Journal of

Hydrology ELSEVIER

Journal of Hydrology 188-189 (1997) 494-515

Evaporation, sensible heat and canopy conductance of fallow savannah and patterned woodland in the Sahel P. K a b a t * , A.J. D o h n a n , J.A. Elbers DLO-Winand Staring Centre, PO Box 125, NL-6700-AC Wageningen, The Netherlands

Abstract The behaviour of evaporation, sensible heat and canopy conductance of fallow savannah and patterned woodland in the Sahel is studied for the HAPEX-Sahel Intensive Observation Period. Both fallow savannah and patterned woodland reach evaporation rates of 4-5 mm day -1 during the rainy part of the lOP and show a decline, after the rains have ceased, to 2 mm day -~. Sensible heat fluxes are different for the two vegetation types. This is also reflected in the behaviour of the evaporative fraction. Analysis of the vegetation surface conductance of the vegetation part shows that maximum values for fallow savannah axe around 10 nun s -I, and for patterned woodland up to 40 mm s -I. The response of the conductance to vapour pressure deficit is different for the two vegetation types. This is attributed to differences in C3 and C4 species composition. The consequences of these differences for modelling vegetation-atmosphere interaction in the semi-add tropics are discussed.

1. Introduction Deterioration of the land surface in the Sahel is thought to play an important role in the observed decline of Sahelian rainfall. Overgrazing by agricultural livestock and overexploitation of the scarce natural resources by man may help to exacerbate a feedback process in which characteristics of the land surface influence the generation of rainfall. Predictions with general circulation models (GCMs) have substantiated this hypothesis, originally put forward by Chamey (1975) (see, for instance, Xue and Shulda (1993)). The model experiments, however, have been performed with relatively simple land surface descriptions, assuming homogeneity over large areas, and with little or no calibration of the land surface schemes against measurements. The HAPEX-Sahel experiment was designed to provide the data needed to calibrate the * Corresponding author. 0022-1694/97/$17.00 © 1997- Elsevier Science B.V. All fights reserved PII S0022-1694(96)03190-3

P. Kabat et al./Journal of Hydrology 188-189 (1997) 494-515

495

land surface descriptions and to study the importance of variability of the land surface in a semi-arid environment on the area average exchanges of heat, water vapour, CO2 and momentum (Goutorbe et al., 1994). The study of spatial variability of surface fluxes is important to be able to define the area average evaporation (and groundwater recharge) and to test aggregation algorithms which may be of use to climate modellers. Variability in the exchange of heat and water vapour may occur owing to differences in the plant physiological behaviour, soil moisture conditions, structural characteristics such as aerodynamic roughness and boundary-layer properties. Within a single supersite in HAPEX-Sahel (Goutorbe et al., 1994), an area of about 400 km 2, all of these differences, except perhaps differences in boundary-layer properties, may occur at spatial scales smaller than 1-2 km. It then becomes important to compare the relative magnitude and diurnal to seasonal behaviour of the surface fluxes of the main vegetation types in the entire HAPEX-Sahel study area, which is compatible with the size of a typical GCM grid box (Goutorbe et al., 1994). The natural vegetation in the Sahel area around 13°N in Niger consists of repeated toposequences from plateaux to valley bottoms, with patterned woodland (Thierry et al., 1995) on the plateaux tops and fallow savannas and millet crops in the valleys (Prince et al., 1995). This paper aims to provide data on the behaviour of the surface fluxes of heat and water vapour of the typical land cover types at the Central West Site (Kabat and Goutorbe, 1995), and representative of large parts of the.Sahel. This is approached by analysing both the diurnal and the seasonal behaviour of the surface fluxes. Also reported is the development of the canopy conductance during the Intensive Observation Period (IOP), which took place from the beginning of August to the beginning of October 1992. Gash et al. (1991) have described micrometeorological measurements of evaporation during a 6 week period in the beginning of the dry season in a fallow savannah site in the Sahel. They observed a pronounced reduction in evaporation from 4.5 to 1.5 mm day -I during these 6 weeks. They also found that the (big leaf) surface conductance showed little diurnal variation and no clear response to solar radiation or humidity deficit. The establishment and subsequent survival of vegetation in the Sahel is strongly dependent on the availability of water. In particular, Culf et al. (1993) have hypothesized that patterned woodland may have developed into its typical striped pattern by using the bare soil areas as harvesting areas for water. This extra water on a unit area basis for the vegetation allows a much denser vegetation growth in the vegetated stripes than, for instance, fallow savannah. There is thus a need for ecologists, meteorologists and hydrologists to understand the evaporation and energy partitioning of typical land covers in the Sahel.

2. Theory The evaporation in (semi) arid regions of the world consist of two components: evaporation from the vegetation and evaporation from the soil, immediately after rainfall. These two components can be of similar order of magnitude for the two vegetation types considered in this paper. Micrometeorological measurements only measure a single integrated flux and it becomes therefore important to distinguish between the two. This

P. Kabat et aL/Journal of Hydrology 188-189 (1997) 494-515

496

can be done in two ways. The first consists of adding up contributions of the soil and vegetation. This is the procedure followed by Schulze et al. (1994) and Kelliher et al. (1995). Several operational GCM land surface schemes (e.g. that of Warrilow et al., 1986) calculate total evaporation by similar schemes. This procedure is allowed in conditions of high leaf area indices and small soil evaporation and is applied here for the fallow savannah. The second way takes into account that the environment in which plants evaporate may be modified by fluxes of heat and water vapour from the soil. Shuttleworth and Wallace (1985) were among the first to note this issue and to develop a theoretical framework for analysis. The Shuttleworth and Wallace model was subsequently modified by Dolman (1993) to take into account the fact that the radiation distribution of very sparse canopies cannot be described by assuming a transparent canopy overlying a soil. This model is used in the analysis for the patterned woodland site. Both models require soil evaporation to be known or estimated. Soil evaporation is calculated in the present paper based on the Ritchie (1972) approach, which describes bare soil evaporation in two distinct phases. During the first phase, which follows immediately after (re)wetting of the soil surface, evaporation proceeds at a potential rate. During the second phase, soil evaporation declines with the square root of time. This procedure for estimating soil evaporation was tested for patterned woodland at the Southern Super Site, using direct measurements of the bare soil evaporation (Wallace and Hollwill, 1997). On the second day after wetting, the measured evaporation rates typically started to depart from potential evaporation rates. Similarly, for fallow savannah, the lysimeter studies at West Central Site (Kabat and Goutorbe, 1995) show a decline from the potential evaporation rate about 1 day after the last rainfall event. Thus, soil evaporation proceeds at its potential rate only during the first day after last rain and can be calculated from the Priestley-Taylor equation (Priestley and Taylor, 1972) as kE,

~+

(R,, - G, )

( 1)

in which A is the slope of the saturated humidity temperature curve, 3, the psychrometer constant, and R..~ and G~ are respectively the net radiation and ground heat flux of the bare soil. After the first day after rain, evaporation declines with the square root of time (Ritchie, 1972):

XE.~=6vS-+l-bv5

(2)

where t is the number of days after rain and ~ a soil-dependent parameter. Ritchie (1972) showed that the magnitude of 6 is approximately proportional to the magnitude of hydraulic conductivity of the soil at soil water suction equal to 100 cm. This means that sandy soils have lower values of/~ compared with the soils with more clay and loam content. Following Wallace and Hollwill (1997), who derived the value of 6 for patterned woodland on a gravelly loamy soil at the Southern Super Site, this was set at 2.1. For fallow savannah a value of 1.0 is used for dt, derived from the lysimeter studies (Kabat and Goutorbe, 1995). This lower value of 6 reflects the much sandier soil at this site. Eq. (2) calculates evaporation on a daily basis, so some assumption about the distribution over the day has to be made. For the present purposes, it was assumed that soil evaporation

P. Kabat et al./Journal of Hydrology 188-189 (1997) 494-515

497

occurred only during daylight and the total daily evaporation was divided by the number of daylight hours to obtain hourly soil evaporation. This assumption may underestimate the soil evaporation during the morning hours, when the surface layer is wetter owing to the night capillary rise and the dew, but it can be generally justified in conditions where evaporation is supply limited rather than (atmospheric) demand limited.

2. I. Fallow savannah model The vegetation contribution to total evaporation is obtained by subtracting the areaweighted soil evaporation from the total evaporation, M~t:

XEt-~XE~ (1-~)

hEy -

(3)

where XE~ is the soil evaporation, M~v the evaporation from the vegetation and el the fractional cover of the soil (u = 0.15 for fallow savannah). In this analysis it is assumed that the grasses and bushes act like a single big leaf. This assumption is justified later in this paper. The surface conductance of the vegetation (gO can now be calculated from the inverted Penman-Monteith equation: 1

A ( R , v _ Gv)+

__ =

gv

pcpD ra

~Ev3'

&---- 1

1 ra

(4)

3'

where R,v and G~ are the net radiation and ground heat fluxes of the vegetation, D is the vapour pressure deficit measured at the reference height of 5.5 m, cp the heat capacity of air, p the density of air and ra the aerodynamic resistance calculated as the reciprocal of g ;

ra -

1

u

ga - u2 t-

Z0h ]

ku.

(5)

where z is the reference height, ZOrnand Z0h are the aerodynamic roughness lengths for momentum and sensible heat transport, u is the horizontal wind speed, u. is the friction velocity, k is von KLrrmin's constant with a value of 0.4 and L is the Monin-Obhukov length. The value of ln(z0/ze) is set at 2.5 (Garrat, 1994) and ~,, and 't% are respectively the momentum and sensible heat flux stability correction functions (Paulson, 1970). The overall resistance network for this single-layer model is shown in Fig. 1.

2.2. Patterned woodland model The procedure used above will introduce some error in the calculation of the surface (canopy) conductance from the vegetation if the sensible heat or latent heat release from the bare soil is substantial and able to modify the humidity deficit around the vegetation (Shuttleworth and Wallace, 1985) and decouple it from the (observed) value at the reference height. Sensible heat release from the bare soil areas will increase the around (within) canopy humidity deficit relative to the reference height value in dry soil

P. Kabat et al./Journal of Hydrology 188-189 (1997) 494-515

498

Z,,,,

l.E,,

h d+z

kE,

1 l/ga

~Et 1g /~ .

d+~

ZE.

Fig. 1. Schematic diagram showing the resistance network of a single (left side) and two source (right side) surface energy balance model. Explanation of symbols in the text.

conditions, which will lead to a apparent decrease in vegetation surface conductance. Similarly, in wet soil conditions the within-canopy deficit will be reduced, leading to a apparent value of gv which will be too high compared with the 'real' value of the vegetation. These effects are better described by a sparse canopy model. In the current paper such a model (Dolman, 1993) is inverted to obtain the canopy conductance of the patterned woodland vegetation. The resistance network needed for this model is also outlined in Fig. 1. In a sparse canopy model the within-canopy deficit, Do, is explicitly calculated from a balance of the fluxes from soil, vegetation and resistances: "y

A a

(6)

D o = D + (Hi - "TkEt) --ra

.ix

pCp

where the subscript t refers to the total, measured fluxes. From Eq. (6) it is obvious that Do may become large compared with D, even if the aerodynamic resistance r~ remains small owing to high roughness of patterned woodland surface. This is especially true for the dry days, where Ht is large. (See also Fig. 13, below, where D and Do are compared for selected wet and dry days during the IOP.) The definition of r~ is different from that used above as it relates only to transfer from canopy to reference height:

1 1 fln(Z-d) h ra-ga-ku,), ( Z _ - - - ~ - - q ' h + ~ x p

[ ((z0 n 1

h

d))]

-1

}

(7)

where z is the reference height, h the height of the canopy (h = 6 m), n an eddy decay coefficient set at n = 1.5 and d is the zero-plane displacement, taken as 2 m (Dolman, 1993). The aerodynamic roughness length of patterned woodland is taken as 0.4 m (Dolman et al., 1992). The difference between heat and momentum transport now does not need to be parameterized but is explicitly modelled through the use of a boundarylayer resistance, r~ in the inverted (vegetation) Penman-Monteith equation for the

P. Kabatet al./Journalof Hydrology188-189(1997)494-515

499

vegetation fraction (e.g. Blyth and Dolman, 1995): - A(R,v - G v ) + l__= gv

pcpD° rv

kEy7

] A 3'

1 rav

(8)

The evaporation from the vegetation, kEy, is calculated from Eq. (3) for o~ = 0.7. Total evaporation of patterned woodland, hEr, was derived from Bowen ratio energy balance measurements. (Note the use of the within-canopy deficit Do rather than the referencelevel deficit in Eq. (8).) The canopy boundary-layer resistance, r~, is calculated from (Shuttleworth and Gumey, 1990; Dolman, 1993)

rV= ~ = lO0 2LAI l _ e x p ( _ 2 )

(9)

where w represents a typical leaf width (w = 0.01 m), uh the windspeed at canopy height extrapolated from the normal logarithmic wind profile corrected for stability, and LAI the leaf area index estimated as four (N.P. Hanan, personal communication, 1996).

3. Site description and measurements A comprehensive measurement programme of surface fluxes was executed for HAPEXSahel (Gash et al., 1997). As part of this programme, flux measurements were obtained at the West Central Super Sites over the main vegetation covers: millet, (fallow) savannah and patterned woodland (Kabat and Goutorbe, 1995; Kabat et al., 1997). This paper is concerned with the analysis of data from the patterned woodland and fallow savannah. The measurements were taken continuously in a period of 2 months from 12 August to 12 October 1992.

3.1. Sites 3.1.1. Fallow savannah The sites used in this paper are located in the West Central Site (Kabat et al., 1997). The fallow site is characterized by a two-layer structure with bushes (Guiera senegalensis) and a herb understorey layer consisting mainly of annual grasses and herbs species. It was estimated that open soil covered 15% of the area. At the beginning of the IOP the understorey was not yet growing, and the bushes contained few green leaves. At around 15 August the understorey emerged and towards the end of September both bushes and grasses reached maximum values of their leaf area index (Fig. 2). This gave a total leaf area index towards the end of the IOP of 1.4~ The average height of the bushes was 1.8 m. At the fallow site the surface fluxes of heat and water vapour were measured by eddy correlation. The device and software are similar to that described by Moncrieff et al. (1997). It consists of a three-dimensional sonic anemometer, an additional fast

500

P. Kabat et al./Journal of Hydrology 188-189 (1997) 494-515 1.4

• • •

1.2 X

total understorey bush layer

••••••A •

• A

•

1.0

"10

._c 0.8 ~ 0.6 _J

0.4

0.2 0.0

!:

,°°ld~llll m m m O m ' ~ m m o o O ° O T,~,~,~,~,T~,~,~,~,I, ,. ,,. ,,,i.,

160

180

200

220

•

,, ,, ,, , i, ,,,,,,,,l,v,.,,,,,

240 260 D a y of y e a r

i,.,

280

,.,

, ,.i

,, . , , ,.,,

300

I

320

Fig. 2. Leaf area indices of fallow savannah bush layer (Guiera) and understorey (herbs). Data provided by

Universityof Maryland.

thermocouple, a Krypton-hygrometer and of a closed-path gas analyser. There was excellent agreement between water vapour fluxes estimated from the open-path (Krypton, Campbell Scientific Inc., Logan, UT) and the closed-path systems as well as between the sensible heat fluxes estimated from the temperature fluctuations measured by the sonic anemometer and by the fast thermocouple (Kabat et al., 1997). For the analysis in this paper, water vapour fluxes from the open-path system were used. The data were processed off line and corrected for frequency loss, sensor path separation, sensor misalignment, etc., according to Moncrieff et al. (1997) and Lloyd et al. (1997). Additional energy balance measurements consisted of shortwave radiation, net radiation, soil and heat flux. Eddy correlation measurements were made at 5.5 m; additional measurements of windspeed by cup anemometers, of air temperature and of relative humidity were taken at 4 m. Heat flux plates were installed at depths ranging from 0.01 to 0.08 m; for the analysis in this paper the top level measurement (an average of four heat flux plates) was taken. Data gaps were filled in by data from an eddy correlation system operated by Wageningen Agricultural University (Jacobs and Verhoef, 1997) over the same site. This was done for less than 10% of the days where some hourly values were missing. The correlation between the two data sets is very high (regression analysis gives r 2 of 0.92 and 0.90 for sensible and latent heat, and slopes of 0.99 and 1.02, respectively) so this procedure introduces minimum errors while creating a continuous flux data set for analysis. 3.1.2. Patterned woodland

The patterned woodland site was located about 5 krn SW from the fallow savannah site (Kabat and Goutorbe, 1995). It is an open natural forest consisting of mostly regular, N E - S W oriented stripes of vegetation interspersed with crusted bare soil (e.g. Thierry et al., 1995). The vegetated stripes are usually 10-25 m wide and 4 - 8 m high; the bare soil stripes are typically 30-80 m wide. The vegetation covers 30% of the area. The leaves of

501

P. Kabat et al./Journal of Hydrology 188-189 (1997) 494-515

the patterned woodland remained green through the whole of the IOP, in contrast to the herb layer of the Savannah. At the patterned woodland site two measurement systems were used, a Bowen ratio system and an eddy correlation system consisting only of a Solent sonic anemometer (Gill Instruments Ltd. Lymington, UK). The interchangeable Bowen ratio system has been fully described by Ogink (1995), and consists of an automated thermometer interchange system, measuring dry and wet bulb temperature at three levels (8.8, 11.1 and 13.3 m), always giving two Bowen ratios, one near, and one away from the canopy level. This system is mounted on top of a scaffolding tower. The sonic anemometer measured only the threedimensional wind field and temperature, from which sensible heat and momentum fluxes were derived. This system was mounted at a height of 14.7 m. Measurements of net radiation and soil heat flux were made separately in both vegetated and bare soil areas. 3.1.3. Representativity of the measurements

To determine the representativeness of the measurements, fetch calculations according to Gash (1986) and Lloyd (1995) were performed. These are shown in Fig. 3. Actual measurements heights were used in these calculations and the results are shown for nearneutral conditions and unstable conditions (z/L -- - 2). The figure shows that for the fallow savannah site a peaked footprint exists with major contributing areas up to 150 m. Given the typical horizontal length scale of the herb understorey and the shrubs at this site (of the order of a few metres), this implies that the fetch was adequate and that a 'homogeneous cover' was sampled. The higher measurement height at the patterned woodland (14.7 m) implies that fluxes are aggregated from a larger footprint. This is required because of the length scale of change of the bush-bare soil pattern in the patterned woodland. The peak under neutral conditions is lower and lies at around 150 m. Given that the typical length scale of this pattern is between 40 and 100 m, the complete pattern was sampled. Under more unstable conditions the footprint becomes smaller, and some sampling problems 0.03

(a)

t.-

....

e-

o~

z/L= -2.0 (unstable) z/L= -0.01 (near stable)

0.02

>

(b)

# # @ P == i= el e= t* == t~ ==

0 0

i

i= = = e |

0.01

0.00

0

200

400

Distance upwind (m)

600 0

200

400

600

Distance upwind (m)

Fig. 3. Footprint analysis of the measurementsmade at the patterned woodlandand fallow savannah.

P. Kabat et al./Journal of Hydrology 188-189 (1997) 494-515

502

may exist. As most of the data were taken in near-neutral conditions, to a good approximation radiative and ground heat fluxes can be weighted according the overall distribution of vegetation and bare soil (30% vegetation, 70% bare soil).

3.1.4. Systemperformance To be able to draw conclusions about the comparative behaviour of the two sites it is important to have a clear understanding of the performance of the systems. Lloyd et al. (1997) investigated the ability of the measurement system at the fallow site to reproduce the theoretical Monin-Obhukov profiles of Ow/U,,and from their results it is apparent that the measurements reproduced the theoretically expected values for both stable and unstable conditions. This implies that the measurements were made in a well-developed surface layer and that the sensors performed well. In Fig. 4 the ability of the eddy correlation system to measure the fluxes of latent and sensible heat is shown by comparing the sum of the turbulent fluxes with the energy balance. The agreement is good, especially on a daily basis. On an hourly basis the scatter is bigger: possible reasons for this have been given by Lloyd et al. (1997). Those workers showed that for the intercomparison periods of 3 days the residual of the energy balance was less than 1.1% of the energy balance; for the whole period of measurements the average absolute error on a daily basis in energy balance closure is 11 W m -2. Measurements over patterned woodland are difficult (e.g. Lloyd, 1995) because of the small-scale heterogeneity in vegetation and soil, and it is therefore important to assess the validity of these measurements. The measurement system set up at the patterned woodland site allowed the measurement of both the Bowen ratio and the turbulent heat flux. Typical minimum daytime values of the humidity gradients measured at the site during the lOP were of the order of 0.03 g kg -] m -l and the temperature gradients did not normally exceed 0.05°C m -t. No problems were experienced with the (ventilated) psychrometers and with the wick system, and the overall performance of the Bowen ratio interchangeable 60

E 40 LU

0

2O

0000 0

O0 O0

• 0•0 00000 • wOO0 6

•

• •

000

•

¢-.

-20 ..Q

-40 LLI

-60

' 230

i

i

i

t

i

i

i

t

i

i

i

i

i

i

J

i

*

i

i

*

i

i

i

I

I

I

I

I

I

240

250

260

270

280

290

Day of year Fig. 4. Energy balance closure for the eddy correlation system at the fallow savannah site.

503

P. Kabat et al./Journal of Hydrology 188-189 (1997) 494-515

system was very good. In Fig. 5 a comparison is shown between the sensible heat measured by the sonic anemometer, using the corrected virtual potential temperature derived from the sonic (Lloyd et al., 1997), and the sensible heat flux calculated from the energy balance measurements and the Bowen ratio. The agreement between the halfhourly values of both systems is good, giving credibility to the values of Bowen ratio energy balance fluxes used to calculate canopy conductances of the patterned woodland.

4. Results and analysis 4.1.

Rainfall

For the interpretation of the measurements of evaporation and energy partitioning in semi-arid regions, it is important to know the rainfall and soil moisture history of the sites. Rainfall in the Sahel is typically generated by squall lines, arriving roughly every 3 days

400

300

• • •

E E

• I

•

•

•

•

•

•

200

•

•

oOo •

ql

"Jl,', 1iLl

(D

% 0

"r, c(D

100

O0

•

0 m

• 0

POI0

Oqj O .~11

"1W~ •

-lOO r[ -100

I

I

I

I

I

0

1O0

200

300

400

H sonic (W m-2) Fig. 5. Comparison of sensible heat flux measured by eddy correlation and obtained from Bowen ratio energy balance measurements.

504

P. Kabat et al./Journal of Hydrology 188-189 (1997) 494-515 70 60

Fallow savannah

50

Total lOP: 218 mm

40 30

g 2o g lO ~ 0 e-

~

~ zo ._> •~ 60 a 50 40 30 20 10 0

lOP

"~

Pattemed woodland Total lOP: 265 mm

,.........,,,],,.,,,,,[,.,,,, 180

190

200

210

220

230

240

250

260

270

280

Day of y e a r

Fig. 6. Daily rainfall distribution for the two sites. Data provided by EPSAT/Niger. throughout the rainy season (Lebel et ai., 1997). Daily rainfall is shown in Fig. 6 for the patterned woodland and fallow site, and is typical of the intermittent pattern of rainfall in the Sahel. Total rainfall received at the two sites during the IOP differs by 47 mm, the fallow site receiving 218 m m and the patterned woodland site 265 mm. This difference is caused primarily by rainfall failing on the patterned woodland from Day 235 to 244. After Day 259 only the patterned woodland site received some further rainfall at Day 277. 4.2. Diurnal behaviour o f evaporation and sensible heat

The diurnal behaviour of the surface fluxes is best examined by looking at typical days under different conditions of soil moisture or antecedent rainfall. For this purpose 2 days were selected, one immediately after the last rains (Day 261) and a day 3 weeks later (Day 283) when soil moisture may start to become limiting: Fig. 7 shows the diurnal course of latent and sensible heat for the 2 days at both sites. On Day 261 (17 September) the latent heat fluxes at the patterned woodland reached a maximum level of 450 W m -2 around noon. This is the combined effect of evaporation from the stripes of woodland and the soil. Using the model described above (Eqs. (1)-(3)) it is estimated that on this day 10.5% of the total evaporation was made up by evaporation from the soil. Sensible heat fluxes are low, with a maximum of around 150 W m -2. In contrast, 3 weeks later the latent heat flux is reduced to 200 W m -2 and the soil evaporation model indicates that the total evaporation is dominated by the evaporation from the vegetation. The sensible heat flux is now increased to a level of 300 W m -2

505

P. Kabat et al./Journa/ of Hydrology 188-189 (1997) 494-515

600

.H

400

" G R ~

:

la, 1 ,

kEt

lO,

E 200 0 __ __

~

.....

~,,i,,,i,,,i,,,I,,,i,,,i

X >,,

800

E ILl

600

W.m"--'r

, , , i , , , i , , , i , ~ , I , , , i , , , i

(c)

(d)

400 200 0 0

4

8

12

16

Time (hour)

20

24

0

4

8

12

16

20

24

Time (hour)

Fig. 7. Hourly fluxes of energy and water vapour for fallow savannah (a, c) and for patterned woodland (b, d) for Day 261 (a, b) and Day 283 (c, d), Day 261 is a day after rain; Day 283 is a 'dry' day. around noon. This shift in Bowen ratio from 0.3 to 1.2 does not necessarily imply that the vegetation experiences substantial soil moisture stresses early in October. From the data of Cuenca et al. (1997) for similar sites at the Central East and Southern HAPEX-Sahel sites, it is concluded that up to a depth of 3 m sufficient water was available to the vegetation for transpiration. In the next section of this paper it is hypothesized that an increase in humidity deficit is the main cause for stomatal closure and the reduction in canopy conductance. The energy partitioning for fallow savannah follows a different pattern. On 17 September (Day 161) the evaporative fluxes from fallow savannah reach a maximum of 350 W m -2, with sensible heat fluxes of up to 100 W m -2. On 17 September the herbaceous understorey and bushes reach a maximum leaf area of 1.4 (Fig. 2). Three weeks later the evaporation has dropped to about 200 W m -2 and the sensible heat flux increased to 250 W m -2. On both days, a similar amount of net radiation (approximately 30%) is converted into soil heat flux. Average water storage in the first 2 m of the soil profile of the fallow savannah declined from 13% to 6% at the end of October (Cuenca et al., 1997). This implies that towards the end of the IOP 120 m m was still stored in the top 2 m profile. The soil water retention curves obtained at this site suggest that a substantial part of this water is still available to plants. It is likely therefore that the grasses with a shallow rooting system were starting to become water limited, but that the bushes with their deeper rooting systems were still able to retrieve water from the soil. It is relevant to note that the grass understorey accounts for 80% of the canopy leaf surface area, thereby being the most important factor determining total evaporation. The evaporation measurement thus reflects mostly processes

506

P. Kabat et al./Journal of Hydrology 188-189 (1997) 494-515

determining grass evaporation rather than bush evaporation. Indeed, this is the single most important reason for using the single-layer approach for analysis. 4.3. Long-term behaviour during l O P

The long-term behaviour of the surface fluxes for fallow savannah and patterned woodland is shown in Fig. 8, where the daily evaporation and sensible heat are plotted for the entire IOP. The patterned woodland evaporation shows a more irregular behaviour than the fallow evaporation immediately after rain. This is a direct consequence of the contribution of soil evaporation after rain, which in patterned woodland can be substantial. Daily evaporation for fallow savannah shows a maximum of 5 mm day -I immediately after rain. The corresponding maximum for patterned woodland is slightly higher at 5.8 mm day -I. Daily evaporation stabilizes at 3 mm towards the end of the IOP for both fallow savannah and patterned woodland, with both vegetation types showing a sharp decline after Day 280. Towards the end of the IOP the evaporation rate has dropped to about 2 mm day -t. These figures agree with those observed previously by Gash et al. (1991) and Culf et al. (1993) for similar vegetation at the Southern Super Site. In contrast to the similarity of the evaporative fluxes, the sensible heat fluxes show larger values for the patterned woodland than for the fallow savannah. This is shown better in cumulative plots of evaporation and sensible heat (Fig. 9). Here the cumulative evaporation of fallow savannah and patterned woodland is very similar, with large, but steady and systematic differences between the sensible heat flux from patterned woodland

56 t

200 175

(a)

150 125

E

E

100 ~_ ILl

75

so |

0

' 'i

iI , I I

i

i

230 240 250 260 270 280 6 5

fl

I

t I,

I

I

~

.

,-

.~

~.

25 0

I

230 240 250 260 270 280 (d)

(c)

g4 E 3 E ~"1-2

1 0

I

I

I

I

I

230 240 250 260 270 280 Day of year

I

I

I

i

I

230 240 250 260 270 280 Day of year

Fig. 8. Dailytotal evaporationand sensible heat for fallow savannah (a, c) and patternedwoodland(b, d) during the lOP (mm).

507

P. Kabat et al./Journal of Hydrology 188-189 (1997) 494-515 200

(a)

Fallow savannah Patterned woodland

.... 150 O" LU

E 100 E

°.°°.°.°°'''°

"r 50 0

...°..--'°''°°"

,

,

,

i

,

,

1

I

,

I

. . . . . . . . .

I

. . . . . . . . .

]

. . . . . . . . .

I

. . . . . . . . .

i

,

,

,

I

. . . .

200

}

I

(b)

150

E

100

50

0

''''i

1.2

(c)

1.0 +

0.8

~T 0.6 0.4 0"2

,, . . . . . . . . .

230

~

. . . . . . . . .

240

I .........

250

I .........

260

I .........

270

I .........

280

I

290

Day of year Fig. 9. Cumulative sensible (a) and latent (b) heat and evaporative fraction (c) for fallow savannah and patterned woodland for the lOP.

and fallow. Total evaporation from the fallow savannah during the measurement period amounts to 175 mm. For patterned woodland the total evaporation is 165 ram. The corresponding sensible heat flux totals are 47 mm and 90 mm, respectively. The consequences of this difference and the similarity in evaporation are explored in the discussion. The apparent difference in energy partitioning between the patterned woodland site and the fallow savannah is also reflected in the evaporative fraction, here defined as the quotient of the latent heat flux to the sum of latent and sensible heat flux. The maximum values of the fallow savannah are around 0.9, whereas the average maximum evaporation fraction of patterned woodland is around 0.7 (Fig. 9(c)). Patterned woodland uses

508

P. Kabat et al./Journal of Hydrology 188-189 (1997) 494-515

therefore about 20% less of its available energy for evaporation, compared with the fallow savannah. The higher total available energy at the patterned woodland, however, compensates for this lower value, thus sustaining similar overall levels of evaporation as for the savannah site. Both sites show a seasonal course, with an increase in evaporative fraction from 0.4 or 0.5 to higher values early in the season, which are then maintained during the wet season, and a subsequent decline after the last rains. The increases in the evaporative fraction for patterned woodland also clearly coincide with rainfall events and thus signify the contribution of soil evaporation to the total. 4.4. Surface conductance

Surface conductances were calculated from Eq. (4) and Eq. (8) for both fallow savannah and patterned woodland. In Fig. 10 the surface conductance is shown for the lOP for both sites. The (linearly) averaged surface conductance during the day for the patterned woodland decreases slowly during the lOP from values of around 25 mm s -I to values of around 10 mm s -I (Fig. 10). These values were calculated with an eddy decay coefficient of 1.5. Considerable uncertainty is associated with this value, but using other values produced more scatter in the resulting conductances, which was thought to be an artefact of the calculation procedure rather than real. Maximum hourly values are very high, around 40 m m s -t, as is shown in Fig. 11 where for a day early during the IOP and late in the IOP the corresponding diurnal behaviour of the surface conductance is shown. Conductances are high immediately after sunrise, with a decline after 10.00 h local time. This decline may be caused either by vapour pressure deficit or possibly by some physiological response expressing the inability of the root systems to supply enough water to the leaves for evaporation. The course of surface 50

o •

Fallow savannah Patterned woodland

40

20

10

~

230

•

240

250

260

270

280

290

D a y of y e a r Fig. 10. Daily average canopy conductance during the lOP for fallow savannah and patterned woodland. The standard error of the mean is also shown.

509

P. Kabat et al./Journal of Hydrology 188-189 (1997) 494-515 50

(a)

(b)

40 i

O'J

E

30

g

20

o}

10 0

I

'

I

'

l

'

I

'

I

'

I

50

'

I

'

I

'

I

'

I

'

I

'

I

'

I

(c)

'

I

'

(d)

40 i

(~ 30 E IE 20 o~ 10

6

8

10

12

14 16

Time (hour)

18

6

8

10

12

14

16

18

Time (hour)

Fig. 11. Hourly canopy conductance for fallow savannah (a, c) and for patterned woodland (b, d) for Day 262 (a, b) and Day 283 (c, d).

conductance suggests that the stomata of the patterned woodland vegetation are open early during the day, with small or negligible influence of radiation levels. Towards the end of the period the conductances have fallen to 20 mm s -I early during the day and to 3 m m s at sunset. For fallow savannah the conductances are lower, but follow a similar trend during the IOP. The values obtained in this study correspond closely to those observed by Gash et al. (1991), who noted values dropping from 5 to 1.5 mm s -I a few days after the last rains. Maximum hourly values obtained in this study are up to 15 mm s -I (Fig. 11), similar to those observed by Gash et al. (1991) immediately after the last rains. The values clearly show a seasonal trend related to growth and decline of the understorey layer (Fig. 2). The diurnal variations are much less pronounced than at the patterned woodland site. Early in the IOP the conductances are high, and stay high during most of the day, whereas later they have reduced in magnitude but still show little variation during the day. At the end of the IOP both the magnitude and diurnal amplitude have decreased, similar to the pattern found in patterned woodland.

4.5. Response to vapour pressure deficit The movement of the Inter Tropical Convergence Zone (ITCZ) southwards towards the end of the IOP causes a large increase in humidity deficit in the lower boundary layer (Dolman et al., 1997). This is shown in Fig. 13(e) (below), where the measured average vapour pressure deficit between 11:00 and 13:00 h GMT is plotted for the entire IOP. This change is associated also with a reduction in rainfall and a corresponding decline in soil

P. gabat et al./Journal of Hydrology 188-189 (1997) 494-515

510

moisture levels. It is therefore interesting to investigate the response of the local vegetation to such a change. The fallow savannah data and patterned woodland data both show a decline in evaporation and surface conductance towards the end of the IOP. In the case of the patterned woodland this reduction is unlikely to be caused by soil moisture stress, as enough water was still available for the deep-rooting forest vegetation. In the patterned woodland also a pronounced daily variation was observed, which needs to be explained. What then, causes the decline? In Fig. 12 the responses of the patterned woodland and fallow savannah conductances are plotted against vapour pressure deficit. In the case of patterned woodland this is the within-canopy deficit, rather than the reference level value. In both cases a response curve may be identified with higher values at low deficits and lower values at high deficits. Hanan and Prince (1997) concluded that vapour pressure deficit was the most significant environmental variable explaining the variation in conductances of the four most common 50-]

o o 0

(a) 0

40

o

o° o "T,

30

(/)

o 0

E E "~

20

•

o

beforeclay261

after day 261

o

0

o

~

o

o O

•

••

o

50

(b)

4O

"T,U )

3O

E E

v

•

20

10

0 0

10

20

30

40

i 50

D O (mbar)

Fig. 12. Relation of surface conductance to specific humidity deficit. In the case of patterned woodland (b), this refers to the in-canopy deficit.

P. Kabat et al./Journal of Hydrology 188-189 (1997) 494-515

511

species of the HAPEX-Sahel study area. C3 species were found to be more sensitive to increasing vapour pressure deficit than C4 species. Similar conclusions may be inferred from the results of Huntingford et al. (1995), who optimized a surface conductance model for fallow savannah. For fallow savannah all data points after Day 260 (Fig. 12) lie on a narrow horizontal band. The fallow savannah consisted of roughly 20% by C 3 bushes (G. senegalensis) and of a continuous herb layer which early in the season contained C 3 species, but during the 50

• L,.

4O

E

30

c~~

2o

D (above canopy)

I

'

I

'

I

(a)

I

I

I

50

O canopy) Oo(above (in canopy)

:

'

I

'

I

'

1

I

(b)

I

'

-1

(d)

(c)

4O 3o

2o a-

10 I

0

'

4

I

I

t

I

I

8

12

16

20

24

0

4

(hour)

Time

8

12

16

20

60o

50

(e)

Fallow savannah Patterned woodland

•

24

Time (hour)

O

4O

Q o*e

o

30 • • O0 go

a

I I oils • 20

•

•0 • • •

•

•

•

•

......... 230

O

0

O0

0

o

OOo o

o 0

o

0 e•

o Coo •

0

o

o

0

0

0

10,9 ...... 240

~)

• OO

•

10

0

o

Oo •

I ......... 250

~......... 260

i ......... 270

I ......... 280

i 290

Day of year Fig. ]3. Vapour pressure deficit for fallow savannah (a, c) and patterned woodland (b, d) for Day 262 (a, b) and

Day 283 (c, d). Average vapour pressure deficit between 11:00 and 13:00 h GMT for the lOP (e).

512

P. Kabat et al./Journal of Hydrology 188-189 (1997) 494-515

season C4 grasses became more important. Vapour pressure deficit at the fallow savannah site shows an increase between Days 262 and 283 (Fig. 13), but, for example, the maximum value of 26 mbar at 16:00 h GMT on Day 283 is reducing the vegetation conductance of the herb understorey only by about 20%, without accounting for possible effects of other regulating variables, such as soil moisture and solar radiation (Huntingford et al., 1995; Hanan and Prince, 1997). Indeed, this suggests that another factor in addition to vapour pressure deficit is causing the decline in fallow savannah conductances. This is most likely to be soil moisture in the top layer, on which the water supply of the grassy understorey with its shallow rooting depth depends. Patterned woodland vegetation mainly consists of the C3 tall shrubs and trees (Prince et al., 1995). The within-canopy vapour pressure deficit at the patterned woodland site increased sharply in the second half of the season (Fig. 13), with a daily maximum of about 45 mbar on Day 283. The corresponding reduction of the conductance for C3 species, derived from Hanan and Prince (1997) and Huntingford et al. (1995) varies between 0.4 and 0.2. Therefore, for the patterned woodland, most of the reduction taking place after Day 260 (roughly when the vapour pressure deficit starts to rise above 20 mbar owing to the ITCZ moving southwards) can be attributed to a vapour pressure deficit response, as, at this time, no substantial moisture stress will occur at the woodland site (see also Fig. 10). This response may also explain part of the diurnal variation in surface conductance as observed in Fig. 11. For patterned woodland, the conductances are highest in the morning and start to decline when the vapour pressure deficit rises above 20 mbar, between 09:00 and 11:00 h local time. Increasing air temperature results in an increase of the vapour pressure deficit in the late afternoon, causing further decrease of the conductance. By the time the vapour pressure becomes less limiting, the stomata are likely to be closing as a result of low radiation levels. The diurnal responses for fallow savannah are much less pronounced. The diurnal variation of vapour pressure deficit for the fallow savannah site (Fig. 13) for Days 262 and 283 shows that the maximum vapour pressure deficit was about 25 mbar, a value not severely limiting the conductances of the C4 understorey. This, in combination with the early morning light saturation (Gash et al., 1991; Hanan and Prince, 1997), explains the lack of diurnal variation in conductances for fallow savannah. The large difference between within-canopy vapour pressure deficit and deficit at the reference height at the patterned woodland towards the end of the IOP with dry conditions (Fig. 13) deserves further mention. Using the reference height of vapour pressure deficit at Day 283, the reduction in canopy conductance owing to the vapour pressure deficit (Huntingford et al., 1995) would be underestimated by no less than 30%. This clearly illustrates the conceptual advantages of the sparse canopy models.

5. Discussion The measurements used in this paper were obtained by what can be considered state-ofthe-art techniques of micrometeorology. Usually these have been applied for relatively short periods at well-managed field sites, and the present longer-term deployment of this

P. Kabat et aL/Journal of Hydrology 188-189 (1997) 494-515

513

instrumentation in a harsh environment such as the Sahel is therefore a considerable SUCCESS.

It is tempting to speculate on the apparent conservative behaviour of both fallow and patterned woodland evaporation. Culf et al. (1993) suggested that the total seasonal evaporation from patterned woodland may approach the total rainfall amount. Were this the case also for fallow savannah--and in the longer term, rainfall represents a natural upper limit to evaporation--then the implication may be that two different, but relatively stable, ecosystems may evolve in a similar climate. Although these systems would use the same amount of water and optimize their strategy on the availability of water, the physiological behaviour of these systems and their surface conductances may be totally different. Expressed on an area average basis they would also support similar amounts of leaf area. This has implications for the modelling of vegetation or biome dependence on climate (e.g. Woodward, 1990). The values of vegetation surface conductance for the patterned woodland site are high (e.g. Schulze et al., 1994) but not unrealistic. Taking the similarity in total evaporation rates between fallow savannah and patterned woodland as a starting point, it is likely this can only be maintained by a high surface conductance of the vegetation and additional supply of energy drawn from the bare soil boosting the within-canopy deficit. Comparing leaf and stand level measurements of CO2 uptake of these two vegetation types in relation to their water use and possible different survival strategies will be the subject of a future paper. Because of the similarity in evaporation rates it is likely that areal groundwater recharge under natural conditions in the Sahel is not very dependent on vegetation type, except perhaps when the different vegetation types generate different surface runoff and overland flow regimes. This may simplify the estimation of groundwater recharge in the Sahel considerably. However, the impact of patterned woodland and fallow savannah on the atmosphere is different. The large patches of bare soil in the tiger bush release a considerable amount of extra sensible heat compared with the fallow savannah, and this may have consequences for boundary-layer growth. Preliminary modelling results indicate that such a response may exist and can give rise to local circulations (Dolman et al., 1995). This small-scale variability would have to be taken into account when modelling boundary-layer growth in the Sahel. The large differences between the conductances of fallow and natural forest may have serious consequences for the parameterization of evaporation and CO2 exchange in the Sahel in large-scale models such as GCMs. The implication is that the parameterization schemes need to take into account the response of physiologically different vegetation types (C3 vs. Ca) to vapour pressure deficit. Also, as is the case for the patterned woodland, more complex parameterization schemes might be needed to model the canopy conductances and evaporation. GCM studies into the effects of land-surface properties in the Sahel on climate (e.g. Xue and Shukla, 1993) should therefore profit from implementation of more complex and more 'physiologically driven' land surface schemes. The change in humidity deficit is associated with the shift in the position of the ITCZ. This large-scale phenomenon feeds back on the local vegetation, which in the case of patterned woodland tries to limit its water loss through closing of the stomata, thereby changing the energy partitioning and the amount of heat and moisture put back into the boundary layer. Wang et al. (1997) speculated that the existence of a north-south gradient

514

P. Kabat et al./Journal of Hydrology 188-189 (1997) 494-515

in soil moisture may uphold the movement of the ITCZ southwards through the generation of a secondary circulation system. On the basis of the present analysis, it may be postulated that the vegetation itself, especially the patterned woodland, through its response to vapour pressure deficit, may contribute to such a phenomenon. Although this hypothesis is largely speculative at this point in time, it suggests the important role vegetation may play in the regional interaction with the climate.

Acknowledgements The authors would like to thank Anne Verhoef from Wageningen Agricultural University for allowing the use of her data, Thierry Lebel for making the EPSAT-Niger rainfall data available, and Steve Prince and Niall Hanan of the University of Maryland for allowing the use of their leaf area index data. This work was partly funded by the European Community under the EPOCH and Environment programmes EPOCH-CT90-0024C(DSCN) and ENVIRONMENT EV5V-CT91-0033, and by the Dutch Ministry of Agriculture, Nature Management and Fisheries. References Blyth, E.M. and Dolman, A.J., 1995. The roughness length for heat of sparse vegetation.J. Appl. Meteoro].,34: 583-585. Chamey, J.G., 1975. Dynamics of deserts and drought in the Sahel. Q. J. R. Meteorol. Soc., 101: 193-202. Culf, A.D., Allen, S.J., Gash, J.H.C., LIo)~d, C.R. and Wallace, J.S., 1993. The energy and water budgets of an area of patterned woodland in the Sahel. Agric. For. Meteorol., 66: 65-80. Cuenca, R.H., Brouwer, J., Droogers, P., Galle, S., Gaze, S., Sicot, M., Stricker, J.N.M., Angulo-Jaramillo, R., Boyle, S.A., Bromley, J,, Chebhouni, A.G., Cooper, J.D., Dixon, A.J., Fries, J.-C., Gandah, M., Gouda, J.-C., Laguerre, L., Soet, M., Stewart, H.J., Vandervaere, J.-P. and Vauclin, M., 1997. Variability of profile and surface soil moisture content and soil physical property measurement during HAPEX-Sahel intensive observation period. J. Hydrol., this issue. Dolman, A.J., 1993. A multiple source land surface energy balance model for use in General Circulation Models. Agric. For. Meteorol., 65: 21-45. Dolman, A.J., Lloyd, (~.R. and Culf, A.D., 1992. Aerodynamic roughness of an area of natural open forest in the Sahel. Ann. Geophys., 10: 930-934. Dolman, A.J., Hutjes, R.W.A., Kabat, P. and Prince, S.D., 1995. The effect of small scale heterogeneity on the area average fluxes during HAPEX-Sahel. Ann. Geophys.o 13(Suppl.): C310. Dolman, A.J., Bessemoulin, P. and Culf, A.D., 1995. Observations of boundary layer structure during HAPEXSahel Intensive Observation Period. J. Hydrol., this issue. Garrat, 1994. The Atmospheric Boundary Layer. Cambridge University Press, Cambridge, 316 pp. Gash, J.H.C., 1986. A note on estimating the effects of a limited fetch on micrometeorological evaporation measurements. Boundary-Layer Meteorol., 35: 409-413. Gash, J.H.C., Wallace, J.S., Lloyd, C.R., Dolman, A.J., Renard, C and Sivakumar, M.V.K., 1991. Measurements of evaporation from fallow Sahelian savannah at the start of the dry season. Q. J. R. Meteorol. Soc., 117: 749-760. Gash, J.H.C., Kabat, P., Amadou, M., Bessemoulin, P., Billing, H., Blyth, E.M., de Bruin, H.A.R., Elbers, J.A., Friborg, T., Harrison, G., Hollwill, C.J., Lloyd, C.R., Lhomme, J.-P., Moncrieff, J.B., Monteny, B., Puech, D., Sogaard, H., Tuzet, A. and Verhoef, A., 1997. The variability of evaporation during the HAPEX-Sahel Intensive Observation Period. J. Hydrol., this issue. 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., 1994.

P. Kabat et al./Journal of Hydrology 188-189 (I997) 494-515

515

HAPEX-Saheh a large scale study of land atmosphere interactions in the semi-arid tropics. Ann. Geophys., 12: 53-64. Hanan, N.P. and Prince, S.D., 1997. Stomatal conductance of West Central Supersite vegetation in Hapex-Sahel: measurements and empirical models. J. Hydrol., this issue. Huntingford, C., Allen, S.J. and Harding, R.J., 1995. An intercomparison of single and dual-source vegetationatmosphere transfer models applied to transpiration from Sahelian savannah. Boundary-Layer Meteorol., 74: 397 -418. Jacobs, A. and Verhoef, A., 1997. Soil evaporation from a sparse natural vegetation estimated from Sherwood numbers. J. Hydrol., this issue. Kabat, P. and Goutorbe, J.P., 1995. HAPEX II-Sahel/Phase I. Final Report EC Contract EPOCH CT90-0024C. DG XII, EPOCH/ENVIRONMENT, Brussels. Kabat, P., Prince, S.D. and Prihodko, L., (Editors), 1997. HAPEX-Sahel West Central Supersite: Methods, Measurements and Selected Results. SC-DLO, Wageningen; University of Maryland, College Park, MD, in press. Kelliber, F.M., Leuning, R., Raupach, M.R. and Schulze, E.D., 1995. Maximum conductances for evaporation from global vegetation types. Agric. For. Meteorol., 73: 1-16, Lebel, T., Taupin, J.D. and LeBarbe, L., 1997. Space-time fluctuations of rainfall during HAPEX-Sahel. J. Hydrol., this issue. Lloyd, C.R., 1995. The effect of heterogeneous terrain on micrometeorological measurements: a case study from HAPEX-Sahel. Agric. For. Meteorol., 73: 209-216. Lloyd, C.R., Bessemoulin, P., Cropley, F.D., Culf, A.D., Dolman, A.J., Elbers, J.A., Heusinkveld, B., Moncrieff, J.B., Monteny, B. and Verhoef, A., 1997. An intercomparison of surface flux measurements during HAPEXSahel. J. Hydrol., this issue. Moncrieff, J.B., Massheder, J.M., de Bruin, H., Elbers, J., Friborg, T., Heusinkveld, B., Kabat, P., Scott, S., Soegaard, H. and Verhoef, A., 1997. A system to measure surface fluxes of momentum, sensible heat, water vapour and carbon dioxide. J. Hydrol., this issue. Ogink, M.J., 1995. Modelling surface conductance and transpiration of an oak forest in the Netherlands. Agric. For, Meteorol., 74: 99- I 18. Paulson, C.A., 1970. The mathematical representation of wind speed and temperature profiles in the unstable atmospheric surface layer. J. Appl. Meteorol., 9: 857-861. Priestley, C.H.B. and Taylor, R.J., 1972. On the assessment of surface heat flux and evaporation using large scale parameters. Mon. Weather Rev., 100: 81-92. Prince, S.D., Kerr, Y.-H., 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., Monteny, B., Said, F., Sellers, P. and Wallace, J.S., 1995. Geographical, biological and remote sensing aspects of the Hydrological Atmospheric Pilot Experiment in the Sahel. Remote Sens. Environ., 51: 215-234. Ritchie, J.T., 1972. Model for predicting evaporation from a row crop with incomplete cover. Water Resour. Res., 8: 1204-1213. Schulze, E.D., Kelliher, F.M., Koerner, C., Lloyd, J. and Leuning, R., 1994. Relationships among maximum stomatal conductance, ecosystem surface conductance, carbon assimilation rate, and plant nitrogen nutrition: a global ecology scaling exercise. Annu. Rev. Ecol. Syst., 25: 629-660. Shuttleworth, W.J. and Gurney, R.J., 1990. The theoretical relationship between foliage temperature and canopy resistance in sparse crops. Q. J. R. Meteorol. Soc., 116: 497-519. Shuttleworth, W.J. and Wallace, J.S., 1985. Evaporation from sparse crops--an energy combination approach. Q. J. R. MeteoroL Soc., I 11: 839-855. Thierry, J.M., D'Herbes, J-M. and Valentin, C., 1995. A model simulating the genesis of banded vegetation patterns in Niger. J. Ecol., 83: 497-507. Wallace, J.S. and Hollwill, C.J., 1997. Soil evaporation from tiger-bush in Niger. J. Hydrol., this issue. Wang, M.M.K., Smith, E.A., Bessemoulin, P., Culf, A.D., Dolman, A.J. and Lebel, T., 1997. Variability in boundary layer structure during HAPEX-Sahel wet-dry transition. J. Hydrol., this issue. Warrilow, D.A., Sangster, A.B. and Slingo, A., 1986. Modelling of land surface processes and their influence on European climate. DCTN 38. National Meteorological Library, Bracknell, UK. Woodward, F.I., 1990. Climate and Plant Distribution. Cambridge University Press, Cambridge, 174 pp. Xue, Y. and Shukla, J., 1993. The influence of land surface properties on Sahel climate. J. Climate, 6: 2232-2245.