Variability of West African monsoon patterns generated by a WRF multi-physics ensemble
- 2.6k Downloads
The credibility of regional climate simulations over West Africa stands and falls with the ability to reproduce the West African monsoon (WAM) whose precipitation plays a pivotal role for people’s livelihood. In this study, we simulate the WAM for the wet year 1999 with a 27-member multi-physics ensemble of the Weather Research and Forecasting (WRF) model. We investigate the inter-member differences in a process-based manner in order to extract generalizable information on the behavior of the tested cumulus (CU), microphysics (MP), and planetary boundary layer (PBL) schemes. Precipitation, temperature and atmospheric dynamics are analyzed in comparison to the Tropical Rainfall Measuring Mission (TRMM) rainfall estimates, the Global Precipitation Climatology Centre (GPCC) gridded gauge-analysis, the Global Historical Climatology Network (GHCN) gridded temperature product and the forcing data (ERA-Interim) to explore interdependencies of processes leading to a certain WAM regime. We find that MP and PBL schemes contribute most to the ensemble spread (147 mm month−1) for monsoon precipitation over the study region. Furthermore, PBL schemes have a strong influence on the movement of the WAM rainband because of their impact on the cloud fraction, that ranges from 8 to 20 % at 600 hPa during August. More low- and mid-level clouds result in less incoming radiation and a weaker monsoon. Ultimately, we identify the differing intensities of the moist Hadley-type meridional circulation that connects the monsoon winds to the Tropical Easterly Jet as the main source for inter-member differences. The ensemble spread of Sahel precipitation and associated dynamics for August 1999 is comparable to the observed inter-annual spread (1979–2010) between dry and wet years, emphasizing the strong potential impact of regional processes and the need for a careful selection of model parameterizations.
KeywordsWRF West African monsoon Multi-physics ensemble Precipitation Parameterization Tropical Easterly Jet African Easterly Jet
The West African monsoon (WAM) is the most prominent feature of the West African climate. It accounts for the majority of the annual precipitation and is therefore of paramount importance for the West African population that primarily relies on agriculture. The WAM is forced by differential heating of the ocean and the land surface which causes a seasonal change of the large-scale wind systems during the boreal summer and results in the migration of the inter-tropical convergence zone (ITCZ) over the West African continent. Relatively cool moist air from the Gulf of Guinea is advected onto the hot dry continent, where the resulting rainband travels from the Guinea Coast to the Sahel and back again, following the movement of the ITCZ. The WAM is generally characterized by a strong intra-annual as well as inter-annual variability especially in the Sahel region (Barbe 2002; Lebel and Ali 2009), driven by a complex and not yet fully understood interplay of several dynamical features and multi-scale factors influencing the WAM intensity and continental coverage (e.g. Sultan and Janicot 2003; Grist and Nicholson 2001; Nicholson and Webster 2008).
For a better understanding of the related processes, regional climate models (RCM) are useful tools in this data sparse region. Nikulin et al. (2012) confirm that RCMs can considerably enhance the representation of precipitation in comparison to their coarser forcing dataset (ERA-Interim) over West Africa. Recent studies demonstrate the ability of state-of-the-art RCMs to represent the WAM (Paeth et al. 2011; Druyan et al. 2010; Sylla et al. 2013; Nikulin et al. 2012). Sylla et al. (2013) conclude that RCMs are suitable for the investigation of the dynamical features and their relation to precipitation patterns, despite some uncertainties related to the generation of convective systems, surface temperatures and differing model sensitivities to the dynamical elements of the WAM. All studies emphasize the progress in recent years in improving the WAM representation, but identify large uncertainties related to model physics. They also point out that a multi-model ensemble approach can help to reduce uncertainties in the simulation of WAM characteristics and variability. Research initiatives like the Coordinated Regional Downscaling Experiment (CORDEX, Giorgi et al. 2009) Africa aim at satisfying the demand for standardized RCM simulations over the West African region with diverse models, which facilitates a joint analysis. However, ensemble means do not always outperform individual models, and computational or temporal constraints often render ensemble approaches unfeasible. An alternative solution therefore is to optimize a given atmospheric model for a particular application.
For this, we use the Weather Research and Forecasting (WRF, Skamarock et al. 2008) model, which is a widely used community model, applicable for numerical weather prediction and regional climate modeling. It incorporates a vast number of physical parameterizations that make it highly adaptable but challenging to set up at the same time. Physical parameterizations are subroutines used to describe physical processes on scales too small or too complex to be represented physically in the model. These parameterizations are usually a key source of model uncertainty. The performance of any physics scheme depends amongst others on the dominant atmospheric processes in the region of interest, the model resolution and the suitability of the parameterization for the particular problem. It is well-known that the results of the model can change considerably with the choice of the model physics. The WRF model therefore incorporates many “different atmospheric models”, which is why a large number of WRF sensitivity studies for many regions of the world can be found in the literature.
So far, only a small number of studies utilize WRF for the West African region. These studies demonstrate the skill of the WRF model and its predecessor, the Mesoscale Meteorology model 5 (MM5, Grell et al. 1994) in representing specific features of the West African climate (Vizy and Cook 2002; Bliefernicht et al. 2013; Hagos and Cook 2007; Sijikumar et al. 2006), for RCM-based climate projections (Jung and Kunstmann 2007; Vigaud et al. 2011), for the investigation of tropical storms triggered over the region (Vizy and Cook 2009; Druyan et al. 2009; Chiao and Jenkins 2010) and for evaporation tagging (Knoche and Kunstmann 2013). These studies either simply mention which model configurations were employed, or include a pragmatic testing of model physics to minimize the bias against observations. However, they rarely discuss uncertainties introduced by their choice of parameterizations. Flaounas et al. (2011a) conducted a first comprehensive study of the sensitivity of WRF for three cumulus and two planetary boundary layer parameterizations during the WAM 2006. They investigate the behavior of the tested schemes by analyzing their capability in representing surface variables and some dynamical monsoon features but they find no consistently best-performing configuration. Noble et al. (2014) compared African Easterly Wave occurrences of 64 WRF configurations with those of two reanalysis datasets and radiosonde observations for 12-days time slices over 10 years. They give valuable insights into the development of these atmospheric disturbances and reveal deficiencies of the model in reproducing them.
However, these two sensitivity studies conclude on very different suggestions for a “best” configuration, which nicely illustrates that any kind of evaluation is subjective, depends on the variables of interest, the focus region and the verification methods. Fersch and Kunstmann (2013) evaluated a large set of different WRF configurations including parameterizations, changing driving data and nudging techniques for four climatological regions of the world, including a domain encompassing West Africa. They concluded that the positive skill of a particular model configuration is often limited to the specific case-study, and that new model applications always require a thorough performance testing. The choice of a feasible configuration is furthermore highly dependent on the chosen (imperfect) reference datasets, which Sylla et al. (2013) identified as a key factor preventing an unambiguous model evaluation.
The goal of this study therefore is to use the advantages of an ensemble approach to generalize the process-based impact of individual parameterization schemes in order to complement existing studies. We use the WRF model in an RCM set-up to investigate the ability of a WRF multi-physics ensemble to represent certain WAM features. All members share the same boundary forcing from reanalysis data. The uncertainties in the representation of a single rainy season introduced by the different parameterizations, are used to explore interdependencies of processes leading to a certain WAM regime. Precipitation is especially sensitive to the model configuration, since it is the result of a complex interplay of parameterizations and therefore combines the uncertainties of the various physics schemes in the ensemble. Thus, we concentrate on the qualitative impact on precipitation and associated monsoon dynamics.
The ensemble members represent all possible combinations of three schemes per parameterization of cumulus convection (CU), microphysical cloud processes (MP) and planetary boundary layer mixing (PBL), totaling in 27 different CU_MP_PBL combinations. These parameterizations modulate the atmospheric moisture distribution and thus can be used to indirectly assess the impact of different regional moisture circulations on the WAM dynamics, while the large-scale forcing at the domain boundaries remains the same for all simulations. Analyzing all possible combinations of parameterizations rather than taking an iterative approach allows us to identify the impact of each scheme, since it reveals robust tendencies for changing parameterization partners. This study therefore gives insights into the sensitivity of the WAM system to local processes, as represented by the model physics schemes. It can further help to trace back bad model behavior to a certain process, and suggest which parameterization scheme to change in order to improve the WAM representation in the WRF model.
Section 2 describes the experimental model set-up and the observational datasets used for model comparison. In Sect. 3, we analyze the model precipitation and uncertainties introduced by the different parameterizations. Section 4 presents the variability of WAM dynamics as represented in the WRF ensemble. Section 5 is devoted to a discussion of our results followed by summary and conclusions in Sect. 6.
2 Experimental set-up and reference datasets
2.1 Model configuration
The simulations in this study are conducted with the WRF/Advanced Research WRF model, version 3.5.1. For our approach, the selected physics schemes should (1) directly be linked to moisture transport and moisture redistribution in the atmosphere, and (2) differ in complexity or methodology to represent a particular process.
Cumulus, microphysics and planetary boundary layer schemes used for the ensemble members
Grell and Freitas (2014)
Kain-Fritsch, convection trigger 2
Lin et al. (1983)
Thompson et al. (2008)
WRF Single Moment 3
Hong et al. (2004)
Planetary boundary layer schemes
Asymmetrical Convective Model V.2
Hong and Lim (2006)
For the CU, MP, and PBL groups, we combine parameterization schemes that follow different approaches to represent the same physical effects. The CU group includes a mass-flux type cloud model (KF), a sounding-adjustment type model (BMJ) and a mass-flux type model based on a stochastic approach, providing an ensemble mean (GF). For the KF scheme, the alternative trigger function (option 2) based on moisture advection was used instead of the default option because of reduced precipitation overestimations in preceding experiments.
The MP schemes used here differ in their classification of hydro-meteors. The WSM3 differentiates between three classes: cloud water/ice, rain/snow and vapor, depending on the temperatures being above or below freezing. LIN and TH take into account all six classes of hydro-meteors: cloud water, cloud ice, rain, snow, vapor and graupel. The more sophisticated TH scheme additionally predicts number concentrations for rain and ice species.
The PBL schemes can be divided into 1.5th order closure schemes based on prognostic turbulent kinetic energy (MYJ), and schemes which treat the turbulent mixing by a first order closure (YSU, ACM2). While the MYJ only considers local mixing into vertically adjacent grid cells, the YSU and ACM2 schemes consider non-local mixing through large convective eddies. In YSU, this is expressed by adding a counter-gradient term to non-local gradients of heat and momentum. ACM2 changes smoothly from local eddy diffusion in stable environments to combined local (downward fluxes) and non-local (upward fluxes) transport for heat, momentum and moisture components in unstable conditions. Further details about these schemes can be found in the literature listed in Table 1. All possible combinations of the schemes are included in the multi-physics ensemble, resulting in a total of twenty-seven members. Analyses of the ensemble or sub-ensembles always refer to the mean value.
Non-variable parameterizations that are shared by all simulations include the short-wave radiation scheme by Dudhia (1989), the Rapid Radiative Transfer Model (RRTM) long-wave radiation scheme (Mlawer et al. 1997), and the Noah land surface model (Chen and Dudhia 2001) with the 21-category MODIS land-use data.
For all simulations, the green vegetation fraction and albedo are taken from the monthly climatology. Lake temperatures are adjusted to correspond to the daily mean skin temperature of the surrounding area, rather than being interpolated from SSTs.
2.2 Simulation period
The rainy season of the wet year 1999 is simulated from March to September, including 1 month of spin-up time. This time span covers the two monsoonal phases as described by Thorncroft et al. (2011): the coastal phase from April to June (AMJ), and the continental phase from July to September (JAS). We choose a wet year to investigate the ensemble spread under a boundary forcing that favors moist conditions to ensure that the monsoon regime simulated by the RCM is not constrained by low incoming moisture fluxes and remote effects that dictate a weak WAM.
Model simulations are compared to satellite and observational data in order to evaluate the skills and physical plausibilities of the different model configurations. For precipitation, the NASA Tropical Rainfall Measuring Mission (TRMM) 0.25° resolution 3B42V7 (3-hourly, daily) and 3B43V7 (monthly) rainfall estimates (Huffman et al. 1995, 1997) and the Global Precipitation Climatology Centre (GPCC) 0.5° gridded gauge-analysis (Schneider et al. 2011) are used. The two datasets clearly show the wet regime at the Guinea Coast and over the Sahel with respect to the climatological mean from 1979–2010 (Fig. 2a, b).
Surface temperatures are compared to the Global Historical Climatology Network (GHCN) 5° monthly gridded temperature product Version 3 (Lawrimore et al. 2011). To address the question to which degree the regional model modifies the large-scale patterns, the ERA-Interim forcing data is taken as reference for the atmospheric dynamics and surface fluxes.
3.1 Seasonality of precipitation
According to Hagos and Cook (2007), the Saharan Heat Low (SHL) reaches its maximum by the end of June when it is positioned around 20°N, and when the Atlantic cold tongue in the Gulf of Guinea is established which induces a pressure gradient strong enough to trigger the “monsoon jump”. This term refers to the sudden relocation of the precipitation maximum from the coast to approximately 10°N and represents the monsoon onset in the Sahel. We define the date of the monsoon jump as the first occurrence of two consecutive days with rainfall amounts within the 0.9 percentile for the period May–July between 9 and 11°N. For TRMM, the monsoon jump takes place on 1st of July, as can be seen in Fig. 3 from the extension and subsequent relocation of the precipitation maximum from the coast to ~12°N. ENS is also capturing the monsoon jump, although 3 days earlier. Most ensemble members are able to capture the monsoon jump close to the observed date with a mean absolute deviation of 4.3 days and a maximum shift of 16–20 days for three of the members.
Intense precipitation events are better represented in ENS than in ERA-I, because of the higher horizontal resolution of WRF. The rainband is slightly shifted to the south over the whole rainy season in comparison to TRMM, but less than for ERA-I. The shift is especially pronounced in August, when the monsoonal rainfall is at its peak and TRMM shows precipitation throughout the whole month in the northern Sahel between 16–20°N, while ERA-I and ENS are not able to capture all of these events. This also applies to all individual ensemble members. The retreat of the rainband to the Guinea Coast sets in by mid-August for TRMM and ERA-I. This movement is delayed in ENS, which results in a too dry coast in the late summer. On the other hand, the observed dry period at the coast during August, i.e., before the retreat of the rainbelt, is not very well represented in ENS, since for some members the rainband remains too far south for the whole period. All ensemble members show a seasonal relocation of the rainband, but strongly differ in the extent, intensity and width, which will be discussed in the following sections.
3.2 Spatial distribution of precipitation
The rainband of ENS shown in Fig. 4a is mostly too narrow with excessive precipitation in its core zone and in Central Africa in comparison to TRMM. A persistent dry bias in the eastern part of the Gulf of Guinea is introduced, which stretches into the continent (Fig. 4b). It is most likely caused by too low SSTs in this area in the NCDC dataset. Further simulations using prescribed SSTs from ERA-I show a reduction of this dry bias in accordance with higher SSTs in that region (not shown).
The intensities of the rainbands in Fig. 4c cover the entire range from dry conditions (ACM2, WSM3) to wet conditions (TH, KF). None of the groups is able to capture the exceptional northward extension of the rainband in 1999, especially visible over Mali, which leads to a dry bias in the northern Sahel. Since this dry bias is also found for ERA-I, one might assume that the bias of the WRF simulations is caused by the bias of the driving data. However, in ERA-I the reason for the dry bias is a shift of the relatively broad rainband to the south, while in WRF it is the small North–South extent of the rainband. For GF and MYJ, the rainband is especially narrow with a daily precipitation of only 1–2 mm at the coast. ACM2 and WSM3 show very low precipitation intensities and induce an overall dry bias. BMJ shows closest rainfall amounts to TRMM, outperforming ENS with more precipitation at the coast and less overprediction in the Sudano-Sahel.
3.3 Parameterization influences on rainband intensity and position
The MP schemes show the same overall tendencies for the three regions, with a clear order (TH being the wettest and WSM3 being the driest). The mean precipitation difference between WSM3 and TH is 1.5 mm day−1 at the Guinea Coast, 2.4 mm day−1 in the Sudano-Sahel and 0.7 mm day−1 in the Sahel. Thus, the average intra-ensemble spread induced by the MP schemes is close to the magnitude of the bias to TRMM, which underlines the considerable impact of the MP schemes.
Looking at the PBL schemes, the picture is more diverse for the different regions: While MYJ is dry at the Guinea Coast and wet in the Sahel, ACM2 behaves the opposite way. This indicates a shift of the monsoon rainband, dependent on the choice of the PBL scheme. YSU shows an almost as strong northward shift as MYJ, but with much wetter conditions at the Guinea Coast due to a generally wider rainband as can be seen from Fig. 4. In terms of northward shift of the monsoon rainband, the order of the PBL schemes is ACM2 < YSU < MYJ. The interquartile ranges of the ACM2 and MYJ parameterization groups do not intersect for any of the three regions, which underlines the opposing impact they have on the position of the rainband. However, the strongest northward shift does not necessarily coincide with largest precipitation amounts: YSU instead of MYJ shows largest values in the Sudano-Sahel, where the core of the rainband is situated.
The CU schemes show the largest interquartile-spreads and on average only weak dry/wet tendencies with respect to ENS. They also show the smallest mean difference in precipitation of only 0.6 mm day−1 at the Guinea Coast, 0.9 mm day−1 in the Sudano-Sahel and 0.4 mm day−1 in the Sahel. This suggests an inferior role for the generation of precipitation in the model. However, this depends on the region: In the Sahel, KF (GF) shows a neutral (restrictive) behavior, but a restrictive (neutral) behavior at the Guinea Coast. BMJ dampens the effect of other schemes in both regions, indicated by the consistently smaller inter-quartile spread.
3.4 Convective and non-convective precipitation
The ability of an atmospheric model to simulate convective processes is crucial and at the same time a limiting factor for the quality of model precipitation in the tropics and sub-tropics. In the WRF model, precipitation from unresolved deep convection is generated by the CU scheme, while the MP scheme produces grid-scale precipitation in case the air is still super-saturated after the instabilities are released. Thus, the convective fraction of the model is artificial and related to the model resolution. Nevertheless, the partitioning into convective and non-convective precipitation helps to identify the impact of either scheme on the representation of convection.
The amount of non-convective precipitation follows the same order we found earlier for the MP schemes (WSM3 < LIN < TH), with a total spread of 93 mm month−1 exceeding the spread of convective precipitation (68 mm month−1). There is no clear correlation between the CU schemes and the amount of convective precipitation. Model configurations with KF or BMJ generate convective precipitation from 40 mm month−1 to 110 mm month−1, depending on the choice of MP and PBL scheme, while GF generates less convective precipitation and never exceeds 80 mm month−1. In particular, the ACM2 scheme leads to small amounts of convective precipitation for all CU schemes and results in a very dry regime. The impact of a particular CU scheme depends on the chosen PBL scheme and vice versa. For example, for the MYJ PBL scheme, maximum convective precipitation is achieved in combination with KF, while for the YSU and ACM2 PBL schemes, BMJ produces almost consistently the largest amounts of convective precipitation. The mean convective fraction over the whole domain of the individual ensemble members varies between 24 and 63 %, with consistently lower values for ACM2 configurations and higher values for BMJ and YSU configurations.
4 Parameterization influences on large-scale dynamics
The WAM precipitation is strongly tied to the characteristics of several dynamical ingredients (see Nicholson 2013 for a comprehensive summary). The differences between the parameterization groups in rainfall raise the question whether these can be related to changes in the dynamics and whether these changes correspond to mechanisms known to cause inter-annual monsoon variability.
4.1 The south-westerly monsoon wind
4.2 The tropical easterly jet
4.3 The African easterly jet
The mid-level AEJ is located at about 600 hPa for ERA-I and for the WRF simulations (Fig. 8). GF, KF and ACM2 are not able to capture the core wind speed, which exceeds 12 m s−1. This jet develops to adjust for thermal wind balance and as such moves with the position of the maximum surface temperature gradient between the monsoon rainband and the periphery of the SHL. This gradient is caused by the different thermal properties of moist/vegetated and desert land surface (Cook 1999). According to Nicholson (2009), the position of the AEJ is typically far to the south (north) for dry (wet) years when the monsoon is weak (strong). The dry and wet ensemble members of the WRF ensemble follow the same pattern: There is a significant correlation (r2 = 0.82, p ≤ 0.01) between the maximum strength of the monsoon winds and the position of the AEJ in Fig. 10e.
The classification by scheme in Fig. 10f reveals the same order as for the velocity of the TEJ. Schemes which favor extreme southward or northward displacements of the AEJ tend to dictate the strength of the monsoon flow, independent of the configuration partners, as indicated by the standard deviation of their respective parameterization group. Moderate schemes show a larger standard deviation and can be pushed to either side.
The impact of a parameterization scheme on the position of the AEJ comes from the displacement of the temperature gradient maximum, which is modified by the strength of the moisture advection from the ocean. The AEJ positions shown in Fig. 9c (depicted as points) correspond well to the regions of maximum temperature gradients for the member groups and ENS. The temperature gradient maximum seems to be shifted northward for ERA-I and is not in agreement with the AEJ position, which corresponds better to GHCN. In accordance to the dry bias in the Sahel (cf. Fig. 4), WRF generally exhibits a southward shift of the maximum temperature gradient compared to GHCN. Furthermore, parameterizations that show a weaker monsoon flow (ACM2, WSM3, BMJ) tend to have larger temperature gradients further to the South and correspondingly show the AEJ and the monsoon rainband further to the South, too. This is especially pronounced for the PBL group with clearly shifted temperature gradient maxima in accordance to their monsoon regime. Our findings are in agreement with Cornforth et al. (2009), who found that moist processes contribute to the meridional extent and intensity of the temperature gradient. We conclude that the position of the AEJ is a result of the northernmost extent of the rainband as described by Cook (1999), who attributes the maintenance of the jet to the negative meridional soil moisture gradient and the associated hydrodynamical response of the atmosphere. In her GCM experiments, the development of the AEJ was suppressed when a uniform soil moisture, corresponding to savanna conditions, was prescribed over the whole continent.
ERA-I exhibits two distinct tracks of AEW at ~20°N and at the Guinea Coast, with a maximum in the west of West Africa. Our WRF ensemble members show a different pattern: The main wave activity follows a single track and originates in the area of maximum rainfall in the eastern Sudano-Sahel. In comparison to ERA-I, the AEWs are overestimated over the Sahel and Sudano-Sahel by parameterizations that over-predict rainfall in the monsoon rainband with respect to TRMM (cf. Fig. 4). Sylla et al. (2013) also reported stronger AEW activity over the Sahel for several RCMs in comparison to ERA-I and attribute this to the internal representation of convection in the RCMs. Therefore, the different wave patterns we see for WRF with respect to ERA-I could be related to a minor importance of the energy transfer between AEJ and AEWs and a higher sensitivity to vertical velocities and associated convective processes.
Our results show that the examined parameterizations can be classified according to their impact on the modeled WAM regime. While the quantitative skill of a certain scheme in comparison to observations might change under different conditions (e.g. time period, driving data, domain size, chosen evaluation criteria), their individual qualitative impact on monsoon dynamics is assumed to be more universal. For overlapping analyses, the model internal tendency to produce more/ less precipitation or to enhance/ weaken the dynamic features with certain parameterizations are in agreement with Flaounas et al. (2011b), which gives us confidence in their robustness.
5.1 Microphysics schemes
Our ranking for the MP schemes is in line with the findings of Hong and Lim (2006), who found that the amount of precipitation is correlated with the complexity of the microphysics scheme. While they suggested that at resolutions of about 25 km, a simple ice-scheme should be sufficient to resolve the mesoscale features, we find that the representation of cloud processes has a strong impact and that the simple 3-class scheme WSM3 consistently leads to drier conditions in the model. During the monsoon season essentially all rainfall is associated with deep convection for which ice processes play a major role in the generation of precipitation. LIN and TH separately include cloud ice, snow and graupel, and TH additionally predicts the number concentration for cloud ice. For these two schemes, cloud particles may penetrate deeper above freezing level. According to Hong and Lim (2006), the conversion from clouds to rain is more efficient at producing precipitation than the ice phase alone in the case of WSM3. Furthermore, Hong et al. (2004) found that the interaction between ice clouds and long-wave radiation has a strong impact on the amount of precipitation because of enhanced radiative heating. In their case, precipitation was decreased with more cloud ice and vice versa.
Different from other WRF sensitivity studies that included microphysics schemes in other regions (e.g. Pohl et al. 2011; Crétat et al. 2012; Evans et al. 2012) we found a major importance of the chosen MP scheme for simulated precipitation amounts. For ENS, non-convective precipitation contributes around 40 % during the pre-monsoon phase and up to 60 % after the monsoon onset in July. During the WAM, MCS contribute most to the precipitation and the fraction of stratiform rainfall increases (Schumacher and Houze 2006), which WRF is able to partly resolve explicitly. This is confirmed by Marsham et al. (2013), who compare two model simulations at 12 km horizontal resolution with explicit and parameterized convection during the WAM. Because of the large fraction of organized convection, they report a better performance of the explicit simulation and relate their findings to a better representation of the diurnal cycle and the associated monsoon dynamics. Our results illustrate that even at a medium horizontal resolution (24 km), the impact of different MP formulations is non-negligible in this region and we expect it to increase with increasing horizontal model resolution.
5.2 Planetary boundary layer schemes
This ultimately attributes the largest spread in monsoon dynamics between the ensemble members to modifications of the incoming radiation, caused by the vertical moisture distribution in the PBL scheme. This is an important result, since the focus often remains on the energy transport via latent heat as main source of monsoon variability. This raises the question whether the inter-annual variation in cloudiness, and especially the amount and prevalence of low-level clouds, are key parameters for the surface energy budget and thus for the monsoon variability, as discussed by Knippertz et al. (2011). Cloud-radiation interactions remain one of the least understood processes and, together with the representation of clouds, are highly difficult to parameterize in atmospheric models with potentially devastating impact on the validity of modeled monsoon dynamics.
5.3 Cumulus schemes
The effects of the cumulus parameterizations are difficult to interpret since they are the result of a complex interplay of processes as illustrated in Fig. 6b. This applies especially to the ensemble approach of GF. However, the dampening effect for precipitation and monsoon dynamics of BMJ could be related to it being a sounding adjusting scheme, which will transform any atmospheric profile it starts from into a plausible, but pre-determined post-convection sounding. This might eliminate special and extreme characteristics of the vertical atmospheric structure.
5.4 Representation of monsoon dynamics
Precipitation (mm month−1)
TEJ (m s−1)
AEJ position (N)
This discrepancy in latent heat flux questions the correct representation of soil moisture, surface runoff and water conservation in the WRF model. As pointed out by several authors (e.g., Sylla et al. 2011; Taylor 2008; Steiner et al. 2009), this potentially has implications on convection and could be one reason for the persistent coastal dry bias. Potential improvements of WAM simulations through the implementation of high-resolution surface information and through coupled hydrological models are a current topic of discussion and should be considered as next steps.
6 Summary and conclusions
In this study, we employ a WRF physics ensemble to investigate the impact of parameterizations on the West African monsoon (WAM) for the rainy season 1999. We use ERA-Interim as forcing data and focus on parameterization schemes that affect the moisture distribution. Three different cumulus (CU), microphysics (MP) and planetary boundary layer (PBL) parameterizations are combined, resulting in an ensemble of twenty-seven members (cf. Table 1). The ensemble reveals strong parameterization related uncertainties but at the same time provides information on the sensitivity of the WAM system to the local dynamics and hence to the parameterization of the sub-grid processes. We analyze the effects of each parameterization group on precipitation and the representation of dynamical WAM features (monsoon wind, Tropical Easterly Jet, African Easterly Jet) and rank the parameterization schemes accordingly.
We find that the MP and PBL schemes introduce the largest spread in total precipitation over the study region. For the ensemble mean, non-convective precipitation generated by the MP schemes contributes 50–60 % of the total rainfall during the WAM, when mesoscale convective systems prevail. Larger amounts of precipitation are associated with more complex MP schemes, which alter atmospheric dynamics by the release of latent heat. Furthermore, we identify a strong influence of the choice of the PBL scheme on the position of the rainband, especially when comparing ACM2 (southward shift) and MYJ (northward shift). This is due to a larger (smaller) fraction of low- and mid-level clouds for ACM2 (MYJ) that weakens (strengthens) the monsoon because of less (more) incoming solar radiation. The choice of the CU scheme has minor influence on the total amount of precipitation over the study region, but alters the spatial distribution and thus the width of the rainband and the location of intense rainfall events. Moreover, the CU schemes have a strong impact on the representation of the diurnal cycle. We detect a complex and non-linear interaction between the CU and PBL schemes with respect to generating convective precipitation.
The differences between the ensemble members illustrate that the WRF model captures the characteristic interdependencies of monsoon dynamics and rainfall that were also found for years with differing monsoon regime. The spread of the ensemble in Sahel precipitation and associated dynamics during August 1999 is comparable to the observed inter-annual spread (1979–2010) in August between dry and wet years in spite of the same boundary forcing.
Our findings emphasize the strong potential impact of regional moist processes on the monsoon dynamics. They also underline the need for a careful selection of model parameterizations and justify the frequent ensemble applications in that region.
This work has been funded by the German Federal Ministry of Education and Research (BMBF) through the West African Science Service Center on Climate Change and Adapted Land Use (WASCAL). We would like to thank the German Climate Computing Center (DKRZ) for providing the computing facilities. We would also like to acknowledge the European Center for Medium-Range Weather Forecasts (ECMWF) for providing the ERA-Interim reanalysis and products, the German Weather Service (DWD) for the GPCC data, the NASA GSFC/DAAC for the TRMM products and the NOAA National Climatic Data Center for the GHCN product. We furthermore thank the international WRF community for the development and support of the WRF model. All analyses were done using IDL with the graphics based on the Coyote graphics library developed by David Fanning. Special thanks go to the two anonymous reviewers who greatly helped to improve our manuscript.
- Bliefernicht J, Kunstmann H, Hingerl L, et al. (2013) Field- and simulation experiments for investigating regional land-atmosphere interactions in West Africa: experimental setup and first results. Clim L Surf Chang Hydrol, Proceedings of H01, IAHS-IAPSO-IASPEI Assembly, IAHS Publication 359:226–232Google Scholar
- Giorgi F, Jones C, Asrar GR (2009) Addressing climate information needs at the regional level: the CORDEX framework. WMO Bull 58:175–183Google Scholar
- Grell GA, Dudhia J, Stauffer, DR (1994) A description of the fifth generation Penn State/NCAR mesoscale model (MM5). Technical report, National Centre for Atmospheric Research, Boulder, Colorado, USA. NCAR/TN-398 + STRGoogle Scholar
- Hong S-Y, Lim JJ (2006) The WRF single-moment 6-class microphysics scheme (WSM6). Korean Meteorol Soc 42:129–151Google Scholar
- Hsieh J-S, Cook KH (2005) Generation of African easterly wave disturbances: relationship to the African easterly jet. Am Meteorol Soc 133:1311–1327Google Scholar
- NCDC > NOAA National Climatic Data Center (2007). GHRSST Level 4 AVHRR_OI global blended sea surface temperature analysis. National oceanographic data center, NOAA. Dataset. [01.09.2014]Google Scholar
- Schneider U, Becker A, Finger P, Meyer-Christoffer A, Rudolf B, Ziese M (2011) GPCC full data reanalysis version 6.0 at 0.5: monthly land-surface precipitation from rain-gauges built on GTS-based and historic data. doi: 10.5676/DWD_GPCC/FD_M_V6_050
- Skamarock W, Klemp JB, Dudhia J, Gill D, Barker D, Duda M, Huang X, Wang W, Powers J (2008) A description of the advanced research WRF version 3. NCAR Technical Note, NCAR/TN-475 + STR. http://www.mmm.ucar.edu/wrf/users/docs/arw_v3.pdfAccessed 17 September 2014
- Sylla MB, Diallo I, Pal JS (2013) West African monsoon in state-of-the-science regional climate models, climate variability—regional and thematic Patterns. Dr. Aondover Tarhule (Ed.), ISBN: 978-953-51-1187-0, InTech, doi: 10.5772/55140
Open AccessThis article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution, and reproduction in any medium, provided the original author(s) and the source are credited.