The latest version of RegCM4 with CLM4.5 as a land surface scheme was used to
assess the performance and sensitivity of the simulated West African
climate system to different convection schemes. The sensitivity studies were
performed over the West African domain from November 2002 to December 2004 at
a spatial resolution of 50 km
Agriculture in West Africa relies mainly on rainfall and is strongly
dependent on the West African monsoon. Therefore, the onset, cessation and
the amount of expected precipitation associated with the West African
monsoon are of great importance for farmers, and accurate simulation and
prediction of rainfall and temperature are crucial for various sectors, such
as agriculture, energy and health, and for decision-makers. Rainfall over
West Africa is strongly related to the meridional migration of the
Intertropical Convergence Zone (ITCZ) and is modulated by successive
active and inactive phases of the monsoon system (Sultan et al., 2003a;
Janicot et al., 2011). After a quasi-stationary position around 5
General circulation models (GCMs) are unable to include the effects of regional features (Xue et al., 2010) due to their relatively coarse resolution. Regional climate models (RCMs) are relevant tools for this purpose since they allow for land surface heterogeneity and fine-scale forcing such as complex topography and vegetation variations (Paeth and al., 2006). Moreover, previous studies have shown that they are able to reasonably simulate the WAM climatology (Kamga and Buscarlet, 2006; Sylla et al., 2009) and its variability (Diallo et al., 2012). RCMs contribute to improving our knowledge of the interactions between atmospheric and surface factors affecting precipitation (Sylla et al., 2011; Browne and Sylla, 2012) and the influence of external forcing such as sea surface temperature (SST, Paeth and A. Hense, 2004), dust (Konare et al., 2008; N'Datchoh et al., 2017) and land-use changes on the dynamic of the monsoon system (Abiodun et al., 2012; Zaroug et al., 2012).
RegCM versions (Giorgi et al., 2012; Pal et al., 2007) are some of the most commonly used among the large range of RCMs to study the climate of West Africa and many regions of the world. Compared with the previous version (RegCM3; Pal et al., 2007), the latest release (RegCM4) has been improved with substantial development of the software code and the physical representations (Giorgi et al., 2012) and with the introduction of CLM (version 3.5 and 4.5) as an option to describe land surface processes. Previously, it was the Biosphere–Atmosphere Transfer Scheme (BATS; Dickinson et al., 1993) only that was used as a land surface model. Many studies have shown that the model performs well when using BATS over West Africa (Sylla et al., 2009; Diallo et al., 2013) but CLM offers improvements in the land–atmosphere exchanges of moisture and energy and in the associated surface climate feedbacks (Steiner et al., 2009). Nonetheless, it was shown over India that CLM use may lead to a weaker performance of RegCM than BATS (Halder and al., 2015). Thus, the performance of RegCM4 when using CLM (RegCM4-CLM4.5) needs to be assessed and sensitivities tests have to be conducted on physical process parameterizations to find the optimal configuration of the RCM for a given region and to give the relevant information to RCM users.
Among the different physical processes in climate models, convective parameterization is usually considered as the most important when simulating monsoon rainfall (Im et al., 2008; Leung et al., 2004). Simulations of regional climate are very sensitive to physical parameterization schemes, particularly over the tropics where convection plays a major role in monsoon dynamics (Singh et al., 2011; Srinivas et al., 2013; Gao et al., 2016). One of the main sources of uncertainties in climate prediction is related to the representation of clouds, which mainly influences the energy response of the models to a disturbance (Soden and Held, 2006; IPCC, 2007). Thus, implementing an appropriate convective scheme in dynamic models is needed for realistic simulations.
Several sensitivity studies using the previous version of RegCM have been conducted over Africa. Meinke et al. (2007) and Djiotang Tchotchou and Kamga Mkankam (2010) showed that in West Africa, monsoon precipitation is sensitive to the choice of cumulus parameterization and closure schemes. Brown and Sylla (2012) performed a sensitivity study of RegCM3 to the domain size over West Africa and showed that a large domain is required to capture the variability of summer monsoon rainfall and circulation features. A recent study by Adeniyi (2014) using version 4 of RegCM indicated that all convective schemes give good spatial representation of rainfall with biases over West Africa. Komkoua et al. (2016) found that the last release of RegCM implementing Grell as a convective scheme with the Arakawa–Schubert closure assumption is more suitable to downscale the diurnal cycle of rainfall over Central Africa. However, none of these studies have attempted to investigate a sensitivity study of the regional climate model RegCM4 to the convective scheme over West Africa with CLM4.5 as the land surface model.
This study investigates the performance of RegCM4-CLM4.5 over West Africa using different convection schemes to identify the “best” configuration option for the region. It is worth noting that the few previous similar sensitivity studies conducted in the region were performed using BATS as a land surface scheme and involved less convective schemes. The paper is structured as follows: the description of the model, data and numerical experiments used to investigate the RegCM4 performance are described in Sect. 2; Sect. 3 analyzes and discusses the model performance under different convection processes; and the main conclusions are summarized in Sect. 4.
The fourth generation of the ICTP RegCM (hereafter RegCM4) is used in this
study. RegCM is a limited-area model using a terrain-following
The soil–vegetation–atmosphere interaction processes are parameterized using Community Land Model (CLM version 4.5; Oleson et al., 2013). CLM4.5 presents in each grid cell the possibility to have 15 soil layers, up to five snow layers, five different land unit types and 16 different plant functional types (Lawrence et al., 2011; Wang et al., 2016). RegCM4-CLM4.5 proposes five different convective schemes (Im et al., 2008; Giorgi et al., 2012): the modified Kuo scheme (Anthes et al., 1987), the Tiedtke scheme (Tiedtke, 1989), the Emanuel scheme (Emanuel, 1991), the Grell scheme (Grell, 1993) and the Kain–Fritsch scheme (Kain-Fritsch, 1990; Kain, 2004) with the possibility to combine different schemes over ocean and land (called “mixed” convection).
The convective precipitation parameterizations used in this study are the Tiedtke (1989), Emanuel (1991) and Grell (1993) schemes.
The Emanuel (1991) scheme assumes that the mixing in clouds is highly episodic and inhomogeneous (in contrast to a continuous entraining plume) and takes into account convective fluxes based on an idealized model of sub-cloud-scale updrafts and downdrafts. Convection is triggered when the level of neutral buoyancy is greater than the cloud base level. Between these two levels, air is lifted and a fraction of the condensed moisture forms precipitation, while the remaining fraction forms the cloud. The cloud is supposed to mix with the air from the environment according to a uniform spectrum of mixtures that ascend or descend to their respective levels of neutral buoyancy. The mixing entrainment and detrainment rates depend on the vertical gradients of buoyancy in clouds. The Emanuel scheme includes a formulation of the auto-conversion of cloud water into precipitation inside cumulus clouds.
In the Grell (1993) scheme, deep convective clouds are represented by an updraft and a downdraft that are undiluted and mix with environmental air only in the cloud base and top. Heating and moistening profiles are derived from latent heat released or absorbed, linked with the updraft–downdraft fluxes and compensating motion (Martinez-Castro et al., 2006). Two types of Grell scheme convective closure assumptions can be found in RegCM4. In the Arakawa–Schubert (1974) closure (AS), a quasi-equilibrium condition is assumed between the generation of instability by grid-scale processes and the dissipation of instability by sub-grid (convective) processes. In the Fritsch–Chappell (FC) closure (Fritsch and Chappell, 1980), the available buoyant energy is dissipated during a specified convective time period (between 30 min and 1 h).
Similarly, the Tiedtke (1989) scheme is a mass flux convection scheme, although it considers a number of cloud types as well as cumulus downdrafts that can represent deep, mid-level and shallow convection (Singh et al., 2011; Bhatla et al., 2016). The closure assumptions for the deep and mid-level convection are maintained by large-scale moisture convergence, while the shallow convection is sustained by the supply of moisture derived from surface evaporation.
Five experiments using the convection schemes of (1) Emanuel over land and Grell over ocean (mix1), (2) Emanuel, (3) Grell, (4) Tiedtke, and (5) Grell over land and Emanuel over ocean (mix2) are conducted using RegCM4-CLM4.5 with 18 sigma levels at 50 km horizontal resolution for the period from November 2002 to September 2004. The first 2 months (i.e., November and December 2002) was considered as spin-up time and not included in the analysis. The years 2003 and 2004 have been selected in this study because they corresponded to a dry and wet year in this region, respectively. The analyses will focus on the rainy season from June to September (JJAS). As quantitative measurements of model skills, we consider mean bias (MB), which is the difference between the area-averaged value of the simulation and the observation, the spatial root mean square difference (RMSD), the spatial correlation called pattern correlation coefficient (PCC) and the distribution of the probability density function (PDF) of the temperature bias. The RMSD, PCC and the PDF provide information at the grid-point level, while the MB does so at the regional level. A Taylor diagram (Taylor, 2001) is used to summarize the assessments above and to show the deviation of different model configuration results from observations.
As assumed in Gao et al. (2016), the temperature bias in JJAS presents a
normal-mode type of distribution. The PDF is expressed as
The PDF is characterized by its bell-shaped curve, and the temperature biases
distribute symmetrically around the mean bias temperature value in
decreasing numbers as one moves away from the mean. The empirical rule
states that for a normal distribution, nearly all of the data will fall
within 3 standard deviations of the mean. The empirical rule can be
broken down into three parts.
68 % of grid points fall within the first standard deviation from the
mean. 95 % of grid points fall within 2 standard deviations from the mean. 99.7 % of grid points fall within 3 standard deviations from the mean.
The rule is also called the 68-95-99.7 rule or the three sigma rule. Thus,
they constitute measurements of model performance and systematic model
errors. These metrics are computed for each of the subregions indicated in
Fig. 1.
Topography of the West African domain. The analysis of the model result has an emphasis on the whole West African domain and the three subregions Guinea coast, central Sahel and west Sahel, which are marked with black boxes.
For this sensitivity study, the model was run at its standard configuration with 18 vertical sigma layers (model top at 50 hPa) and with initial and boundary conditions provided by the European Centre for Medium-Range Weather Forecasts reanalysis ERA-Interim (Simmons et al., 2007; Uppala et al., 2008) at a horizontal resolution of 50 km and a temporal resolution of 6 h (00:00, 06:00, 12:00 and 18:00 UTC). Sea surface temperatures (SSTs) were from NOAA optimal interpolation weekly SST data (Reynolds et al., 2007). The terrain characteristics (topography and land use data) were derived from United States Geological Survey (USGS) Global Land Cover Characterization (GLCC; Loveland et al., 2000) at 10 min horizontal resolution.
We focus our analysis on the precipitation and the air temperature at 2 m
in the summer of June–July–August–September (JJAS) over mainland West
Africa. To reduce uncertainty due to lack of surface climate observations
over the region (Nikulin et al., 2012; Sylla et al., 2013a), the simulated
precipitation is validated using two observational datasets: the GPCP
product (
Averaged 2003–2004 JJAS 2 m temperature (in
Pattern correlation coefficient (PCC) and root mean square difference (RMSD) for JJAS 2 m temperature for model simulations and observations (UDEL) with respect to CRU over the subregions Guinea coast, central Sahel, west Sahel and the West African domain during the period 2002–2003
The spatial distribution of averaged temperature during JJAS over 2003–2004 from CRU and UDEL observations (Fig. 2a, b) is compared to the temperature simulated by RegCM4 using the convection schemes Mix1, Emanuel, Grell, Tiedtke and Mix2 (Fig. 2c–g). Figure 3 shows the associated mean model biases with areas statistically significant at the 95 % confidence level (the dotted area denotes differences that are statistically significant at a significance level of 0.05) relative to CRU for observations (UDEL; Fig. 3a) and the model simulations (Fig. 3b–f). Table 1 reports the PCC and the RMSD between the simulated and observed temperature calculated for Guinea coast, central Sahel, west Sahel and the entire West African domain.
JJAS 2 m temperature bias (in
The CRU temperatures present a zonal distribution in West Africa with
a maximum (
PDF distributions (%) of temperature bias in JJAS over Guinea,
central Sahel, west Sahel and West Africa derived from the model
simulations using different convection schemes (land only; units:
For a quantitative evaluation of the performance of these sensitivity tests, the PDF statistical tool was used. The PDF distributions of the temperature bias in JJAS are shown in Fig. 4 for Guinea coast, central Sahel, west Sahel and the entire West African domain. The PDF distribution shows a generally dominant cold bias (see Fig. 4a–d) in model simulations over most of the study domain, except with the Tiedtke configuration in the central Sahel region.
Over the Guinea coast region, the Grell configuration presents a colder bias with
the maximum of temperature bias distribution centered around
Averaged 2003–2004 JJAS precipitation (in millimeters per day) over West
Africa from
JJAS precipitation bias (in %) over West Africa with respect
to GPCP from
The spatial distribution of mean JJAS precipitation (2003–2004) over West
Africa is shown in Fig. 5 for observations from GPCP and TRMM (Fig. 5a–b) and for
RegCM4 simulations with the following convective schemes: Mix1,
Emanuel, Grell, Tiedtke and Mix2 (Fig. 5c–g). Sylla et al. (2013a)
argued that over Africa, GPCP is more consistent with gauge-based
observations, while Nikulin et al. (2012) found a significant dry bias over
tropical Africa in TRMM compared to GPCP. We therefore select GPCP as our main observational reference for
precipitation in this paper.
Figure 6 shows the corresponding precipitation mean biases that are statistically
significant at the 95 % confidence level (the dotted area denotes
differences that are statistically significant at a significance level of
0.05) relative to GPCP for TRMM (Fig. 6a) and for the different simulation
configurations (Mix1, Emanuel, Grell, Tiedtke and Mix2; Fig. 6b–f,
respectively). GPCP depicts a zonal band of rainfall decreasing from north
to south (see Fig. 5a). Precipitation maxima are found in orographic regions
of the Guinea highlands, Jos Plateau and Cameroon mountains. Figure 6 shows
that the spatial distribution of the precipitation biases is statistically
significant at 0.05 levels over almost the entire domain studied. Differences
between TRMM and GPCP observation products (Table 2) can reach up to
Mean bias (MB) and the pattern correlation coefficient (PCC) for JJAS precipitation for model simulations and observations (TRMM) with respect to GPCP for the subregions Guinea coast, central Sahel and west Sahel as well as the West African domain. The PCC is calculated only for the West African domain during the period 2002–2003.
The
In order to understand the origins of the model rainfall biases, we analyzed
the JJAS mid-level (850–300 hPa) vertically integrated water vapor mixing
ratio and the 650 hPa low-level wind (African easterly jet, AEJ) over West
Africa averaged over the 2003–2004 period (Fig. 7). The AEJ is the most
prominent feature affecting the West African monsoon through its role in
organizing convection and precipitation over the region (Cook, 1999;
Diedhiou et al., 1999; Mohr and Thorncroft, 2006; Sylla et al., 2011). Areas
with larger water vapor mixing ratio correspond to the areas of maximum
precipitation in observations (see Fig. 5a–b). Around 9
Taylor diagram showing the pattern correlation and the standard deviation (normalized) for JJAS precipitation with respect to GPCP from Mix1, Emanuel, Grell, Tiedtke and Mix2 over Guinea coast, central Sahel, west Sahel and West Africa during the period 2002–2003.
The underestimation of vertically integrated water vapor mixing ratio is larger in the Grell and Mix2 simulations (Fig. 7c, e) over the Guinea coast and Atlantic Ocean compared to those of Mix1, Emanuel and Tiedtke (Fig. 7a, b, e). Mix1 and the Emanuel configurations better reproduce the spatial extent of the moisture convergence than the other model configurations (Fig. 7b, c). All model configurations simulate a stronger easterly wind flux (AEJ) than observed, in particular over the Guinea coast and Atlantic Ocean, inducing a negative impact on simulated precipitations in the subregions (see Fig. 5c–g). Another possible explanation for model rainfall biases is further discussed in Brown and Sylla (2011), whereby a sensitivity study on the domain size with RegCM3 over West Africa showed that RegCM3 simulates drier conditions over a default domain (RegCM-D1) quite similar to our domain size used in this study.
Hovmöller diagram of monthly precipitation (mm d
A Taylor diagram is used to give a combined synthesized view of the pattern correlation coefficient and the JJAS standard deviation of precipitation from the different sensitivity studies with respect to GPCP over Guinea coast, central Sahel, west Sahel and West Africa. Model standard deviations are normalized by the observed value from GPCP (indicated by REF; see Fig. 8). For the entire West African domain, the diagram shows that Tiedtke and Emanuel outperform the other configurations with values of normalized standard deviation much closer to 1. However, the Emanuel configuration present a better spatial correlation, reaching 0.8 compared to the Tiedke configuration. Over the Guinea coast subregion, Grell and Emanuel present better values of normalized standard deviation. However, regarding the spatial correlation, a value of about 0.7 as in the Emanuel configuration is the best. For west and central Sahel, Mix1 and Emanuel are closer to observations. However, Emanuel outperforms the Mix1 configuration with a good spatial correlation score between 0.7 and 0.8 over the central and west Sahel subregions. From the Taylor diagram, it can be inferred that Emanuel performs better regarding the normalized standard deviation and the pattern correlation over the entire West African domain and its subregions.
Based on previous experience and studies, Gao et al. (2016) noted that the use of the Emanuel convection scheme in RegCM3 and RegCM4 over China tends to simulate too much precipitation when using BATS as the land surface scheme. They explained that it is mainly due to the fact that the Emanuel scheme responds quite strongly to heating from the land surface compared to the Grell and Tiedtke convection schemes once convection is triggered. BATS with only two soil-level depths maximizes this response; this is why Emanuel is too wet when using BATS compared to Grell and Tiedtke. By contrast, CLM uses several soil layers down to a depth of several meters; therefore, the upper soil temperatures respond less strongly to the solar heating. The precipitation amount is much reduced when using CLM, which is good for Emanuel but not good for Grell and Tiedtke (Gao and al., 2016), while the combination of BATS with Grell and Tiedtke shows good performance (Gao et al., 2012; Ali et al., 2015).
In conclusion, although RegCM4-CLM4.5 shows some weaknesses, such as a dry bias over most of the central Sahel and Guinea coast region, its performance in replicating the spatial distribution of rainfall appears in line with that documented in previous studies using the previous version RegCM3 (Sylla et al., 2009; Abiodun et al., 2012).
In this section, we examine the effect of the convection scheme on the
characterization of the three distinct phases of the West African monsoon:
the onset, the high rain period and the southward retreat of the monsoon
rain band (Sultan et al., 2003). Such behavior is best represented by a
meridional cross section (time–latitude Hovmöller diagram). This diagram
provides a robust framework to assess RCM skill in simulating seasonal
and intraseasonal variations of the WAM and thus the mechanisms of the
region's rainfall (Hourdin et al., 2010). Figure 9 shows the time–latitude
diagrams of rainfall averaged over the region between 10
The Mix1, Emanuel, and Grell model configurations (Fig. 9c–e) capture the three phases of the seasonal evolution of the WAM, while the Tiedke and Mix2 simulations fail to reproduce them, in particular the rainy season over central Sahel. However, the Emanuel and Mix1 model configurations (Fig. 9c, d) overestimate rainfall amounts during the two rainy seasons over Guinea coast, mostly as a result of an overestimate of the precipitation over the orographic regions of Guinea highlands, Jos Plateau and Cameroon mountains. The Mix1 and Mix2 configurations are respectively wetter and drier compared to the other model configurations (Fig. 9c, g). Generally, the three monsoon phases are shown well by the Grell simulation, although it is drier compared to the other model simulations.
Annual cycle of monthly precipitation (mm d
Another analysis of the annual cycle consists of considering the area-averaged (land-only grid points) value of monthly rainfall and temperature over the Gulf of Guinea, central Sahel and the entire West African domain (Figs. 10, 11). This allows for better identification of rainfall and temperature minima and peaks. Figure 10a–d show the annual cycle of precipitation averaged over Guinea coast, central Sahel, west Sahel and the entire West African domain, respectively. Over the Guinea coast (Fig. 10a), both GPCP and TRMM observations show a primary maximum in June and a secondary one in September. The Mix1 and Tiedtke model configurations simulate an early first peak in May, while the Emanuel, Grell and Mix2 configurations capture the observed peak in June well. We note that all model configurations reproduce the timing of the midsummer break and second rainfall peak in September well, but they underestimate its magnitude, although the Mix1 simulation result is higher and much closer to observations compared to the other model simulations.
In both central Sahel and west Sahel, observations (GPCP and TRMM) display a dry spring (from January to March) and winter (from October to December) and a wet summer (from June to September) with a well-defined peak occurring in August. Model configurations reproduce both the phase of the annual cycle and the observed rainfall peak in August except the Emanuel configuration, which shifts it in September over the west Sahel region. Model simulations underestimate the peak intensity compare to observations. However, the Mix1 configuration rainfall peak is much closer to observations for both the central Sahel and west Sahel regions (Fig. 10b, d) compared to the other model simulations. Over the entire West African domain, the annual cycle (Fig. 10c) is smoother with a notable shift of the peak in September in the different model configurations. All the model configurations underestimate the rainfall peak and shift it in October. However, the Mix1 and Emanuel model simulations are much closer to the observed annual cycle of precipitation compared to the others. The Mix1 simulation, compared to the others, better reproduces the observed annual cycle of precipitation over the subregions and the entire West African domain.
Annual cycle of 2 m temperature (
The annual cycles of temperature for central Sahel, west Sahel and the entire West African domain of the Mix1, Emanuel, Grell, Tiedtke and Mix2 convection schemes are shown in Fig. 11b–d. The observations (CRU and UDEL) indicate a cooler winter from December to February and warmer pre- and post-monsoon periods with relative minima occurring during August, while over Guinea coast both winter and post-monsoon are cooler and only the pre-monsoon phase is warmer (Fig. 11a). Model configurations present a similar seasonal variation of the mean monthly temperature at 2 m compared to observations, but do exhibit some differences.
Vertical cross section of the JJAS mean zonal wind (in m s
Over Guinea coast model simulations underestimate the magnitude of the temperature compared to observations. However, the Tiedtke configuration is higher and much closer to observations compared to the other model simulations throughout the year (Fig. 11a). Over the central Sahel region, Grell and Tiedtke capture the seasonal variation from November to June well, in particular the first peak in August, compared to the other model simulations. During the summer (JJAS) Emanuel and Mix1 reproduce the observed precipitation annual cycle quite well (Fig. 11b). Therefore, model simulations underestimate the seasonal variation of temperature over the entire West African domain. Although the Tiedtke simulation overestimates the midsummer break period, it is much closer to the observed annual cycle of temperature throughout the year compared to the other model simulations. Over west Sahel, model simulations reproduce the annual cycle of temperature quite well except the Grell and Mix2 configurations, in particular during the summer (JJAS). In summary, the Tiedtke simulation better reproduces the observed annual cycle of temperature throughout the year over the subregions and the entire West African domain compared to the other model configurations.
The divergences in the RCM annual cycles arise mostly from their different abilities to simulate the main features responsible for triggering and maintaining the WAM precipitation (Gbobaniyi et al., 2013). Among them, we have the monsoon flow, the African easterly jet (AEJ), the tropical easterly jet (TEJ) and the African easterly waves (AEWs; Diedhiou et al., 1999; Sylla et al., 2013b).
Atmospheric circulations and their interactions with the ITCZ play an
important role in the WAM system (Nicholson, 2013). Thus, this section aims
to analyze the impact of the choice of convection scheme in simulations
of zonal wind features, including the near-surface westerly component (the
West African monsoon, WAM), the African easterly jet (AEJ) and the tropical
easterly jet (TEJ) in the middle and upper troposphere. Figure 12
depicts the vertical cross section of the JJAS mean of the zonal wind
averaged between 10
Model simulations Mix1, Emanuel, Grell and Tiedtke present a strong core of
monsoon flow compared to Era-Interim (reaching 6 m s
The latest release of RegCM4 has been applied over West Africa for 2 years (2002–2003) to assess its performance using five convective parameterizations: (a) the Emanuel scheme, (b) Emanuel over land and Grell over ocean (Mix1), (c) the Grell scheme, (d) the Tiedke scheme, and (e) Grell over land and Emanuel over ocean (Mix2). The sensitivity of the model to different convection schemes was validated using observations. The main findings and conclusions can be summarized as follows.
Compared with the previous version of RegCM, RegCM4-CLM also shows a general cold bias over West Africa. However, in the central Sahel region, the Tiedtke simulation presents a warm bias. This warms bias tends to displace the meridional temperature gradient southward relative to its observed position. An overall better performance with respect to temperature is obtained when using the Emanuel scheme.
With respect to precipitation, the dominant feature in model simulations is a dry bias, which is more pronounced when using the Tiedtke convection scheme. Considering the good performance over the entire West African domain and its subregions in the temperature and precipitation simulations, we suggest the Emanuel convection scheme when using RegCM4-CLM4.5 over West Africa.
Simulations using the Mix1 and Emanuel schemes reproduce the spatial
extent of moisture convergence of the ERA-Interim reanalyses well compared to the
other convection schemes. However, in the mid-levels of the atmosphere,
model simulations show an easterly wind flux (AEJ) stronger than observed, in
particular over the Guinea coast and Atlantic Ocean below the latitude
4
In the time evolution of simulations using the Grell convection scheme, rainfall matches well with the observed evolution, including the timing of the discontinuous northward jump of the main rainfall band in late June, although it is drier compared to the Mix1 and Emanuel convection scheme.
Over central Sahel and west Sahel, the mean annual cycle of precipitation and temperature, with the single-peak rainy season, is especially well captured in terms of timing despite the fact that all model simulations underestimated the magnitude. However, simulations using Mix1 better reproduce the annual cycle of precipitation compared to the other schemes.
Over Guinea coast, the Mix1 and Tiedtke model simulations failed to reproduce the double-peak rainy seasons, while the Emanuel, Grell and Mix2 simulations reproduce them well but underestimate their amplitude. The bimodal nature of rainfall associated with the Guinea subregion is not so well defined when averaging rainfall over the entire West African domain. This emphasizes the importance of separating regions into homogeneous precipitation subregions for evaluation analyses.
The mean annual cycle of temperature is reproduced well in simulation when using the Tiedtke convection scheme throughout the year over the subregions and the entire West African domain compared to the other model simulations.
As a more advanced package compared to the previous version of RegCM with BATS, CLM4.5 can be considered as the primary land surface process option in RegCM4. Therein, the use of the Emanuel scheme is recommended over the West African region. We plan to use this configuration in long-term multi-decadal simulations to further evaluate the model capability to reproduce the mean climatology. To make this study more complete, we will examine the sensitivity of temperature and precipitation extremes simulated by RegCM4-CLM4.5 to different convective schemes.
Data from these simulations are shared for free on request via email to arona.diedhiou@ird.fr.
BK conducted this research under the supervision of AD and NET. MBS, FG and SA were the advisors. Methodology and original draft preparation: BK, AD and and NET. Writing, review and editing: MBS, FG, SA, AB and ATK.
The authors declare that they have no conflict of interest.
This work is dedicated to the memory of Prof. Abdourahamane Konaré with whom we started this assessment. The authors thank the Institute of Research for Development (IRD, France) and the Institute of Geosciences for Environment (IGE, University Grenoble Alpes) for providing the facility (the Regional Climate Modelling Platform) to perform these simulations and the IT support funded by an IRD/PRPT contract at the University Felix Houphouet Boigny (Abidjan, Côte d'Ivoire). The authors are grateful to all students, technicians, engineers and researchers involved at ICTP (Abdus Salam International Centre of Theoretical Physics; Trieste, Italy) in the development and improvement of the regional climate model RegCM. The research leading to this publication is co-funded by the NERC/DFID “Future Climate for Africa” program under the AMMA-2050 project, grant number NE/M019969/1, and by IRD (Institut de Recherche pour le Développement; France) grant number UMR IGE imputation 252RA5. Edited by: Zhenghui Xie Reviewed by: two anonymous referees