Journal of
Hydrology ELSEVIER
Journal of Hydrology 188-189 (1997) 482-493
Flux heterogeneity and evapotranspiration partitioning in a sparse canopy: the fallow savanna A. Tuzet*, J-F. Castell, A. Perrier, O. Zurfluh Bioclimatologie INRA-INAPG, 78850 ThivervaI-Grignon France
Abstract This paper focuses on in situ measurements obtained during the intensive observation period of the HAPEX-Sahel experiment. Micrometeorological measurements and trunk sap flow monitoring were combined to analyse transfer characteristics of a fallow savanna site within the East Central Supersite. Results show that the shrub canopy heterogeneity induces a large spatial variability of solar irradiance, soil heat flux and "sensible and latent heat fluxes at the grassland level. This variability is induced by both a "shade effect" and a "wake effect". Both shrubs and grassland provide sources of vapour, but the partitioning of evapotranspiration between these two components varies considerably with soil surface water availability.
1. Introduction A good understanding of the overall transfer behaviour between the atmospheric boundary layer and the underlying surface will be provided by determination of transfer characteristics for specific surface types occurring in nature. In the HAPEX-Sahel experimental domain, substantial areas correspond to sparsely-vegetated semi-arid rangelands. This tropical grassland with a more or less even scattering of small trees, shrubs and bare soil constitutes a heterogeneous surface. This paper deals with mass and energy exchange characteristics over such a surface, the fallow savanna. Mass and energy exchanges above homogeneous surfaces have been the subject of many studies, and are well documented (Thom, 1975; Perrier, 1982; Finnigan and Raupach, 1987; Wilson, 1989), but less attention has been given to the question of transfer characteristics in partially vegetated areas. First attempts to analyse the microclimatic * Corresponding author. 0022-1694/97/$17.00 © 1997- ElsevierScience B.V. All rights reserved PII S0022-1694(96)03189-7
A. Tuzet et aL/Journal of Hydrology 188-189 (1997) 482-493
483
effects of a sparse shrub canopy on the underlying grassland were proposed by Tuzet et al. (1995a, 1995b). Several models based on the Penman-Monteith equation (Monteith, 1965) calculated evaporation from sparse canopies using resistance networks and energy combination formulas (Shuttleworth and Wallace, 1985; Kustas, 1990; Lafleur and Rouse, 1990), and attempts to estimate aerodynamic roughness parameters for sparse surfaces from experimental measurements have been made (Matthias et al., 1990; Lloyd et al., 1992; Raupach, 1992). However, none of these studies offers a detailed approach which takes into account the major effects of the vegetation heterogeneity. These studies are particularly difficult because the structure of heterogeneous surfaces imposes a random source/sink distribution differing for each transferred quantity (momentum, heat and moisture). Trees and shrubs are the main roughness elements influencing the vertical momentum transfer. They contribute to a large part of the total drag exerted by the surface. Between the obstacles, the herbaceous vegetation is rather smooth and homogeneous. But far away from the surface, the profiles of temperature, humidity and wind are influenced by the smooth herbaceous vegetation as well as by the sparsely distributed shrubs and trees. Both shrubs and grass provide sources of water vapour, but the partition of evapotranspiration between these vegetation elements can vary with soil water status. The objective of this study was to examine the effect of sparsely distributed shrubs on heat and momentum fluxes using field data from fallow savanna. In this paper, we present experimental results revealing the shrub canopy heterogeneity effects. We also describe the evapotranspiration partition between shrubs and grassland and its interaction with soil water status.
2. Materials and methods
2.1. Site and vegetation The experiment was carded out in an area of fallow savanna over the East Central Supersite (Monteny et al., 1997), from 14 August to 3 September 1992. The site consists of a ground layer of annual herbs and grasses, with scattered bushes (almost entirely Guiera senegalensis). The area had not been cropped for about 7 - 8 years, allowing the vegetation to regenerate. Since all the oldest stems are of the same age (depending on the date of fallowing) their mean height is fairly homogenous (2-3 metres), but the volume occupied by each shrub can vary considerably depending on the number and age of the shoots of which it is formed. A shrub crown map for the experimental area is given in Fig. 1. From this map we estimated that about 18% of the ground area was covered by bushes. The mean values of bush diameter, height, shrub shoot number and shoot diameter are given in Table 1.
2.2. Micrometeorological measurements 2.2.1. Microclimatological data Standard meteorological data were measured at the top of a six metre tower, which was set up in the middle of a clearing (see Fig. 1). Incoming global solar radiation was
484
A. Tuzetet al./Journal of Hydrology 188-189 (1997) 482-493 60
Nerth
oo 50
°o
....
o.,.
.~o.
~o
c~.~6~...+ o
...
°-% o %.+~
o
0
.
" '
40
o-,-
oo
o
6 ° 2°0"°
30
e o
+o
o
"• 0 o °`4> 0 2 bo
20
~.Q'
iO --
©
o@o -iO
,
-iO
[
©
0
I I0
.
o~ CoN -I
20
distance
L
I
I
I
30
40
50
60
(m)
Fig. 1. Shrub crown projection map. The location of the meteorological tower (T), and the transect of eddy correlation measurement masts (1, 2, 3) are also plotted. measured by a pyranometer (Eppley Scientific Instruments, Newport, RI, USA). At this level, net radiation was also measured using a net pyrradiometer (model S - l , Swissteco Instruments, Oberriet, Switzerland). A i r temperature and wet bulb temperature were measured using c o p p e r - c o n s t a n t a n thermocouples. A n aspiration rate of about 4 m s -t was provided by 24 V fans. W i n d speed was measured using a cup anemometer (model CE 155, Cimel Electronique, Paris, France). The outputs from these and all other micrometeorological sensors were measured by a data-logger (CR10, Campbell Scientific, Logan, UT, USA) every 5 s and then recorded as 30-min averages. 2.2.2. Canopy microclimate To derive an estimation of the microclimate within the canopy, air temperature, wet bulb temperature, net radiation and wind velocity were measured on the same tower, 1.5 m Table 1 Mean value and standard deviation of bush crown diameter, bush height, shoot number per bush and shoot diameter
Mean Std.
Crown diameter m -z
Bush height m -~
Shoot number bush-j
Shoot diameter mm 4
2.3 0.9
2.1 0.8
5.0 2.6
26.8 14.4
A. ruzet et aL/Journal of Hydrology 188-189 (1997) 482-493
485
above the surface. At the ground level, measurements of solar radiation were made at four different positions by linear pyranometers (sensors designed and constructed at the Department of Bioclimatology, INRA, France) to characterize the spatial variability (illuminated and shaded areas).
2.2.3. Eddy correlation measurements The eddy correlation technique (Kaimal, 1975; Verma et al., 1989) was used to measure sensible and latent heat fluxes. The instrumentation included four one-dimensional sonic anemometers (model CA27, Campbell Scientific), fine wire (0.025 mm) thermocouples (model 127, Campbell Scientific) and two Krypton hygrometers (model KH20, Campbell Scientific), with a 10-ram pathlength. Sampling and recording of the data were carried out with a data-logger (model 21X, Campbell Scientific). The signals were sampled at 10 Hz. Fluxes were obtained from covariances computed over 30-min averaging periods. The latent heat flux values were corrected for the variation in air density due to simultaneous transfer of sensible heat and water vapour following the method of Webb et al. (1980). Vertical velocity, temperature and moisture fluctuations were measured at two levels, above (6 m) and within the shrub canopy (mast location is given on Fig. 1). The two other one-dimensional sonic anemometers were put within the canopy along a transect including the mast (Fig. 1). These three measurement locations were used to characterize the spatial variability of sensible heat flux within the vegetation, at different heights ranging from 0.75 to 2 m. 2.3. Soil and plant measurements 2.3.1. Soil heat flux A set of seven heat flux plates (model 610, CW Thornthwaite Associates, Elmer, NJ, USA) were inserted below the surface (at a depth of about 2 mm) to analyse the spatial variability of soil heat flux. The location of the sensors was chosen in relation to illuminated and shaded areas. Three sensors (G1, G5, G7) were in the sun in the morning and shaded in the afternoon; others (G2, G4) were more illuminated in the afternoon. The other two (G3, G6) were located under shrubs, and remained shaded almost all day long. The sensors were placed only 2 mm below the surface to minimize heat storage above the sensors. Unfortunately, measuring G just below the soil surface leads to significant error because the plates can act as vapour barriers and distort the heat flow pattern significantly (Brutsaert, 1982). In such conditions, the heat flux data can only be used for a qualitative study of the spatial variability. 2.3.2. Shrub sap flow Constant-power heat balance gauges (model SGB-50, Dynamax, Houston, TX, USA) were placed on five Guiera senegalensis trunks. Stem diameters ranged from 45 to 55 mm. Each gauge was installed as recommended by the manufacturer and covered with aluminium foil. To prevent large temperature variations resulting from direct solar radiation heating the device, an additional shield (polystyrene foam covered by aluminium foil) was placed around each device. An air space allowing natural ventilation was left between this shield and the sap flow gauge. Three stems were located in a relatively dense clump. They
A. Tuzet et aL/Journal of Hydrology 188-189 (1997) 482-493
486
were chosen in order to get three different exposures of the major part of their foliage (east, west and south). The two other stems were chosen in relatively isolated bushes. Their foliage was respectively exposed to the east and to the west.
2•3.3. Stomatal conductances Shrub stomatal conductances were measured using an automatic porometer (model MK3, Delta T Devices, Ltd, Cambridge, UK). Measurements were made every hour for six days, on both shaded and illuminated leaves on gauged stems. Each hourly data set comprised forty measurements made on the eastern and western parts of the shrubs, at two height levels (lower and upper parts of the crown).
3. Results and discussion
3.1. Shrub canopy heterogeneity effects 3.1.1. Shade effect The spatial heterogeneity of the shrub canopy leads to a non-uniform radiation interception. In Fig. 2(a), values of global radiation measured below and between shrubs are shown• Also shown in this figure are the values of global radiation measured above the canopy• Global radiation at each measurement position on the ground is a function of shrub shading. The areas exposed to sunlight change with time of day. The result, Rsl, which corresponds to a sensor located under a shrub, shows nearly all-day-long radiation values of about half those measured at the reference level (6 m). At surface positions in direct sunlight, the measured values of global radiation are always lower (about -10%) than those measured above the canopy. This attenuation could be explained by the fact that the presence of shrubs modifies the amount of diffuse radiation received at the ground. If this is the case, this result suggests that the downward radiation reflected by shrub foliage is always less than the incoming diffuse radiation.
1000
800
R* IW/m2) "~-
Rslla
~
Ral •
SO0
b
G (W/m2) 160
100
Ra2
.... Rd Re4
400
20G
1 2 $ 4 $ 8 7 8 9 10 1112 t:11418 te171819 20 21~J n 24
Time (Houre UT)
'i!
0 1 2 3 4 6 8 7 8 g 10 11121~ t418 m171819 ~0 212Z ~8 24
Time (Houre UT)
Fig. 2. Spatial variability of global radiation (a) and soil heat flux (h) at the soil surface on 1 September 1992. (a) Daily cycle of global radiation Rs measured above the canopy (6 m) and at four locations below and between shrubs (R si-R ~). (b) Daily cycle of ground heat flux G measured at seven locations below and between shrubs.
A. Tuzet et aL/Journal of Hydrology 188-189 (1997) 482-493
487
The large positional variations of solar irradiance which occur beneath the shrub layer suggest that soil heating and cooling will be non-uniform. The diurnal courses of soil heat flux measured at seven locations below and between shrubs are plotted in Fig. 2(b). Maximum soil heat flux occurred at a location exposed to solar heating. In the morning, there is a factor of almost 2.5 between the values of soil heat flux measured in illuminated or shaded areas. This difference is not so pronounced in the afternoon. These results suggest that the ground heat flux rapidly increases when the position is illuminated by direct beam irradiance, then decreases more slowly after becoming shaded later in the day. During the night, the least negative soil heat flux occurs below the shrubs, where the radiative losses are reduced.
3.1.2. Wake effect Within the shrub layer, shrubs interact with the mean flow by extracting momentum from the wind, producing turbulence in the form of wakes behind the shrubs, and breaking down large-scale turbulent eddies into smaller scale motions. Within a wake region, the wind speed is less than the surrounding region, and the exchanges are reduced. Therefore, the shrub canopy heterogeneity leads to a spatial variability of vertical transfers (momentum, heat and water vapour). Most authors recommend that care should be taken when making eddy correlation measurements within the canopy since eddy size varies directly with distance from obstacles (ground or shrubs) (Tanner et al., 1985; Wyngaard, 1991). The distribution of eddy sizes contributing to vertical transport creates a range of frequencies that has to be taken into acount for eddy correlation measurements. The sensors must have a small enough response time to measure the frequencies at the higher end of the range, while flux averaging times must be long enough to include frequencies at the lower end. A nondimensional frequency bandwidth combining these desired features and recomended by many investigators (McBean, 1972; Kaimal et al., 1972) is given by 10 -3 --
