Abstract

Millet production has been monitored in the northern part of the Sahelian zone at the village Petakole located near Dori, NE Burkina Faso. Different methods, ranging from interviewing the local farmers to eddy covariance measurements of CO₂-fluxes, have been applied for estimating the total biomass and grain yield. Using field measurements conducted in 1997, a mechanistic photosynthesis model was calibrated to suit the current, local environment. A comparison of the daily net assimilation measured by the eddy covariance method and modelled by the photosynthesis model shows that more than 85% of the day-to-day variation in assimilation can be described by the model. An analysis of the diurnal variation in CO₂-fluxes over the growing season reveals that midday radiation load and leaf temperature are the most important climatic-hydrological parameters for predicting the carbon assimilation. At the end of the growing season when the soil moisture content is low (<8 v01.%) the rate of carbon assimilation is also limited by the stomatal control. For 1996 and 1997 the total CO₂-uptake is compared to millet yield estimates (in kg m'²) obtained through interviews with the local farmers and using aerial photos and GPS for determining the field sizes. The estimated millet yields are found to be much higher for 1997 than for 1996 and in both years with a 25% difference between the two yield estimation methods.

Keywords

Photosynthesis, millet, CO2-modelling, grain yield.

Henrik Soegaard, Mads Rossing Lund & Simon Bolwig: Institute of
Geography, University of Copenhagen, Øster Voldgade 10, 1350
Copenhagen K., Denmark.

Email: hs@geogr.ku.dk., mrl@geogr.ku.dk, sb@geogr.ku.dk

Geografisk Tidsskrift, Danish Journal of Geography
Special Issue, 2: 69-80, 1999

The Sahelian landscape and its vegetation differ in a number of ways from what is found in most parts of the industrialized world. Both the natural and cultivated vegetation found in this landscape are much more sparse than what is found in the temperate part of the world, and because irrigation and artificial fertilization only play a minor role, the harvest depends much more on the natural environment. In the Sahel, the climate in particular fluctuates strongly (Nicholson, 1994; Thornton et al., 1997). The coefficient of variation in annual rainfall is high (22% over the last 30 years within the study area of Dori, Burkina Faso) and this fact combined with the low annual rainfall of 400-600 mm creates high risk of crop failure and famine. Much recent research has focused on early warning of famine (Thornton et al., 1997) and on methods that enable biomass, especially of cultivated millet, to be esti

mated at an early stage in plant development. Many of these studies (Bartolomé, 1988; Rasmussen, 1992) have employed the Normalized Difference Vegetation Index (NDVI) as a proxy for the absorbed radiation and by regression analysis linked it to the observed yield. Detailed ground-based radiation studies (e.g. Bégué et al., 1991) have made it possible to estimate yield production based on absorbed radiation. In connection with large-scale field experiments (e.g. the HAPEX-Sahel, Goutorpe et al., 1997) and the development of the eddy covariance technique, our general knowledge on the functioning of the Sahelian ecosystems, manipulated as well as natural, has recently been greatly improved (Levy et al., 1997; Hanan et al., 1998; Boegh et al., 1999A).

The aim of the present project is to examine to what extent
the above-mentioned improved modelling of the

photosynthetic processes can be implemented and adjusted for the conditions in northern Burkina Faso. By applying such a model it would then be possible to examine some of the limiting factors affecting crop growth (e.g. radiation, soil water availability or temperature). A comparison of yield estimates must consider the question of scale and thus the heterogeneity of the Sahelian landscape. When monitoring the conditions for plant growth at the village/ landscape scale, methods which are adoptable to and can be expanded by the use of satellite data are clearly preferable. In order to focus on the objectives of the present study, the main emphasis will be placed on estimating the production of the most important crop in the Sahel, namely millet. The study includes the following steps:

a) Long-term monitoring of carbon dioxide exchange
above a millet stand in the Dori/Petakole area in the
Séno province of Burkina Faso

b) Elaboration of a photosynthesis model that incorporates physical and biochemical factors affecting plant growth in the Sahel, and a comparison of measured and modelled carbon assimilation with yield estimates obtained by interviews with the local farmers.

c) Analysing the dependence of carbon assimilation on
radiation, temperature and soil moisture through a sensitivity
study using the elaborated photosynthesis model.

Methods

The experimental work was carried out near the village of Petakole (14°01' N, o°o3'E) which is located skmSEof Dori in the Séno province of northeastern Burkina Faso. The data set consists of two sets of micro-meteorological observations collected during the peak growing season. In 1996, the measurement period lasted from Day 240 to Day 280, and in 1997 from Day 242 to Day 277. In both years, the same millet species (Pennisetum gambiense) was sown and the instrumentation was mounted approximately in the middle of a 4.8 ha cultivated field. Except for patches of sorghum covering less than 15% of the area, the neighbouring fields, however, were also sown with millet allowing fetches in all directions to more than 500 m.

The soil type of the millet field is sandy soil originating from ancient aeolian deposits. The soil is low in both carbon (0.2%) and nitrogen (0.04%) content, manure is applied but at a lower level than the village average. The field is flat and the main physical obstacle consists of 3-4 acacia trees growing near the boundary of the field. Whereas germination and growth of the millet were successful in 1997, the 1996 crop suffered from a dry spell early in the season which caused whitening of the young shoots and re-sowing was not undertaken until the last week of July. Consequently, the two crops developed rather differently; in 1996 the maximum Leaf Area Index (LAI) was only 0.4 whereas in 1997 it was 0.7.

Canopy development, biomass and grain yield

Leaf Area Index (LAI) was estimated approximately every week by measuring the transmission of diffuse light through the canopy with an optical sensor (LAI-2000 Plant Canopy Analyzer, LI-COR Inc., Nebraska, USA). A porometer (AP4, Delta-T devices, Cambridge, UK) was used for manual measurement of stomatal resistance on 75 leaves. Based on the porometer data and visual inspection the leaves were divided in two categories green leaves and non-functioning senescent leaves. The Green Leaf Area Index (GLAI) was then calculated as LAI times the ratio of the number of green leaves to the total number of leaves. Within an area of 100 m by 100 m surrounding the mast both plant height and LAI were measured in 25 regularly spaced grid cells. For the 1996 growing season, the total above-ground biomass was estimated by harvesting and weighing all the millet plants within the 25 grid cells each measuring 20 m by 20 m.

Harvest data were obtained by interviewing 39 farmers who cultivated a total of 85 village fields with millet. The recall period was 3 months for the 1997 and 1996 harvest, and 15 months for the 1995 harvest. Millet is harvested in bundles and counted in such by the farmers. All farmers could recall the size of their total harvest at least two years back, while some could only remember field specific yields one year back. The aggregated harvest of village fields was calculated on the basis of average bundle weights times the number of harvested bundles for each farmer. The size of each field was measured by a trained assistant using 'the steps and compass' method. The size of the total cultivated area was measured by use of a GPS, and controlled by aerial photographs.

For converting the CO₂-fluxes into yield, it is used that it takes 2.1 g of CO₂ to produce l g of dry matter (Penning de Vries et al., 1989). To be able to separate between above-ground biomass and total biomass, the ratio between

shoots and roots was studied during the HAPEX-Sahel (Hanan et al., 1997) by sieving the soil samples taken around two millet clumps down to a depth of 2 m. It was found that the weight of the roots was 328 g while the above-ground biomass was weighted to 642 g, giving a shoot:root ratio of 2:1. For converting above-ground biomass into yield estimates a harvest index (HI) of 0.2 was used(Bégué et al., 1991).

Measurement of CO2 fluxes and meteorological data

The fluxes of carbon dioxide and water vapour above the vegetation were measured using the eddy covariance technique (Baldocchi et al., 1988). The measuring mast was equipped with a 3-D sonic anemometer (Gill Solent, UK). An IRGA (Infraßed Gas Analyzer, LI-COR 6262, LI-COR Inc., USA) with a built-in sensor for pressure corrections was used for measuring carbon dioxide and water vapour concentrations. Due to the heterogenous growth of the millet in 1996, the sonic anemometer and the inlet to the IRGA sampling tube were raised from 3 m to 6 m above ground level in the middle of the campaign. The sampling tube (Bev-A-Line, LI-COR Inc., USA) was 15 m in length with an inner diameter of 3.2 mm, and the air was sucked through the sampling tube by a membrane pump at a flow rate of 6/ mm"l. Due to an unfortunate malfunction of the pump, the CO2-flux data from 1996 proved unreliable. In 1997 the same setup was used, with a constant measuring height of 6 m. More than 768 hours of flux data were collected. Data acquiring and processing were based on the EdiSol software package (Moncrieff et al., 1997). The flux mast was also equipped with sensors for the measurements of basic micro-meteorological parameters including; net radiation (Q*6, REBS inc., USA), soil heat flux (HFT-3, REBS inc., USA), short wave radiation (CM-7, Kipp & Zonen, NL), soil and air temperatures (thermocouple), and IR surface/ canopy temperature (KT-17, Heimann, D). Incident and reflected photosynthetically active radiation (PAR) were measured by quantum sensors (LI-190, LICOR Inc.,USA) while soil moisture was recorded by ThetaProbes which responds to the dielectrical properties of the soil (Delta-T devices, UK). Supplementary meteorological data (rainfall and temperature) were provided by L 'Institute de Meteorologie de Burkina Faso which operates a meteorological station in Dori. Intercepted rainfall during the growing season was calculated according to Casenave & Valentin (1989) where superficial runoff is assumed when the rainfall rate is more than 25 mm/d for dry soil or 10 rnm/d for wet soils. The infiltration capacity for the excessive rainfall is equal to 67%. The onset of the rainy season is defined as the date for the first major rainfall (>25 mm) which is not followed by dry spells lasting for more than 10 days. The end of the rainy season is the date for the last rainfall event larger than 5 mm.

Carbon dioxide model

The applied photosynthesis model is based on the "big leaf concept (Farquhar et al., 1980), elaborated by Collatz et al. (1992) for use on C₄-species. The general idea is that the net photosynthetic rate (A_n) is calculated as the minimum of three potential photosynthetic rates (J_E, J_c, J_s) less the dark respiration. J_E is the rate controlled by absorbed PAR, J_c is the rate controlled by the stomata, and J_s is the rate controlled by removal of assimilates. Each of the limitations is scaled from leaf level to the actual canopy level (with Leaf Area Index=L ) using the area of sunlit leaves per unit ground 11= (l-e^kL )/k where kis the average extinction coefficient (=0.7). The three specific equations are as follows:

(1)

where a is the leaf absorptance of PAR (0.86) , a is the initial slope of the light response curve, Q_p is the photosynthetic active radiation (jamol m"² s"¹), P; is the intercellular CO₂ partial pressure (Pa), k_s is the initial slope of the photosynthetic response to p_js and Pis the air pressure (Pa). The photosynthetic rate is related to the Rubisco capacity, V_m (i.e. the catalytic capacity of Rubisco, the CO₂ fixing enzyme). Because V_m is temperature dependent (Q_lo=2), it is for reference scaled to a common temperature 25 °C and denoted V_m2s. The net assimilation is then calculated by:

(2)

Eqn. 2 indicates a sharp transition from one limitation to
another. However, when two or more limitations are nearly
equal co-limitation occur and the result, which is

slightly lower, is found by solving a quadratic function as
described by Campbell & Norman (1998).

The dark respiration R_/d per leaf area is calculated as 0.02V_m (Collatz et al., 1992). The bulk stomatal resistance (r_s) in (s m2m² mol"¹) is calculated by use of the Ball & Berry model (Collatz et al., 1992):

(3)

where b (mol m"² s"¹) and m are empirical constants (Table 1), h_s is the relative air humidity at the surface of the leaf (%), c_s is the mole fraction of CO₂ at the surface of the leaf, and f_w is the effect of limited soil water content (6_S) on stomatal resistance. It is scaled using maximum (6,^) and minimum (Ö^J soil water content (vol %) at a depth of 5 cm applying the following empirical function (Wang & Leuning, 1998):

(4)

A listing of model parameters is given in Table 1. Due to the feedback coupling between A_n and l/r_s the Eqns. 1-4 are solved through iteration. The photosynthesis model was designed to use a single boundary layer resistance to control the exchange above the leaf. In the present canopy version, micro-meteorological measurements at reference height are used as input parameters and the single boundary layer resistance is substituted by the sum of the atmospheric resistance and the boundary layer resistance. According to Monteith & Unsworth (1990), the atmospheric resistance (r_a) is calculated from wind speed (u) and frictional velocity (u,), see Eqn. 5.

<5)

In accordance with Boegh et al. (1999A) the boundary layer resistance for water vapour is found as the minimum of the two resistances calculated assuming respectively molecular diffusion in laminar forced convection (Jones, 1992) and free convection (Brenner & Jarvis, 1995). Due to the larger molecules the resistances for carbon dioxide is found by multiplying by 1.4 (Jones, 1992):

(6)

where w is the width of leaves (0.05 m), u is the wind speed (m s"¹), vis kinematic viscosity, a, is the coefficient of thermal expansion, g is the gravity acceleration, D_h is the thermal diffusivity, and (T,-T_a) is the difference between leaf and air temperature.

Calculation of canopy flux and net ecosystem flux

In accordance with Ruimy et al. (1995), the eddy covariance CO2-fluxes measured above the canopy during daytime (06:00- are denoted by Fd. The eddy covariance flux for the remaining 12 hours (night-time) is denoted by Fn. The sign convention used here for the carbon dioxide fluxes originates from ecophysiology, in which a flux of carbon dioxide from the atmosphere to the canopy is defined as positive, whereas an upward flux is negative. This

means, that during daytime, the net assimilation (An) can
be set equal to the canopy flux (Fc) (Ruimy et al., 1995)
defined as:

(7)

where the soil respiration (F_s) is directed upward and is thus negative. By definition F_s consists of heterotrophic microbial decomposition, and root respiration, but no separation of these two components is made here (see also Lloyd & Taylor, 1994). The temperature dependence of F_s is expressed using an exponential function adopted from Lloyd & Taylor (1994). Previous studies from the Sahel (Friborg et al., 1997) have shown that soil respiration also depends on the soil moisture content, and to take this into account the ratio between actual soil moisture 0_a and the soil moisture at field capacity O_FC at a depth of scm has been included, leading to the following equation:

where R₁₀is the soil respiration at 10 °C and T_soil is the soil temperature in the upper 5 cm. As will be discussed later R_lO can be estimated from nocturnal fluxes. The night-time eddy covariance flux (F_n) originates from soil respiration, and dark respiration from the plants as defined by:

(9)

The dark respiration is included in the An-term, and during
nighttime when the rate of photosynthesis is nil RldL is
equal to -An (Eqn. 2).

Results

CO₂ fluxes day and night

Figure 1A and IB show two examples (Day 262 and Day 275) of the daytime courses in the CO₂ fluxes both measured and modelled. The three individual limitations (J_E, J_c and J_s) are also included on the graphs. By adopting the parameters obtained from a similar study in Niger (Boegh et al., 1999A) (Table 1), it is possible to simulate F_c with reasonable accuracy both at the middle and at end of the growing season. On Day 262 the two controlling limitations are light (J_E) and removal of assimilates (J_s) whereas stomatal control does not cause any limitation in the CO₂ assimilation because J_c is above 40 jamol m"² s'¹ all day. On Day 275, when the soil is drying out and the air humidity has become low, limitations in carbon assimilation due to increased stomatal resistance become important (J_c). The shape of the A_n-curve is also influenced by inhibitions as the leaf temperatures above 40 °C cause an afternoon secondary minimum in J_s. The implications of a low stomatal control (i.e. the relatively high J_c values) have previously been reported (Levy et al., 1997; Boegh et al., 1999A), and this aspect will be discussed later.

Figure 2 shows the calibration of the soil respiration model based on nocturnal fluxes. For each night, the mean flux (F_n) is calculated. Soil respiration (F_s) is also calculated by subtracting plant respiration as predicted by the photosynthesis model according to Eqn. 9. It was found that neither soil temperature nor soil moisture used as sole

parameters could be used in predicting the soil respiration flux. However, when using the two-parameter model (Eqn.B), the accuracy was greatly improved. The y-axis shows the model estimate using reference respiration (R₁₀) equal to -0.8 umol m"² s'¹ and a soil moisture content at field capacity set equal to the maximum measured value of 17 vol. %. The x-axis shows the nocturnal fluxes corrected for leaf respiration.

As found in a number of other studies (e.g. Soegaard & Thorgeirsson, 1998), there is a considerable scatter of data points when modelling nocturnal fluxes, but since 22 out of the 29 points deviate less than 0.5 umol m"² s"¹, the model was consequently accepted.

By combining the photosynthesis model with the soil respiration model, according to Eqn. 7 and 9, it is now possible to simulate the CO₂ exchange above the canopy and calculate the net ecosystem production which is equal to the time-integrated eddy covariance measurements. The level of agreement between these two estimates is shown

developed. For the senescence phase, it is obvious that the daytime fluxes are much lower, and the diurnal variation is characterized by its midday upper ceiling, as imposed by the combined action of the J_c and the J_s limitations (confer Figure IB).

Figure 4 compares the daytime average CO₂ assimilation as calculated by the photosynthesis model and the canopy flux calculated by Eqn. 7 for all days from Day 245 to Day 277. On days with missing half-hourly observations (see Figure 3) the daily values are found as the simple average of the actual observations. The agreement between model and observations is satisfactory as all model parameters are predefined and the slope is close to the 1:1- with R2R² = 0.86. To examine the change in the CO₂ function at the time of senescence, the temporal variation in soil moisture and air and leaf temperatures are shown in Figure 5 together with measured and modelled Stomatal resistance. As long as the soil is moist, the evapotranspiration will cause both leaf and air temperatures to remain low (i.e. below 35 °C), but as the soil dries out, leaf temperatures exceed 40 °C inhibiting plant functions and causing increased leaf senescence (Figure 3). The measured and modelled midday Stomatal resistance (r_st) show the same trend; for sunny days in the growing phase r_st is around 80 sm"¹ and roughly twice as high during senescence. This rather stepwise change in r_st is in agreement with Soegaard & Boegh (1995). For the modelled data (Eqn. 3) r_st is calculated from bulk Stomatal resistance r_s as r_st =1.6 -GLAIrs, •r_s, where the factor 1.6 takes account for stomata on both sides of millet leaves (Boegh et al., 1999A). The versatility of Eqn. 3 is confirmed when the modelled Stomatal resistance is compared with those measured by porometry.

Different approaches have been used when modelling the lower photosynthetic activity during the senescence phase. In a study of natural vegetation in Niger Hanan et al. (1997) used a decrease in rubisco capacity to account for lower CO₂ assimilation at the end of the season. Measurements of leaf nitrogen during the 1997-campaign show, however, a nearly constant nitrogen content (2-2.5%) which does not support the hypothesis concerning decreasing Rubisco capacity. In general it is likely that several physiological features alter more and less simultaneously when the senescence occurs, as argued by Boegh et al. (1999A).

Comparing carbon assimilation from subsequent years

Employing the photosynthesis model in the form elaborated on the basis of data from Niger (Boegh et al., 1999A) and verified in Figure 4, it is now possible to simulate the seasonal course in carbon assimilation in 1996 and compare it to the measurements and modelling from 1997. The results of the calculations are shown in Figure 6A and reveal a large difference in the assimilation between the two years. The growing season in 1997 already starts around Day 220 and attains a daily assimilation rate of 15 g CO2 rn2d'¹ on Day 242 increasing to 35 g CO2 m'² d'¹ by Day 255. Thereafter there is a rapid decrease until the end

of the campaign (Day 278). By comparison hereto the 1996 growing season simulation is different in both timing and magnitude. When comparing Figure 6 A and Figure 6 B the difference in carbon assimilation between the two years seems to be associated with the development in soil moisture. In fact there is a number of environmental conditions which changes more or less simultaneously with the drying of the soil. The lower soil moisture content will thus cause an increase in bulk stomatal conductance (Eqn. 4) which for millet may result in a reduction green LAI and at the same time cause leaf temperatures to increase. In Figure 6 B it is shown that until day 265 the soil moisture content in 5 cm is one-third lower in 1996 than in 1997. In 1996, the canopy develops much later and reaches a maximum of 20 g CO₂ m"² d"¹ only at the end of September (Day 272) when the soil moisture integrated over the upper 50 cm is still above 10 vol. %.

Comparing yield estimates from interviews and carbon dioxide fluxes

The grain yield estimates are shown in Table 2 based on two different methods; i) the photosynthesis modelling and ii) local interviews with the farmers. The total CO₂ assimilation is converted into total dry-mater (dm), aboveground dm and finally yield by applying the set of conversion factors presented in the method section. From the interviews the number of millet bundles was counted to be 2195 in 1996, and 4283 in 1997, for the whole village, comprising 120 ha of fields. The average grain content of one bundle was found to differ from one year to the next. The average grain content of 47 bundles in 1996 was 12.1 kg, but as high as 18.1 kg (n=4o) in 1997. The grain content per bundle for 1995 was calculated by averaging the 1996 and 1997 estimates.

Using these figures the millet grain yield pr m2m² is calculated and the results are presented in Table 2. The table shows that the two totally independent methods give results which are of the same magnitude. The difference in 1996 is thus 7g m"². In 1997, the harvest result is 15 g m"² higher than that reached by the photosynthesis model. This larger discrepancy is properly due to fertilization. In dry years (1996) the low degree of fertilization applied to the study field (relative to the surrounding fields) is less important because the millet production is mostly controlled by water availability, whereas in wet years millet production benefits from a higher level of fertilization, and the yield from the study field is probably less than on the more

fertilized surrounding fields. In 1996, further comparison can be done because the above-ground biomass was weighed for each of the 25 (20 m by 20 m) grid cells surrounding the mast. The harvesting results give 151 g m"² ±52g m"²; which is quite close to the 166 g m"² aboveground dry matter reported in Table 2. In this comparison it is assumed that the water content (10-15 v01.%), at harvest time is more or less counterbalanced by the remaining biomass not collected during the harvest. In Table 2 yield data is also shown for 1995 together with intercepted rainfall and the length of the rainy season, and there are obvious differences among the three years. In 1997 the rainfall was equally distributed over a 90-days period which should be optimal for the millet variety grown in the region. In both 1995 and 1996 dry spells around Day 200 (late part of July) shortened the length of the coherent rainy season to less than 70 days. The amount of inter

cepted rainfall was only 82% and 58% of the 1997-value,
respectively in the two years.

Dependence of carbon assimilation on environmental conditions

By use of the photosynthesis model it is possible to examine
the influence of changing environmental conditions on
the carbon assimilation of the millet crop.

Already from Figure 1A and IB it was shown than during the wet part of the growing season the major limitations for carbon assimilation were radiation and the removal of the assimilates largely controlled by the rubisco capacity and the leaf temperature. In the last part of the growing season the stomata control becomes increasingly important (Figure IB) especially when the soil is drying out as shown in Figure 5.

In Table 3 the major results of simulation are given for three important environmental properties namely photosynthetic active radiation (PAR), soil moisture and midday leaf temperature. In the simulation average weather and soil moisture data from Day 269 to Day 277 have been used. Severe water stress is assumed to result in withering of the leaves and thus reduction of GLAI rather than increased stomatal resistance (confer Figure 5). The response factor is defined as the percentage change in daily assimilation divided by the percentage change in the specific parameter. It is assumed that for plant growth 0 °C can be considered as a "true" zero point thus allowing calculation of percentage changes from the Celsius scale. The carbon assimilation is found to respond rather linearly on any change in the amount of PAR as long as J_E is the major limitation. When the change becomes larger the response is reduced as the other limitations for plant growth (J_c and J_s) become more important.

The effect of low soil moisture content is expressed
through the role of the water stress function (Eqn. 4). As
long as the soil moisture content is above approximately 8

vol. % water stress does not play any role but for lower
soil moisture content the function becomes very important
especially through the associated lowering of the GLAI.

The temperature dependence of carbon assimilation changes with the actual leaf temperature, and the assimilation rate decreases for both increasing and decreasing temperature as shown in Table 3 and Figure 7. For comparison with other results the data have been scaled relative to the maximum assimilation rate. The comparable results include simulations based on measurements from Niger, (Boegh et al., 1999A) where an optimal leaf temperature of around 36 °C was found. For both increasing and decreasing leaf temperature, the net CO₂ assimilation rate was found to decrease. The third plot in Figure 7 is based on a study from India (Mclntyre eta 1.,1993) showing that during the growing phase both very high and very low air temperatures reduced the daily stem extension with respect to the maximum value of 130 mm d"¹. For comparison, the original daily mean air temperatures (T_m) are converted to daytime leaf temperatures (T,) by use of linear regression:

T/=1.86 T_m-19.3 (R²=0.83) derived from the Petakole data. When plotting the data from India in Figure 7, there is an encouraging agreement among all three estimates confirming that the growth rate of millet depends strongly on temperatures.

Discussion

In order to estimate net carbon assimilation in a millet stand, a mechanistic photosynthesis model elaborated by Collatz et al. (1992) and modified by Boegh et al. (1999A) has been adopted. Originally this photosynthesis model did not include limitations in the assimilation rate due to water stress. To take account of the water stress situation observed at the end of the 1997 growing season, a water stress factor proposed by Wang and Leuning(l99B) has been included. This seems to be a useful step because it improves the simulation at the end of the growing season (primo October) when the moisture content in the upper part of the soil approaches 3-4%. In combination with the use of GLAI rather than LAI in the modelling it seems that the present model is effective in capturing the temporal variation in CO₂-assimilation throughout the season. For the two years with micro-meteorological measurements (1996 and 1997) the yield estimates based on interviews and those based on carbon flux measurements and photosynthesis modelling were found to be within the same magnitude (± 25%).

The present simulation results confirm the findings by Boegh et al. (1999A) who used a similar photosynthesis model to study the net CO₂ assimilation dependence on the change in various environmental parameters. Among the meteorological parameters Boegh et al. (1999A) found that leaf temperature and photosynthetic radiation had the strongest influence on the millet growth rate, while other meteorological parameters such as air temperature, water vapour, and CO₂ partial pressure had only a marginal influence. In another simulation study from Niger, Agne w (1990) found that the annual rainfall had a significant contribution when predicting millet yield. From three years of data from Petakole village it was found that the millet yield increased with increasing amount of intercepted rainfall (Table 2), the data material was, however, too limited to draw any statistical conclusions.

The biomass measurements conducted within the grid
surrounding the mast underlines the fact that there is considerable

variation in primary production within the individual millet fields. For large scale monitoring this is in favour of sampling methods which are integrating the data over larger areas (e.g. eddy correlation technique / remote sensing). For the biomass modelling it is also important that the models can be applied to larger scales. Thornton et al. (1997) have by use of their CERES model demonstrated that they are able to predict the millet production for the whole of Burkina Faso using standard weather data combined with information on land surface properties such as soil type and drainage conditions but without the use of data on vegetation density (i.e. LAI). By contrast, Boegh et al. (1999A) have shown that the amount of data necessary to run the mechanistic photosynthesis model can be limited to three parameters: LAI, midday leaf temperature, and irradiances; through which there is a natural link up to remote sensing because both LAI and leaf temperature are satellite measurable quantities (Boegh et al., 19998; Troufleau & Soegaard, 1998; Prihodko & Goward, 1997). In Burkina Faso the use of satellite remote sensing for estimating vegetation density, not only at the landscape level but also at a regional scale, has recently been the object of a pilot study (Lind & Fensholt, this issue). It should be stressed, however, that leaf area is not solely dependent on weather conditions, but also depends on farming practices; such as soil preparation, sowing time, sowing distance between millet pockets, etc.

Consequently, when predicting plant production in the northern part of the Sahel zone, as found within the Dori area, it seems necessary to combine information on land use (including vegetation type), land management practice, vegetation density and vegetation and water status (remotely sensed). This data integration can be achieved by running the mechanistic photosynthesis model in a GIS environment; a technique previously applied in other environments (e.g. Soegaard et al., 1999).

Acknowledgement

The present study was conducted as part of the SEREIN project within the Danish Strategical Environmental Programme. The SEREIN activities were coordinated by professor Anette Reenberg, Institute of Geography, University of Copenhagen and the authors want to express their gratitude for comprehensive assistance both in planning and conducting the experimental work. Thanks are also due to

associate professor Lars Krogh, Institute of Geography, University of Copenhagen for providing soil and 1996 harvest data from the research grid surrounding the micrometeorological mast.

References

Agnew, C. (1990): Spatial aspects of drought in the Sahel.
Journal of Arid Environment 18: 279-293.

Baldocchi, D.8., Hicks, 8.8., & Meyers T.P. (1988): Measuring
biosphere-atmosphere exchange of biologically related gases
with micrometeorological methods. Ecology, 69: 1331-1340.

Bartholomé, E. (1988): Radiometrie measurements and crop yield forecasting. Some observations over millet and sorghum experimental plots in Mali. International Journal of Remote Sensing 9: 1522-1539.

Bégué, A., Desprat, J.F., Imbernon, J., & Baret, F. (1991): Radiation
use efficiency of pearl millet in the Sahelian zone. Agricultural
and Forest Meteorology 56: 93-110.

Boegh, E., H. Soegaard, H. Friborg,T. & Levy, P. (1999A): Models
of CO2 and water vapour fluxes from a sparse millet crop in the
Sahel. Agricultural and Forest Meteorology 93: 7-26.

Boegh, E., Soegaard, H., Kabat, P., Hanan, N. & Lenz, C. (1999B): A remote sensing study of theNDVI-TS relationship and the transpiration from sparse vegetation in the Sahel based on high-resolution satellite data. Remote Sensing of Environment (Accepted)

Brenner, A.J. & Jarvis, P. (1995): A heated leaf replica technique
for determining of leaf boundary layer conductance in the field.
Agricultural and Forest Meteorology 72: 261-275.

Campbell, S.G. and Norman, J.M. (1998): An introduction to
environmental biophysics. 2nd ed. Springer-Verlag.

Casenave, A. & Valentin, C. (1989): Les etats de surface de la
zone Sahelienne. ORTSOM. Paris.

Collatz, G.J, Ribas-Carbo, M., & Berry, J.A. (1992): Coupled
photosynthesis-stomatal conductance model for leaves of C4 -
plants. Austalian Journal of Plant Physiology 19: 519-538.

Farquhar, G.D., von Caemmerer, S., & Berry, J.A. (1980): A
biochemical model of photosynthetic CO2 assimilation in
leaves of C3 plants. Planta 149: 78-90.

Friborg, T.,Boegh, E. & Soegaard, H. (1997): Carbon dioxide
flux, transpiration and light response of millet in the Sahel.
Journal of Hydrology 188-189; 623-640.

Goutorbe, J.P., Lebel, T., Dolman, A. J., Gash, J.H.C., Kabat, P.,
Kerr, Y.H., Monteny, B. Prince, S.D. Strieker, J.N.M., Tinga,
A., & Wallace, J.S. (1997): An overview of HAPEX-Sahel: a

study in climate and desertification. Journal of Hydrology 188-189:
4-12.

Hanan, N. P., Prince, S. D. & Bégué, A. (1997): Modelling vegetation primary production during HAPEX-Sahel using production efficiency and canopy conductance model formulations. Journal of Hydrology 188-189: 651-675.

Hanan, N., Kabat, P., Dolman, H. & Elbers, J. (1998): Photosynthesis
and carbon balance of a Sahelian fallow savanna. Global
Change Biology 4: 523-538.

Jones, H.G. (1992): Plants and microclimate. A quantitative approach
to environmental plant physiology. Cambridge University
Press.

Levy, P., Moncrieff, J.B. Massheder, J.M. Jarvis, P.G., Scott, S.L. & Brouwer, J.(1997): CO₂ fluxes at leaf and canopy scale in millet, fallow and tiger bush vegetation at the HAPEX-Sahel southern super-site. Journal of Hydrology, 188-189: 612-632.

Lind, M. & Fensholt, R. (This issue): The spatio-temporal relationship
between rainfall and vegetation development in Burkina
* '
Faso.

Lloyd, J. & Taylor, A. (1994): On the temperature dependence of
soil respiration. Functional Ecology 8: 315-323.

Mclntyre, 8.D., Flower, D.J. & Riha, S.J. (1993): Temperature and soil water status effects on radiation use and growth of pearl millet in a semi-arid environment. Agricultural and Forest Meteorology 66: 211-227.

Moncrief, J.8., Monteney, 8., Verhoef, A., Friborg,T., Elbers, J., Kabat, P., de Bruin, H., Soegaard, H., Jarvis, P.G. & Billing, H. (1997): Spatial and temporal variations in net carbon flux during HAPEX-Sahel. Journal of Hydrology 188-189:553-578.

Monteith J.L. & Unsworth M.H. (1990): Principles of environmental
Physics. London. Edward Arnold.

Nicholson, S.E. (1994): Recent rainfall fluctuations in Africa and
their relationship to past conditions over the continent. The
Holocene4: 121-131.

Penning de Vries, F.W.T., Jansen, D.M. ten Berge, H.F.M. &
Bakema, A. (1989): Simulation of ecophysiological processes

of growth in several annual crops. Wageningen, Pudoc.

Prihodko, L. & Goward, S.N. (1997): Estimation of air temperature
from remotely sensed surface observations. Remote Sensing
of Environment 60: 335-346.

Rasmussen, M.S. (1992): Assessment of millet yields and production in northern Burkina Faso using integrated NDVI from the AVHRR. International Journal of Remote Sensing 13: 3431-42.

Ruimy A., Jarvis P.0., Baldocchi D., & Saugier, B. (1995): CO2
fluxes over plant canopies and solar radiation: A review. Advances
in Ecological Research 26: 1-65.

Soegaard, H. & Boegh, E. (1995): Estimation of evapotranspiration from a millet crop in the Sahel combining sap flow, Leaf Area Index and Eddy Correlation Technique. Journal of Hydrology 166: 265-282.

Soegaard, H. & Thorgeirsson, H. (1998): Carbon dioxide exchange
at leaf and canopy scale for agricultural crops in the
boreal environment. Journal of Hydrology 212-213: 51-61.

Soegaard, H, Friborg, T., Hansen,B.U. Nordstroem, C & Christensen, T.R. (1999): Trace gas fluxes in a high arctic environment, NE-Greenland IV. Scaling of COyfluxes from a canopy to landscape level. Submitted to Global Biochemical Cycles.

Thornton, P.K., Bowen, W.T., Ravelo, A.C. Wilkens, P.W., Farmer, G., Brock, J., & Brink, J.E. (1997): Estimating millet production for famine early warning: An application of crop simulation modelling using satellite and ground-based data in Burkina Faso. Agricultural and Forest Meteorology 83: 95-112.

Troufleau, D. & Soegaard, H. (1998): Deriving surface water
status in the Sahel from the Pathfinder AVHRR Land data set.
Physics and Chemistry of the Earth 23(4): 421-426.

Wang, Y.-P. & Leuning, R. (1998): A two-leaf model for canopy conductance, photosynthesis and partitioning of available energy I: Model description and comparison with a multilayered model. Agricultural and Forest Meteorology 91: 89-111.