15
Agricultural and Forest Meteorology 154–155 (2012) 84–98 Contents lists available at SciVerse ScienceDirect Agricultural and Forest Meteorology jou rn al h om epa ge: www.elsevier.com/locate/agrformet Long-term observations of turbulent fluxes over heterogeneous vegetation using scintillometry and additional observations: A contribution to AMMA under Sudano-Sahelian climate Adrien Guyot a,b,, Jean-Martial Cohard c,∗∗ , Sandrine Anquetin a , Sylvie Galle d a LTHE, CNRS, Université Grenoble 1, France b National Centre for Groundwater Research and Training, The University of Queensland, Australia c LTHE, Université Grenoble 1, France d LTHE, IRD, Université Grenoble 1, France a r t i c l e i n f o Article history: Received 10 April 2011 Received in revised form 15 October 2011 Accepted 17 October 2011 Keywords: Composite landscape Energy balance Evaporative fraction Scintillometry Sudano-Sahelian climate Turbulent fluxes a b s t r a c t Based on a 3-year period of infrared scintillometry, soil and meteorological measurements, this study presents an analysis of the surface energy balance partitioning over a heterogeneous savannah, in the Sudano-Sahelian region. The site is located in Northern Benin, meso-site of the African Monsoon Multi- disciplinary Analyses (AMMA) project. The 3-year period enables an analysis of several alternate dry and wet periods, as well as the intermediate dry-to-wet and wet-to-dry periods. Infrared scintillometry, coupled with measurements of the available energy (net radiation minus ground heat flux) and a careful analysis of the aerodynamic properties of the scintillometer footprint, are employed to provide robust estimates of the turbulent (sensible and latent heat) fluxes over complex terrain, in terms of the topography and in terms of the spatially and temporally heterogeneous vegetation cover. A characterization of the uncertainties on each term of the energy balance is given at the scale of the scintillometer footprint. These uncertainties strongly depend on the season for the residual latent heat flux. Results point out that the climate of the Sudano-Sahelian region is characterized by a strong seasonal cycle and inter-annual variability, related to changing atmospheric and land surface conditions. The evap- orative fraction is found to be relatively constant during the wet period (0.67) and more variable during the dry and intermediate periods. In addition, sensible heat flux and net radiation are well correlated during the dry season. The diurnal cycle shows a predominance of evaporation during the wet season and sensible heat during the dry season. Results point a significant latent heat flux during the dry period, signature of persistent vegetation in the Sudano-Sahelian region. Finally, that data set at hourly time step would provide useful information for modelling and the parameterization of the associated processes for this region. © 2011 Elsevier B.V. All rights reserved. 1. Introduction West Africa is known to be a vulnerable area exposed to climate changes where large uncertainties remain, concerning rain amount tendencies and induced water resources. In particular, the impact of climate changes on the African surfaces and its feedback through their energy balance are still poorly understood (Boko et al., 2007), Corresponding author at: The University of Queensland, School of Civil Engi- neering, St. Lucia 4072, Brisbane, Australia. Tel.: +61 733653887. ∗∗ Corresponding author at: Laboratoire d’étude des Transferts en Hydrologie et Environnement, LTHE, 1023, Rue de la Piscine, BP53, 38041 Grenoble Cedex 9, France. E-mail addresses: [email protected] (A. Guyot), [email protected] (J.-M. Cohard). despite the fact that spatial and temporal energy partitioning vari- ability is though to play a major role in the whole water and energy cycle of the West African monsoon (Charney et al., 1975; Zheng and Eltahir, 1997; Wang and Eltahir, 2000). From a hydrological point of view, Sahelian region had suffered, in the past decades, from dra- matic/severe droughts, which had considerably changed land use and surface energy partitioning over a wide area. West African stud- ies have then primarily focused on Sahelian surfaces, more exposed to climate variability (Wallace et al., 1991; Verhoef et al., 1996; Gash et al., 1997; Kabat et al., 1997; Lloyd et al., 1997). Kabat et al. (1997) and Gash et al. (1997) have speculated on the possible role of the vegetation gradient in the control of the monsoon. In this context, they asked the question of the energy partitioning depending on the vegetation types. This has been explored in the recent AMMA experiment (African Monsoon Multidisciplinary Analyses) in 2006 0168-1923/$ see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.agrformet.2011.10.008

Agricultural and Forest Meteorology€¦ · 4/10/2017  · pressure (Pa) 2m Vaisala WXT510 Station 10s 30min Air temperature (T) (Vaisala Oyj, Helsinki, Finland) 10s 30min Relative

  • Upload
    others

  • View
    0

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Agricultural and Forest Meteorology€¦ · 4/10/2017  · pressure (Pa) 2m Vaisala WXT510 Station 10s 30min Air temperature (T) (Vaisala Oyj, Helsinki, Finland) 10s 30min Relative

Agricultural and Forest Meteorology 154– 155 (2012) 84– 98

Contents lists available at SciVerse ScienceDirect

Agricultural and Forest Meteorology

jou rn al h om epa ge: www.elsev ier .com/ locate /agr formet

Long-term observations of turbulent fluxes over heterogeneous vegetation usingscintillometry and additional observations: A contribution to AMMA underSudano-Sahelian climate

Adrien Guyota,b,∗, Jean-Martial Cohardc,∗∗, Sandrine Anquetina, Sylvie Galled

a LTHE, CNRS, Université Grenoble 1, Franceb National Centre for Groundwater Research and Training, The University of Queensland, Australiac LTHE, Université Grenoble 1, Franced LTHE, IRD, Université Grenoble 1, France

a r t i c l e i n f o

Article history:Received 10 April 2011Received in revised form 15 October 2011Accepted 17 October 2011

Keywords:Composite landscapeEnergy balanceEvaporative fractionScintillometrySudano-Sahelian climateTurbulent fluxes

a b s t r a c t

Based on a 3-year period of infrared scintillometry, soil and meteorological measurements, this studypresents an analysis of the surface energy balance partitioning over a heterogeneous savannah, in theSudano-Sahelian region. The site is located in Northern Benin, meso-site of the African Monsoon Multi-disciplinary Analyses (AMMA) project. The 3-year period enables an analysis of several alternate dry andwet periods, as well as the intermediate dry-to-wet and wet-to-dry periods.

Infrared scintillometry, coupled with measurements of the available energy (net radiation minusground heat flux) and a careful analysis of the aerodynamic properties of the scintillometer footprint,are employed to provide robust estimates of the turbulent (sensible and latent heat) fluxes over complexterrain, in terms of the topography and in terms of the spatially and temporally heterogeneous vegetationcover. A characterization of the uncertainties on each term of the energy balance is given at the scale ofthe scintillometer footprint. These uncertainties strongly depend on the season for the residual latentheat flux.

Results point out that the climate of the Sudano-Sahelian region is characterized by a strong seasonalcycle and inter-annual variability, related to changing atmospheric and land surface conditions. The evap-orative fraction is found to be relatively constant during the wet period (0.67) and more variable duringthe dry and intermediate periods. In addition, sensible heat flux and net radiation are well correlatedduring the dry season. The diurnal cycle shows a predominance of evaporation during the wet seasonand sensible heat during the dry season. Results point a significant latent heat flux during the dry period,signature of persistent vegetation in the Sudano-Sahelian region. Finally, that data set at hourly time stepwould provide useful information for modelling and the parameterization of the associated processes forthis region.

© 2011 Elsevier B.V. All rights reserved.

1. Introduction

West Africa is known to be a vulnerable area exposed to climatechanges where large uncertainties remain, concerning rain amounttendencies and induced water resources. In particular, the impactof climate changes on the African surfaces and its feedback throughtheir energy balance are still poorly understood (Boko et al., 2007),

∗ Corresponding author at: The University of Queensland, School of Civil Engi-neering, St. Lucia 4072, Brisbane, Australia. Tel.: +61 733653887.∗∗ Corresponding author at: Laboratoire d’étude des Transferts en Hydrologie et

Environnement, LTHE, 1023, Rue de la Piscine, BP53, 38041 Grenoble Cedex 9,France.

E-mail addresses: [email protected] (A. Guyot),[email protected] (J.-M. Cohard).

despite the fact that spatial and temporal energy partitioning vari-ability is though to play a major role in the whole water and energycycle of the West African monsoon (Charney et al., 1975; Zheng andEltahir, 1997; Wang and Eltahir, 2000). From a hydrological pointof view, Sahelian region had suffered, in the past decades, from dra-matic/severe droughts, which had considerably changed land useand surface energy partitioning over a wide area. West African stud-ies have then primarily focused on Sahelian surfaces, more exposedto climate variability (Wallace et al., 1991; Verhoef et al., 1996; Gashet al., 1997; Kabat et al., 1997; Lloyd et al., 1997). Kabat et al. (1997)and Gash et al. (1997) have speculated on the possible role of thevegetation gradient in the control of the monsoon. In this context,they asked the question of the energy partitioning depending onthe vegetation types. This has been explored in the recent AMMAexperiment (African Monsoon Multidisciplinary Analyses) in 2006

0168-1923/$ – see front matter © 2011 Elsevier B.V. All rights reserved.doi:10.1016/j.agrformet.2011.10.008

Page 2: Agricultural and Forest Meteorology€¦ · 4/10/2017  · pressure (Pa) 2m Vaisala WXT510 Station 10s 30min Air temperature (T) (Vaisala Oyj, Helsinki, Finland) 10s 30min Relative

A. Guyot et al. / Agricultural and Forest Meteorology 154– 155 (2012) 84– 98 85

and documented by Ramier et al. (2009), Timouk et al. (2009) andEzzahar et al. (2009).

The Sudano-Sahelian region, a large East-Western band(approximately between 7◦N and 12◦N), with an intermediate cli-mate between the dry Sahel and the Guinean Coast, is also animportant zone in the West African climatic context. In particu-lar, the role of this South-North West African vegetation gradienton Monsoon cycle is not fully understood. In the Sudano-Sahelianregion, annual rainfall amounts range from 800 mm to 1200 mm,which also confer to this region an important role of agriculturalproduction. The uncertainty in rainfall tendency and the demo-graphic growth caused by Sahelian migration is then susceptibleto increase anthropic pressure on land use and water resources.The need for long term combined hydrological and earth surfacestudies becomes a priority to prepare water management policy.

Sudano-Sahelian region is composed of a wide range of veg-etation cover, a mixture of cultivated areas and savannah. In thisregion, energy fluxes have been recently documented in Ivory Coast(Touré, 2004), in the Upper Volta (Schuettemeyer et al., 2006)and in Burkina-Faso (Bagayoko et al., 2007). More recently, inBurkina Faso, close to the Sahelian landscapes, a whole energy-biogeochemical study of a savannah has been undertaken byBruemmer et al. (2008) and Groete et al. (2009). Bruemmer et al.(2008) published the first set of pluri-annual data for a Sudano-Sahelian savannah landscape undisturbed by human presence.These studies emphasize the need for more long-term observa-tions of the energy balance partitioning in the West African contextwhere temporal variability of energy fluxes is high. These savan-nah areas are deforested little by little and changed to crop fields.This adds another difficulty to study Sudano-Sahelian areas becauseof the patchiness of the land cover. This spatial heterogeneitymakes the comparison between local data and satellite estima-tions or Soil Vegetation Atmosphere (SVAT) models outputs veryquestionable. Although the need of energy partitioning observa-tions is required, Timouk et al. (2009) proposed an aggregationscheme from punctual data to document the energy partitioningover patchy areas, which have been used for SVAT validation (Booneet al., 2009).

Several authors have shown the relevance of the use of infraredscintillometer measurements to estimate sensible heat flux overpatchy areas, and latent heat flux when combined with an esti-mation of the other terms of the energy budget (Meijninger et al.,2002b; Schuettemeyer et al., 2006; Hoedjes et al., 2007; Ezzaharet al., 2009). Most of these authors estimated the net radiation asa local measurement, which does not represent the whole surfacewhere the energy budget should be done. Furthermore, the groundheat flux is often estimated as a proportion of the net radiation.

Recently, Guyot et al. (2009) proposed a similar methodologywhere they emphasized the necessity of using aggregated aero-dynamical parameters, net radiation and ground heat flux at thescintillometer footprint scale. These estimations lead to a compre-hensive energy budget, and therefore a robust estimation of thelatent heat flux (LE) as a result of the energy balance over a com-posite Sudano-Sahelian landscape. However, Guyot et al. (2009)studied only a short period of three months of the dry season,period with few atmospheric and surface conditions variability.The study of multi-annual energy partitioning cycles requires tak-ing into account the evolution of vegetation, in terms of surfacereflectivity and surface aerodynamic properties.

In the present study, we propose to extent Guyot et al. (2009)methodology to take into account the temporal variability andspatial heterogeneity of the surface. It is applied to a Sudano-Sahelian landscape, part of the AMMA-CATCH network. The site,the experimental setup and the climatological context are pre-sented in Section 2. The methodology includes temporal variabilityfor albedo, roughness length and displacement height, which are

known to be the most sensible parameters to derive sensible heatfluxes from scintillometric measurements. A footprint temporalvariability is also included in the sensible heat flux calculation. Thisis detailed in Section 3. Then, Section 4 presents three years ofmeasurements, encountering a strong seasonal and inter-annualvariability of both atmospheric and surface conditions. Finally, afirst daily, seasonal and annual analysis is proposed in Section 5 inthe light of composite daily cycles and annual cycle of evaporativefraction. The major conclusions are reviewed in Section 6.

2. Experimental setup and study period

2.1. Experimental setup

The 12 km2 Ara watershed (Fig. 1) and its experimental setuphave been presented extensively in Guyot et al. (2009).

The base map in Fig. 2 presents the vegetation distribution inthe Ara catchment. This classification is derived from SPOT 4-HRVIRscenes acquired during 2005 and 2006 (I. Zin, personal communi-cation). Vegetation has been classified as: bare soil/fallows/crops(32%), shrub savannah (61%) and woody savannah (7%). This dis-tribution is similar to the distribution observed at the Dongacatchment scale, i.e. bare soil/fallows/crops (31%), shrub savannah(57%) and woody savannah (12%).

The instrument locations are presented in Fig. 2 and Table 1gives all sensors specifications. In addition to the setup from Guyotet al. (2009), two radiation stations have been installed in December2008 and are equipped with CNR2 radiation budget sensors to mon-itor the net radiation in long wave and short wave over: (i) a woodysavannah RWS and (ii) a shrub savannah (RSS).

2.2. Study period and climatological context

This paper focuses on the years 2006, 2007 and 2008 (Fig. 3).The seasonality of the monsoon is driven at the surface by the fluc-tuation of the ITCZ (Inter Tropical Convergence Zone) (Eltahir andGong, 1996). The climate is then characterized by a succession ofdry and wet seasons when most of the rain falls in a 5-monthsperiod. As they drive soil moisture, vegetation dynamic and theenergy partitioning, rainfalls are described in their climatic context.Fig. 4 presents the cumulative rainfalls at Djougou (10 km awayfrom the Ara catchment). Most of the total precipitation occursbetween mid-April (DOY 130) to the end of October (DOY 300) withthe peak of the monsoon season in July–August (DOY 180–240) (LeLay and Galle, 2005). The mean annual rainfall for the 1950–2009period at Djougou is 1303 mm, with a standard deviation of 223 mm(Fig. 4).

Rainfalls over the Ara catchment are estimated by a Thiessenpolygons interpolation of the three rain gauge stations. Annual rain-fall was 1254 mm (resp. 1339 mm at Djougou) in 2008, 1184 mm(resp. 1109 mm at Djougou) in 2007 and only 889 mm (resp.1005 mm at Djougou) in 2006. The year 2006 is one of the driestyears of the 1950–2009 period (Fig. 4). This year is also character-ized by a very early rain event at the beginning of the season. Theyear 2007 is slightly under the mean, but within the range of vari-ability of the 1950–2009 period. On the other hand, the year 2008is in the mean of the last 60 years (Fig. 4).

Dry and wet periods are generally defined according to thezonal wind (Sultan and Janicot, 2004; Lothon et al., 2008). We pro-pose here to distinguish dry and wet periods according to the VPDparameter (Fig. 3(e)) because it is more appropriate for energy par-titioning since it takes into account indirectly vegetation and soilmoisture. Indeed, there is a close relationship between stomatalsensitivity and the vapor pressure deficit VPD (Oren et al., 1999),making this parameter of high interest to distinguish vegetation

Page 3: Agricultural and Forest Meteorology€¦ · 4/10/2017  · pressure (Pa) 2m Vaisala WXT510 Station 10s 30min Air temperature (T) (Vaisala Oyj, Helsinki, Finland) 10s 30min Relative

86 A. Guyot et al. / Agricultural and Forest Meteorology 154– 155 (2012) 84– 98

Table 1Characteristics of the instruments.

Variable Height or depth Instrument Executioninterval

Storageinterval

Atmospheric measurementsAtmospheric pressure (Pa) 2 m Vaisala WXT510 Station 10 s 30 minAir temperature (T) (Vaisala Oyj, Helsinki, Finland) 10 s 30 minRelative humidity (RH) 10 s 30 minWind speed (v) 10 s 30 minWind direction (Dirvent) 10 s 30 min

Radiation bare soil/fallows (SWin , SWout , LWin , LWout) 2 m CNR1 Radiometer (Kipp and Zonen, Delft, TheNetherlands)

10 s 30 min

Radiation shrub savannah (SW, LW) 5 m CNR2 Radiometer (Kipp and Zonen) 10 s 30 minRadiation woody savannah (SW, LW) 10 m CNR2 Radiometer (Kipp and Zonen) 10 s 30 min

EM signal fluctuations 5 m towers BLS 900 Scintillometer 1 s 1 min!2

x heq ∼ 19 m (Scintec AG, Rottenburg, Germany)3D wind components (u′ , v′, w′) 4.95 m Solent R3-50 Gill Sonic Anemometer (Gill

Instruments, Lymington, UK)60 Hz 20 Hz

Rainfalls (P) 1.20 m Weighting precipitation gauge (OTTHydrometry Ltd., Dublin, Ireland)

5 min 30 min

Soil measurementsSoil temperature (Ts) − 0.1, − 0.2, − 0.4,

− 0.6, and − 1 mT108 Temperature probe (Campbell Scientific) 5 min 60 min

Soil moisture (") − 0.1, − 0.2, − 0.4,− 0.6, and − 1 m

CS 616 water content reflectometer (CampbellScientific)

5 min 60 min

Temporary atmospheric measurementsAerodynamic properties, d, z0 Site dependent Sonic Anemometer CSAT3 (Campbell Scientific) 60 Hz 15 Hz

stress/unstressed periods. This relationship has been described inthe region by Verhoef (1995), Kabat et al. (1997) and Hanna et al.(2005). In the dry season, when dry air and high temperatures arepredominant, daily VPD reaches value higher than 3 kPa with max-ima at 5 kPa. In contrast, during June to September, daily VPD can belower than 1 kPa, with a conjunction of high relative humidity andlow temperatures. In the following we will refer: (i) to dry periodsfor VPD > 3 kPa; (ii) to intermediate periods for 1 kPa < VPD < 3 kPa;(iii) to wet periods for VPD < 1 kPa (as suggested by Oren et al., 1999).

These periods are reported in Figs. 3, 7, 8 and 9 with shaded areaand appropriate label dry or wet.

Fig. 3 shows meteorological variables, including VPD, from thebeginning of 2006 to the end of 2008. The dry season lasts approxi-mately a month and a half (from December to mid-January). Duringthis period, the wind direction remains east-northeast (Harmat-tan). Around mid-January, a transition period from dry to wet startsand lasts for several weeks. It is characterized by sudden changes inwind direction and air humidity (Fig. 3(c) and (e)), often observed

Fig. 1. Location of the Ara catchment (12 km2): (a) in West Africa, and inside the AMMA experimental framework; (b) in the Ouémé catchment (14 300 km2); (c) in the Dongacatchment (586 km2).

Page 4: Agricultural and Forest Meteorology€¦ · 4/10/2017  · pressure (Pa) 2m Vaisala WXT510 Station 10s 30min Air temperature (T) (Vaisala Oyj, Helsinki, Finland) 10s 30min Relative

A. Guyot et al. / Agricultural and Forest Meteorology 154– 155 (2012) 84– 98 87

Fig. 2. (a) Map of the Ara catchment. Isocontours are 10 m based on the SRTM digital elevation model. Vegetation is indicated in the top-right legend. Main wind directionsare plotted on the left hand side. Instruments are indicated as: T, transmitter; R, receiver; M, meteorological station; RWS , CNR2 radiation sensor on woody savannah; RSS ,CNR2 radiation sensor on shrub savannah. The soil stations are named as Upper Station (US), Middle Station (MS) and Lower Station (LS). The outlet of the catchment isindicated by a square. (b) Orthogonal projection of the scintillometter beam, and underlying topography. Instruments from (a) are indicated with the same typology.

at daily time scales (Sultan and Janicot, 2003; Lothon et al., 2008).In Fig. 3(c), this trend is well marked from February to April in2006. From mid-January to mid-June, these nighttime inflows allowmoist air to gradually humidify the atmosphere in the lower layersat night, which are mixed during the day with the higher layersby convection(Lothon et al., 2008). Once the ground track of theITCZ has moved to the Northern basin of the Altar, the wind sys-tem remains south-west, day and night, leading to a significantmoistening of the atmosphere, and a joint decrease in mean tem-perature (difference of 6 ◦C between the maximum of the wet anddry periods). The wet season runs from the “onset” of the monsoon(Sultan and Janicot, 2004) until the last rains. Then, a short tran-sition period occurs (Lothon et al., 2008), for which the prevailingwind alternates again from east-northeast to southwest.

The inter-annual variability of atmospheric variables (T , RH, andwind direction) depends essentially on the date of onset of the mon-soon, given by the sudden displacement of the ITCZ northward(Sultan and Janicot, 2003). The wetting phase that precedes the“pre-onset” of the monsoon is more or less pronounced dependingon the years. The year 2007 is, for instance, in advance compared toother years. The rains appear earlier in 2007 (Fig. 4). On the otherhand, in 2006 the onset of the monsoon occurred later, despite an

early rain event in the beginning of February (Fig. 3(a)–(f); Lothonet al., 2008; Lohou et al., 2010).

The next section introduces the improvement in the theoryapproaches to evaluate the different terms of the energy balance,at the footprint scale, over a long period, and therefore taking intoaccount the dynamics of the height of the vegetation.

3. Theory and calculation

3.1. Turbulent heat fluxes with scintillometry

Guyot et al. (2009) presented an extensive study to derivesensible heat flux and latent heat flux from scintillometer measure-ments. One can refer to this paper to have a complete descriptionof the methodology.

Because the underlying cover is changing under the scintillome-ter path, we extend the methodology from Guyot et al. (2009) inorder to take into account the change of effective instrument heightalong the year. Latent heat flux is estimated through the energybudget made at the scale of the scintillometer footprint, based onthe net radiation <Rn>, the ground heat flux <G>, and the sensible

Page 5: Agricultural and Forest Meteorology€¦ · 4/10/2017  · pressure (Pa) 2m Vaisala WXT510 Station 10s 30min Air temperature (T) (Vaisala Oyj, Helsinki, Finland) 10s 30min Relative

88 A. Guyot et al. / Agricultural and Forest Meteorology 154– 155 (2012) 84– 98

Fig. 3. Time evolution of (a) daily precipitation over the Ara catchment; (b) daily soil water content at the three soil stations (US, dashed line; MS, grey line and LS, blackline); (c) half-hourly wind direction at Nalohou; (d) half-hourly temperature at Nalohou; (e) half-hourly relative humidity at Nalohou; (f) half-hourly vapor pressure deficitat Nalohou. Lines in (d), (e) and (f) are 10 days moving averages.

heat flux <H>. The brackets represent the spatial average over acomposite area corresponding of the footprint.

Because under the dominating unstable conditions, larger mea-surements heights result in less influence of turbulent friction onthe derivation of sensible heat flux from scintillation measure-ments, the scintillometer has been installed well above the ground.In this way, the influence of uncertainties in the wind velocity andthe roughness length are reduced. In the present study, the estima-tion of <H> from scintillometer measurements described in Guyotet al. (2009) is improved by taking into account the annual change inthe vegetation height below the beam along the year, and by mod-elling the footprint of the instrument where the energy balance isconsidered.

Indeed, according to Hartogensis et al. (2003) and Hartogensis(2006), the uncertainty on <H> largely depends on the uncertaintyof the estimated instrument height. Hartogensis et al. (2003) andHartogensis (2006) showed that these uncertainties arise from boththe error in estimating the height of the scintillometer along thepath, and from the method used to integrate the height value.

The estimations of the aerodynamical parameters (i.e. the dis-placement height d and the roughness length z0) are discussed inthe first section. The second section describes the footprint mod-elling approach. The last section investigates the effective height ofthe scintillometer.

3.1.1. Estimation and aggregation of aerodynamical parametersThe estimations of z0 and d are needed at the scale of the scin-

tillometer device to estimate the flux in the calculation describedabove at least for three objectives: (i) to quantify the friction veloc-ity and its effect on the scintillation measurements with knownwind velocities; (ii) in the estimation of the effective height (z − d)eq(Hartogensis et al., 2003; Hartogensis, 2006); (iii) and to estimatethe footprint of the scintillometer, which will be described in thenext section.

In this study, we use the method proposed by Martano (2000) toestimate z0 and d from a sonic anemometer serie at a single height.This method is applied to a sonic anemometer dataset from a fal-low site (Fig. 2 and Table 1) for the year 2006, considering databy decades. Data analyses have been done using the EdiRe soft-ware (University of Edinburgh, UK) and comprise despiking, angleof attack, double rotation and frequency corrections (Aubinet et al.,1999).

A linear fit is then applied between points to obtain daily values.Fig. 5(a) and (b) shows the annual variations of d and z0 estimatedat this site. These figures show that both aerodynamical parametersfollow the seasonal cycle of the vegetation, and d is roughly foundequal to 0.7 × h, where h is the height of the vegetation that hasbeen measured separately (Guyot, 2010). This is in agreement withthe empirical estimation proposed by Stanhill (1969).

Page 6: Agricultural and Forest Meteorology€¦ · 4/10/2017  · pressure (Pa) 2m Vaisala WXT510 Station 10s 30min Air temperature (T) (Vaisala Oyj, Helsinki, Finland) 10s 30min Relative

A. Guyot et al. / Agricultural and Forest Meteorology 154– 155 (2012) 84– 98 89

Fig. 4. Cumulative rainfall at Djougou during the 1950–2009 period (grey font, meanand standard deviation), and for the 2006, 2007 and 2008 years (thick lines).

For the woody savannah, we supposed d and z0 to be con-stant during the year, because trees grow slowly and keep theirleaves all the year long, except during a 1-month period inDecember–January. z0 and d are found according to Martano (2000)for data collected during a specific field mission in March 2008 withthe sonic anemometer CSAT3 (Table 1) installed successively on ashrub savannah, and on a woody savannah. Values of 4 m (resp.

Fig. 5. (a) Displacement height d for the three vegetation class (woody savannah,black; shrub savannah, grey and bare soil/herbaceous, dashed black); (b) same as(a) for the roughness length z0.

0.4 m) and 0.76 m (resp. 0.88 m) are found for d and z0 for the woody(resp. shrub) savannah (Fig. 5(a) and (b)).

For the shrub savannah, since herbaceous exceed the height ofthe shrubs during the wet season, we characterized d and z0 duringthe dry season when shrubs stand alone. Then, d and z0 are calcu-lated as a combination of values from dry season shrub savannahand fallow sites. This makes the herbaceous values replace step bystep the shrub savannah values, as it can be seen in Fig. 5(a) and(b). The cycle for d and z0 evaluated for 2006 is used for the threeyears.

Then, we apply an aggregation scheme to estimate the averageaerodynamical parameters over a given area. Our approach differsfrom Verhoef (1995) as we are not considering a drag partitionmodel (Raupach, 1992). In our study, the strong heterogeneity ofthe vegetation makes an estimate of the vegetation height inappro-priate for each of the classes.

Therefore, an average roughness length <z0> is calculated fol-lowing Schmid and Bunzli (1995), as explained in Guyot et al.(2009). In the present study, the average <z0> is calculated over themodelled footprint surface. An average displacement height <d> isestimated as a logarithmic mean of local estimations within thefootprint area. The methodology used to determine the footprintarea is described below.

3.1.2. Footprint modellingIn this study, we use the approximate analytical 1D footprint

model of Hsieh et al. (2000). This model is coupled with the weightfunction of the scintillometer following Bastiaanssen (1999) andMeijninger (2003), to obtain a 2D footprint model, which enables todescribe the contribution of the source area to the measured fluxes.Footprints are calculated for the two major winds directions, i.e. NEand SW.

Fig. 6(a) shows the stability classes for the year 2006. Duringdaytime, i.e. 6–18 h (white and clear grey areas), the atmosphereis unstable (class 1–6), while during night-time (dark grays), it ismostly stable (class 7–11). More than 25% are near-neutral cases(− 0.125 < # < 0.125). For a given stability, the changes in the foot-prints are explained by (i) the aerodynamical properties of thesurface, and (ii) the wind direction.

In our case, because changes in stability conditions during theday does not influence much <z0> and <d>, and so footprint areas,footprints are not calculated at each time step. Thus, we considertwo major footprints repartitions, both for neutral atmosphere, andfor (i) ENE, and (ii) SW cases. As <z0> is needed for footprint calcu-lation, footprints are calculated iteratively taking 0.6 m as an initialcondition for <z0>. This gives a vegetation distribution and a new<z0> value, following Schmid and Bunzli (1995). Footprints for ENEand SW winds are represented in Fig. 5(b). The SW footprint haslarger proportions of savannah (bare soil/fallow 11%, shrub savan-nah, 68% and woody savannah 21%) than the ENE footprint (baresoil/fallow 50%, shrub savannah, 45% and woody savannah 5%).

These two footprints enable to use the average <z0>, using theaerodynamical parameters for <z0> depending on the ENE footprintor the SW footprint, as input in the calculation of <H>. In the fol-lowing sections, the energy budget is studied for these footprintareas.

3.1.3. Effective height of the scintillometerBoth topography and vegetation require to introduce the notion

of effective height ((z − d)eq) to estimate <H> from scintillation mea-surements (Chehbouni et al., 2000). We calculate (z − d)eq) as in Eq.(1):

(z − d)eq =[∫ 1

0(zx− dx)− 4/3W(x)dx

]− 3/4

(1)

Page 7: Agricultural and Forest Meteorology€¦ · 4/10/2017  · pressure (Pa) 2m Vaisala WXT510 Station 10s 30min Air temperature (T) (Vaisala Oyj, Helsinki, Finland) 10s 30min Relative

90 A. Guyot et al. / Agricultural and Forest Meteorology 154– 155 (2012) 84– 98

Fig. 6. (a) Stability (# = (z − d)/Lmo) distributions for the year 2006 at the Nalohou EC station; (b) LAS footprints for East-North-Eastern, and South-Western conditions, for zm

= 19 m, neutral atmosphere and aggregated z0 (the basemap is the same as for Fig. 3).

where zx is the distance between the beam and the underlyingground, and dx the displacement, both defined at the normalizeddistance x. W(x) is the weighting function of the scintillometer.

As the underlying topography perpendicular to the beam isnot strictly flat (the mean slope perpendicular to the beam is2.8 ± 0.6%), one needs to take into account the lateral slope to esti-mate (z − d)eq). Thus, we calculated the effective height for virtualbeams shifted from 20 m on each side of the actual beam. 20 m waschosen because this is the resolution of the vegetation grid. Wemade the hypothesis that the mean flow was parallel to the slope.Finally, we were able to estimate the effective height, an estimateof its uncertainty, and its impact on the uncertainty on the sensibleheat flux <H>.

3.2. Ground heat flux

The ground heat flux is calculated using a fast Fourier trans-form analysis of the measured temperature at depth z1 = 10 cm(Horton and Wierenga, 1983). The solution T(z, t), at any depth zand time t, of the heat diffusion equation, as well as ∂T/∂ z(z, T), canbe expressed by Eqs. (2) and (3):

T (z, t) =∑

k

2Ck exp(

− z − z1

)(2)

sin(

ωkt + &k − z − z1

)(2)

∂T

∂z(z, t) =

k

2√

2Ck

z˛exp

(− z − z1

)(3)

sin(

'4

+ ωkt + &k − z − z1

)(3)

Ck and &k are, respectively the norm and the phase of the kthcomplex Fourier coefficient. ωk = 2'k/Nıt is the period of the kthharmonic. N is the total number of data considered in the Fourieranalysis and ıt, their time step. z˛ is the diffusion equivalent depthand is equal to (2˛T/ωk)1/2 where ˛T is the temperature diffusioncoefficient of the ground. The ground heat flux is given by Eq. (4):

G = − (∂T

∂z|z=0 (4)

( = )scs˛T with known values for )scs. Soil moisture changes andthus affects )scs. A daily value for )scs is estimated from a linearcombination (Eq. (5)) of dry soil and water characteristics (Hillel,1998).

)scs = )dcd + )wcw* (5)

* is a daily average of volumetric soil humidity measured byan in situ soil moisture sensor (CS616, Campbell Scientific, Shep-shed, UK) at 10 cm, and indices w and d represent, respectively thewater and dry soil properties ()d = 1620 kg m− 3, )w = 1000 kg m− 3,cd = 1250 J kg− 1 K− 1 and cw = 4250 J kg− 1 K− 1).

A daily average ˛T is estimated by a fit of the Harmonic solu-tion T(z = z2, t) with a soil temperature measurement at z2 = 20 cm.We follow here the ‘amplitude method’ to estimate ˛T as it wasdescribed and noted to be one of the most robust methods byVerhoef et al. (1996). Considering an homogeneous soil layer overthe first tens of centimeters, ˛T is derived from amplitude extinc-tion between T(z1) and T(z2) using Eq. (6) in which +Tz1 and +Tz2are the daily amplitudes of temperature series at z1 and z2.

˛T = ωN+z2

2(ln(+Tz1/+Tz2)2 (6)

Page 8: Agricultural and Forest Meteorology€¦ · 4/10/2017  · pressure (Pa) 2m Vaisala WXT510 Station 10s 30min Air temperature (T) (Vaisala Oyj, Helsinki, Finland) 10s 30min Relative

A. Guyot et al. / Agricultural and Forest Meteorology 154– 155 (2012) 84– 98 91

Table 2Correlation coefficient, slope of the regression and RMS for the incoming short wave(SWin) and long wave (LWin) at Bira-Bira and Nalohou sites (20 km from each other).The period is November 15 to December 12, 2006 (half-hourly time step).

SWin LWin

Correlation coefficient 0.997 0.955Slope of the linear regression 0.98 1.00

RMS [W/m2] 24 6.5

Mean (Bira) [W/m2] 217 361Mean (Nalohou) [W/m2] 221 361

For dry conditions, values of ˛T are typically 0.6 × 10− 6 m2 s− 1 andrise to 3.0 × 10− 6 m2 s− 1 for wet conditions after rain. Thermal con-ductivity is ∼ 1 W m− 1 K− 1 for dry conditions and ∼ 3.5 W m− 1 K− 1

for wet soils.Three soil stations with temperature and humidity profiles

make possible three estimates of the ground heat flux. Analysingof the dispersion of G among the station permits an estimate of theuncertainty on this term of the energy balance.

3.3. Net radiation

As the net radiation for the three classes of land cover is notavailable for the period 2006–2008, this paragraph presents themethodology used to estimate the net radiation at the scale of thescintillometer footprint.

Table 1 presents the availability of the data for the three years forthe CNR1 installed on the bare soil/fallow/crops class (full periodavailability), and the CNR2 installed on the shrub savannah site(December 2008 to December 2009), and the woody savannah site(December 2008 to December 2009).

The strategy is: (i) to investigate the relevance of homogene-ity hypothesis for incoming radiation in both short wave (SW) andlong wave (LW), by comparing with incoming radiation at Bira-Bira (25 km away from Nalohou) where a second CNR2 sensor isinstalled (Table 2); (ii) to calculate the albedo and the out cominglong wave radiation at each site, using an uniform incoming radia-tion over the footprint area (as the CNR2 only measure LW and SWradiative budgets); (iii) to calculate corrections coefficients for thetwo savannah classes, in order to be able to estimate the net radi-ation on the two savannah classes for the years 2006–2008; (iv)finally to calculate the net radiation in the scintillometer footprint,by a linear combination of local net radiation over the three classes,using the vegetation partition given in the preceding paragraph; (v)to investigate the uncertainty on the net radiation at the footprintscale.

4. Results

The dataset presented in this study is of high interest for land-surface processes parameterization used in meteorological models,thus a focus is made on the uncertainties on each term of theenergy balance. Each term is presented separately and discussed,and finally, the residual latent heat flux is presented as itself and asan evaporative fraction.

4.1. Sensible heat flux

The uncertainty on the sensible heat flux <H> calculated fromthe scintillation index comes mostly from the effective height(z − d)eq as discussed previously. When aerodynamical parame-ters are constant, i.e. from day 1 to day 150. (z − d)eq = 19 m forthe actual beam, (z − d)eq = 18.5 m for a virtual beam shifted to theNorthwest and (z − d)eq = 19.6 m for a virtual beam shifted to theSoutheast. The residual uncertainty on (z − d)eq = 19 m is therefore

approximately 1 m. This introduces a relative uncertainty of 6% on<H>. This corresponds to more than half of the uncertainty on <H>(total uncertainty of 10%, calculated as the sum of the individualuncertainties, and as shown by Kleissl et al., 2008).

Sensible heat flux measurements retrieved from scintillome-try observations cover three months of the dry season, and twomonths of the wet season in 2006, five months of the dry seasonand the “onset” of the monsoon in 2007, and quite the entire year for2008 (except two months of the wet season). The seasonal, day-to-day, and inter-annual variability of daily (0–24 h) <H> estimated bythe scintillometer are examined for the whole 2006–2008 period.Fig. 9(b) shows the seasonal cycle of <H> at daily time step, themean and standard deviation for <H> are given in Table 3 for thedifferent periods.

It is possible to distinguish a strong seasonal contrast in <H>.<H> is maximum and relatively constant at 77 ± 26 W/m2 for dryconditions (Table 3). The day-to-day variability is low. During thisperiod, the surface is dry (average daily soil moisture at 10 cm equalto 5% [v/v]) and the soil is warm (average daily soil temperature at10 cm equal to 38 ◦C). Therefore, these conditions are favourablefor convection to develop.

The intermediate periods are best documented in 2007 and2008. During these years, <H> has a higher day-to-day variability,and undergoes a gradual decrease from dry-to-wet periods, and agradual increase from wet-to-dry periods, both with a high day-to-day variability. <H> is equal to 73 ± 20 W/m2, values close to oneobserved during the dry season.

During the wet period, <H> reaches minimal values, down to48 ± 17 W/m2 (Table 3). The day-to-day variability presents largevalues at the end of the season, when cloud covering and rainfallsbecome sparser, for example in September 2006 and 2007.

4.2. Ground heat flux

The seasonal variability of G, and its spatial variability are exam-ined for the year 2008, to finally present the spatial aggregation of<G> across the footprint of the scintillometer.

Daily rainfall over the Ara catchment, soil water content, mean,maximal and minimal values of ground heat flux G for the three sitesalong the hill slope (US, MS and LS) are presented in Fig. 10(a)–(d)for the year 2008.

During the dry period, the daily amplitudes increase and reachmaximum values (almost 400 W/m2 of amplitude). The daily aver-age is close to zero. During this period, the soil is dry and its thermalconductivity minimum. When first rain appears, the amplitudesremain high because surface dries rapidly and the surface temper-ature recovers high values when irradiated by the sun.

However, the day-to-day variability of G is large because eachrain event produces an energy release. The daily means can reach− 60 W/m2 after a rain, when the surface is cold and the deeperground remains warm. This means that the ground gradually trans-fers energy to the atmosphere through the surface water. From thewet season until the burn of the vegetation in December, the sur-face is less warm because of vegetation shadow, and higher soilmoisture. It results a decrease in amplitude (amplitudes equals to200 W/m2), with a daily mean close to zero, despite the increaseof heat conductivity. The intermediate wet-to-dry period presentsa very low variability, and daily averages close to zero. The ampli-tudes start to grow gradually until the end of the next dry season.

In general, whatever the period, GHS is smaller than GMS and GBS.Stronger G at the lower (GBS) and meddler (GMS) hill slope station,reflects a better thermal conductivity of the soil, which can storeheat more easily. Fig. 10(c) and (d) shows that ground heat fluxeshave low spatial variability at daily time step. However, maximaand minima show large station-to-station discrepancies caused byheterogeneities in ground properties.

Page 9: Agricultural and Forest Meteorology€¦ · 4/10/2017  · pressure (Pa) 2m Vaisala WXT510 Station 10s 30min Air temperature (T) (Vaisala Oyj, Helsinki, Finland) 10s 30min Relative

92 A. Guyot et al. / Agricultural and Forest Meteorology 154– 155 (2012) 84– 98

Table 3Mean and standard deviations of <Rn> and <H> regarding to VPD for 2006, 2007 and 2008. <RnD> (resp. <RnW>) corresponds to the estimated <Rn> associate for the dry period(resp. wet period). The daily averages are calculated for days whose number of data was greater than 20 over 24 h.

Wet conditions Intermediate conditions Dry conditionsVPD< 1 kPa 1 kPa < VPD< 3 kPa VPD> 3 kPa

Mean [W/m2] <RnD> 130 ± 40 144 ± 37 98 ± 21± Std dev <RnW> 117 ± 36 128 ± 34 87 ± 18No. of days verifying nd > 20/24 297 510 332

Mean [W/m2] <H> 48 ± 17 73 ± 20 77 ± 26± Std devNo. of days verifying nd > 20/24 97 152 62

The uncertainties on <G> could be important at hourly time step,because of the large heterogeneity between stations. However, ifwe consider the daily time step, the standard deviation of <G> isless than 3% on average for the year 2008 (Fig. 10(d)). It is in thesame order of magnitude as sensor accuracy and calibration, as wellas sensor positioning in depth.

Finally, the spatial average <G> calculated as an equal mean ofthe three stations is used to enter in the energy budget at the scaleof the scintillometer footprint, for the three years.

4.3. Radiative fluxes and albedo

Fig. 8(a)–(e) presents for the year 2009, the daily rainfalls, thehalf-hourly wind direction at Nalohou, the daily LWin, albedo andLWout over the bare soil/fallow, the shrub and woody savannah.

LWout series show a well correlated annual cycle over the 3 sites.During wet seasons, LWout series converge in a common series asherbaceous invade all places, even under the woody canopy. Dur-ing dry season, LWout is weaker over shrub and woody savannahs

because vegetation shades limit the increase of surface tempera-ture. Albedo series over shrub and woody savanna follow a morecomplex annual cycle. During the wet season albedo values tendto converge with the fallow values around 0.15. The other peri-ods are much more different. One can first distinguish that shruband woody savannah albedo series are quite well correlated. Theirchanges are controlled by the leaves cycle and meteorologicalevents (rain, wind, aerosols that change the partition betweendirect and scattered radiations). Shrub and tree leaves are renewedin a short period in December-January (1 month). This affects thealbedo series drastically as leaves colors change from green tobrown and then fall and cover the ground with a dark color. Thisproduces very weak albedo values. When these leaves are blown bya high wind event (flag 1 in Fig. 8), albedo retrieves bare soil values.Rain events affect also albedo (flags 2 and 3). In March leaves arecompletely renewed and herbaceous have not started yet. Duringthis period one can observe the albedo differences of different veg-etation with different LAI values. Mid November, herbaceous areburnt when wind direction change (flag 4). It is associated with a

Fig. 7. Daily evolution of four radiative components (a)–(d), and the albedo (e) at the Nalohou site. Dashed line in (a) is the potential solar radiation RSO . Line in (c) is the LWin

following Culf and Gash (1993). Lines in (b), (d) and (e) are 10 days moving averages. Shaded areas correspond to dry and wet periods as noticed at the top of the figure.

Page 10: Agricultural and Forest Meteorology€¦ · 4/10/2017  · pressure (Pa) 2m Vaisala WXT510 Station 10s 30min Air temperature (T) (Vaisala Oyj, Helsinki, Finland) 10s 30min Relative

A. Guyot et al. / Agricultural and Forest Meteorology 154– 155 (2012) 84– 98 93

Fig. 8. Daily evolution for the 2009 year of (a) rainfall; (b) wind direction at the Nalohou site; (c) incoming long wave at the Nalohou site; (d) outgoing long wave for thethree radiation sites (woody savannah, open circles; shrub savannah, grey cross and bare soil/fallow, black dots); (e) albedo for the three radiation sites (same legend as for(d)). Shaded areas correspond to dry and wet periods as noticed at the top of the figure. Events indicated by numbers from 1 to 5 are detailed in the text.

Fig. 9. Daily evolution of (a) the net radiation <Rn>; (b) the sensible heat flux <H>; (c) the residual latent heat flux <LE>; (d) the evaporative fraction <EF>. Black dots indicatefluxes calculated with <Rn > D , and white dots with <Rn > W .

Page 11: Agricultural and Forest Meteorology€¦ · 4/10/2017  · pressure (Pa) 2m Vaisala WXT510 Station 10s 30min Air temperature (T) (Vaisala Oyj, Helsinki, Finland) 10s 30min Relative

94 A. Guyot et al. / Agricultural and Forest Meteorology 154– 155 (2012) 84– 98

Fig. 10. Daily evolution for the year 2008 of (a) rainfall at Nalo 1 and 3; (b) soil water content at the three hill slope sites (US, clear line; MS, dashed line and LS, dark line);(c) daily maxima and minima of G at the three hill slope sites (same legend as for (b)); (d) daily mean of G (same legend as for (b)).

sharp decrease in LWin because humidity and clouds are pushedaway by Northern entries. Flag 5 indicates a new fall.

4.3.1. Seasonal evolution of the aggregated net radiation <Rn>The aggregated net radiation <Rn> at the scale of the scintillome-

ter footprint is calculated as a linear combination of net radiationat Nalohou site, and net radiation for shrub and woody savan-nah. As only 2009 data are available, shrub and woody savannahseries are rebuilt using correlation coefficients between <RnWS>and <RnDS> site and M site (Table 3). We choose to create two setsof data because during intermediate periods, albedo and surfaceconditions are driven by random meteorological events. They cor-responds to (i) dry conditions from December to January, and (ii)wet conditions from June to August. Then, the differences betweenboth can be viewed as the incertitude on both net radiation series.As a result, the uncertainty on <Rn> is thus more important for inter-mediate periods, when <RnWS> and <RnDS> estimates give upperand lower bounds. The relative uncertainty on <Rn> could reach13% in the less favorable cases (i.e. for the intermediate periods), ifwe take into account both sensor accuracy (3%) and the aggregationscheme.

Finally, the average net radiation is calculated using vegetationdistribution within the footprint area. Fig. 9(a) shows the seasonalcycle of the aggregated net radiation at the footprint scale <Rn>.Table 3 gives the mean and the standard deviation of <Rn> for thedry, the intermediate and the wet periods. <Rn> displays distinct

seasonal patterns. During the dry period, <Rn> encounter weak day-to-day variability, with the lowest mean value of 98 ± 21 W/m2

(Table 3). This period corresponds to clear sky days. On the otherhand, the wet period is subject to strong day-to-day variability, dueto rapid changes in cloud covering (with a mean of 117 ± 36 W/m2,Table 3). Intermediate periods correspond to regular increase ordecrease of <Rn>. The intermediate wet to dry period does notencounter day-to-day variability, whereas the dry-to-wet periodis subject to rapid changes due to South-West entries which comewith clouds. These period have the highest average net radia-tion values. The inter-annual variability of <Rn> is slightly marked.Indeed, 2006 faced to a sudden increase of <Rn> after the rain earlyFebruary, while 2007 and 2008 present similar patterns. This is notsurprising, as the main factors affecting the variability of <Rn> arethe incoming radiation on one hand (controlled by the variabilityof cloud cover and temperature and humidity of the atmosphere atlow levels), and the albedo on the other hand. Finally, differencesbetween <RnDS> and <RnWS> are small if compared with the annualvariability.

4.4. Residual latent heat flux and evaporative fraction

The seasonal, day-to-day, and inter-annual variability of latentheat flux and the evaporative fraction are given in Fig. 9(c) and(d), respectively. Fig. 11 gives the quartile distribution of the

Page 12: Agricultural and Forest Meteorology€¦ · 4/10/2017  · pressure (Pa) 2m Vaisala WXT510 Station 10s 30min Air temperature (T) (Vaisala Oyj, Helsinki, Finland) 10s 30min Relative

A. Guyot et al. / Agricultural and Forest Meteorology 154– 155 (2012) 84– 98 95

Fig. 11. Quartiles of the evaporative fraction <EF> for 2006, 2007 and 2008. Dryperiod corresponds to VPD >3 kPa, intermediate period to 1 kPa < VPD < 3 kPa andwet period to VPD < 1 kPa. <Rn > D has been used for calculating <EF> for dry period,<Rn > W for the intermediate and the wet periods.

evaporative fraction for the periods based on VPD criteria (dry, wetand intermediate periods) for each year.

The uncertainty on <LE> can be estimated as the sum of theuncertainties of each term of the energy balance. ConsideringTable 3, we can estimate the relative uncertainties on <LE> forthe three periods: for the wet period, +< LE >/< LE > = 30%, for theintermediate period +< LE >/< LE > = 40% and for the dry period+< LE >/< LE > = 100%. The dry period relative uncertainty is highbecause the mean latent heat flux is small (<LE > = 21 W/m2 and+ < LE > = 21 W/m2). These estimates give an idea of the uncertaintyon the residual latent heat flux, and have to be considered as upperbounds.

As expected in Fig. 9(c), <LE> is minimal during the dry periods,and maximum during wet periods. The large variability of <Rn> dur-ing the wet periods leads to a large day-to-day variability of <LE>.Weak values during the dry periods correspond to ENE winds, whenfootprints include more bare soil conditions. <LE> is slightly largerfor SW conditions, after the change of wind regime. This is particu-larly clear in 2008 at the beginning of March. During intermediateand wet seasons, different behaviors can be observed. Light rainthat evaporates during the following days leads to a small increaseof <LE>, while heavier rain events are followed by several days ofsurface drying which produces detached peaks in the <LE> series.This is particularly clear in 2006, but also for some early rain eventsin 2007 and 2008.

The <EF>, normalizing the amount of energy used for evapora-tion by the available energy is well less noisy than <LE>, especiallyduring the wet season. It reaches an average value around 0.7 and isalmost constant during the wet season. However, the inter-annualvariability is non negligible (see Fig. 11). It depends on the quan-tity of rain that have fallen during the considered season. In 2007,the <EF> increases regularly during intermediate and wet periodsbecause of quite regular rain (Fig. 4) and reaches values close to 0.7before the period defined as wet by the VPD criterion. Inter-annualvariability is larger in dry season and is largely affected by isolatedrainy events and changes in wind directions. Then, dry season 2006behaves more like an intermediate season. The most stable dry sea-son is 2007 because most of the available data are sampled for SWconditions.

In dry season, measured <EF> has a more larger variability(Fig. 11) because of varying wind conditions from NE to SW whichcauses large changes in footprint areas. Vegetation distributionseen by the scintillometer contains more woody land cover whenwinds come from SW. It makes <EF> drop from 0.1 to 0.25. Itis clear in 2008 when winds change at the beginning of March(Figs. 4 and 9). It is also certainly the case in 2006, but the signal is

masked by a heavy rain in February which causes the major part ofthe <EF> variability (Guyot et al., 2009). The rain event during thedry season of 2006, quickly leads to a large increase of <EF> (Guyotet al., 2009).

5. Discussion

5.1. Interactions between components

In this section, we look at the energy budget partitioning duringan annual cycle, as defined by the VPD criteria. The daily compositefluxes of the surface energy balance are plotted in Fig. 12 for thethree periods (dry, intermediate and wet).

In order to characterize the co-evolution of <H> and <Rn> for theconditions identified by the VPD, statistical criteria are calculatedon a circadian cycle (statistical criteria applied to 24 values for oneday). Table 4 gives the average statistical parameters for dry, wetand intermediate periods, and their standard deviation.

From beginning of December to the pre-monsoon period (end ofMarch), (i.e. the dry period identified by VPD >3 kPa), <H>, <G> andLWout (Figs. 9(b) and 7(d)) present the same positive trend. Theirvalues reach a plateau at the end of March, while <LE> (Fig. 9(c))remains small. The regressions at the origin, show that <H> is equalto 0.69× < RnD > for dry conditions, with a correlation of 0.86 ± 26(Table 4). Net radiation drives the sensible heat flux during thisperiod. Bare soil, covering a large part of the catchment, is heatedand return the energy mostly in sensible heat flux, ground heatflux and LWout. In Fig. 12(a), <H > hourly reaches 300 W/m2 just aftermidday. <G > hourly is quite important with values of 200 W/m2 forthe maxima. The bell shape of <Rn > hourly shows the predominanceof sunny days without clouds. However, the energy brought by<Rn > hourly is the smallest of the year (Fig. 9(a)), because of atmo-spheric aerosols and large LWout.

The first intermediate period corresponds to a moistening phase,with a strong and rapid decrease of the VPD, i.e. a conjunction of airtemperature fall and air moistening. This period encounters spo-radic rainfalls, transforming progressively the surface conditions.These important variations of atmospheric and surface conditionslead to a strong competition of the different components in theenergy budget. When water is available, the energy is used forevaporation. When water is missing, and vegetation sparse, <H>is stronger, due to the heating of the surface. As rain occurs pref-erentially at night, <Rn > hourly keep a bell shape but sharper thanin the dry season. Evapotranspiration is larger in the morning dueto light rain events that produce low volume, which is evaporatedduring the following day. Out of phase series are not excluded, butthe early <LE> hourly signal seems high enough to be explained byphysical processes. The second intermediate period corresponds toa drying phase, with a strong and rapid increase of the VPD.

From the Onset of the monsoon (i.e. the end of June), to the end ofthe monsoon (i.e. the end of October), the VPD is roughly constant,under 1 kPa. Surface conditions encounter vegetation growth, andhigher frequency of rainfalls. The water is in excess at the surfaceduring this period, and the growing herbaceous vegetation occupiesall the available surfaces previously considered as bare soil condi-tions during the dry season. The energy input by the net radiation isthe largest (more than 600 W/m2 for the maximum). <LE > hourly isthe first beneficiary of this energy, with a maximum at 300 W/m2.Half of the energy is used for the evaporation, whereas sensibleand ground heat flux are reduced to small but non-negligible pro-portion of the available energy (daily maxima ∼ 150 W/m2). Thesensible heat flux is no more coupled with <Rn>, with a correlationcoefficient equal to 0.35× < RnW >. The squared errors are on aver-age two times larger for wet than for dry conditions, characterizingthe low proportion of <H> in the energy balance for wet conditions.

Page 13: Agricultural and Forest Meteorology€¦ · 4/10/2017  · pressure (Pa) 2m Vaisala WXT510 Station 10s 30min Air temperature (T) (Vaisala Oyj, Helsinki, Finland) 10s 30min Relative

96 A. Guyot et al. / Agricultural and Forest Meteorology 154– 155 (2012) 84– 98

Table 4Correlation coefficient, slope of the linear regression and RMS of <Rn> versus <H> for VPD intervals, calculated for 2006, 2007 and 2008. Values of <H> corresponding to themost suitable corrections for <Rn> are in bold.

Wet conditions Intermediate conditions Dry conditionsVPD < 1 kPa 1 kPa < VPD < 3 kPa VPD > 3 kPa

Correlation coefficient <H> vs. <RnD> 0.78 ± 0.22 0.89 ± 0.12 0.86 ± 0.26<H> vs. <RnW> 0.78 ± 0.22 0.89 ± 0.12 0.86 ± 0.26

Slope of the linear regression <H> vs. <RnD> 0.32 ± 0.16 0.59 ± 0.21 0.69 ± 0.21<H> vs. <RnW> 0.35 ± 0.18 0.64 ± 0.23 0.76 ± 0.24

RMS <H> vs. <RnD> 224 ± 59 166 ± 44 129 ± 44<H> vs. <RnW> 202 ± 53 145 ± 42 111 ± 38

<LE> clearly controls the energy balance, with a constant EF duringthis period (Fig. 11).

5.2. Inter-annual variability of the evaporative fraction

The years 2006, 2007 and 2008 differ significantly by: the begin-ning of the first rainfall, the date of the Onset of the Monsoon, theintensity of rainfalls, and the occurrences of rainfall along the year.Therefore, surface conditions, i.e. soil moisture and vegetation differas well.

The annual cycle described above is clearly visible on the evap-orative fraction in Fig. 11. For the dry periods, <EF> is low but notzero, which confirms vegetation activity even during dry periods.The inter-annual variability is important (0.16–0.31). <EF> reaches0.31 on average in 2006, with a strong dispersion because of astrong isolated rainy event. The dispersion during the dry seasonis related to: (i) the wind variability. In 2007, SW entries are not asnumerous as in 2006 and 2008. The footprint area is then more con-stant and it results in a weaker dispersion. In 2008, there is a wide

dispersion of the <EF>. This year encounters a sudden change ofwind, from ENE to SW, late February, which induces a change offootprint. The increase of the <EF> in late February is therefore prob-ably the fact that the sampled surface (SW footprint) presents morewoody strata that the ENE footprint (Fig. 6(b)); (ii) the uncertaintieson <LE> and therefore on <EF> during this period, that increase thedispersion.

The intermediate periods show a constant year-to-year valuefor <EF> around 0.4, but a large dispersion. This dispersion is dueto: (i) the great variability of atmospheric and surface conditions;(ii) the variability of sampling surfaces with winds alternatingfrom Northeast to Southwest. One can notice that during thisperiod, uncertainties on <LE> can also contribute to the disper-sion.

The year 2006 appears different from the two other years: the<EF> dispersion is approximately the same during the dry and inter-mediate periods. The results show that it is quite difficult to pointout the dry period for this year. It has to be noted that very few dataare available for this 2006 intermediate period.

Fig. 12. Hourly composite fluxes (<G > hourly , <H > hourly , <Rn > hourly and the residual <LE > hourly) for the periods based on the VPD criterion. (a) Dry period (VPD > 3 kPa); (b)intermediate period (1 kPa < VPD < 3 kPa) (for this period <Rn > D and <Rn > W as well as the residual <LE > hourly are indicated); (c) wet period (VPD < 1 kPa).

Page 14: Agricultural and Forest Meteorology€¦ · 4/10/2017  · pressure (Pa) 2m Vaisala WXT510 Station 10s 30min Air temperature (T) (Vaisala Oyj, Helsinki, Finland) 10s 30min Relative

A. Guyot et al. / Agricultural and Forest Meteorology 154– 155 (2012) 84– 98 97

During wet period, <EF> remains relatively constant (larger in2007 (0.7) and 2008 (0.68), and slightly lower in 2006 (0.64)). Thedispersions are very low for the three years during this period, com-pared to what can be observed in dry and intermediate periods. Theevaporative fraction in 2006 is a little bit weaker compared with2007 and 2008. It can be related with the low cumulative rainfallfor this year on the region (861 mm at Nalohou) causing a lowerefficiency in the evaporative processes.

These observations of turbulent and radiation fluxes are inagreement with the values observed by other authors at almostequivalent latitudes, but on quite different land covers (Touré,2004; Schuettemeyer et al., 2006; Bagayoko et al., 2007; Bruemmeret al., 2008). The values of sensible heat flux and latent are simi-lar to Schuettemeyer et al. (2006) and Bagayoko et al. (2007) forthe periods studied. In particular, this study confirms an observa-tion of significant latent heat flux in the dry season. As shown bySchuettemeyer et al. (2006), it reaches about 100 W/m2 for dailymaximum. This is not observed by Ramier et al. (2009) or Timouket al. (2009) in the Sahel. This significant latent heat flux in the dryperiod is the signature of persistent vegetation in Sudano-Sahelianareas.

6. Conclusion

This paper presents an improved methodology for long-termobservations using scintillometer in an energy budget. Time andspatial variations of land cover are taken into account in the estima-tion of the aggregated resultant latent heat flux. This methodologyprovides an estimation of turbulent (sensible and latent heat) fluxesover complex terrain (complex both in terms of the topography andin terms of the spatially and temporally heterogeneous vegetationcover). This study shows the difficulties encountered by the need totake into account the variability of vegetation over long term. Weestimate an uncertainty on <H> around 10%, while the uncertaintyon <LE> depends on the period of the year (30% in the wet period,40% in the intermediate period and 100% in the dry period).

The climate of the Sudano-Sahelian region is characterized by astrong seasonal cycle and inter-annual variability related to chang-ing atmospheric and land surface conditions. <H> and <Rn> are wellcorrelated during dry periods, that become much less during wetperiods. The composite daily cycles show a predominance of <LE>in wet periods, and <H> in dry periods. This analysis points out theconsiderable role of evaporation in the energy budget partition-ing in the wet period. <LE> is not zero during the dry seasons inthis region, but can encounter large variability observed by turn-ing winds and opposite footprint areas. The evaporative fractionis found to be relatively constant during the wet period for thethree years (around 0.7). The 2006–2008 period shows a stronginter-annual variability. It is particularly true for dry and dry-to-wet intermediate periods, when atmospheric conditions alternatebetween the Northeastern regimes, to the monsoon Southwesternregime.

These hourly time resolution data represent unique and veryvaluable information for such a long period in the Sudano-Sahelianregion, and for such a composite surface. They provide estimateof latent heat flux at a relevant scale for hydrological issues, foridentifying processes (Descloitres et al., 2011; Séguis et al., 2011),or for future hydrological modelling at catchment scale. They alsoprovide estimates of fluxes at relevant scales for atmospheric mod-elling, by testing land surface processes parameterization used inmeteorological models (ALMIP, Boone et al., 2009).

Acknowledgements

The authors wish to thank Moussa Doukouré for his help forthe fieldwork, and Simon Allonganvinon for his close survey of the

scintillometer and data collection. The authors also wish to thanksIsabella Zin for providing land cover maps, and Joris Pianezze forhis collaboration as part as an internship at LTHE. The authors areparticularly grateful to two anonymous reviewers for their criti-cal examination of the manuscript and helpful suggestions. Thiswork has been performed within the AMMA project, and the ‘Ser-vice d’Observation’ AMMA-CATCH and financed by the INSU-CNRSand IRD agencies. This study was made possible by the financialsupport of BLS 900 scintillometer by ‘Observatoire des Sciences del’Univers de Grenoble’. This work has been founded by a BDI-CNRSgrant.

References

Aubinet, M., Grelle, A., Ibrom, A., Rannik, U., Moncrieff, J., Foken, T., Kowalski, A.S.,Martin, P.H., Berbigier, P., Bernhofer, C., Clement, R., Elbers, J., Granier, A., Green-wald, T., Morgenstern, K., Pilegaard, K., Rebmann, C., Snijders, W., Valentini, R.,Vesala, T., Raffaelli, R., 1999. Estimates of the Annual Net Carbon and WaterExchange of Forests: The EUROFLUX methodology. Adv. Ecol. Res. 30, 113–175.

Bagayoko, F., Yonkeu, S., Elbers, J., van de Giesen, N., 2007. Energy partitionningover the West African savanna: multi-year evaporation and surface conductancemeasurements. J. Hydrol. 334, 549–559.

Bastiaanssen, W., 1999. SEBAL-based sensible and latent heat fluxes in the irrigatedGediz Basin, Turkey. J. Hydrol. 229, 87–100.

Boko, M., Nyong, C., Vogel, A., Githeko, A., Medany, M., Osman-Elasha, B., Tabo, R.,Yanda, P., 2007. In: Parry M.L., Canziani, O.F., Palutikof, J.P., van den Linde, P.J.,Hanson, C.E. (Eds.), Africa: Climate Change 2007: Impacts, Adaptation and Vul-nerability. Contribution of Working Group II to the Fourth Assesment Report ofthe Intergovernmental Panel on Climate Change.

Boone, A., de Rosnay, P., Balsamo, G., Beljaars, A., Chopin, F., Decharme, B., Delire, C.,Ducharne, A., Gascoin, S., Grippa, M., Guichard, F., Gusev, Y., Harris, P., Jarlan, L.,Kergoat, L., Mougin, E., Nasonova, O., Norgaard, A., Orgeval, T., Ottlé, C., Poccard-Leclercq, I., Polcher, J., Sandholt, I., Saux-Picart, S., Taylor, C., Xue, Y., 2009. TheAMMA Land Surface Model Intercomparison Project (ALMIP). Bull. Am. Meteorol.Soc. 90, 1865–1880.

Bruemmer, C., Falk, U., Papen, H., Szarzynski, J., Wassmann, R., Brueeggemann, N.,2008. Diurnal, seasonal, and interannual variation in carbon dioxide and energyexchange in shrub savannah in Burkina Faso (West Africa). J. Geophys. Res. F:Earth Surf. 113, G02030.

Charney, J., Stone, P., Quirk, W., 1975. Drought in the Sahara: a biogeophysical feed-back mechanism. Science 187, 434–435.

Chehbouni, A., Santiago, F., Dedieu, G., Goodrich, D.C., Unkrich, C., 2000. Estima-tion of heat and momentum fluxes over complex terrain using a large aperturescintillometer. Agricul. Forest Meteorol. 105, 215–226.

Culf, A.D., Gash, J.H.C., 1993. Longwave radiation from clear skies in Niger: a com-parison of observations with simple formulas. J. Appl. Meteorol. 32, 539–547.

Descloitres, M., Seguis, L., Legchenko, A., Wubda, M., Guyot, A., Cohard, J., 2011.The contribution of MRS and resistivity methods to the interpretation of actualevapo-transpiration measurements: a case study in metamorphic context innorth Bnin. Near Surf. Geophys. 9 (2), 187–200.

Eltahir, E.A.B., Gong, C., 1996. Dynamics of wet and dry years in West Africa. J. Clim.9, 1030–1042.

Ezzahar, J., Chehbouni, A., Hoedjes, J., Ramier, D., Boulain, N., Boubkraoui, S., Cap-pelaere, B., Descroix, L., Mougenot, B., Timouk, F., 2009. Combining scintillometerand aggregation schemes to estimate area-averaged latent heat flux duringAMMA experiment. J. Hydrol. 375 (1–2), 217–226.

Gash, J., Kabat, P., Monteny, B., Amadou, M., Bessemoulin, P., Billing, H., Blyth, E.,deBruin, H., Elbers, J., Friborg, T., Harrison, G., Holwill, C., Lloyd, C., Lhomme, J.P.,Moncrieff, J., Puech, D., Soegaard, H., Taupin, J., Tuzet, A., Verhoef, A., 1997. Thevariability of evaporation during the HAPEX-Sahel Intensive Observation Period.J. Hydrol. 188–189, 385–399.

Groete, R., Lehmann, E., Bruemmer, C., Brueeggemann, N., Szarzynski, J., Kunstmann,H., 2009. Modelling and observation of biosphere-atmosphere interactions innatural savannah in Burkina Faso, West Africa. Phys. Chem. Earth 34, 251–260.

Guyot, A., 2010. Estimation de l’évapotranspiration sur un couvert complexe parutilisation de la scintillométrie infrarouge (Application à un bassin versant dezone soudano-sahélienne). Ph.D. Thesis. Université de Grenoble.

Guyot, A., Cohard, J.M., Anquetin, S., Galle, S., Lloyd, C.R., 2009. Combined analysisof energy and water balances to estimate latent heat flux of a Sudanian smallcatchment. J. Hydrol. 375 (1–2), 227–240.

Hanna, R., Onzo, A., Lingeman, R., Yaninek, J.S., Sabelis, M.W., 2005. Seasonal cyclesand persistence in an acarine predator–prey system on cassava in africa. Popul.Ecol. 47, 107–117.

Hartogensis, O., 2006. Exploring Scintillometry in the stable Atmospheric SurfaceLayer. Ph.D. Thesis. Wageningen University.

Hartogensis, O.K., Watts, C.J., Rodriguez, J., De Bruin, H., 2003. Derivation of an Effec-tive Height for Scintillometers: La Poza Experiment in Northwest Mexico. J.Hydrometeorol. 4, 915–928.

Hillel, D., 1998. Environmental Soil Physics. Academic Press.Hoedjes, J., Chebouni, A., Ezzahar, J., Escadafal, R., de Bruin, H., 2007. Comparison of

large aperture scintillometer and eddy covariance measurements: can thermal

Page 15: Agricultural and Forest Meteorology€¦ · 4/10/2017  · pressure (Pa) 2m Vaisala WXT510 Station 10s 30min Air temperature (T) (Vaisala Oyj, Helsinki, Finland) 10s 30min Relative

98 A. Guyot et al. / Agricultural and Forest Meteorology 154– 155 (2012) 84– 98

infrared data be used to capture footprint induced differences. J. Hydrometeorol.8, 194–206.

Horton, R., Wierenga, P.J., 1983. Estimating the soil heat flux from observations ofsoil temperature near the surface. Soil Sci. Soc. Am. J. 47, 14–20.

Hsieh, C.I., Katul, G., Tze-wen, C., 2000. An approximate analytical model for footprintestimation of scalar fluxes in thermally stratified atmospheric flows. Adv. WaterResour. 23, 765–772.

Kabat, P., Dolman, A., Elbers, J., 1997. Evaporation, sensible heat and canopy con-ductance of fallow savannah and patterned woodland in the sahel. J. Hydrol.188–189, 494–515.

Kleissl, J., Gomez, J., Hong, S., Hendrickx, J., Rahn, T., Defoor, W., 2008. Large aperturescintillometers intercomparison Study. Bound. Layer Meteorol. 128, 128–133.

Le Lay, M., Galle, S., 2005. How changing rainfall regimes may affect the water bal-ance: a modelling approach in West Africa. Int. Assoc. Hydrol. Sci. Publ. 296,203–210.

Lloyd, C.R., Bessemoulin, P., Cropley, F.D., Culf, A.D., Elbers, J., Heusinkveld, B., Mon-crieff, J.B., Monteny, B., Verhoef, A., 1997. A comparison of surface fluxes at theHAPEX-Sahel fallow bush sites. J. Hydrol. 188–189, 400–425.

Lohou, F., Said, F., Lothon, M., Durand, P., Sera, D., 2010. Impact of the boundary-layerprocesses on surface turbulence characteristics in the frame of the West AfricanMonsoon. Bound. Layer Meteorol. 136, 1–23.

Lothon, M., Said, F., Lohou, F., Campistron, B., 2008. Observation of the diurnal cyclein the low troposphere of West Africa. Mon. Weather Rev. 136, 3477–3500.

Martano, P., 2000. Estimation of the surface roughness length and displacementheight from single-level anemometer data. J. Appl. Meteorol. 39 (5), 708–715.

Meijninger, W.M.L., 2003. Surface fluxes over natural landscapes using scintillome-try. Ph.D. Thesis. Wageningen University, The Netherlands.

Meijninger, W.M.L., Green, A.E., Hartogensis, O.K., Kohsiek, W., Hoedjes, J.C.B., Zuur-bier, R.M., De Bruin, H.A.R., 2002b. Determination of area-averaged water vapourfluxes with large aperture and radio wave scintillometers over a heterogeneoussurface—Flevoland Field Experiment. Bound. Layer Meteorol. 105, 63–83.

Oren, R., Sperry, J., Katul, G., Pataki, D., Ewers, B., Phillips, N., Shaeffer, K., 1999.Survey and synthesis of intra- and interspecific variation in stomatal sensitivityto vapour pressure deficit. Plant Cell Environ. 22, 1515–1526.

Ramier, D., Boulain, N., Cappelaere, B., Timouk, F., Rabanit, M., Lloyd, C., Boukraoui,S., Métayer, F., Descroix, L., Wawrzyniak, V., 2009. Towards an understanding ofcoupled physical and biological processes in the cultivated Sahel. 1. Energy andwater. J. Hydrol. 375 (1–2), 204–216.

Raupach, 1992. Drag and drag partition on rough surfaces. Bound. Layer Meteorol.60, 375–395.

Schmid, H.P., Bunzli, B., 1995. The influence of surface texture on the effective rough-ness length. Quart. J. Roy. Meteorol. Soc. 121, 1–21.

Schuettemeyer, D., Moene, A.F., Holtslag, A.A.M., De Bruin, H.A.R., Van De Giesen,N., 2006. Surface fluxes and characteristics of drying semi-arid Terrain in WestAfrica. Bound. Layer Meteorol. 118, 583–612.

Séguis, L., Boulain, N., Cappelaere, B., Cohard, J., Favreau, G., Galle, S., Guyot, A., Hier-naux, P., Mougin, E., Peugeot, C., Ramier, D., Seghieri, J., Timouk, F., Demarez, V.,Demarty, J., Descroix, L., Descloitres, M., Grippa, M., Guichard, F., Kamagaté, B.,Kergoat, L., Lebel, T., Le Dantec, V., Le Lay, M., Massuel, S., Trichon, V., 2011. Con-trasted land-surface processes along the West African rainfall gradient. Atmos.Sci. Lett. 12, 31–37.

Stanhill, G., 1969. A simple instrument for field measurement of turbulent diffusionflux. J. Appl. Meteorol. 8, 509–513.

Sultan, B., Janicot, S., 2003. The West African monsoon dynamics. Part II. The “Pre-onset” and the “Onset” of the summer monsoon. J. Clim. 16, 3407–3425.

Sultan, B., Janicot, S., 2004. La variabilité climatique en Afrique de l’Ouest aux échellessaisonnière et intra-saisonnière. I: mise en place de la mousson et variabilitéintra-saisonnière de la convection. Sécheresse 15, 321–330.

Timouk, F., Kergoat, L., Mougin, E., Lloyd, C., Ceschia, E., Cohard, J.M., de Rosnay, P.,Hiernaux, P., Demarez, V., Taylor, C., 2009. Response of surface energy balanceto water regime and vegetation development in Sahelian landscape. J. Hydrol.375, 178–189.

Touré, S., 2004. Etude des densités de flux d’énergie de la parcelleagrométéorologique à l’échelle du bassin versant. Ph.D. Thesis. Universitéde Liège.

Verhoef, A., 1995. Surface energy balance of shrub vegetation in the Sahel. Ph.D.Thesis. Wageningen University.

Verhoef, A., van den Hurk, B.J.J.M., Jacobs, A.F.G., Heusinkveld, B.G., 1996. Thermalsoil properties for vineyard (EFEDA-I) and savannah (HAPEX-Sahel) sites. Agric.For. Meteorol. 78, 1–18.

Verhoef, A., Allen, S.J., De Bruin, Henk, A.R., Jacobs, C.M.J., Heusinkveld, B.G., 1996.Fluxes of carbon dioxide and water vapour from a Sahelian savanna. Agric. For.Meteorol. 80.

Wallace, J., Wright, I., Stewart, J., Holwill, C., 1991. The Sahelian Energy BalanceExperiment (SEBEX): ground based measurements and their potential extrapo-lation using satellite data. Adv. Space Res. 11, 131–141.

Wang, G., Eltahir, A.B.E., 2000. Ecosystems dynamics and the sahel drought. Geophys.Res. Lett. 27, 795–798.

Zheng, X., Eltahir, A., 1997. The response to deforestation and desertification in amodel of West African Monsoons. Geophys. Res. Lett. 24, 155–158.