Geografisk Tidsskrift, Bind 93 (1993)

Water Use and Millet Production in the Sahel

Eva Bøgh

Side 80

Bøgh, Eva: Water Use and Millet Production in the Sahel.
Geografisk Tidsskrift 93:80- Copenhagen 1993.

Water-availability is an important and often critical factor for plant-production in the Sahel. The most important crop of this region is millet, which has a low water requirement and a high resistance to drought. In this article attention is drawn to the importance of properly understanding the water balance in the Sahel, and the relationship between biomass production and transpiration is discussed. An acquaintance with the magnitude of transpiration and a knowledge of the transpiration process may be important in the evaluation of plant production and water management planning. A method for estimating transpiration from a heterogeneous vegetation cover is tested, and the results are compared to Shuttleworth & Wallace's model of transpiration for heterogeneous vegetation.

Keywords: Sahel, evapotranspiration, transpiration, biomass production.

Eva Bøgh, M.Sc., Research fellow, Institute of Geography, University
of Copenhagen, Øster Voldgade 10, DK-13 50 Copenhagen

During the 1970's and 1980's most of the West African countries have experienced a serious decline in economic and environmental resources. The reasons are many and illustrate an inter-related and complicated process of degradation resulting in the weakening of human resources due to increased poverty and malnutrition among large population groups.

In the semi-arid areas of West Africa which typify the Sahelian region (Fig. 1), the living conditions of the majority of the population are dependent on a short, seasonly determined, unstable precipitation. Livestock have traditionally fulfilled the dominant role in the local strategy for survival. This type of landuse is advantageous, because livestock are mobile, thus allowing the inhabitants to avoid regions which are temporarily drought-prone. Shifting cultivation has generally been the most common form of agricultural system.

Today a change is taking place in this traditional landuse system. The increased population pressure causes a need for the intensification of food production. The population pressure leads to conflicts between nomads and peasants, and the continuous use of traditional landuse systems contributes to the ecological degradation of large areas due to overgrazing, felling of trees, shorter fallow periods and periods with drought. The degradation is caused by a currently inadequate use of existing resources.

PROSPECTS FOR IMPROVED WATER CONTROL IN THE SAHEL

Drought is a central factor of importance often discussed in relation to the degradation of Sahelian resources (Adefolalu,D.0., 1983; Beran, M.A. & Rodier, J.A., 1985; Agnew, C.T., 1989). During recent decades a trend towardsdeclining precipitation has been observed, and it may be that an irreversible degradation-process is taking place due to global climatic change (Parry, M.L. et al., 1988). Whatever the case, it is clear that the unstable


DIVL2462

Fig. 1. Map showing the Sahelian Region and the location of the HAPEX-Sahel-experiment.

Side 81

precipitation has great importance for the low investment
in agriculture as well as for its direct influence on plantphysiologicalproduction.

As the majority of the Sahelian population is involved in dry-land farming, an analysis of regional sustainability should be combined with investigations on the influence of available water resources for plant production and the possibility of improving water use efficiency and the control of water resources. A complete understanding of the different water balance components is therefore essential.

In reality there are many ways of improving water use efficiency, and in many areas of the Sahel simple regulatory systems - such as the construction of small dikes and the damming of gullies - have a vital importance for agricultural production (Chleq, J.-L., 1988). In the Northern Sahel, water use efficiency is very high. Millet, which develops deep roots and has a high resistance to drought, is cultivated with only 200-250 mm of precipitation (Kowal, J.M. & Kassam, A.H., 1978). This is possible because water is collected in micro-catchments. Other methods for optimizing water use is to reduce evaporation losses by introducing windbreaks or by introducing cropping systems with a high water use efficiency - such as intercropping. Intercropping is widely used and advantageous because different crops have different periods of maximum water requirement. Windbreaks and intercropping have, moreover, many other advantages, which may be combined in alley-cropping or agro-forestry.

PROSPECTS FOR SPATIAL INFORMATION ON WATER BALANCE

A poor physical infra structure and a faulty information service between city and country side make it difficult to form a general picture of the problems and the potential for agricultural production in the Sahel. One of the largest obstacles for the sustainable development of Sahelian resources is indeed the lack of information. This is due to the fact that data collection from this large area is costly, time-consuming, and therefore of low priority (Prince, S.D. etal., 1990).

Regional and national monitoring is necessary for efficient landuse planning and optimal exploitation of water resources. Run-off may for instance constitute an important potential water resource, but knowledge of the correlations between regional landforms, soil types and vegetation is required for optimal utilization. Recession watering is widespread in the Sahel and often dikes are constructed upstream to retain water. Such systems may in extreme cases cause a lack of water for the population living downstream. Development plans for an area should therefore be based upon the distribution of resources within single watersheds.

Processing and analysis of satellite images could be very useful to provide a broad view of regional and national resources. Today many techniques already exist which are ready to be used in the monitoring and evaluation of regional resources. Monitoring of landuse, biomass production, erosion and population density is to some extent possible at the present time and is of particular current interest to the Sahel which, just now, is characterized by changes in landuse and biomass production due to the population pressure and the droughts in the 1970's and 1980's.

Satellite image processing is still a technique under rapid development which furthermore shows promising prospects for the physical modelling of biomass production and hydrological/atmospheric processes. Combined with geographical information systems (GIS) and standard meteorological data, satellite image processing may be a strong dynamic information and analysis facility in improving our understanding of the relationship between climate, hydrology, biomass production and landuse. This may lead to the identification of regional water resources and eventually the realization of potential agricultural

Physical modelling with GIS has been used for estimating potential and low water budget crop production in the EC (Bulens, J.D. et al., 1990). In temperate climates, such models are simpler than they will be in the heterogeneous Sahelian areas. Evaporation from bare soil complicates water use estimations, and intercropping complicates biomass estimates. It is therefore necessary to understand how water use influences plant production; to improve knowledge of the influence of micro-meterological conditions on transpiration in the Sahel - and to explore how these processes may be modelled on the basis of satellite images and GIS.

THE THEORETICAL RELATIONSHIP BETWEEN BIOMASS PRODUCTION AND TRANSPIRATION

Potential biomass production may be expressed as a function of a series of specifically climate and plant conversion coefficients of global radiation (Monteith, J.L., 1972). Low water budget plant production may accordingly be modelled by including a water-stress index such as ET/EP in this expression, where ET is actual evapotranspiration and EP is potential evapotranspiration (Begue,A.,

The relationship between biomass production and plant water use has frequently been represented as linear until a closed vegetation cover exists or light saturation takes place (Rosenberg, N.J. et al., 1983; Onken & Wendt, 1989). In the Sahelian region a closed vegetation cover is

Side 82

seldom present and the dominant crop, millet, is, like other C4-plants, able to exploit high radiation intensities. It is therefore likely that the biomass of millet may be linearly related to transpiration.

Transpiration takes place through stomatas simultaneously with the assimiliation of CO2. As CO2 forms the basic component in the photosynthetic process, the analogy between the transpiration and assimiliation processes causes the stated linear relationship between transpiration and biomass production.

The transpiration (T) and photosynthesis (N) for leaves
have been described as (Gregory, P.J., 1989):


DIVL2493

where Pst and cst are respectively the absolute humidity (kg/m3) and CO2 concentration (kg/m3) in stomata; Pa and ca are the absolute humidity and CO2 concentration in the air; rst and rst c are the stomatal resistance to transpiration and CO2 assimiliation, and rb and rb c are the laminary boundary resistance to vapour and CO2 fluxes respectively.

The transpiration-efficiency is accordingly defined as:


DIVL2499

The relationship between the resistances of vapour and CO2 is correlated with the diffusion coefficients and mole-weights of the two gases. It may therefore be assumed

The transpiration efficiency is thus expressed as:


DIVL2505

Often it is assumed that Pst is saturated, and that cst is at or near zero, so that if the leaf temperature, air humidity and CO2 density in the air is known, the gradient may be easily established. It has been calculated, however, that a range of cst-values is dependent upon species and the response to irradiation (Rosenberg, N.J. et al., 1983).

Nevertheless, it has been shown for some species that cst/ca is regulated by the stomata function and is almost constant (Gregory, P.J., 1989). Therefore, it is often assumed, that the difference ca-cst is nearly constant, and the transpiration efficiency may then be expressed as (Gregory, P.J., 1989):


DIVL2511

The relation between biomass production and transpiration is consequently determined by Pst-Pa. Accordingly, the biomass production is controlled by the plant water use when it is assumed that there are no nutrient-deficiencies or plant-diseases.

TECHNIQUES FOR ESTIMATING PLANT WATER USE

The measurement and modelling of plant water use is much more complicated in the heterogeneous Sahelian region than in temperate regions where a complete and homogeneous vegetation cover usually exists. Most techniques for estimating water use are however almost entirely developed in the temperate regions of the world.

Techniques for measuring plant water use

There are two major methods for estimating evapotranspiration: 1) measurements of vapour fluxes above the surface, 2) calculations on the soil water balance. These techniques are suitable for estimating plant water use in the temperate regions where evaporation from bare soil only makes up a minor part of the total evapotranspiration. In the Sahelian region it is necessary to distinguish between evaporation and transpiration to evaluate plant water use. Plant-physiological techniques are available to measure transpiration. So far most progress has been accomplished since the introduction of porometers. A porometer measures a plant's ability to conduct vapour through its epidermis. Measurements of conductivity for the individual leaves may then be combined with the total leaf-area to calculate transpiration. This technique is time-consuming and not suitable for continous recordings. Recently a technique for measuring the heat-balance of the stem has attracted new attention. This technique has been used to measure sapflow during an experimental period in Niger, 1992.

The measurement of plant water use during the experiment in Niger

During the HAPEX-Sahel Experiment in Niger (17th August - 10th October 1992) transpiration was measured continuously from 6 single millet plants with the stem heat balance technique (Dynagage Inc., Models no. SGAlOws and SGA-13ws). The recording period started 30 days after sowing and lasted until the harvest started (84 days after sowing).

The principle of the method is based on a continuous heating of a stem section over a short vertical distance. The total measurement system is insulated by a foam cylinder as seen in Fig. 2, and the partition of the radial (Qr) and vertical conduction (Qv) is measured with thermocouples.The

Side 83

DIVL2559

Fig. 2. Photo of Sapflow-gauges installed on millet.

mocouples.Theconvective heat loss through the sapflow
(Qf) is then calculated as the residual in the following
energy balance equation:


DIVL2538

where P = energy-input.

Due to problems with estimating extent of heat storage in the stem at night, only daytime measurements are presently included in the analysis. Consequently, evapotranspiration at night is assumed to be equal to zero. In Fig. 3 the daily sapflow data during the experimental period are shown. It should be noted that the gauges were moved to new plants several times during the experiment owing to plant diseases and broken stems. This is the reason for the sudden irregularities between the measurements from the 6 gauges. In the last week of the experiment arose some "disorder" in the data because some of the plants had matured, and the gauges were consequently moved to late developing plants, which had a higher transpiration.


DIVL2556

Fig. 3. Daily Transpiration (gram/day) during the experimental period, 17th Aug.- llth Oct. 1992.

There is a high spatial variability in the state and height of millet plants within single fields in the Sahel. The 6 chosen millet plants were found to be representative on the basis of a statistical analysis of height measurements for more than 200 millet-clumps. It was, however, not possible to install the sapflow gauges on minor plants.

To allow an analysis of the sapflow measurements against above-canopy fluxes, the leaf-area of the 6 plants and leaf-area index were continuously measured. The average sapflow per leaf-area unit was then calculated for the 6 plants and hereafter multiplied with the leaf-area index (LAI).

The leaf-area was estimated by measuring the length and width of single leaves. A relationship between "length • width" and "leaf-area" was obtained by measuring the length, width and leaf-area of 30 collected leaves from the field. Leaf-area was measured with a scanning area measurement system (Delta-T Device; Model Mark 2 Area Meter). As seen in Fig. 4 the two different types of measurements have a high correlation.

LAI was estimated by measuring the transmission of diffuse radiation (blue spectra) through the vegetation cover with an optical sensor (LICOR LAI-2000). This technique has not usually been applied on heterogeonous vegetation. It was therefore tested against manual measurements and found satisfactory.

LAI are given in Fig. 5. It is clearly illustrated that the
vegetation cover is very sparse since maximum LAI only
reaches 0.37.

Side 84

DIVL2562

Fig. 5. Leaf-area index, 17th Aug.-11th Oct. 1992. Fig. 4. Correlation between "leaf-width multiplied by leaflength" and "scanned leaf-area".


DIVL2565

The diurnal variation of transpiration is seen in Fig. 6a. Fig. 6b illustrates the important control by irradiation of stomata resistance and transpiration. The estimated daily transpiration is seen in Fig. 7. Note that the transpiration does not decline, as in Fig. 3. This is caused by diminished leaf-area of the 6 plants giving a high transpiration per leaf-area unit. At the same time, LAI does not decline simultaneously with the leaf-area of millet since alternative crops are evolving and included in the LAI measurements.

Models for estimating plant water use

Modelling evapotranspiration is usually based upon atmospheric parameters. The most widespread model used is Penmann's Formula (1948) for potential evapotranspiration


DIVL2574
Side 85

where Rn is net radiation; QG is soil-heat flux; em-ea is the vapour pressure deficit of the air; ra is the aerodynamic resistance, which is expressed as an empirical funcion of the wind-speed derived above grass, Ca is the specific heat of air, Lv is the latent heat of vaporization, A is the saturated vapour pressure gradient and y is the psychrometric constant. Actual evapotranspiration is then calculated from EP, crop coefficients and the soil water balance (Doorenbos, J. & Pruitt, W.0., 1977; Agnew, C.T., 1989). The Penmann Formula has been criticized for its lack of aerodynamic effect and plant resistance to transpiration, but it is still widely used because of its simplicity and modest demand on input parameters (Agnew, C.T., 1991; Wallace,!., 1991).

Monteith (1965) modified the Penmann Formula by introducing a canopy resistance and an aerodynamic resistance based on existing roughness to substitute the empirical wind-function derived by Penmann. The Penmann-Monteith Formula is used for calculating actual evapotranspiration (ET) (Monteith, J.L., 1973):


DIVL2580

One practcal problem is the estimation of the canopy
resistance, rc, which is to be calculated from measurements
of stomata resistance and leaf-area index.

Modelling of evapotranspiration from a hetereogenous vegetation cover requires a distinction to be made between evaporation from bare soil and transpiration from plants. Evapotranspiration here takes place under altered micro-metereological conditions due to the larger turbulence around the tall well-spaced plants. Furthermore the available energy for evapotranspiration has to be partitioned between soil and plants, and separate surface resistances should be included for soil and plants.


DIVL2610

Fig. 8. Sketch of resistances to evapotranspiration in heterogeneous vegetation. See text for nomenclature and description.

Shuttleworth & Wallace (1985) introduced a model based on these concepts. The model is a 2-layer model based on the Penmann-Monteith Formula and applied to soil and plants respectively.

For the partition of the net radiation, Rn, between plants
and soil, it is assumed that absorption by soil, Rns can be
calculated from Beers' Law:


DIVL2590
Side 86

where the crop-dependent coefficient of 0.4 is estimated from an empiric relationship between LAI and radiation absorption for millet given by Begue, A. et al. (1991). The absorption by plants is calculated as Rn-Rns.

A number of flux resistances is introduced as illustrated in Fig. 8. The aero-dynamic resistance between the soillevel and the height for mean canopy conditions, ra s, and the aerodynamic resistance between this canopy-height and screen-height, ra a, is calculated from the wind-profile as described by Monteith, J. (1973): ra a is calculated from the prevailing wind conditions above the canopy, and ra s is calculated by introducing an exponentially decreasing eddy diffusion coefficient in the canopy-layer. It is not known how ra a and ra s respond to differing leaf-area indices. Shuttleworth & Wallace therefore assume a model where both ra a and ra s vary linearly with the leaf-area index between the values associated with their two limits, namely bare soil and a complete canopy. The height of mean canopy conditions is calculated from (d+zo), where d is the zero plane displacement (d = 0.63 h; Implantheight), and z0 is the roughness length (z0 = 0.13 h) (Monteith, J.L., 1973).

rs s, the soil resistance, defines the resistance to evaporation of a wet soil underlying a dry soil layer of increasing thickness. In dry climates the soil resistance may be expressed as a function of moisture in the surface layer (Wallace,!., 1991). Various estimates of rs s are given in Wallace,!. (1991).

ra c is the bulk boundary layer resistance of the vegetative elements in the canopy. It is calculated as rb/(2 LAI), where rb is the mean boundary layer resistance. Typical values of rb are in the order of 25 s/m (Shuttleworth & Wallace, 1985), which is of limited numerical importance compared to the much larger stomatal resistance.

rsc is the canopy resistance and calculated as rst/(2 LAI),
where rst is the stomatal resistance.

The model by Shuttleworth & Wallace reveals that the available energy for transpiration may exceed the fraction of intercepted energy. This may happen when a very dry soil is present and the soil-evaporation is very low or zero, because soil-heat is transferred as sensible heat to the canopy. This implies the difficulty of defining a waterstress index (ET/EP) for modelling biomass production in a heterogeneous vegetation cover as suggested by Begue, A.eta1.(1991).

A COMPARISON OF MEASURED AND MODELLED TRANSPIRATION

Measurements of sapflow are transferred to field values of
transpiration as described above. The calculation of transpiration
is performed by applying Shuttleworth & Wallace's
lace'smodel to micro-metereological data collected during
the experiment.


DIVL2632

Fig. 9. Comparison between measured and calculated transpiration (mm/day) during the experimental period, 17th Aug. -11th Oct. 1992. The calculation has been based on Shuttleworth and Wallace's model.

Net irradiance (REBS Q*6 radiometer), soil heat fluxes below millet and bare soil (REBS HFT-3), air-humidity (AANDERAA hygrometer), temperatures (thermocouples) and windspeed (VECTOR anemometers) were measured at 2 second intervals and averaged over 10 minutes. Stomatal resistance (Delta-T-Device, AP4-Porometer) were measured continuously during one "wet" day in the vegetative period (55 DAS) and one "dry" day (77 DAS) in the senescent period.

In the period 30-72 days after sowing, a stomatal resistance of 220 s/m is used as the input in the model, and for the rest of the period 500 s/m is used. It is not unusual to witness values of minimal stomatal resistance within this range for cereal crops in the vegetative and reproductive periods respectively (Taconet, O. et al., 1986; Ben Mehrez,M.,

In Fig. 9 measured ("sap LAI") and calculated ("S & W") transpiration are seen in Fig. 9 to correlate well. In the end of the period, the model, however, tends to underestimatethe measured transpiration. Serious uncertaintiesin the method of transferring the sapflow measurementsto a representative data-set of transpiration for the total field arose in the last 2 weeks of the experiment. In Fig. 5, it is seen, that LAI is reduced in the senescent period, but it is still quite high at harvest-time. This is due to the fact that weeds, cowpeas and hibiscus grow very fast during this time. The LAI of millet is therefore overestimated.Furthermore, the estimation of the leaf-area of

Side 87

DIVL2638

Fig. 10. Comparison of Daily Evapotranspiration, estimated by eddy correlation technique, and Transpiration during a period with presumably no soil evaporation. Last day of precipitation was 63 days after sowing.

the 6 senescent plants became more difficult. The area of yellow spots and gaps in the leaves were estimated roughly by eye and subtracted from the calculated leaf-area. At the same time, the leaf-area of millet diminishes rapidly givinga transpiration per leaf-area unit, which is too very high. Uncertainties in this period are increased even furthersince transpiration per leaf-area unit is calculated and measured on the basis of the largest plant of a millet clump, while LAI includes leaf-areas of smaller and later developed plants. As the stomatal resistance is dependent upon both the age of leaves and plant height (lower plants are shadowed), the measurements will not represent transpirationfrom lower and late-maturing plants correctly.

To examine the validity of the transpiration data (T), these were compared to fluxes of evapotranspiration (ET) during a period when soil-evaporation is assumed to be at zero. Evapotranspiration was estimated by use of the eddy-correlation-technique (GILL SOLENT 3D anemometer combined with a KRYPTON hygrometer). The results are seen in Fig. 10. Despite the above stated uncertainties at the end of the growing season, the transpiration and evapotranspiration data seem to correlate very well.

Evapotranspiration has been calculated from the Penmann-Monteith Formula. These results are shown in Fig. 11, which clearly demonstrates that the SW-Model has introduced some concepts of utmost importance to the modelling of transpiration above heterogeneous vegetation.

Potential evapotranspiration has been calculated using
Penmann's Formula and the results are seen in Fig. 11.


DIVL2635

Fig. 11. Transpiration compared to potential evapotranspiration calculated from Penmann's Formula and actual evapotranspiration calculated from the Penmann-Monteith-Formula.


DIVL2641

Fig. 12. Relationship between Transpiration (gram/day) and plant temperature (brightness-temperature) (°C).

Side 88

plant temperature coincides with low transpiration. This happens in the last 10 days of the measurement period and is caused by high stomatal resistance and low evaporativecooling. At moderate plant temperatures the transpirationincreases with an increase of leaf temperature which, in turn, increases the leafs internal vapour pressureand causes an increased est-ea.

APPLYING CONCEPTS TO REGIONAL ANALYSIS

The division of radiation and aerodynamic resistance in the canopy and the introduction of separate surface resistances for soil and vegetation have proved to be necessary concepts for modelling transpiration from heterogeneous vegetation. Unfortunately it is a very cumbersome process to estimate the applied surface resistances. If these components could be estimated by remote sensing techniques, it would be a assistance to the model.

Nemani & Running (1988) demonstrate that the expression TS/NDVI may represent a collective surface resistance for aerodynamically rough surfaces (Ts = surface temperature; NDVI = Normalized Difference Vegetation Index). Other investigations suggest that the soil resistance may be estimated from remote sensing analysis of microwaves (Chanzy & Bruckler, 1991). Further examination of Ts, NDVI and microwave reflection may induce techniques for the estimation of separate surface resistances.

The division of radiation-absorption is based upon LAI and may therefore be determined from NDVI. Rn may be estimated from the analysis of Ts and surface-albedo by remote-sensing techniques as shown by Soegaard (1988).

The influence of the spatial heterogenity of vegetation density to aerodynamic resistance is poorly known. The heterogenity of the millet fields results in large aerodynamic roughness and therefore contributes to little aerodynamic resistance. The roughness of heterogeneous surfaces is however at present insufficiently expressed as an increasing linear function of LAI. This relationship was originally derived above homogeneous vegetation (Monteith,J.L., 1973). Its application to low-density spatially heterogeneous vegetation such as millet probably leads to an under-estimated roughness and an over-estimated aerodynamic resistance.

Another uncertainty in the estimation of aerodynamic
resistance is the assumption of similarity between the
resistance of momentum fluxes and heat fluxes.

Ben Mehrez et al. (1992) estimated a flux-division parameter in the canopy on the basis of NDVI. The calculation of this factor is based on an empirical relationship given by Thorn (1972). More knowledge about intra-canopy opyaerodynamic transfer is required for the improvement of this method.

CONCLUSION

In the semi-arid areas of the world it is necessary to distinguish between soil evaporation and transpiration to evaluate plant water use. Recently the stem heat balance technique has attracted fresh attention for the analysis of transpiration in the Sahel. Combined with measurements of leaf-area and LAI, a method for obtaining continuous data on transpiration from a millet field was tested during two consecutive months in Niger, 1992. The transpiration data were compared to measurements of vapour-fluxes during a period with presumably no soil evaporation. Despite several uncertainties in exactly this period, the transpiration data correlate very well with the data of evapotranspiration.

The transpiration data were furthermore compared to calculations of transpiration by the SW-model and evapotranspiration by the Penmann-Monteith Formula. It was shown that a division of net radiation and flux resistances in the canopy and the introduction of separate surface resistances for soil and vegetation led to considerable consistency between measured and calculated transpiration. Towards the end of the experimental period the modelled transpiration (S & W) was under-estimated. This may have been caused by the input of a stomata resistance which was too high, since this parameter was only measured on the senescenting millet plants. One important obstacle for the use of the SW-model is the estimation of surface-resistance of soil and plants, while all other input parameters may be provided by a standard micro-metereological station.

The transpiration data were further compared to Penmann's potential evapotranspiration. The difficulty of utilizing this formula for the inclusion of a water stress index in a plant production model for a heterogenous vegetation cover has been indicated. Instead it is advised that plant production modelling be based on an input of actual transpiration.

ACKNOWLEDGEMENTS

The HAPEX-Sahel-experiment (Hydrological AtmosphericPilot Experiment in the Sahel) is being conducted on the recommendation of the World Climatic Research Programme. The programme is mainly financed by the EC (No. EVSV-CT9I-0033), but has also received valuablefunding from DANIDA (No. 104.Dan.8/554). It is carried out in cooperation with several research institutionsfrom England, France, Germany, the Netherlands, Niger and USA. The author wish to thank Dr. Pavel

Kabat, Winand Staring Centre, Wageningen, the Netherlandsfor providing stomata data for producing Fig. 6b. Special thanks are further owed to Henrik Søgaard, ThomasFriborg Jacobsen, Karen Juul Møller and Henrik Steen Andersen from Institute of Geography, Copenhagen,Denmark for their valuable contribution of materialand good advice in connection with the making of this article.

References

Adefolalu, D.O. (1983): "Desertification of the Sahel". In Natural
Resources in Tropical Countries. Singapore University Press.

Agnew, C.T. (1989): "Sahel Drought: Meteorological or Agricultural?".
In International Journal of Climatology (p.371-382),
V01.9.

Agnew, C.T. (1991): "Evaluation of a Soil Water Balance Model for the Analysis of Agricultural Drought in the Sahel". In Soil Water Balance in the Sudano-Sahelian Zone (p.583-593), lAHS Publication No. 199.

Begue, A. et al. (1991): "Radiation Use Efficiency of Pearl Millet
in the Sahelian Zone". In Agricultural and Forest Metereology
(p.93-110),n0.56.

Ben Mehrez, M. et al. (1992): "Estimation of Stomatal Resistance and Canopy Evaporation during the HAPEX-MOBILHY-Experiment". In Agricultural and Forest Metereology (p.285-313),N0.58.

Beran, M.A. & Rodier, J.A. (1985): "Hydrological Aspects of
Drought". UNESCO/WMO, France.

Bulens, J.D. et al. (1990): "GIS Supporting Crop Production Potential Research for the European Communities". In Proceedings EGIS 1990. First European Conference on Geographical Information Systems, Amsterdam.

Chanzy, A. & Bruckler, L. (1991): "Estimation de l'evaporation å partir d'informations accessibles par télédection microonde: Approches mécaniste, semi-empirique et empirique". In Soil Water Balance in the Sudano-Sahelian Zone (p. 149-159), lAHS Publication No. 199.

Chleq, J.-L. et al (1988): "Vanishing Land and Water. Soil and
Water Conservation in Dry Lands". CTA/Macmillian
Publishers/Terres et Vie, London/Basingstoke/Hong Kong.

Doorenbos, J. & Pruitt, W.O. (1977): "Crop Water Requirements".
FAO Irrigation and Drainage Paper 24, Rome.

Gregory, P.J. (1989): "Water use-efficiency of Crops in the Semiarid
Tropics", in Soil, Crop and Water Management in the
Sudano-Sahelian Zone (p.85-99), ICRISAT, Niamey.

Kowal, J.M. & Kassam, A.H. (1978): "Agricultural Ecology of
Savanna. A Study of West-africa". Oxford.

Monteith, J.L.(\ 972): "Solar Radiation and Productivity in Tropical
Ecosystems". In Journal of Applied Ecology (p. 747-766),
9.

Monteith, J.L. (1980): "Principles of Environmental Physics".
Edward Arnold Publ., London, 1973 (1980).

Nemani, R.R. & Running, S.W. (1988): "Estimation of Regional Surface Resistance to Evapotranspiration from NDVI and Thermal-IR AVHRR Data". In Journal of Applied Meteorology (p.276-285), V01.28.

Onken, A.B. & Wendt, C. W. (1989): "Soil Fertility Management and Water Relationships". In Soil, Crop and Water Management in the Sudano-Sahelian Zone (p.99-107), ICRISAT, Niamey.

Parry, M.L., Carter, T.R. & Konijn, N.T. (eds.) (1988): "The Impact of Climatic Variations on Agriculture. V01.2: Assessments in Semi-arid Regions". UNEP/Kluwer Academic Publishers, Dordrecht/Boston/London.

Prince, S.D. et al. (1990): "Remote Sensing of the Sahelian Environment.
A Keview of the Current Status and Future Prospects".

Rosenberg, N.J., Blad, B.L. & Verma, S.B. (1983): "Micro-climate.
The Biological Environment". John Wiley & Sons, N.Y./
Chichester/Brisbane/Toronto/Singapore.

Shuttleworth, W.J. & Wallace, J. (1985): "Evaporation from Sparse Crops - an Energy Combination Theory". I Quarterly Journal of the Royal Metereological Soc. (p.839-857), Vol.lll, N0.469.

Soegaard, H. (1988): "Comparison between Satellite-derived Evapotranspiration and Normalized Difference Vegetation Index in the Sahelian Zone". 22nd International Symposium on Remote Sensing of Environment, Abidjan, Ivory Coast, October

Taconet, O. et al. (1986): "Evapotranspiration over an Agricultural Region using a Surface Flux/Temperature Model based on NOAA-AVHRR Data". In Journal of Climate and Applied Metereology (p.284-308), V01.25, N0.3, March.

Thorn, A. (1972): "Momentum, Mass and Heat Exchange of
Vegetation". In Quarterly Journal of Royal Metereological Soc.
(p.124-134), N0.98.

Wallace, J. (1991): "The Measurement and Modelling of Evaporation
from Semiarid Land". In Soil Water Balance in the
Sudano-Sahelian Zone (p. 131 -149), lAHS Publication No. 199.