Journal of
Hydrology ELSEVIER
Journal of Hydrology188-189 (1997) 869-877
Representing heterogeneity at the Southern Super Site with average surface parameters E.M. Blyth Institute of Hydrology, Wallingford OXIO 8BB, UK
Abstract
Evaporation data from the three sub-sites of the Southern Super Site of HAPEX-Sahel are used to assess whether spatial average values of surface resistance or conductance can represent the mixture of vegetation types present in this region. The result of the analysis shows that a linear average of the surface conductances is a good estimate of the effective surface conductance of this area.
1. Introduction One of the central aims of HAPEX-Sahel (Hydrologic Atmospheric Pilot Experiment) is to integrate local-scale measurements (100 m) up to the General Circulation Model (GCM) scale (100 km) (Goutorbe et al., 1994). There is spatial heterogeneity at almost any scale, each with a different underlying cause. For instance, the semi-arid nature of the climate means that full vegetative cover cannot be supported. Therefore, at the scale of a few metres, there can be large variations in the amount of exposed bare soil. The effect that this heterogeneity has on the evaporation fluxes averaged over a few hundred metres has been studied by Blyth and Harding (1995). While the rainfall can be considered spatially uniform at the scale of a few metres, smallscale undulations cause the surface water to form into ponds. Therefore, the infiltration of rainwater into the soil is heterogeneous and this source of small-scale heterogeneity is the subject of another study (see Wallace et al., 1994). The surface fluxes measured by the 'Hydra' at the Southern Super Site in HAPEX-Sahel are the area average fluxes over a few hundred metres. But, at the scale of a few kilometres, there are changes in the vegetation type between tiger bush, fallow savannah and millet. The variation in surface conditions caused by this difference in vegetation is the subject of this paper. Methods are investigated to integrate from the surface flux measurements over each vegetation type up to the 10-km scale. 0022-1694/97/$17.00 © 1997- Elsevier Science B.V. All rights reserved PII S0022-1694(96)03175-7
870
E.M. BlythlJournal of Hydrology 188-189 (1997) 869-877
Rainfall variability also exists at the scale of a few kilometres and is the subject of another study (see Goutorbe et al., 1994), while Taylor et al. (1997) look at the effects of rainfall variability at the scale of 10 km and above. In the absence of relevant data, aggregation research has so far concentrated on theoretical heterogeneous landscapes. For instance, Ranpach (1991) and Lhomme (1992) developed formulae for effective surface and aerodynamic resistances which would correctly predict the area average evaporation and the area average surface temperature, respectively. However, neither of the proposed methods could predict both area-average evaporation and surface temperature simultaneously. Dolman (1992) and Bonan et al. (1993) studied the potential errors in using simple averages of surface parameters to represent heterogeneous terrain where the surface roughness does not change. They both found that the greatest errors in predicted evaporation occurred when the average surface resistance to evaporation was high. This conclusion is particularly relevant to modelling represent
E.M. Blyth/Journal of Hydrology 188-189 (1997) 869-877
871
resistance can be used directly, with no correction needed for their dependence on environmental variables. The first criterion for filtering the data to obtain a data set with uniform conditions is that solar radiation should not vary spatially. Also, since spatial variability is the key to this study, data were only used when the full complement of measurements (XE, H, u., q, T and u) was available at all three sites. The data was further restricted to those hours when the available energy, A (in this study the sum of the sensible and latent heat fluxes), was high. The lower limit at the fallow savannah site was 400 W m 2 and at the millet site was 300 W m 2. This filter ensured that relative errors in the measured evaporation are low. This latter restriction effectively isolates the cases where there was no rain so that the spatial variation in evaporation is probably due to differences in vegetation type rather than in rainfall differences. After filtering the data as described above, there were 41 h of simultaneous data at the three sites. There is a gap in the filtered data between September 23rd and October 3rd due to missing data. Conveniently, this gap marks the end of the rainy season and the period before (including 35 data points) and after (including six data points) this gap will be referred to as the rainy season and the dry season, respectively. Fig. 1 shows the values of available energy, Fig. 2 shows the humidity deficit corrected for the height difference of the measurements as described below and Fig. 3 and Table 1 show the evaporation rates. 700
60o
®
~
x
x
x
x x
x
x x x
,a--
500
x
~x
x
x ~j~ ~ X~ x ~~
xx
x
x
~ x
x Ax
~ 400 A
AA
A
A AA
A
~
~ A
x A
AA
A
A
A
3oo
;>
