Simulating the impact of varying vegetation on West African monsoon surface fluxes using a regional convection‐permitting model

Abstract This study assessed the sensitivity of the West African climate to varying vegetation fractions. The assessment of a such relationship is critical in understanding the interactions between land surface and atmosphere. Two sets of convection‐permitting simulations from the UK Met Office Unified Model at 12 km horizontal resolution covering the monsoon period May–September (MJJAS) were used, one with fixed vegetation fraction (MF‐V) and the other with time‐varying vegetation fraction (MV‐V). Vegetation fractions are based on MODIS retrievals between May and September. We focused on three climatic zones over West Africa: Guinea Coast, Sudanian Sahel, and the Sahel while investigating heat fluxes, temperature, and evapotranspiration. Results reveal that latent heat fluxes are the most strongly affected by vegetation fraction over the Sahelian and Sudanian regions while sensible heat fluxes are more impacted over the Guinea Coast and Sudanian Sahel. Also, in MV‐V simulation there is an increase in evapotranspiration mainly over the Sahel and some specific areas in Guinea Coast from June to September. Moreover, it is noticed that high near‐surface temperature is associated with a weak vegetation fraction, especially during May and June. Finally, varying vegetation seems to improve the simulation of surface energy fluxes and in turn impact on climate parameters. This suggests that climate modelers should prioritize the use of varying vegetation options to improve the representation of the West African climate system.


| 135
BAMBA et al. evapotranspiration, soil moisture content, radiation flux partitioning, and aerodynamic roughness. These factors can in turn modify the weather and climate by interacting with atmospheric processes such as mesoscale circulation, initiation and development of convection, cloud formation, and subsequent precipitation (Weiss et al., 2004).
In turn, the climate exerts a dominant control on the spatial distribution of the major vegetation types from local to global scales through variation in rainfall and temperature. However, the socioeconomic activities across West Africa are heavily dependent on rainfall-fed agriculture, and the population of the region is projected to have a strong increase during the twenty-first century. Thus, according to the United Nations, Department of Economic and Social Affairs, Population Division (2019), sub-Saharan Africa will account for most of the growth of the world's population over the coming decades. As discussed by Strengers et al. (2010), the land-use and land-cover change (LULCC) have been shown to be one of the most important drivers of changes in land surface properties in the past, and they are likely to trigger further changes in the future. Climate models have recently been used to investigate various processes such as land-atmospheric interactions (Bamba et al., 2019), climate dynamics (Kouadio et al., 2015), and aerosols radiative impacts (N'Datchoh et al., 2018). Furthermore, these climate models have been useful tools to understand the climate and its interaction.
Thus, Walters et al. (2019) argue that the exchange of fluxes between the land surface and the atmosphere is an important mechanism for heating and moistening the atmospheric boundary layer.
The different processes at the land-atmosphere interface are hardly reproduced by regional climate models (RCMs). According to Betts et al. (1997) and Bounoua et al. (2002) modeling, studies on global scales show that vegetated land interacts with the atmosphere to produce significant effects on regional climate. Also, Lu et al. (2001) find that the variability in vegetation phenology (timing of biological events) influences the regional climate through changes in surface moisture and energy balances.
In general, climate models are mainly limited by the use of prescribed annual vegetation which does not allow a realistic interaction between vegetated surfaces and atmospheric processes. For instance, the annual or biannual rains in Africa regularly transform the transitional regions between the desert and vegetated land from almost bare-soil pre-monsoon to lush green vegetation post-monsoon (Mougin et al., 2014). In common with many similar models, these variations are not generally represented in the UK Met Office Unified Model (UM), the climate model used in Future Climate For Africa (FCFA) and the focus of this study, resulting in large systematic biases in its surface fluxes (heat, moisture, momentum, and dust), and providing largely unknown likely important errors. The UM normally uses fixed vegetation fractions that cannot capture the abovementioned changes (Best et al., 2011). This is an important assumption that likely weakens the role of vegetation in the physical processes of landatmosphere interactions as the vegetation seasonality is generally not or only weakly taken into account. This can increase biases between observations and simulations, thus affecting the ability of a model setup to perform weather or climate prediction. Given the key role of the land surface and dust, the FCFA Improving Model Processes for African cLimAte (IMPALA) project includes improvements to dust and the land surface in the UM but did not plan to address this model deficiency. Since IMPALA was conceived, the UK's National Environment Research Council (NERC), Saharan West African Monsoon Multi-Scale Analysis project (SWAMMA) has demonstrated that this gap provides a fundamental error in dust modeling.
Also, efforts within IMPALA to improve the dust-generating winds via a haboob parametrization will not lead to the expected improvement unless it is addressed (Roberts et al., 2018). The land surface plays a key role in the West African monsoon, and so the seasonal variation in moisture fluxes introduced by seasonally varying vegetation fractions would also be expected to substantially affect regional climate (Nogherotto et al., 2013;Steiner et al., 2009).
Yet, most climate models typically assume a fixed annual vegetation fraction. This VegFlux project, therefore, addresses a key gap in FCFA, which, if not addressed, will limit model developments being made in IMPALA. It will therefore facilitate a better fundamental understanding of land-atmosphere interaction, a better understanding of limits to current projections and allow improvement of future projections. In light of all above, the present study assesses the impact of the vegetation seasonal variation in surface features such as surface winds and temperature, evaporative fraction, and energy at the continental scale in the UM so we investigate the benefits of introducing seasonally varying vegetation cover in climate models. This is realized through UM simulations broadly following the model setup used in the SWAMMA (Section 2.2) and the Seasonally varying vegetation impacts on surface fluxes (VegFlux) project simulations In Section 2, we describe the model setup and the experiments performed as well as the observations used to validate the model. Results on the variations in temperature, latent and sensible fluxes, and evapotranspiration fraction are presented and discussed in Section 3. The conclusion is given in Section 4.

| Model description
The UM has been widely used and improved throughout studies and research programs (Marsham et al., 2016(Marsham et al., , 2013. Detailed descriptions of the UM version 8.2 used for the SWAMMA project are provided in several research papers (Kealy et al., 2017;Roberts et al., 2018), and a summary is given below (Walters et al., 2019).  Walters et al., 2019) to simulate processes at the land surface and in the subsurface soil. JULES also uses a canopy radiation scheme to represent the penetration of light within the vegetation canopy and its subsequent impact on photosynthesis. Soil processes are represented using a four-layer scheme for the heat and water fluxes with hydraulic relationships. Sub-grid-scale heterogeneity of soil moisture is represented using the large-scale hydrology approach. A tile approach is used to represent sub-grid-scale heterogeneity, with the surface of each land point subdivided into five types of vegetation (broadleaf trees, needle-leaved trees, temperate C3 grass, tropical C4 grass, and shrubs) and four nonvegetated surface types (urban areas, inland water, bare soil, and land ice).

| Experiment design
Vegflux simulations are identical to the 12 km grid-spaced convection-permitting SWAMMA simulations apart from changes to the representation of vegetation type and fraction (described below). The simulations are run at a horizontal grid spacing of 12 km using convection-permitting (Klein et al., 2021)  Using a directional Splitting method-application to GEOstationary data) AOD were used and separately; the first one was performed using a fixed vegetation fraction referred to hereafter as MF-V.
For this simulation, the MODIS Leaf Area Index (LAI) was used to represent the vegetation fraction, which was averaged over May-September period. The second simulation was performed using a time-varying LAI from MODIS which we refer hereafter as MV-V. The daily varying LAI data cover the period from May 1 to September 30. A full description of the SWAMMA simulations is provided by Roberts et al. (2018). Thus, VegFlux is using the same configuration such as SWAMMA except the vegetation fraction treatment, which is invariant in SWAMMA and varies in VegFlux.

| Methods
Three subregions have been used in this work to account for the latitudinal variation of the regional climate, due to the annual migration of the Intertropical Convergence Zone (ITCZ), from tropical to semiarid climate zones across West Africa. The Sahel (SL) is located between 12° N and 15° N; the Sudanian zone (SD) is located between 8° N and 12° N; and, the Guinea coast zone (GC) is between 4° N and 8° N. All these subregions are located within 5° W-5° E (  The sensitivity of the UM for the changes in vegetation cover on EF is also assessed. EF is one of the most widely used methods to estimate the daily evapotranspiration (ET; Nutini et al., 2014). It is defined as the ratio between LH and the total heat leaving the Earth's surface. Daily ET is estimated as the product of the daily available energy estimated from SEVIRI/MSG data and the instantaneous evaporative fraction (ET frac ) estimated from Terra MODIS data.
Further detail on ET frac estimation can be found in Sun et al. (2012).
The ET frac is computed based on Equation (1) The significance of the changes induced by the variation of the vegetation fraction is evaluated through the comparison between the monthly mean of MV-V and MV-F (two independent samples). A ttest with a 10% significance level is applied to MV-V and M-VF simulations in which the sample sizes are 5 months. This seasonal trend in the vegetation fraction is consistent with the observations of Tucker et al. (1985), who noted a strong correlation between the rainfall seasonality and the dynamics of the vegetation cover, using the spectral vegetation index for different climatic zones over West Africa. Thus, the comparison of the two vegetation states reveals two main phases in vegetation variation. The first phase, from May to July, is characterized by a rapid increase in the vegetation fraction up to the value of the seasonal (MJJAS) average (0.133) as shown in Figure 2f. The second phase is characterized by an increasing vegetation fraction above the seasonal average, from July to August, where it reaches the maximum values. The most important vegetation fraction seasonal variation is recorded between 10° N and 15° N. This area represents the Sudanian Savannas and Sahelian region, a transition zone of semiarid grasslands, savannas, steppes, and thorn shrublands lying between the Sahara desert and the Sudanian Savannas (Huber & Fensholt, 2011). Previous studies focusing on the seasonal variability of the vegetation dynamics over West Africa showed that the evolution of NDVI in the Sahel region is closely related to rainfall seasonality (Anyamba & Tucker, 2005).

| Vegetation fraction
Moreover, there is also a good correlation between the rainfall variations and the normalized difference vegetation index (NDVI) at seasonal and interannual time scales for areas where mean annual rainfall ranges from approximately 200 to 1200 mm . This explains close links between vegetation seasonality and the displacement of the ITCZ (Anyamba et al., 2001).  To summarize, the variation is mainly affecting the LH compared with SH and this occurs in general over SL and SD regions.

|
Previous studies have shown that forests influence climate through the exchange of energy, water cycle, surface temperature, and SH increase with drought (Bonan, 2008). give weak values of surface LH flux in June and July, and then, the trend is inverted in August and September where the surface LH flux biases are 5 w m −2 and above 10 w m −2 , respectively. Also, over SL region the surface LH flux is higher than in SD and GC regions. So, the SL region seems to be more sensitive to surface LH flux during daytime contrary to GC regions. As we are moving into the deep monsoon from May to August, the rainfall increases therefore more moisture is accumulated in the soil during this period. Precipitation and soil moisture play a key role in LH availability and variability (Guo et al. (2011) and Song et al. (2009)). Finally, when the difference in vegetation fraction is high, the variation in LH flux is more pronounced and the daytime biases amplitude becomes more evident.
The variation of evapotranspiration can also affect the LH flux. Zheng and Eltahir (1998) have linked surface evaporation to surface water availability. Thus, the decrease in the surface water availability also reduces the surface evaporation. This trend can be explained by seasonal precipitation variability, which is accompanied by vegetation fraction growth. As the region is moving into the deep monsoon, the rainfall increases; therefore, more moisture is accumulated in the soil. Also, some of the observed bimodality could be caused by daily variation of phenology retrievals from seasonal variability (Vrieling et al., 2013). Contrary to the SD and SL regions, over GC region variations in surface SH flux seem to not follow any trend. This could be due to the fact that the GC region in general is permanently wet with annual rainfall amount above 1200 mm. This may have likely an impact on vegetation fraction, which seems to be stable with very less variation (Bamba et al., 2015). The surface SH flux is high during the dry period due to a dry ground surface, but decreases during the rainy period (Guo et al., 2011;Song et al., 2009).
When the surface LH flux increases during this period, the surface SH flux decreases due to increasing precipitation from monsoon. Also, the sensible heat flux is controlled in part by surface temperature. A wet surface from which water is evaporating is cooled down and so the sensible heat flux is suppressed. This is in agreement with Guo et al. (2011), who have shown that the SH flux is large during the dry ground period and decreases when the ground is wet. In contrast, LH flux increases with the wet ground but decreases with the dry ground period. Based on previous analysis, the two experiments have shown sometimes the anticorrelation relationship between surface LH and SH fluxes, which characterize these two patterns. Thus, these variations differ from one region to another. This important phenomenon may not be seen using only fixed vegetation within the model. However, the anticorrelation relationship/behavior of the LH and SH fluxes is not always fulfilled such as the case in May and August where the diurnal range of the two patterns evolves in the same direction. This could be due to the influence of other patterns such as cloudy sky, vegetation heterogeneity, and model resolution.

| Varying vegetation sensitivity to temperature
The difference between the monthly mean air temperature at 1.5 m, between the runs using MV-V and MF-V, is shown in Figure 6. The highest differences are observed over SD and SL regions contrary to GC region where this variation is the weakest. Based on the monthly variation, from May to July the temperature variation seems to be regular in terms of amplitude. Over all the climatic regions, these variations in air temperature increase from May to June and then decrease from June to July and then from July to September; variations in air temperature are weak even negative from August to September. As mentioned previously in Section 3.1 in MV-V The temperature over SL and SD regions is the more sensitive (between 0.12% and −0.10% of variation, respectively) to the vegetation variation compared with GC temperatures variation (0%-0.05%). In general, the air temperature decreases from May to September.
The spatial distribution of temperature variations averaged over MJJAS is shown in Figure 7. Some significant variations are noted over SD precisely in Cote d'Ivoire, Ghana, and Togo. As mentioned above, varying vegetation is slightly impacting the temperature variation at a rate of 0.6 to 1°C. The air temperature decreases when vegetation fraction increases which in turn will affect the evapotranspiration. variations between the three climatic zones can be explained vegetation fraction variation in the model. The vegetation type appears to reflect soil moisture availability and that water-use efficiency , indicating that these areas respond differently to multiple factors of local climate variability.
Based on the temperature seasonal cycle over West Africa which is bimodal, the monsoon period is characterized by two peaks with the MJJAS corresponding to the monsoon period. According to Afiesimama et al. (2006), the peak in temperature associated with the first mode occurs in May before the monsoon onset, and the peak of the second mode usually occurs in October as the monsoon retreats. . This is also in agreement with some previous studies which found that, during the rainy period, there was no long-term trend (almost constant) in the evaporative fraction (Gash et al., 1997). Therefore, MV-V has a slight impact on evaporation. As shown in Section 3.1, in terms of vegetation variation, the GC region is characterized by a marginal variation in vegetation fraction during this period.

| Varying vegetation sensitivity to evaporative fraction
Over SL and SD regions, the lack of significant impact of vegetation variation seems inconsistent based on the strong sensitivity of this region to rainfall variability and seasonal vegetation variation.
The primary factor determining the variability of the evaporation during the wet season was the rainfall pattern. Thus, the low impact of vegetation variation in rainfall could explain the behavior of evaporation. Surface evaporation rates display distinct ranges and spatial structures, which are related in various ways to the daytime rainfall (Guichard et al., 2010).

| CON CLUS ION
This study assessed the sensitivities of the UM to variations in vegeta- This trend is reversed in August-September when vegetation fraction is higher. The LH increases with vegetation fraction; this may not be seen when vegetation fraction is fixed in the climate models.
The analysis suggested an impact of the varying vegetation fraction on the temperature seasonal cycle. High temperature is associated with a lower vegetation fraction, especially during May and June. This is followed by a decrease/increase in temperature as the vegetation fraction increases/decreases. It observed some differential responses between climatic variables and vegetation. They may have major implications for energy balance, vegetation response to climate anomalies, and climate variation. We conclude, therefore, that the seasonal cycle of vegetation over West Africa should be a priority for inclusion in weather prediction and climate models. Therefore, for accurate assessment of the model, long-term data and the filtration of some phenomena should be taken into account.

CO N FLI C T O F I NTER E S T S TATEM ENT
The authors declare they have no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are freely available