Dry deposition of nitrogen compounds ( NO 2 , HNO 3 , NH 3 ) , sulfur dioxide and ozone in West and Central African ecosystems using the inferential method

Introduction Conclusions References


Introduction
Deposition of chemical species onto the Earth's surface plays an essential role in controlling the concentration of gases and aerosols in the troposphere.The study of deposition thus allows for tracing the temporal and spatial evolution of atmospheric chemistry and is a pertinent indicator for evaluating natural and anthropogenic influences.The deposition of atmospheric nitrogen (N) species constitutes a major nutrient input to the biosphere.On a long-term scale, the increase of N inputs into terrestrial or aquatic ecosystems leads to important environmental consequences such as a loss of biodiversity, eutrophication, acidification of seminatural ecosystems, leaching of nitrate into groundwater and increased carbon storage (Vitousek et al., 1997;Rodhe et al., 2002;Bobbink et al., 1998;Bouwman et al., 2002b;Liu et al., 2013).N deposition on the terrestrial surface thus impacts both atmospheric chemistry and ecosystem dynamics.Ammonia (NH 3 ), nitrogen dioxide (NO 2 ) and nitric acid (HNO 3 ) are the most important contributors to N dry deposition (Trebs et al., 2006).
Sulfur dioxide (SO 2 ) is one of the important species considering the acid deposition issues, and it is also the precursor of sulfate.Dry deposition estimation of SO 2 is essential to assess ecological impact research, crop growing and air quality research (Tsai et al., 2010).Tropospheric ozone (O 3 ) is known to harm human health, damage vegetation and lead to deterioration of materials.The dry deposition of O 3 is one of the most important sinks in the boundary layer ozone budget (Rummel et al., 2007).
Monitoring networks have been established around the world to measure wet and dry deposition.The international program DEBITS (Deposition of Biogeochemically Important Trace Species) was initiated in 1990 as part of IGAC/IGBP (International Global Atmospheric Chemistry/International Geosphere-Biosphere Programme) "core project" in order to study wet and dry atmospheric deposition in tropical regions (Lacaux et al., 2003).The DEBITS network collects data from 25 stations that are distributed within the tropical belt in Africa, Asia and South America, and results are presented in the new IGAC structure or DEBITS II (Pienaar et al., 2005;Bates et al., 2006; http://debits.sedoo.fr).For tropical Africa, the IDAF (IGAC/DEBITS/AFRICA) project started in 1994 and was implemented in partnership with INSU (Institut National des Sciences de l'Univers, in France) and the CNRS (Centre National de la Recherche Scientifique, in France) as part of the Environmental Research Observatory (ORE, in France) networks.
The main objectives of IDAF are to measure wet and dry deposition fluxes and to identify the relative contribution of natural and anthropogenic sources, as well as the factors regulating these fluxes.IDAF activity is based on high-quality measurements of atmospheric chemical data (gaseous, precipitation and aerosol chemical compositions) on the basis of multiyear monitoring (http://idaf.sedoo.fr).Within the framework of IDAF, several studies of precipitation chemical composition representative of great African ecosystems have been recently published (Galy- Lacaux and Modi, 1998;Galy-Lacaux et al., 2001, 2009;Al-Ourabi and Lacaux, 2002;Lacaux et al., 1993Lacaux et al., , 2003;;Sigha et al., 2003;Yoboue et al., 2005;Mphepya et al., 2004;2006;Laouali et al., 2012).
To complement these studies, it is appropriate to study and quantify dry deposition fluxes.Direct methods (eddy correlation, chamber method) and indirect methods (inferential method, gradient method) are available to determine dry deposition fluxes (Seinfeld and Pandis, 2006).The DEB-ITS committee in charge of deposition studies in IGAC has decided to use indirect dry deposition fluxes determination in tropical sites because of difficulties in operating sophisticated direct methods of flux measurements in remote sites (Wolff et al., 2010;Sutton et al., 2007).In this study, dry deposition fluxes are estimated using the inferential method, which is a combination of gaseous concentration measurements and modeling of deposition velocities according to the resistance analogy (Wesely, 1989;Zhang et al., 2003b;and references therein).Bidirectional exchange of NH 3 and NO 2 have been frequently observed over different canopies (Dorsey et al., 2004;Trebs et al., 2006;Walker et al., 2006;Wichink Kruit et al., 2007).There has been some effort in the development of bidirectional exchange models (e.g., Sutton et al., 1998;Flechard et al., 1999;Trebs et al., 2006;Massad et al., 2010;Zhang et al., 2010, Wichink Kruit et al., 2012;Bash et al., 2013;Hamaoui-Laguel et al., 2012); however, the application of these models remains difficult over the different canopies of African ecosystems owing to the lack of necessary input parameters.Nevertheless, the two-layer bidirectional model of Zhang et al. (2010) is applied in this study to estimate NH 3 surface-atmosphere exchange fluxes.
In the present paper, realistic dry deposition velocities according to the sites and the species involved are calculated in order to estimate dry deposition fluxes.The big-leaf model of Zhang et al. (2003b) is used to simulate dry deposition velocities representative of major African ecosystems.The results are compared to previous estimates from the literature.For NH 3 net fluxes, the bidirectional model of Zhang et al. (2010) is used.Then, we present an estimate on a long-term basis (10 yr) of dry deposition fluxes of gases (NO 2 , HNO 3 , NH 3 , O 3 and SO 2 ) at the scale of major African ecosystems.The monthly, seasonal and annual mean variations of gaseous dry deposition fluxes are analyzed.

Presentation of measurement sites
Figure 1 presents the location of the seven IDAF measurement stations displayed on the map of African biomes adapted from the land cover product of Mayaux et al. (2004).The IDAF sites of west and central Africa are located to represent a transect of ecosystems, i.e., dry savannas (Agoufou, Banizoumbou, Katibougou), wet savannas (Djougou, Lamto) and equatorial forests (Zoetele, Bomassa).The geographical, ecological and climatic characteristics of the study sites are presented in Table 1.Dry savannas are characterized by a long dry season from October to May and a short wet season from June to September.The mean wet season extends from April to October in wet savannas and from March to November in forests; other months are the dry season.A detailed description of IDAF monitoring stations can be found in Adon et al. (2010).

Dry deposition estimate
The inferential method, which combines measured air concentrations and modeled exchange rates, was employed in this study to estimate the dry deposition fluxes of different gaseous species.The inferential technique has been widely used in other studies for different types of ecosystems (Shen et al., 2009;Pineda and Venegas, 2009;Jin et al., 2006;Zhang et al., 2005Zhang et al., , 2009;;Delon et al., 2010Delon et al., , 2012;;and Pan et al., 2012).This approach is more suited when routine monitoring data are available but the values of the derived fluxes are clearly dependent on the validity of the dry deposition velocity calculation.

Atmospheric concentration measurements
Atmospheric concentrations of NO 2 , HNO 3 , NH 3 , O 3 and SO 2 are measured by passive samplers on a monthly ba-    Brook et al. (1999) the validation method according to international standards, have been widely detailed in Adon et al. (2010).To give an indication of the precision of this sampling technique, the covariance of all duplicate samples over the studied period were found to be 20, 9.8, 14.3, 16.6 and 10 % for HNO 3 , NO 2 , NH 3 , SO 2 and O 3 , respectively.Furthermore, Adon et al. (2010) presented the evolution of NO 2 , HNO 3 , NH 3 , O 3 and SO 2 concentrations along the period 1998-2007 for each station.Table 2 presents a synthesis of the mean annual gas concentrations for the IDAF sites of west and central Africa (Adon et al., 2010).

Modeling of the dry deposition velocity for each IDAF site
Dry deposition velocities (V d ) were calculated using the big-leaf dry deposition model of Zhang et al. (2003b).The main parameterizations of the resistances in the model are briefly presented in Appendix A. The main input parameters of this updated deposition model are physiological (parameters of land-use categories), biophysical (LAI) and meteorological data.
In general, parameters of inferential models have been largely derived from European and North American studies and may not necessarily be adequate for African (or tropical) climate, vegetation and soil conditions (Fowler et al., 2009).Based on ecological and climatic characteristic of African ecosystems, we tried to adapt the parameters of landuse categories (LUC) described in Zhang et al. (2003b) to IDAF-specific sites for the calculation of canopy resistances, i.e., cuticle resistance (R cut ), stomatal resistance (R st ), incanopy aerodynamic resistance (R ac ) and soil resistance in the big-leaf model (Table 3).Note that the types of LUC described in Zhang et al. (2003b) have one vegetation layer, while the savannas (shrub, tree or woody) have two vegetation layers.As a first approximation, the shrub and tree savannas (Lamto, Djougou and Katibougou) were assigned to grassland LUC (long grass), which is the dominant vegetation type in these areas.However, for the calculation of the stomatal resistance, we used the savanna parameters described in Brook et al. (1999).These parameters are r smin (minimal stomatal resistance), b rs (empirical light response coefficient), b vpd (water-vapor-pressure-deficit constant), (leaf water potential), T min , T max and T opt (minimum and maximum temperatures at which stomatal closure occurs and optimum temperature for maximum stomatal opening).
For the soil resistance (R g ), values suggested by Zhang et al. (2003b) were used for IDAF sites excepted in the Sahelian domain.Considering the semiarid climate of the Sahel, close to the Sahara (desert) and the steppe vegetation, we assumed some values based on published measurement in the literature (Wesely et al., 1989;Brook et al., 1999;Ganzeveld and Lelieved, 1995) (Table 3).
Roughness length (Z 0 ) is needed for calculating friction velocity, which subsequently affects aerodynamic, quasilaminar and non-stomatal resistances (Zhang et al., 2003b).We have used Z 0 values already simulated at the savanna sites in the framework of a previous study performed with the SVAT (Soil Vegetation Atmosphere Transfer) model ISBA (Interactions between Soil, Biosphere and Atmosphere - Noilhan and Mahfouf, 1996;Noilhan et al., 1989) and explained in Delon et al. (2010) (Table 3).
LAI is an important parameter for calculating canopy resistances.For the IDAF sites, LAI was obtained from MODIS (MODerate Resolution Imaging Spectroradiometer) satellite data for the period 2000-2007.The MODIS LAI is the ratio of one-sided green foliage area per unit horizontal ground area in broadleaf canopies, or the projected needleleaf area per unit ground area in conifer canopies, and is given in m 2 m −2 (Myneni, 1999).To that end, we used the 1 km MODIS LAI values that are processed over an 8-day period (Yang et al., 2006).The MODIS LAI product has already been validated with field measurements on many sites over the western African region (Fensholt et al., 2004;Samain et al., 2008).The mean seasonal variations of the monthly LAI during the period 2000-2007 for IDAF sites are shown in Fig. 2 and the range values are displayed in Table 3.The MODIS LAI of IDAF specific sites are well comparable to the ones of the corresponding land covers of the ECOCLIMAP database (Kaptue et al., 2010).For our simulation of dry deposition velocity, we used the monthly LAI averaged over 8 yr (2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007) for savanna sites and a constant value of 5 m 2 m −2 for tropical forest sites.We note that Zhang et al. (2003b) and Brook et al. (1999) have used a constant LAI of 6 and 4.5 m 2 m −2 , respectively, to simulate dry deposition velocity for tropical forests.
Meteorological data required for the simulation are wind speed and temperature at a reference height in the surface layer, surface temperature, solar irradiance at the surface, precipitation, surface pressure, relative humidity and cloud cover fraction.With the exception of cloud cover fraction, meteorological conditions are provided by the forcing developed in the frame of ALMIP (AMMA Land Surface Model Intercomparison Project) from satellite data, described in    Boone et al. (2009).The spatial resolution of this forcing is 0.5 • × 0.5 • with a 3 h temporal resolution.For the simulation of dry deposition velocity, we used a database of 6 yr from 2002 to 2007.Validations of surface temperature and moisture have been performed within the ALMIP project (Boone et al., 2009;De Rosnay et al., 2009).However, the forcing in ALMIP is available at 10 m, whereas the concentrations of gases are measured at around 2 m for the savanna sites and 3 m for the forested ecosystems within IDAF network.As a first approximation, a logarithmic decrease of the wind forcing from 10 to 2 m, depending on the rugosity of the site, has been applied to calculate deposition velocities at 2 m for the savanna sites, as described in Delon et al. (2010Delon et al. ( , 2012)).This approach to reduce the wind from 10 m to the height of 2 m is based on the constant flux assumption (Baldocchi, 1988;Zhang et al., 2009).However for the forested ecosystems, the constant flux assumption cannot be applied below the canopy due to extra sink terms; hence another approximation is done and described below in Sect.2.2.4.
Surface wetness controls non-stomatal resistances for soluble trace gases.The big-leaf model of Zhang et al. (2003b) predicts surface wetness semimechanistically and distinguishes dew from rain based on precipitation data for rain and on nighttime cloud cover and friction velocity for dew formation (Janssen and Romer, 1991;Brook et al., 1999).In this study, as the cloud cover information was missing, we performed sensitivity tests for the presence of dew by using the constant relative humidity (RH) threshold method for estimating dew condensation (Sentelhas et al., 2008).Various RH thresholds have been used as proxies to determine canopy wetness over forests and grasslands (Tsai et al., 2010;Wichink Kruit et al., 2008, 2010;Flechard et al., 2011).Thus, we have performed sensitivity tests using a threshold of 90 % for forest, 81 % for wet savanna and 71 % for dry savanna to test the presence of dew.Results showed that the variation of monthly V d is negligible when it is assumed that dew has occurred; for example the RMSE (root-mean-squared error) of V d (SO 2 ) is 2, 1 and 4 % for dry savanna, wet savanna and forest, respectively (Adon, 2011).Therefore, in this study, we do not take into account the assumption of the occurrence of dew.
Previous studies of nitrogen budget in wet and dry savanna ecosystems have proposed an estimate of N compounds deposition velocities (Delon et al., 2010(Delon et al., , 2012)).In those studies, V d were calculated according to the Wesely (1989) resistance analogy in the SVAT big-leaf model ISBA, with modifications of the cuticle and ground resistances adapted from Zhang et al. (2003b).The results were obtained at the regional scale at 0.5 • resolution for the period 2002-2007, and focused on the reactive nitrogen compounds budget (with estimates of emission and deposition (dry + wet) fluxes).The present study is focused on dry deposition only, and integrates all the parameterizations developed by Zhang et al. (2003b), which gives a different but consistent and coherent estimate of V d at the local scale, and for more species (same N compounds + O 3 and SO 2 ).Indeed, we investigate each specific vegetal cover representative of each IDAF measurement sites in this study.Differences between the two approaches are mainly due to the different resolutions of models and to the degree of details involved in the estimation of input parameters such as the ones detailed in the above paragraph.The previous studies of Delon et al. (2010Delon et al. ( , 2012) ) give a point of comparison for N compounds' V d calculated in the present work, as very few studies are available for African ecosystems.Furthermore, as explicated below, the NH 3 bidirectional exchange is applied in the present study, which was not the case in the previous studies of Delon et al. (2010Delon et al. ( , 2012)).
We present in this paper the calculation, over the period 2002-2007, of monthly means (from 3-hourly values) deposition velocities for O 3 , SO 2 , NO 2 , HNO 3 and NH 3 for the IDAF sites.

NH 3 bidirectional exchange and canopy compensation point
The NH 3 net fluxes were calculated using the bidirectional air-surface exchange model of Zhang et al. (2010), which is a modification from the original big-leaf model of Zhang et al. (2003b).The main parameterization of this new bidirectional exchange model and the method of calculation of the NH 3 net fluxes in this study are briefly presented in Appendix B.
In this modified model, the main input parameters are stomatal ( st ) and ground ( g ) emission potentials, leading to stomatal and soil compensation points, respectively, and thus the canopy compensation point (X cp ) (Eq.B3).For each of the LUCs, Zhang et al. (2010) derived representative input values based on literature data.In our simulation and for the seminatural forested ecosystem, we used the median values of st (300) and g (20) suggested for tropical forest LUC.The daytime canopy compensation point values of NH 3 were less than 2 µg m −3 , as predicted by Zhang et al. (2010).
For the savannas sites, a sensitivity test showed that the median values suggested for the long grass or short grass and forbs LUC, considered as fertilized, ( g = 2000) would give too high daytime values of X cp (50-300 µg m −3 at Banizoumbou) in the dry season due to very high ground surface temperatures in the afternoon (38-50 • C).However, the African savannas remain either slightly or not fertilized (Bouwman and van der Hoek, 1997;Mosier et al., 1998).Delon et al. (2010) estimated the mean total nitrogen input from animal manure for the Sahelian sites from 11 to 23 kg N ha −1 yr −1 , lower when compared to fertilized vegetation input in temperate ecosystems (Loubet et al., 2002;Massad et al., 2010).Due to the lack of data for typical tropical savanna sites, we chose a range of lower-end values reported for grass or unfertilized ecosystems in the literature.Thus, we have chosen values of 100 (low scenario) and 200 (high scenario) for st (Spindler et al., 2001;Horvath et al., 2005;Trebs et al., 2006;Personne et al., 2009) and values of 200 (low scenario) and 360 (high scenario) for g (Massad et al., 2010;Zhang et al., 2010).
Figure 3 presents the mean diurnal variation of X cp (NH 3 ) for a Sahelian site (Banizoumbou) and a wet Guinean savanna site (Lamto) in the dry season (January 2006).In our simulation, the maximum ground surface temperature was fixed at 40 • C to avoid too high values of X cp for the "high scenario" in dry savannas.Although low values of st and g were used to run the model, the estimated X cp (NH 3 ) values are within a reasonable range of values determined for grassland in other studies (Langford et al., 1992;Hesterberg et al., 1996;Spindler et al., 2001;Loubet et al., 2002;Trebs et al., 2006;Zhang et al., 2010).This is due to high surface temperatures mainly at Sahelian sites.Note that the X cp diurnal values of the Sudano-Guinean site of Djougou are comparable to those of Sahelian sites (result not shown).During the wet season, the X cp diurnal values are lower due to lower surface temperatures: 0.1-1 µg m −3 during nighttime and 0.3-5 µg m −3 during daytime for the savanna sites.
For the NH 3 net fluxes, we used the two scenarios (lower and upper estimates) presented above compared to the "dry deposition only" scenario (X cp = 0).

Corrections for within-canopy concentration data for forests
The inferential method requires atmospheric concentrations and turbulence intensity above the canopy to predict rates of dry deposition over the forest.In the IDAF network, gas concentrations are measured at about 3 m in forests.Moreover, there are very few published within-canopy vertical gas concentration profiles (HNO 3 , NH 3 ) in the literature for forest (Flechard et al., 2011).Thus, we have carried out a pilot experiment by measuring simultaneously gas concentrations at 10 and 3 m (or 2 m) from the ground in the forested ecosystem of Zoetele (and in the wet savanna of Lamto and dry savanna of Banizoumbou) over the period of September 2010 to December 2011.
For the forested ecosystem, NO 2 , HNO 3 and SO 2 monthly concentrations measured simultaneously at 10 and 3 m show no significant trend and are on the same order.Over this period, mean annual concentrations -at 3 and 10 m, respectively -are 0.81 and 0.80 ppb for NO 2 , 0.2 and 0.3 ppb for HNO 3 , and 1.8 and 1.6 ppb for SO 2 .However for NH 3 and O 3 , monthly concentrations measured at 10 m are higher and the mean ratio was 1.5 for NH 3 and 1.3 for O 3 (mean annual concentration are 3.6 and 6.1 ppb for NH 3 and 5.4 and 7.0 ppb for O 3 at 3 and 10 m, respectively).This observation is consistent with the approximation made in Flechard et al. (2011) showing that NH 3 concentrations measured in clearings and below canopy are consistently smaller than above treetops; they thus applied a constant correction factor of 1.3 for NH 3 concentrations measured below trees.We note that in the dry savanna of Banizoumbou, NH 3 concentrations measured at 2 m (annual mean 8.1 ppb) are higher than at 10 m (6.3 ppb), indicative of NH 3 sources in the ground and in the leaf litter in grasslands (Nemitz et al., 2009).For O 3 , our observations are consistent with ozone vertical profile concentrations that decrease towards the ground, as observed by Krzyzanowski (2004) by measuring ozone concentrations at different heights in a forest canopy.Contrary to our observation for HNO 3 and SO 2 , Hicks (2006) observed a ratio of 1.34 and 1.26, respectively, between concentrations measured above the treetops and within the canopy of forests.
In a first approximation, for forested sites of IDAF (Zoetele and Bomassa), we thus applied a constant correction factor of 1.5 for NH 3 and 1.3 for O 3 concentrations measured at 3 m in order to calculate the dry deposition flux at 10 m.For the other gases (NO 2 , HNO 3 and SO 2 ) no correction was applied.The dry deposition fluxes of HNO 3 and SO 2 using the forest-clearing concentrations may be underestimated as discussed by Hicks (2006).In addition, the mean covariance of samplers exposed simultaneously at 3 and 10 m were found to be 13, 25.5 and 21 % for NO 2 , HNO 3 and SO 2 respectively, and these values are comparable of the mean reproducibility of IDAF passive samplers calculated from 1998 to 2007 (9.8, 20 and 16.6 % for NO 2 , HNO 3 and SO 2 , respectively).Thus, the small difference in measured concentrations between the two heights (3 and 10 m) for NO 2 , HNO 3 and SO 2 could be included in the global uncertainty for the passive sampler method.

Uncertainties in dry deposition fluxes estimates
Uncertainties in the estimated dry deposition fluxes result from combined uncertainties in measured gaseous concentrations and in modeled exchange rates.In this section, rather than quantifying the total uncertainty of the inferential method in the study, we focus on addressing the uncertainties for each contribution of the dry deposition estimates.Part of uncertainties linked to the measurement of gas concentration using IDAF passive samplers have been given by the covariance of duplicates (reproducibility), between 10 and 20 % according to the species (Adon et al., 2010).Other parts are related to the measurement techniques of passive samplers.One of the uncertainties of the dry deposition velocities is related to the wind forcing.The difference between the wind in the forcing and the wind measured in situ is between 5 and 35 % depending on the site.A mean rate of 20 to 30 % depending on the site is applied for the wind speed uncertainty, leading to an uncertainty rate for the dry deposition velocity between 10 and 20 % for all sites.Others uncertainties are related to the representation of the soil type in the model indirectly dependent on the roughness length and the soil resistances (Zhang et al., 2003b) as well as the choice of plant physical parameters.
In this study, concentrations are measured with passive samplers and are monthly integrated.Dry deposition velocities are simulated 3-hourly and then monthly averaged.As we use monthly means for concentrations and deposition velocities, the covariance between the two may induce an additional uncertainty (the missing covariance term), especially for species having strong diurnal variations, in the range of ∼ 20 % (Matt and Meyers, 1993;Zhang et al., 2005).For the NH 3 bidirectional exchange, in addition of the uncertainties related to the model itself (Zhang et al., 2010), the ground emission potential remains the most uncertain in our simulations due to the lack of detailed investigations although flux measurements have often shown significant emission from soils and leaf litter (Sutton et al., 2013;Flechard et al., 2010).The unidirectional approach (Zhang et al., 2003b) used for the other gases could induce other sources of uncertainty.
Additional uncertainties are associated with different parameterizations of dry deposition used in deposition models.The physical, biological and chemical exchange mechanisms involved in deposition processes are too complex to be explicitly and completely modeled, and as such parameterizations tend to be empirical in the models (Flechard et al., 2011;Schwede et al., 2011).Moreover, multiple species model intercomparison show factor of 2-5 differences in exchange rates between models depending on the chemical species (Flechard et al., 2011).

Dry deposition velocities
Monthly means (from 3-hourly values) of deposition velocities for O 3 , SO 2 , NO 2 , HNO 3 and NH 3 have been calculated for the period 2002-2007 in order to reproduce the seasonal cycle of the deposition processes at each site (Fig. 4).We note a fairly clear distinction between the different ecosystems and climatic domains.The monthly V d values of each gas increase from the semiarid savannas (Agoufou, Banizoumbou) to the equatorial forested ecosystems (Zoetele, Bomassa) indicating that the dry deposition velocity increases with the vegetation density.As a first approximation, along the African ecosystem transect, the V d gradient thus follows the climatic gradient.In fact, there exists a marked latitudinal gradient over the west and central African zone.Areas with the highest rainfall have the greatest volume of biomass or primary productivity; as a consequence, surface deposition (cuticle) and stomatal uptake become important, and the canopy resistance (R c ) is the main factor determining the deposition rate in vegetative zones (Tsai et al., 2010).
The monthly variation of NO 2 V d follows the same pattern as O 3 but with slightly smaller values.In fact, in this model of deposition, the NO 2 V d is parameterized similar to the one of O 3 due to similar behavior for a variety of conditions and the importance of their stomatal uptake (Zhang et al., 2002a).However, some studies pointed out the importance of the non-stomatal deposition fluxes in the case of O 3 (Fowler et al., 2001;Stella et al., 2011).NH 3 V d is similar to SO 2 but slightly higher due to its higher molecular diffusivity.NH 3 and SO 2 are reasonably soluble gases in pure water and are effectively removed at higher rates under moist conditions (Erisman et al., 1993a, b;Erisman and Wyers, 1993).Even if the chemical characteristics of NH 3 are not the same as SO 2 , the NH 3 V d is parameterized similar to SO 2 in this deposition model (Zhang et al., 2002a(Zhang et al., , 2003b)).HNO 3 presents the highest V d among all the chemical species in this study because of its high solubility and reactivity.Note that in the big-leaf model of Zhang et al. (2003b), SO 2 and O 3 are used as base species to scale the dry deposition rate for other chemical species.The monthly means of dry deposition velocities range from 0.16 to 0.84 cm s −1 for SO 2 , from 0.14 to 0.40 cm s −1 for O 3 , from 0.16 to 0.99 cm s −1 for NH 3 , from 0.13 to 0.37 cm s −1 for NO 2 and from 0.48 to 2.56 cm s −1 for HNO 3 on the transect dry savannas-wet savannas-forests.In addition, V d are higher in the wet season for each ecosystem, especially in dry savannas, where the dry season is well marked.It is considered that higher deposition velocities in the wet season are mainly caused by non-stomatal uptake of wet canopies (Matsuda et al., 2006;Tsai et al., 2010).The interannual variability of V d for the 7 yr period (2002)(2003)(2004)(2005)(2006)(2007) is low and range from 3 to 12 % for NH 3 and SO 2 , from 2 to 11 % for NO 2 and O 3 , and from 2 to 17 % for HNO 3 on the transect of ecosystems.These variations could be attributed to the spatio-temporal variations of meteorological data.
The comparison between our modeled V d and previous studies is shown in Table 4.The V d of O 3 and SO 2 lie well within the range of values determined for other tropical forests (Matsuda et al., 2006, Rummel et al., 2007;Tsai et al., 2010) and grasslands (Takahashi et al., 2001;Sorimachi et al., 2003).The modeled V d of NO 2 and NH 3 are within a reasonable range compared to reference values (Zhang et al., 2005(Zhang et al., , 2009;;Trebs et al., 2006;Kirkman et al., 2002;Delon et al., 2010;Endo et al., 2011).The modeled V d of HNO 3 on forest and grass were the lowest values of those reported by Endo et al. (2011), Huebert and Robert (1985) and others studies (e.g., Hanson and Lindberg, 1991;Duyzer and Fowler, 1994;Zhang et al., 2002a) mainly due to lower wind velocities in African ecosystems.

Dry deposition fluxes
Monthly dry deposition fluxes of NO 2 , NH 3 , HNO 3 , SO 2 and O 3 were estimated using the inferential method at west and central African sites of the IDAF network over the study period (1998-2007).By using the monthly V d from each year and the monthly 6 yr average V d (2002)(2003)(2004)(2005)(2006)(2007) for the years that do not have V d data, the monthly deposition fluxes have been calculated for each year and then averaged over the study period in order to estimate the mean range values of gaseous deposition fluxes representative of African ecosystems.

Nitrogen compounds (NO 2 , HNO 3 , NH 3 )
Table 5 presents a synthesis of the mean seasonal and annual exchange fluxes of nitrogen compounds (NO 2 , HNO 3 , NH 3 ) for all the IDAF sites.

Nitrogen dioxide (NO 2 )
Figure 5 presents the monthly evolution of NO 2 dry deposition flux (NO 2 _dd) estimated at the dry savannas of Niger and Mali (Fig. 5a: Banizoumbou, Agoufou, Katibougou), at the wet savannas of Côte d'Ivoire and Benin (Fig. 5b: Lamto, Djougou) and at the evergreen equatorial forests of Cameroon and Congo (Fig. 5c: Zoetele, Bomassa).Vertical bars indicate the standard deviation calculated over the study period (1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007).Average monthly evolution of NO 2 concentrations over the same study period is superimposed on the monthly dry deposition fluxes.The axes of fluxes are inverted for all figures.We note a good correlation between gas concentrations and dry deposition fluxes on the transect dry savannas-wet savannas-forests.
In the dry savannas, maximum flux values are observed in May/June as for concentration values (Fig. 5a).In general, higher deposition fluxes in the wet season are due to both higher measured NO 2 concentrations and its higher dry deposition velocity calculated in this season.In the Sahelian sites (Agoufou, Banizoumbou), seasonally averaged dry deposition fluxes are 2 times higher in the wet season than in the dry season (Table 5).Annual average dry deposition fluxes of NO 2 are around −0.4 to −0.7 kg N ha −1 yr −1 in the dry savannas.These values calculated in these pastoral areas are in the lower range of deposition fluxes calculated for a tropical pasture site in Brazil (Rondonia) (annual NO 2 deposition flux: −0.76 to −2.4 kg N ha −1 yr −1 ), which is considerably higher, mainly due to higher NO 2 mixing ratios observed (Trebs et al., 2006).
In the wet savannas, higher deposition flux values are obtained in the dry season with maximum fluxes in December at Djougou and January at Lamto, such as NO 2 concentrations (Fig. 5b).Biomass burning is the most important source of NO x (NO+NO 2 ) during the dry season in the wet savannas, where the vegetation density is more important than in the dry savannas.Galanter et al. (2000) showed that more than 75 % of the NO x at the surface near equatorial Africa is the result of biomass burning that occurs from December to February.In the dry season, deposition flux values are 1.5 times higher at Lamto and around 3 times higher at Djougou than in the wet season (Table 5) due to higher NO 2 concentrations.Annual average dry deposition fluxes of NO 2 are on the same order (−0.4 kg N ha −1 yr −1 ) for the two wet savanna sites (Table 5).In forests, higher deposition fluxes are observed in February and March, i.e., at the end of the dry season and at the beginning of the wet season (Fig. 5c).Seasonal average fluxes are on the same order between the two seasons (Table 5).NO 2 concentration measurements tend to show that the biomass burning source during the dry season is equivalent to soil emissions buffered by canopy uptake in forests during the wet season (Adon et al., 2010).Annual average dry deposition fluxes of NO 2 in forests (−0.5 to −0.8 kg N ha −1 yr −1 ) are on the same order as in dry savannas (Table 5) due to higher V d values in forests and higher concentrations measured in dry savannas (Table 2).These values are well within the range of measurements derived by Hanson et al. (1989), who showed that N deposition from NO 2 was between −0.008 and −1.9 kg N ha −1 yr −1 for natural forests.They are also comparable to NO 2 deposition fluxes estimated by Zhang et al. (2005) that range from −0.1 to −1.5 kg N ha −1 yr −1 at seven eastern Canadian rural sites for a 1 yr period.
Note that the NO 2 deposition fluxes estimated in this study could be overestimated since the photochemical reaction of NO, NO 2 and O 3 and the NO 2 compensation point approach are not included in the deposition model.In fact, soil emission of NO that reacts rapidly with O 3 to produce NO 2 can greatly diminish the magnitude of the downward NO 2 flux and sometimes cause it to be directed upward (Bakwin et

Nitric acid (HNO 3 )
Monthly evolution of HNO 3 dry deposition fluxes is similar to that of HNO 3 concentrations on the transect of ecosystems (Fig. 6a, b, c).In the dry savannas, HNO 3 dry deposition fluxes are very low in the dry season, especially at Agoufou and Banizoumbou, and much (4-12 times more) higher in the wet season due to both higher HNO 3 concentrations and dry deposition velocities in this season for the three sites.Seasonal average fluxes range from −0.1 kg N ha −1 yr −1 in the dry season to −1.8 kg N ha −1 yr −1 in the wet season (Table 5).In the wet savannas, the difference in deposition fluxes between the two seasons is low with higher values in the dry season at Lamto (−0.9 kg N ha −1 yr −1 ) and in the wet season at Djougou (−0.8 kg N ha −1 yr −1 ).The deposition fluxes are less important in the dry season at Djougou due to smaller HNO 3 V d in this season.In forests, seasonal average deposition fluxes of HNO 3 are on the same order between the two seasons (−1 kg N ha −1 yr −1 ) at Bomassa and higher (−1.6 kg N ha −1 yr −1 ) in the dry season of Zoetele due to higher values of HNO 3 concentrations.Annual mean dry deposition fluxes of HNO 3 are around −0.7 kg N ha −1 yr −1 in the dry and wet savannas and −1.0 kg N ha −1 yr −1 in forests (Table 5).Annual mean HNO 3 deposition fluxes show a low variability according to the ecosystems, with seasonal differences especially pronounced in the dry savannas.Although HNO 3 concentrations (0.3-0.5 ppb) were lower than NO 2 (0.9-2.4 ppb) on the African ecosystems transect (Table 2), HNO 3 dry deposition fluxes were typically as important as NO 2 dry deposition fluxes owing to higher V d values for HNO 3 .

Ammonia (NH 3 ) bidirectional fluxes
Figure 7 presents the monthly evolution of estimated low and high surface-atmosphere exchange fluxes of NH 3 on the transect of ecosystems.For the forests, only one scenario has been estimated (Sect.2.2.3).Bidirectional NH 3 flux scenarios complemented by a "deposition only" scenario (X cp (NH 3 ) = 0) are presented in Table 5.
In the dry savannas, seasonal mean fluxes of NH 3 range from a deposition of −1.0 kg N ha −1 yr −1 to an emission of +0.3 kg N ha −1 yr −1 in the dry season and a deposition of −1.9 to −4.6 kg N ha −1 yr −1 in the wet season (Table 5).The NH 3 net emission fluxes observed in the dry season for our higher scenario estimate are due to higher X cp (NH 3 ) in this season (Fig. 7a, Fig. 3).This is consistent with observations made in some seminatural ecosystems where air concentrations are low under very dry condition or at high temperatures (Langford and Fehsenfeld, 1992;Erisman et al., 1994;Sutton et al., 1995;Flechard and Fowler, 1998).In the wet season, NH 3 dry deposition fluxes are more important due to higher measured NH 3 concentrations and higher NH 3 V d simulated during the short vegetative period.Dry savannas are generally pastoral areas; the main sources include bacterial decomposition of urea in animal excreta, which is very active in the wet season with the hydrolysis of urea, and emissions from natural soils (Schlesinger and Hartley, 1992;Bouwman et al., 1997Bouwman et al., , 2002a)).Because of the short lifetime of NH 3 (1-5 days or less), low source height and relatively high V d , a substantial fraction (20-40 %) would likely deposit near its source (Aneja et al., 2001).
In the wet savannas, seasonal average net fluxes of NH 3 range from −3.1 kg N ha −1 yr −1 (deposition) to +0.1 kg N ha −1 yr −1 (emission) (Table 5).In the Guinean savanna of Lamto, the estimated NH 3 deposition fluxes are more important at the beginning of the dry season (January-March) where concentrations are higher (Fig. 7b).However, in the Sudano-Guinean site of Djougou, lower NH 3 deposition fluxes were estimated in the dry season, although higher measured NH 3 concentrations.This is mainly due to higher simulated X cp (NH 3 ) in the dry season that led to net emission fluxes, like in the dry savannas.Generally, higher concentrations of NH 3 in the dry season are due to savanna fires, which are a significant source of ammonia in tropical regions (Lobert et al., 1990;Delmas et al., 1995).For the forested ecosystems, only dry deposition occurs due to low compensation point and no significant difference between wet and dry seasons (Fig. 7c).Seasonal NH 3 dry deposition fluxes are around −8 kg N ha −1 yr −1 in the dry and wet seasons (Table 5).
Over the study period, annual average dry deposition fluxes of NH 3 range from −1.6 to −2.5 kg N ha −1 yr −1 in the dry savannas and from −1.3 to −2.6 kg N ha −1 yr −1 in the wet savannas for our lower scenario estimate, and reduced by 26 to 62 % in the case of our higher scenario (Table 5).In the forests, the mean annual NH 3 net fluxes are around −8 kg N ha −1 yr −1 .If NH 3 were assumed to be deposited only (i.e., no bidirectional exchange, X cp (NH 3 ) = 0), then the mean annual dry deposition fluxes would be more important and would range from −2.2 to −3.9 kg N ha −1 yr −1 in dry and wet savannas and would be around −10 kg N ha −1 yr −1 in forests (Table 5).
These estimates of nitrogen dry deposition from NO 2 , HNO 3 and NH 3 on the transect of African ecosystems has allowed us to make a partial assessment of the total quantity of nitrogen deposited in gaseous form.

Total nitrogen dry deposition fluxes from NO 2 , HNO 3 and NH 3
Using the two bidirectional NH 3 exchange scenarios, the annual total N (NO 2 +HNO 3 +NH 3 ) dry depositions fluxes are estimated to range from −1.8 to −3.9 kg N ha −1 yr −1 in dry savannas and from −1.6 to −3.8 kg N ha −1 yr −1 in wet savannas and are estimated to be around −10 kg N ha −1 yr −1 in forests.On the transect, annual total N dry deposition is more important in forests due to higher NH 3 deposition fluxes, and is on the same order in wet savannas and dry savannas.In savannas, N dry deposition is more important at Katibougou for the Sahelian sites and at Lamto for the wet savanna sites due to its higher vegetation density than the other sites in the same climatic domain.Higher NH 3 V d are related to important non-stomatal uptake of wet canopies, which lead to higher NH 3 deposition fluxes.Our annual N dry deposition estimates for the savanna sites are well comparable with those estimated at a tropical pasture site (−1.57to −3.68 kg N ha −1 yr −1 ) by Trebs et al. (2006), although they considered more species (NO 2 , HNO 3 , NH 3 , HONO and aerosols (NO − 3 , NH + 4 )).In our study, particulate dry deposition (NO − 3 , NH + 4 ) is not taken into account in this partial assessment due to very low concentrations measured at IDAF sites (Delon et al., 2010).
If NH 3 were considered to be deposited only (no bidirectional exchange), the annual total N dry deposition would range from −4.0 to −5.3 kg N ha −1 yr −1 in dry savannas, from −3.2 to −4.6 kg N ha −1 yr −1 in wet savannas and from −11.2 to −11.7 kg N ha −1 yr −1 in forests.
Over the study period (1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007), the interannual variability of the dry deposition fluxes ranged from 23 to 63 % for NH 3 (low scenario), from 13 to 30 % for NO 2 and from 21 to 43 % for HNO 3 on the transect dry savannas-wet savannasforests.No specific trend in the variability is observed, as already concluded in the study of Delon et al. (2012).The interannual variability is largely attributable to the variability of gaseous concentrations due to the potential variation of the intensity of atmospheric sources and the variability of meteorological data.
To give the relative contribution of dry to total deposition fluxes, we calculated total nitrogen wet deposition fluxes from the IDAF database (http://idaf.sedoo.fr)for the same study period (1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007) for each site (Adon, 2011).The wet deposition flux has been calculated as the product of the ammonium and nitrate concentrations in rain (annual volume-weighted mean VWM) by the annual rainfall (Galy- Lacaux and Modi, 1998;Galy-Lacaux et al., 2009, and references therein).The total N wet (NO − 3 +NH + 4 ) deposition fluxes estimated range from −1.6 to −3.2 kg N ha −1 yr −1 in dry savannas, from −3.5 to −5.2 kg N ha −1 yr −1 in wet savannas and are around −4.6 kg N ha −1 yr −1 in forests.These values are well comparable with those reported by previous studies (Yoboue et al., 2005;Sigha et al., 2003;Laouali et al., 2012) but not exactly in the same period.Using the bidirectional scenarios, the total N (dry + wet) deposition fluxes were estimated to range from −3.4 to −7.1 kg N ha −1 yr −1 in dry savannas, from −5.1 to −9.0 kg N ha −1 yr −1 in wet savannas and be around −14.4 kg N ha −1 yr −1 in forests.We estimate that dry deposition process in gaseous form contributes 31-68 % of the total N deposition fluxes over African ecosystems.

Sulfur dioxide (SO 2 )
The monthly evolution of SO 2 dry deposition flux (SO 2 _dd) on the transect of ecosystems is generally comparable to SO 2 concentrations over the period 2002-2007 following the same gradient (Fig. 8a-c).We note that the standard deviation of monthly mean fluxes is important.In fact, monthly SO 2 concentrations measured at IDAF sites are low and vary considerably from one year to another for the same month.
If we consider the season, average SO 2 dry deposition fluxes range in the wet season from −0.8 to −1.0 kg S ha −1 yr −1 in dry and wet savannas and from −0.7 to −1.1 kg S ha −1 yr −1 in forests, and in the dry season from −0.3 to −0.5 kg S ha −1 yr −1 in dry savannas and from −0.8 to −0.9 kg S ha −1 yr −1 in wet savannas and forests.Adon et al. (2010) observed the same order of SO 2 concentrations in dry and wet seasons in each ecosystem with higher values in the wet season for dry savannas.This suggests a contribution of soil emission, biosphere and biomass burning sources (Van Breemen, 1982, 1993;Macdonald et al., 2004;Bates et al., 1992;Arndt et al., 1997).The sulfur content of vegetation is lower compared to carbon and nitrogen elements and SO 2 emissions factors for combustion processes are lower than those for carbonaceous or nitrogen species (Lacaux et al., 1995).The 6 yr mean annual dry deposition fluxes of SO 2 are −0.7 ± 0.3 kg S ha −1 yr −1 at Agoufou, −0.5 ± 0.2 kg S ha −1 yr −1 at Banizoumbou and −0.7 ± 0.3 kg S ha −1 yr −1 at Katibougou in dry savannas; −0.9 ± 0.2 kg S ha −1 yr −1 at Djougou and −0.8 ± 0.4 kg S ha −1 yr −1 at Lamto in wet savannas; and −0.8 ± 0.3 kg S ha −1 yr −1 at Zoetele; and −1.0 ± 0.5 kg S ha −1 yr −1 at Bomassa in forests.Along the ecosystem transect, the annual mean of SO 2 dry deposition fluxes presents low values and a small variability (−0.5 to −1 kg S ha −1 yr −1 ).The interannual variability over the 6 yr period is between 25 and 49 % for all the sites and no specific trend is observed.
In general, dry deposition fluxes of SO 2 estimated at African ecosystems are lower compared to other tropical ecosystems (Takahashi et al., 2002;Sorimachi et al., 2003) although the V d modeled are on the same order (Table 4).This is due to low measured concentrations (on the order of 1-2 ppb).Indeed, the remote measurement sites of the IDAF network have not yet been impacted on by anthropogenic activities or industrial emissions.

Ozone (O 3 )
The monthly evolution of O 3 dry deposition fluxes (O 3 _dd) is similar to that of O 3 concentrations on the transect of ecosystems over the period 2001-2007, as in the case of NO 2 (Fig. 9a-c).In the semiarid savannas, the seasonal cycle is clear with maximum O 3 deposition fluxes in the heart of the wet season due to higher values of both O 3 concentrations and V d .In fact, high ozone concentrations during the wet season are the result of active photochemical production of O 3 in the boundary layer associated with high NO x concentrations (Stewart et al., 2008).In the Sahelian region, the ozone production in the wet season is mainly related to natural biogenic precursor sources (Adon et al., 2010).The seasonal average deposition fluxes are −6.4±4.1 and −21.0 ± 3.2 kg ha −1 yr −1 at Agoufou, −7.2 ± 2.0 and −20.4 ± 3.6 kg ha −1 yr −1 at Banizoumbou and −13.0 ± 2.5 and −23.0 ± 4.1 kg ha −1 yr −1 at Katibougou, in dry and wet seasons, respectively.In the wet savannas, the seasonal average O 3 deposition fluxes are on the same order with values of −20.3 ± 1.5 and −18.6 ± 3.3 kg ha −1 yr −1 at Lamto, and −16.0 ± 3.6 and −17.6 ± 3.8 kg ha −1 yr −1 at Djougou, in dry and wet seasons, respectively.The difference between the seasonal ozone deposition fluxes is reduced with higher values of V d in the wet season and higher O 3 concentrations in the dry season as a consequence of strong regional biomass burning activities.For forested ecosystems, the mean seasonal deposition fluxes of O 3 are −18.8 ± 0.3 and −11.3±0.9 kg ha −1 yr −1 at Zoetele and −10.1±1.1 and −10.8 ± 0.8 kg ha −1 yr −1 at Bomassa, in dry and wet seasons, respectively.We note that these seasonal values are on the same order between the two seasons at Bomassa but they are higher in the dry season at Zoetele due to higher O 3 concentrations measured in this season.
Few observational studies of O 3 deposition fluxes for a long-term period are available over tropical areas, although several studies concerning the diurnal evolution have been conducted in the rainforests (Cros et al., 1992(Cros et al., , 2000;;Andreae et al., 1992;Rummel et al., 2007;Matsuda et al., 2005).The modeled O 3 V d in our study are within the range of values determined in other tropical ecosystems (Table 4).In general, the O 3 deposition fluxes estimated in African forests are lower compared to other tropical forests (Mikkelsen et al., 2004;Zeller and Nikolov, 2000); this is due to lower concentrations measured.

Conclusions
In this study, we estimated dry deposition fluxes of nitrogen compounds (NO 2 , HNO 3 , NH 3 ), sulfur dioxide (SO 2 ) and ozone (O 3 ) in major African ecosystems, represented by IDAF sites.
NO 2 , HNO 3 , SO 2 and O 3 were considered to be net deposited, while surface-atmosphere exchange of NH 3 is considered to be bidirectional for all ecosystems.Monthly dry deposition fluxes have been estimated by the inferential technique, using air concentrations measured monthly by passive samplers for a long-term period (1998-2007) and modeled exchange rates.Surface and meteorological conditions specific to IDAF sites have been used in the deposition models.Along the transect of ecosystems, simulation results show that V d increases with the vegetation density.Thus, the lower values of V d for all gases have been obtained in the dry savannas and the higher values in the forests.For each ecosystem, the seasonal and annual mean variations of gaseous dry deposition fluxes have been analyzed.Dry deposition fluxes are more important in the wet season for all the gases in the dry savannas due to higher values of both concentrations and V d in this season.For the wet savannas and forested ecosystems, seasonal dry deposition fluxes are generally on the same order between the dry and wet seasons for all the gases except for NO 2 in the wet savannas, and for HNO 3 and O 3 at the Zoetele forested site, where flux values are higher in the dry season due to much higher concentrations.
Along the latitudinal transect of ecosystems, the annual mean dry deposition fluxes of nitrogen compounds range from −0.4 to −0.8 kg N ha −1 yr −1 for NO 2 , from −0.7 to −1.0 kg N ha −1 yr −1 for HNO 3 , and from −0.7 to −8.3 kg N ha −1 yr −1 for NH 3 over the study period (1998-2007).The total nitrogen dry deposition flux (NO 2 +HNO 3 +NH 3 ) is more important in forests (∼ −10 kg N ha −1 yr −1 ) than in wet and dry savannas (−1.6 to −3.9 kg N ha −1 yr −1 ).If NH 3 was considered to be deposited only to the ecosystems, the annual total N dry deposition would range from −3.2 to −5.3 kg N ha −1 yr −1 in dry and wet savannas, and from −11.2 to −11.7 kg N ha −1 yr −1 in forests.Along the ecosystem transect, the annual mean of SO 2 dry deposition fluxes presents low values and a small variability (−0.5 to −1 kg S ha −1 yr −1 ).For ozone, the annual mean dry deposition fluxes range from −11 to −19 kg ha −1 yr −1 in dry and wet savannas and from −11 to −13 kg ha −1 yr −1 in forests.The lower O 3 dry deposition fluxes in forests are due to low measured O 3 concentrations despite higher V d .Over the study period, the interannual variability of gaseous dry deposition fluxes showed no specific trend.Over African ecosystems, our study assumed that gaseous dry deposition contributes to 31-68 % of the total (dry + wet) N deposition fluxes.
This study allowed for estimates of the mean range of gaseous dry deposition fluxes representative of major tropical African ecosystems in west and central Africa.This is one of the major scientific objectives of the IDAF program.It is based on original and unique data from remote and seldomexplored regions.To improve this work, it is important to not only address the uncertainties in the determination of dry de-position velocities but also use the bidirectional approach for other gases such as NO 2 ; more investigation on the ground emission potential in the case of NH 3 surface-atmosphere exchange is needed.Within the IDAF network, we suggest to perform an experimental determination of dry deposition fluxes by other methods (e.g., gradient method, eddy correlation) on the measurement sites to be compared with fluxes estimated by the inferential method.Furthermore, we plan to use the model RegCM4 (regional climatic model) (Giorgi et al., 2012;Shalaby et al., 2012) to simulate the regional trends of gaseous dry deposition fluxes and to compare these results to long-term IDAF observations.This work will allow for providing a high-resolution map of dry deposition at regional scales of the African ecosystems useful for impact studies.
where ν is the kinematic viscosity of air and D j is the molecular diffusivity of a species j in air.
R c is the surface or canopy resistance, and characterizes the surface affinity for pollutant uptake.A large part of the uncertainty of the inferential method might be attributed to the parameterization of R c .Zhang et al. (2003b) proposed a revised parameterization of R c by including non-stomatal resistance (R ns ) parameterizations based on study results over five different vegetation types in North America (i.e., Zhang et al., 2002bZhang et al., , 2003a)): where the sub-resistances R st , R m , R cut , R ac and R g are respectively stomatal, mesophyll, cuticle, in-canopy aerodynamic and soil resistances.w st is the fraction of stomatal blocking under wet conditions.One of the improvements to the model of Zhang et al. (2003b) includes more realistic treatment of cuticle resistance, which is parameterized as functions of leaf wetness (dry vs. wet; dew vs. rain), relative humidity, leaf area index (LAI), friction velocity and landuse-specific reference values.R ac is also a function of the LAI, the friction velocity and the land-use specific reference value.Note that R g and R cut are calculated for SO 2 and O 3 and then scaled for other gases.R st is calculated using a sunlit/shade stomatal resistance sub-model (Zhang et al., 2002a).Thus, in this improved parameterization, R c depends on the type of canopy, the chemical species and the meteorological conditions.It is important to note that R c of HNO 3 is calculated in this model and constrained to a lower limit value of 10 s m −1 , although previous studies showed that dry deposition of HNO 3 is mostly controlled by aerodynamic resistances (Zhang et al., 2002a, and references therein).The parameterizations of all these sub-resistances, the land-use categories (LUC) and all related parameters as well as more details can be found in Zhang et al. (2003b).

Appendix B
The net bidirectional flux at a reference height above the canopy can be expressed as where X a and X c are the ambient concentrations at the reference height and at the canopy top, respectively.In the model, X c itself is expressed as a function of X a (Zhang et al., 2010).
The formulas of resistances are the same as in the original big-leaf model (Zhang et al., 2003b).Zhang et al. (2010) demonstrated that the differences in the fluxes ( F ) between the new bidirectional model (F t ) and the only dry deposition big-leaf model (F dry , Zhang et al., 2003b) can be estimated from the following equation: with X cp the canopy compensation point and V d the deposition velocity calculated from the original big-leaf model.X cp can be estimated from Eq. (B3): where x st and x g are stomatal and soil compensation points, respectively.x st is a function of stomatal emission potential ( st ) and the temperature of the leaf stomata, and then x g is a function of the ground emission potential ( g ) and the temperature of the ground surface.The compensation point increases exponentially with increasing temperature (Zhang et al., 2010).
In this study, as we use monthly measurement concentrations for a long period, we infer the NH 3 net fluxes (F t ) from the following equation: where F (> 0) is calculated monthly (from 3-hourly values) directly in the bidirectional model (Eq.B2) and F dry (< 0) calculated monthly by Eq. (A1).The canopy compensation point (X cp ) is the atmospheric NH 3 concentration for which the fluxes between the surface and the atmosphere change directions from emission (F t > 0) to deposition (F t < 0) (or vice versa).
More details on the parameterizations of this new bidirectional model can be found in Zhang et al. (2010).
since 1998, while measurement of O 3 started in 2001 and that of SO 2 in 2002.As part of the Long-term Observation Period of the AMMA (African Monsoon Multidisciplinary Analysis) program, the Djougou and Agoufou sites started operating in 2005.All measurements are still continuing at all the IDAF sites.

Fig. 1 .
Fig. 1.Vegetation and location map of the seven measurement stations of the IDAF network.Only the seven IDAF stations of west and central Africa included in the present study are represented. 38

Fig. 2 .
Fig. 2. Monthly variation of MODIS LAI averaged over the period 2000-2007 for IDAF sites.Vertical bars depict the standard deviation over the study period.

Fig. 2 .
Fig. 2. Monthly variation of MODIS LAI averaged over the period 2000-2007 for IDAF sites.Vertical bars depict the standard deviation over the study period. 39

Fig. 3 .
Fig. 3. Mean diurnal variation of the simulated NH3 canopy compensation point Xcp (NH3) for two scenarios (high, low), associated to the variation of the relative humidity (RH) at a Sahelian site (a, Banizoumbou) and a Guinean savanna site (b, Lamto ) in the dry season (January, 2006).

Fig. 3 .
Fig. 3. Mean diurnal variation of the simulated lower (low flux) and upper (high flux) NH 3 canopy compensation point X cp (NH 3 ) associated with the variation of the relative humidity (RH) at a Sahelian site ((a) Banizoumbou) and a Guinean savanna site ((b) Lamto) in the dry season (January, 2006).

Fig. 4 .
Fig.4.Monthly variation of dry deposition velocity (V d ) of NO 2 , NH 3 , HNO 3 , O 3 and SO 2 averaged over the period 2002-2007 for IDAF sites.Vertical bars depict standard deviation over the study period.

Fig. 4 .
Fig. 4. Monthly variation of dry deposition velocity (V d ) of NO 2 , NH 3 , HNO 3 , O 3 and SO 2 averaged over the period 2002-2007 for IDAF sites.Vertical bars depict standard deviation over the study period.
of the Sahelian zone (thus, low V d ).The low values of O 3 dry deposition in forests are correlated to low values of O 3 concentrations that were 2 to 3 times lower than those measured in the other ecosystems (Table

Table 1 .
Geographic, ecologic and climatic characteristics of the western and central Africa IDAF sites.WS: wet season; DS: dry season.

Table 3 .
Adaptation of land-use categories (LUC) used in the big-leaf model to IDAF sites for the calculation of canopy resistances (R ac , R cut , R st ) in V d and range of LAI and Z 0 (roughness length).

Table 4 .
Comparison of dry deposition velocities (cm s −1 ) for ozone, sulfur and nitrogen compounds between this study and previous studies.

Table 5 .
Mean seasonal and annual dry deposition fluxes and standard deviation (F dry ) of NO 2 and HNO 3 as well as NH 3 net bidirectional fluxes (F t , from high to low scenario estimates) at IDAF sites of west and central Africa over the period 1998-2007.Annual NH 3 dry deposition fluxes are added for comparison.Fluxes are expressed in kg N ha −1 yr −1 .