Journal of

Hydrology ELSEVIER

Journal of Hydrology 188-189 (1997) 633-650

Carbon dioxide flux, transpiration and light response of millet in the Sahel T. Friborg*, E. Boegh, H. S o e g a a r d Institute of Geography, Universityof Copenhagen, Oester Voldgade 10, 1350 Copenhagen K, Denmark

Abstract Within the framework of the HAPEX-Sahel experiment carried out in Niger during the rainy season of 1992, measurements of fluxes defining the vegetation-atmosphere interaction were conducted over a millet field, for a period of nearly 2 months. These measurements comprised continuous recording of solar radiation, atmospheric carbon dioxide fluxes using the eddy correlation technique, and sap flow through millet plants. Based on biom~tric measurements of the millet plants comprising height, spacing and leaf area index, the solar radiation is converted to absorbed photosynthetically active radiation (aPAR). The coupling between the three parameters is examined in pairs. The diurnal and seasonal variations are analysed in relation to plant development. A strong linear relationship between aPAR and carbon dioxide assimilation can be established from the measurements, giving a quantum yield of 0.03 mol CO2 moi -1 quanta. A comparison between CO2 flux and transpiration shows that this relationship is affected by the water vapour pressure deficit of the atmosphere, but corresponds to the results found for other drought-tolerant C4 crops.

I. Introduction The study of interactions between the land surface and the atmosphere has received much attention during the last decade. The present study is taking place within the framework of the HAPEX-Sahel experiment (see Goutorbe et al. (1994) and Goutorbe et al. (1997)), which was carried out in Niger to obtain a better understanding of the interaction between the land surface and the atmosphere in a region which is currently threatened by desertification. Microclimatic measurements, such as atmospheric fluxes, provide direct information about the exchange of atmospheric constituents between the land surface and the * Corresponding author. 0022-1694/97/$17.00 © 1997- Elsevier Science B.V. All rights reserved PII S0022-1694(96)03196-4

634

635

T. Friborg et al./Journal of Hydrology 188-189 (1997) 633-650

out on a sandy millet (Pennisetumglaucum (L.) R. (Br.) field (13°32.10'N, 2°30.83'E) near the village of Foundou Beri, approximately 40 km ENE of Niamey. The site was located near the centre of the HAPEX square and was referred to as West Central Supersite b (WCSb). The vegetation cover at the site was heterogeneous and further characterized by a low plant density. The millet was sown around 15 July, in pockets of 3 - 1 0 plants at about 1100-2500 pockets per hectare. At the beginning of the experiment the millet was 0.5 m high with a leaf area index (I.AI) of 0.17. Five weeks later the millet had reached its maximum height of 2 m, and LAI had increased to 0.37. Intercropping with Hibiscus added to the heterogeneity in the last part of the measuring period, thus being representative of the traditional crop pattern on Sahelian cultivated soils. During the first part of the experimental period the Hibiscus crop was barely visible. However, from the last week of September it grew rapidly and reached a height of 0.5 m during the last days of the experiment. The area is characterized by a yearly precipitation of approximately 600 mm, which occurs almost exclusively during 3 months (July-September). The variation of yearly and daily precipitation is, in general, high. In 1992, well-distributed rainfall throughout the growing season (80-90 days) secured a good millet harvest from the field despite the relatively low total precipitation amount of 483 mm. During the experiment, the precipitation was 228 mm. The last precipitation amount recorded was on the sixty-third day after sowing (Day Of Year 263). Rainfall and leaf area index (LAI) during the experiment are given in Fig. 1.

Rainfall and LAI Aug. 17 to Oct. 9 Foundou Beri

60

0.4

/, /

50

......

?

0.32

,4/

I

40

0.24 ~" E

E 30 ~ 20 10

o2~6,

i

0.16 .~

0.08

,,!!:,ii,,I!,°, ,,.... ,,°r

.....

~6

.......

29 .%0

Day of year 1992 Rainfall

LAI

Fig. 1. Rainfall and leaf area index(LAI)during the experimentalperiod, 17 August-9 October.

636

T. Friborg et aL/Journal of Hydrology 188-189 (1997) 633-650

2.2. Materials and methods 2.2.1. C02 fluxes Measurements of sensible and latent heat fluxes, and CO2 flux, were obtained by the use of the eddy correlation technique. For fluctuations in wind speed, a three-dimensional Solent sonic anemometer (Gill Instruments, Lymington, UK) was used. A fast-responding closed-path, IR gas analyser (Li-6262, Li-Cor, Inc., Lincoln, NE, USA) was used for measuring fluctuations in CO2 concentrations. Data were stored in a personal computer at 21 Hz, and fluxes were calculated and corrected as described by Monerieff et al. (1997a). The eddy correlation system was mounted at 2.5 m above the vegetation height, near the leeward edge of the millet field to obtain maximum fetch (approximately 200 m) in prevailing wind directions. CO2 flux (Fc) measured above the vegetation as in this experiment, gives a good indication of CO2 exchange between surface and atmosphere. However, these fluxes do not equal the net assimilation of CO2 in the vegetation, because CO2 from soil respiration can be directly assimilated by the plants. On a diurnal basis the CO2 fluxes that should be taken into account are given in the gross photosynthesis (gPS), expressed by Rosenberg et al. (1983) as g P S - F c +R e +Rg +R r

(1)

where R c is crop respiration, Rg is soil respiration and Rr is root respiration. R c and RT are both respiratory losses of CO2 from the vegetation, which, during the daytime, as a response to the photosynthetic process, are re-assimilated by the vegetation. During the hours of photosynthesis, the net amount of CO2 assimilated by the vegetation is expressed by the apparent photosynthesis (aPS): aPS = Fc +Rg

(2)

To obtain aPS it is necessary to distinguish between soil respiration and vegetation-related respiration. This is not directly possible solely from above-canopy measurements. As a consequence, a simple model, described in Section 3, was applied to evaluate soil respiration. 2.2.2. Transpiration The transpiration was measured with six sap flow gauges (Models SGA-10ws and SGA13ws, Dynagage Inc., Houston, TX, USA) placed on the stems of six millet plants which were considered to be representative of the crop. The method is based on the stem heatbalance technique involving the continuous heating of a short vertical distance of a stem section, as described by Sakuratani (1981), Baker and Van Bavel (1987) and Boegh (1993). Transpiration per unit leaf area was calculated from the sap flow through the individual gauges. The sap flow per unit leaf area was then multiplied by the LAI of the field to extrapolate the transpiration from the individual plants to the field as a whole (see Soegaard and Boegh (1995)). 2.2.3. Leaf area index The areas of the individual leaves were measured manually with a ruler, using an empirical relation between the product of length and width of leaf and leaf area (Soegaard

1". Friborg et aL/Journal of Hydrology 188-189 (1997) 633-650

637

and Boegh, 1995). The LAI for the millet field was estimated approximately every fifth day by measuring the transmission of diffuse light through the canopy with an optical sensor (Li-cor, I_AI-2000 Plant Canopy Analyzer). Each set of measurements consisted of 60 recordings at marked points located along three transects of 30 m each. Quality checking of the LAI estimates has been discussed by Soegaard and Boegh (1995). 2.2.4. Solar radiation Measurements of incoming and reflected solar radiation were carried out with two Kipp and Zonen (Delft, Netherlands) CM-5 radiometers mounted I m above the millet canopy. The global radiation data were converted into absorbed photosynthetically active radiation (aPAR) using a model developed by B6gu6 (1992). As the absorption efficiencies for diffuse and direct solar radiation are not equal, the global radiation was separated in two components, namely, diffuse and direct radiation. For clear sky conditions both the direct and the total irradiance can be calculated from astronomical parameters (Sellers, 1965) using an atmospheric transmissivity of 0.85. For clear sky conditions the diffuse radiation (R dxlea0can thus be found simply by subtraction. On a daily basis the diffuse radiation was found to be approximately 20% of measured global radiation. Cloudy and hazy conditions were found to increase the diffuse radiation load by the same amount that the global radiation decreased (AR) compared with the theoretical clear sky radiation. R d could thus simply be calculated from Rd =Rd, clear+ A R

(3)

In instances of very dense cloud and rainfall, the global radiation is found to be less than Rd,d,ar, but here Rd is simply set to be equal to measured global radiation. This diffuse radiation model was tested against measured diffuse radiation at a site operated by Winand Staring Centre located approximately 10 km east of the millet site. The model was run on 10 min time intervals. In Fig. 2 the daytime mean values (07:00-17:00 h GMT) of modelled diffuse radiation are plotted against the measured data. Allowing for some scattering owing to different cloud conditions at the two sites, 68% of the total variance is accounted for by the 1:1 relationship. By combining the radiation data with biometric measurements comprising crop height, spacing and LAI, the aPAR could then be calculated applying a model adopted from B6gu6 (1992). In this model, the canopy is approximated by an array of porous cylinders representative of a millet field. The model is split into two submodels which are applied at different levels: (1) a macro-structure model, which is based on geometrical optics theory, and which has been adapted to porous, opaque cylinders; (2) a micro-structure model, which is based on canopy transmission theory. As such, it takes into account the interaction between the clumps, i.e. mutual shading. The interception efficiency (ei) in B~gu6's model is calculated separately for diffuse and direct radiation, and the final ei is calculated as e,i • (13isRb / gs) + ( ~idR d / Rs)

(4)

where Rs is total incoming solar radiation, Rb is incoming direct solar radiation, Rd is the

T. Friborg et al./Journal o.fHydrology 188-189 (1997) 633-650

638

Day-Time Mean Diffuse Radiation 400

~

350 J

c O

300 m

a 250 O

~200

150 1 150

200

•

250

300

Measured Diffuse Radiation (WIm 2)

Day-time mean value

1:1 line

350

400

I

Fig. 2. Modelled diffuse radiation vs. diffuse radiation measured by Winand Staring Centre.

incoming diffuse solar radiation, and eis and eid are the interception efficiencies for direct and diffuse sunlight, respectively, eis depends on shadowing, and it is modelled as a daytime variable function of the solar zenith and azimuth angles. The applied geometrical model has been discussed by BEgu6 (1992). The diffuse eid is, o n the other hand, independent of shadowing. It depends only on the geometry of the plant cover, and it is calculated as a daytime integral of the interception efficiency of direct radiation. The porosity of the clumps is computed from the micro-model as a function of the leaf area density, the leaf inclination distribution type, the dimensions of the cylinders (height and radius), and the transmission of green leaves in the appropriate spectral band. The PAR absorption efficiency is then modelled as £:a = [Otl/(O/l + Pl)][~:i +

(1 - ~i)Pseid]

(5)

where cq is the leaf absorption coefficient, p~ is the leaf reflection coefficient and Ps is the soil reflection. (o~l/(cq + Pl)) expresses the fraction of PAR which is absorbed by the leaf and intercepted by a fraction el. The non-intercepted f l u x (1 - el) is transmitted to the soil, reflected by a factor Ps and lastly intercepted as diffuse light with the efficiency Rid. The absorption efficiency of millet was modelled at 10 min intervals using inputs of LAI, average plant height, spacing between millet clumps (3 m in the present case) and leaf distribution type (erectophile). The quantity of absorbed PAR was estimated at 10 min intervals as a multiple of R, (W m2), a spectral factor defining the ratio of PAR to global radiation (equal to 0.446) and the modelled absorption efficiency.

639

T. Friborg et aL/Journal of Hydrology 188-189 (1997) 633-650

3. Modelling soil respiration In a number of studies bare soil and above-ground respiration have been measured directly through the use of chambers (e.g. Biscoe et al., 1975; Mogensen, 1977). Through these studies an exponential relationship between respiration and temperature has been found to be valid for both bare soil and vegetation related respiration losses. For soil respiration this relationship can be written as (Mogensen, 1977) r, ~(r~- r,,,f)/lo g ="ref~10

R

(6)

where R s and Tg are soil respiration and soil temperature at any time of day, and R =f and Tref are the soil respiration and soil temperature at a specific time of day. Q to is the factor by which respiration increases for a temperature increase of 10°C. A similar dependence on temperature has been found valid for plant respiration (Re), using air temperature instead of soil temperature (Da Costa et al., 1986). To account for the effect of soil respiration on the daytime apparent photosynthesis (aPS), an empirical model based on the relationship in Eq. (6) has been elaborated. At night-time there is no photosynthesis, and Fc therefore equals Re + Rg + R r (Anderson et al., 1984). In Fig. 3 the seasonal variation in night-time (17:00-07:00 h GMT) Fc is shown together with the daily mean temperatures of the air and topsoil. It is obvious that Fc, and thus the respiration loss, approaches zero by the end of the

Average night-time Fc,Tair and Tsoil Day 230-283 -

.] 38 36

2.5

I I

i

34

t 32 k

- -

O 3O o_. I-28

oE1.5 o

E:

1

O 14.

26 0.5 i

24

i i

t 01

i

240

230

i

. . . . . . . . . .

250

~ . . . . . .

260 Day

i

22 290

r

270

280

i

~_ ,,

Fc=(Rc+Rg+Rr)

Tair_

,~ Tsoil(-0.1m)

i

il

Fig. 3. Mean night-time CO2 flux (Fc), air temperature (Tair) at 2 m amd soil temperature (T,oii) at 0.1 m depth, 17 August-9 October 1992 (Days 230-283).

640

T. Friborg et al./Journal of Hydrology 188-189 (1997) 633-650

Night-time Fc and Model Day 230-283

A

"e

mm

2 Ir •

•

.

m

a

m

"5 E 1.5 o. u_ E 1

,

X 7~ ¢:

0 o

0.5 0 I

-0,5 230

I .

.

.

240

.

.

250

260

Day

270

280

290

.

Fig. 4. Measurednight-timeCO2 flux and modelledrespirationcomponents, 17 August-9 October. campaign. By comparing Fig. 1 and Fig. 3 it is seen that the temperatures of both the soil and the air have a seasonal variation closely linked to the rainfall distribution given in Fig. 1. The more frequent the rainfall, the lower the temperature. As increasing temperature would normally lead to higher respiration rates, the decrease in night-time respiration cannot be explained by the temperature variation. It can be assumed that the plant-related respiration rate (Re + R,) is closely linked to the LAI. However, when examining Fig. 1, it should be noted that the LAI for the millet field decreases only slowly at the end the

T. Friborg et aL/Journal of Hydrology 188-189 (1997) 633-650

64l

measured night-time fluxes for the wet part of the period (until Day 260). Assuming no change occurs in the organic matter content of soils, a constant Q10 value can be used to calculate diurnal variations in R g directly from Eq. (6). The average night-time R g found in this way is approximately 0.8/xmol m -2 s -1, which is used as R ref in Eq. (6). The expression is multiplied by a soil moisture factor, which is unity up to Day 270 and then decreases linearly to zero by Day 284. The magnitude of the Q10 factor is, of course, important to the calculation, but in this case, when the different components of respiration are not measured, it is impossible to determine precisely. However, looking at the model in Fig. 4 a magnitude of two seems to be appropriate. This value is also often reported in the literature (Da Costa et al., 1986; Raich and Schlesinger, 1992). Despite the number of assumptions used in the calculation of the respiration, as described above, the modelled values correspond reasonably well to the measured night-time respiration. It should also be stressed that as the total night-time respiration in this experiment is approximately 10% of the daytime fluxes, where some of it is plant respiration (R c + R r), the error owing to soil respiration in the daytime fluxes must be less than 10%. Fig. 5 shows Fc, Rg and Fc +Rg for Day 269, and is an example of the magnitude of the modelled soil respiration. As Rg follows the diurnal variations in soil temperature, Rg reaches its maximum value during late afternoon. In the following, aPS represents the

Diurnal variation in Fc,Rg and Fc+Rg Day 2 6 9

4 2 0

"E -2 o

-4

T: -e -10 -12 -14

i

0

~

4

8

•

Fc

12 GMT

Rg

x

16

20

"

--

24

Fc+Rg

[

Fig. 5. Diurnal variation in measured CO2 flux (Fc) and modelled soil respiration (Rs), 25 September 1992 (Day 269).

642

T. Friborg et al./Journal of Hydrology 188-189 (1997) 633-650

sum of the C02 flux above the vegetation measured by the eddy correlation system and the soil respiration calculated by the model.

4. Results and discussion

4.1. Light response The relationship between aPS and incoming radiation illustrates the photosynthetic light response. Typical light response curves for unstressed plants show a nearly linear relationship between Fc and Rs for low to moderate radiation loads (100-500 W m2), whereas at higher radiation loads aPS is less dependent on radiation. Studies conducted during the HAPEX-Sahel experiment confirm this variation pattern (Moncrieff et al., 1997b), which behaves according to a rectangular hyperbolic function. To examine the millet light response, 10 min average values of aPS have been plotted against global radiation in Fig. 6. To represent the seasonal variation, 4 days have been selected evenly throughout the growing season for the analysis. The first example, Day 239, is from the early part of the growing season, when the millet was at the vegetative stage. The average plant height was 1.25 m and the millet was growing at roughly 5 cm day -~. The LAI was 0.25. The cloud cover was limited to a few cumulus clouds around noon. By Day 256 the rainy season had nearly ended. However, it had rained the previous night so there was no shortage of soil moisture, and atmospheric evaporative demand was low. During the afternoon some large cumulus clouds developed. The millet was near the end of the vegetative stage, with a mean height of 1.90 m and an LAI close to the maximum of 0.36. Day 269 represents the beginning of the dry season. The sky was clear throughout the day and the millet had just reached its maximum mean height of 2.00 m. On Day 283 the millet was at the senescent stage, just before harvest. The day was clear, with only a few convective clouds developing in the early afternoon. Even though some scattering occurs, a change in slope in the scatterplot at around 400 W m 2 can be identified in all four examples. Most C4 plants do not show light saturation under normal daylight conditions (Rosenberg et al., 1983), and the change in slope of the curve is most likely to be due to a less efficient absorption of direct irradiance in the canopy, as discussed below. Furthermore, it can be seen that the aPS/Rs ratio is clearly different from one day to the next. In Fig. 7 the same aPS data are plotted against aPAR as calculated by the model. In contrast to Fig. 6, the relationship between aPAR and aPS is close to linear. In all four graphs a line which has a slope of 0.175/~mol W -I m -2 is shown. It can be seen that the slope of the aPS/aPAR ratio is nearly identical on all 4 days. A linear relationship between Fc and aPAR has also been found by Baldocchi (1994) for both maize (C4 species) and wheat (C3 species). The main reason for the difference in the relationships shown in Fig. 6 and Fig. 7 is related to the light absorption mechanism. At low sun angles, the diffuse radiation with the highest absorption efficiency predominates, leading to an increase in the aPAR/R s ratio.

T. Friborg et aL/Journal of Hydrology 188-189 (1997) 633-650 TO~

] .......

.m+

i

~oo;

Di

0

643

,

i 0

nn~

~0 0 Oo

In l'Y

,,.2

2~

U)

i

O

1::3

0

~m 03

i

++

00 I

TM

>~,

~m

o~

m+

,,,el

i

E

D

1.I.

0

~t

q

O

ii

Di

+ O U_

c~

s~ D

E

D

08

8

ra~ ('-4

c) o

m '

o

m

'7,

"7

u'~

m,

~ ,

(:3

m

"7

~

==