-’, estimated from %C3)
’
Total resistance to water flow in the plant and at the soil-root
SLAG
S
interface, normalized to units used (1.03 bar kg-’ m2 day)
Specific leaf area (0.0007 ha kg-’ DM)
Senescence rate (0.003 day- )
Canopy temperature at maximum photosynthesis (38°C)
Soil moisture at saturation (0.3821, estimated from soil texture (clay =
6%, sand = 85%), Cosby et al., 1984)
Volumetric soil moisture content at wilting point (at 16 bar) and at
field capacity (at 0.1 bar) (0.0363, 0.1388)
Fraction of the total mass of carbon used, incorporated into new
structural tissues (0.75)
Wind speed and eddy diffusivity attenuation constant in the canopy
(2.5, dimensionless)
Canopy albedo (0.22)
Canopy emissivity and ground surface emissivity (0.98, 0.96)
Growth efficiency in optimal conditions of water availability and
temperature (3.6 g DM W-‘ day-’ or 4.17 g DM MJ-’)
Canopy water potential corresponding to 50% stomatal closure (15
bar)
Soil water potential at saturation (-0.00216 bar, estimated from soil
texture (clay = 6%, sand = 85%), Cosby et al.; 1984)
’
D. Lo Seen et al./Agrìcultural and Forest Meteorology 83 (1997) 49-74
73
References
Ben Mehrez, M., 1990. Etalonnage et validation d’un modkle de flux de surface dans le cas de I’exp6rience
Hapex-Mobilhy. Application ?
l’estimation
i
de la resistance de couvert. Ph.D. Thesis, Universid Paris VII.
Bhumralkar. C.M., 1975. Numerical experiments on the computation of ground surface temperature in an
atmospheric general circulation model. J. Appl. Meteor., 14: 1246-1258.
Bmtsaen, W., 1975. On a derivable formula for longwave radiation from clear skies. Water Resour. Res., 11:
742-744.
Camillo, P.J. and Gumey, RJ., 1986. A resistance parameter for bare-soil evaporation models. Soil Sci.,
141(2): 95-105.
Camillo, PJ. and Schmugge, T.J., 1983. Estimation of soil moisture storage in the root zone from surface
measurements. Soil Sci., 135: 245-264.
Chehbouni, A., Lo Seen, D., Njoku, E. and Monteny, B.M., 1996. Examination of the difference between
radiative and aerodynamic surface temperatures over sparsely vegetated surfaces. Remote Sensing Environ.,
in press.
Choudhury, B.J. and Monteith, J.L., 1988. A four-layer model for the heat budget of homogeneous land
surfaces. Q.J.R. Met Soc., 114 373-398.
Cosby, B.J., Hornberger, G.M., Clapp, R.B. and Ginn, T.R., 1984. A statistical exploration of the relationships
of soil moisture characteristics to the physical properties of soils. Water Resour. Res., 2 0 682-690.
Deardroff, J.W., 1978. Efficient prediction of ground surface temperature and moisture, with inclusion of a
layer of vegetation. J. Geophys. Res., 83: 1889-1903.
DeVries, D.A., 1963. Thermal properties of soils. In: N.R. Van Wijk (Editors), Physics of Plant Environment.
Holland Publishing, Amsterdam.
Goutorbe, J.P., Lebel, T., Tinga, A., Bessemoulin, P., Brouwer, J., Dolman, A.J., Engman, E.T., Gash, J.H.C.,
Hoepffner, M., Kabat, P., Kerr, Y.H., Monteny, B., Prince, S., Said, F., Sellers, P. and Wallace, J.S., 1993.
HAPEX-Sahel: a large scale study of land-atmosphere interactions in the semiarid wopics. Ann.
Geophysicae, 1 2 53-64.
Ham, J.M. and Heilman, J.L., 1991. Aerodynamic and surface resistances affecting energy transport in a
sparse crop. Agric. For. Meteorol., 53: 267-284.
Kanemasu, E.T., Rosenthal, U.D., Raney, R.J. and Stone, L.R., 1977. Evaluation of an evapotranspiration
model for com. Agron. J., 6 9 461-464.
Kondo, J., Saigusa, N. and Sato, T., 1990. A parameterization of evaporation from bare soil surfaces. J. Appl.
Meteorol., 2 9 385-389.
Lebel, T., Sauvageot, H., Hoepffner, M., Desbois, M., Guillot, B. and Huben, P., 1992. Rainfall estimation in
the Sahel: the EPSAT-NIGER experiment. Hydrol. Sci. J., 37(3): 201-205.
‘Lloyd, C.R., Bessemoulin, P., Cropley, F.D., Culf, A.D., Dolman, A.J., Elbers, J., Heusinkveld, B., Moncrieff,
J.B., Monteny, B. and Verhoef, A., 1996. A comparison of surface fluxes at the HAPEX-Sahel fallow
bush sites. In: HAPEX-Sahel Special Issue. J. Hydrol., in press.
Mahfouf, J.F. and Noilhan, J., 1991. Comparative study of various formulations of evaporation from bare soil
using in situ data. J. Appl. Meteorol., 3 0 1354-1365.
Mahrt, L. and Ek, M., 1984. The influence of atmospheric stability on potentia1 evaporation. J. Clim. Appl.
Meteorol., 23: 222-234.
Monteny, B.A., Lhomme, J.P., Chehbouni, A., Troufleau, D.,Amadou, M., Sicot, M., Verhoef, A., Galle, S.,
Siid, E and Lloyd, C.R., 1996. The role of the sahevan biosphere on the water and the CO, cycle during
the HAPEX-Sahel experiment. In: HAPEX-Sahel Special Issue. J. Hydrol., in press.
Mougin, E., Lo Seen, D., Rambal, S., Gaston, A. and Hiemaux, P., 1994. A regional sahelian grassland model
to be coupled with satellite multispectral data. 1. Model description and validation. Remote Sensing
Environ., 5 2 181-193.
Nichols, W.D., 1992. Energy budgets and resistances to energy transport in sparsely vegetated rangeland.
Agric. For. Meteorol., 6 0 221-249.
Noilhan, J. and Planton, S., 1989. A simple parameterization of land surface processes for meteorologica1
models. Mon. Weather Rev., 117: 536-549.
Pielke, R.A. and Avissar, R., 1990. Influence of landscape structure on local and regional climate. Landscape
E d . , 4 133- 155.
74
D.to Seen et ul./Agriculrurul
und Forest Meteorology 83 (1997) 49-74
Pinty, J.P., Mascart, P., Richard, E. and Rosset, R., 1989. An investigation of mesoscale flows induced by
vegetation inhomogeneities using an evapotranspiration model calibrated against HAPEX-MOBILHY
data. J. Appl. Meteorol., 28: 976-992.
Prince, S.D., Kerr, Y.H., Goutorbe, J.P., Lebel, T., Tinga, A., Bessemoulin, P., Brouwer, J., Do1man:’A.J..
Engman, E.T., Gash, J.H.C., Hcepffner, M., Kabat, P., Monteny. B.A., Said, F., Sellers, P. and Wallace,
J.S., 1995. Geographical, biological and remote sensing aspects of the hydrologic atmospheric pilot
experiment in the Sahel (HAPEX-Sahel). Remote Sensing Environ., 51: 215-234.
Raupach, M.R., 1991. Vegetation-atmosphere interaction in homogeneous and heterogeneous terrain: some
implications of mixed-layer dynamics. Vegetatio, 91: 105- 120.
Shuttleworth, W.J. and Gumey, R.J., 1990. Evaporation from sparse crops - an energy combination theory.
Q.J.R. Met. Soc., 111: 839-855.
Shuttleworth, WJ. and Wallace, J.S., 1985. The theoretical relationship between foliage temperature and
canopy resistance in sparse crops. Q.J.R. Met. Soc., 116 497-519.
Taconet, O., Bemard, R. and Vidal-Madjar, D., 1986. Evapotranspiration over an agricultural region using a
surface flux/temperature model based on NOAA-AVHRR data. J. Clim. Appl. Meteorol., 25: 284-307.
Wilson, M.F., Henderson-Sellers, A., Diclrison, R.E. and Kennedy, P.J., 1987. Investigation of the sensitivity
of land-surface parameterization of NCAR Community Climate Model in regions of tundra vegetation. J.
Climatol., 7 319-343.
d
,
3
01
PI