A regional climate model simulation over West Africa: parameterization tests and analysis of land-surface fields

Abstract

The West African Monsoon has been simulated with the regional climate model PROMES, coupled to the land-surface model ORCHIDEE
and nested in ECMWF analysis, within AMMA-EU project. Three different runs are presented to address the influence of changes
in two parameterizations (moist convection and radiation) on the simulated West African Monsoon. Another aim of the study
is to get an insight into the relationship of simulated precipitation and 2-m temperature with land-surface fluxes. To this
effect, data from the AMMA land-surface model intercomparison project (ALMIP) have been used. In ALMIP, offline simulations
have been made using the same land-surface model than in the coupled simulation presented here, which makes ALMIP data particularly
relevant for the present study, as it enables us to analyse the simulated soil and land-surface fields. The simulation of
the monsoon depends clearly on the two analysed parameterizations. The inclusion of shallow convection parametrization affects
the intensity of the simulated monsoon precipitation and modifies some dynamical aspects of the monsoon. The use of a fractional
cloud-cover parameterization and a more complex radiation scheme is important for better reproducing the amplitude of the
latitudinal displacement of the precipitation band. This is associated to an improved simulation of the surface temperature
field and the easterly jets. However, the parameterization changes do not affect the timing of the main rainy and break periods
of the monsoon. A better representation of downward solar radiation is associated with a smaller bias in the surface heat
fluxes. The comparison with ALMIP land-surface and soil fields shows that precipitation and temperature biases in the regional
climate model simulation are associated to certain biases in land-surface fluxes. The biases in soil moisture seem to be driven
by atmospheric biases as they are strongly affected by the parameterization changes in atmospheric processes.

KeywordsRegional climate model-Land-surface model-West African Monsoon

Full text

Page 1

A regional climate model simulation over West Africa:
parameterization tests and analysis of land-surface fields
M. Domı ´nguez•M. A. Gaertner•P. de Rosnay•
T. Losada
Received: 13 November 2008/Accepted: 12 February 2010/Published online: 6 March 2010
? Springer-Verlag 2010
Abstract
with the regional climate model PROMES, coupled to the
land-surface model ORCHIDEE and nested in ECMWF
analysis, within AMMA-EU project. Three different runs
are presented to address the influence of changes in two
parameterizations (moist convection and radiation) on the
simulated West African Monsoon. Another aim of the
study is to get an insight into the relationship of simulated
precipitation and 2-m temperature with land-surface fluxes.
To this effect, data from the AMMA land-surface model
The West African Monsoon has been simulated
intercomparison project (ALMIP) have been used. In
ALMIP, offline simulations have been made using the
same land-surface model than in the coupled simulation
presented here, which makes ALMIP data particularly
relevant for the present study, as it enables us to analyse the
simulated soil and land-surface fields. The simulation of
the monsoon depends clearly on the two analysed para-
meterizations. The inclusion of shallow convection para-
metrization affects the intensity of the simulated monsoon
precipitation and modifies some dynamical aspects of the
monsoon. The use of a fractional cloud-cover parameteri-
zation and a more complex radiation scheme is important
for better reproducing the amplitude of the latitudinal
displacement of the precipitation band. This is associated
to an improved simulation of the surface temperature field
and the easterly jets. However, the parameterization
changes do not affect the timing of the main rainy and
break periods of the monsoon. A better representation of
downward solar radiation is associated with a smaller bias
in the surface heat fluxes. The comparison with ALMIP
land-surface and soil fields shows that precipitation and
temperature biases in the regional climate model simula-
tion are associated to certain biases in land-surface fluxes.
The biases in soil moisture seem to be driven by atmo-
spheric biases as they are strongly affected by the para-
meterization changes in atmospheric processes.
Keywords
Land-surface model ? West African Monsoon
Regional climate model ?
1 Introduction
The West African Monsoon (WAM) is a complex system,
where severalcomponentsof theclimaticsystem
This paper is a contribution to the special issue on West African
Climate, consisting of papers from the African multidisciplinary
monsoon analysis (AMMA) and West African monsoon modelling
and evaluation (WAMME) projects and coordinated by Y. Xue and
P. M. Ruti.
M. Domı ´nguez (&)
Environmental Sciences Institute, University of Castilla-La
Mancha, Toledo, Spain
e-mail: Marta.Dominguez@uclm.es
M. A. Gaertner
Environmental Sciences Faculty, University of Castilla-La
Mancha, Toledo, Spain
e-mail: Miguel.Gaertner@uclm.es
P. de Rosnay
Centre National de la Recherche Scientifique (CNRS/CESBIO),
Toulouse, France
e-mail: Patricia.Rosnay@ecmwf.int
P. de Rosnay
European Centre for Medium-Range Weather Forecasts
(ECMWF), Reading, UK
T. Losada
Physical Sciences Faculty, Complutense University of Madrid,
Madrid, Spain
e-mail: tldoval@fis.ucm.es
123
Clim Dyn (2010) 35:249–265
DOI 10.1007/s00382-010-0769-3

Page 2

(atmosphere,ocean,land-surface)interact,generatingahigh
temporal and spatial variability in the associated precipita-
tion. The drought that affected the Sahel during the 1970s
and 1980s, after several rainy decades (Nicholson et al.
2000), is a good example of such variability. The monsoon
precipitation varies also strongly on interannual and intra-
seasonal scales (Le Barbe ´ et al. 2002; Sultan et al. 2003).
This variability has been linked to different factors, like sea
surfacetemperatures(VizyandCook2001)andland-surface
feedbacks (e.g. Koster et al. 2004), in addition to possible
effects of global climate change (Paeth and Hense 2004).
Simulations of the West African Monsoon with coupled
global climate models (CGCMs) have significant problems
in reproducing adequately the West African Monsoon
(d’Orgeval et al. 2006). Climate change projections are
very different among the different CGCMs, that do not
agree even in the sign of the precipitation change over the
Sahel and Guinean Coast (Christensen et al. 2007).
Several regional climate models (RCMs) have been used
during the last years in order to analyse the monsoon and
try to improve its simulation (Vizy and Cook 2002; Galle ´e
et al. 2004; Paeth et al. 2005; Afiesimama et al. 2006).
They have been applied to analyse specific aspects of the
monsoon system, like the onset (Ramel et al. 2006;
Sijikumar et al. 2006), or the sensitivity to regional SSTs
(Messager et al. 2004). The RCMs have higher spatial
resolution than CGCMs, and use a prescribed SST evolu-
tion. Together with the lateral boundary conditions, this
limits the degrees of freedom in the model in comparison to
CGCMs, highlighting the role of land-surface-atmosphere
interactions and convective processes. The use of adequate
physical parameterizations is an important issue for
improving the simulation of the monsoon. Changes in the
moist convective and the cloud parameterizations, as well
as in the land-surface models, have been found to be par-
ticularly important in this respect (Messager et al. 2004;
Hourdin et al. 2006).
Land-surface processes are a major component of the
West African Monsoon system (Redelsperger et al. 2006,
Koster et al. 2004). Over Sahel, soil moisture conditions
have been shown to strongly affect the dynamics of the
mesoscale convective systems (Taylor et al. 2007).
Land-surface modelling is a suitable approach to address
and to quantify land-surface-atmosphere interactions from
local to regional and continental scales. Within the past few
years, several coordinated land-surface modelling activities
were conducted. Among them, the project for the inter-
comparison of land-surface parameterization schemes
(PILPS, Lettenmaier 2003; Henderson-Sellers et al. 1995)
was conducted in the framework of the global energy and
water experiment (GEWEX) as part of the global land
atmosphere system study (GLASS). Global estimates of
soil moisture and land-surface fluxes were provided within
the global soil wetness project-2 (GSWP-2) (Dirmeyer
et al. 2006). More recently, the African Monsoon multi-
disciplinary analysis (AMMA, Redelsperger et al. 2006)
Land-surface model intercomparison project (ALMIP) has
been focusing on the West African region (Boone et al.
2009, and http://www.cnrm.meteo.fr/amma-moana/amma_
surf/almip/index.html).
The simulations described in the next sections form part
of the European project AMMA. We have performed
parameterization tests with the regional climate model
PROMES to study the influence of the shallow convection
and radiation parameterizations on the simulated monsoon.
ALMIP forcing data (precipitation and 2-m temperature)
are used in this analysis. The simulations are studied also
from the point of view of land-surface fluxes and soil fields.
To this end, soil variables and land-surface fluxes from a
corresponding ALMIP offline simulation are used as ref-
erence values. Most of the analyses are done for the 3-year
period where ALMIP data are available (2004–2006), as
the main focus of the study is not a long period evaluation
of the model ability to simulate the monsoon, but to get an
insight into the relationship of simulated precipitation and
2-m temperature with land-surface and atmospheric
processes.
In the following section, the PROMES climate model is
described briefly, as well as the simulations setup and the
ALMIP data. Section 3 describes the changes in the
parameterizations and how they affect the monsoon simu-
lation. After this, land-surface and soil fields are analysed
in more detail. Section 4 summarizes the main findings.
2 Methodology and data
2.1 Description of PROMES model
We have used a climatic version of the model PROMES
(Sa ´nchez et al. 2004). PROMES has been developed
by modelizacio ´n para el medio ambiente y el clima
(MOMAC) research group at the Complutense University
of Madrid (UCM) and the University of Castilla-La
Mancha (UCLM). It is a hydrostatic model with sigma
levels as vertical coordinates and Lambert projection
for the horizontal coordinates. The spatial arrangement of
variables follows the so called Arakawa-C grid.
The needed lateral boundary conditions are updated
every 6 h from analysis or GCM data. An eight point con-
tour band is used to relax the model variables to the external
information, following Davies (1976). The vertical inter-
polation of the large scale variables to model levels follows
the method described in Gaertner and Castro (1996).
With respect to the version described in Sa ´nchez et al.
(2004), the physical parameterizations of PROMES have
250M. Domı ´nguez et al.: Regional climate model simulation
123

Page 3

been changed with the aim of better simulating tropical
climate processes like the West African Monsoon. The
resolved-scale cloud formation and its associated precipi-
tation processes are modelled now according to Hong et al.
(2004). Unlike the previous parameterization (Hsie et al.
1984), it includes ice microphysics processes and ice
crystals sedimentation.
The sub-grid scale convective clouds and their precipi-
tation are calculated following the Kain and Fritsch (1993)
scheme, but with some updates proposed by Kain (2004).
We have activated the shallow convection option, which
has turned out to be an important contribution for the
monsoon simulation. Hereby the formation of shallow
convection clouds is contemplated when the modelled
convective clouds don’t reach the minimal thickness nee-
ded for the deep convection. Its inclusion is important as
they modify the vertical profiles of moisture and
temperature.
The radiation parameterization has been modified too.
Previously, the absorption and scattering of short-wave
radiation by clouds was based on the method proposed by
Anthes et al. (1987), and long wave radiation processes
were parameterized according to Stephens (1978) and
Garand (1983). The shortwave and longwave radiation
processes are modelled now with the parameterizations
described in ECMWF (2004).
In the previous version, the cloud-radiation interaction
was modelled through the resolved-scale cloud water and
ice. In the case of convective clouds, its impact on radiation
was taken into account indirectly, by adding the convective
cloud water and ice to the resolved-scale variables. No
fractional cloud cover parameterization was applied. A
fractional cloud cover parameterization (Chaboureau and
Bechtold 2002, 2005) has been introduced in the new model
version and coupled to the new radiation module.
Regarding the turbulent vertical exchange of the prog-
nostic variables in the planetary boundary layer (PBL), it’s
modelled now following Cuxart et al. (2000), with turbu-
lent kinetic energy (TKE) as a prognostic variable (1.5-
order closure).
PROMES has been coupled to the land-surface model
organizing carbon and hydrology in dynamic ecosystems
(ORCHIDEE) (Krinner et al. 2005). ORCHIDEE is a
dynamic model of the terrestrial biosphere composed by
two existing modules, the surface-vegetation atmosphere
transfer (SVAT) model SECHIBA (De Rosnay and Polcher
1998) and the dynamical vegetation model LPJ (Sitch et al.
2003), together with one additional module developed
recently, the carbon cycle model Saclay Toulouse Orsay
model for the analysis of terrestrial ecosystems (STO-
MATE), which describes photosynthesis, carbon cycle and
phenology. SECHIBA represents the processes of water
and energy exchange between atmosphere and biosphere
that take place in short time scales, describing completely
the diurnal cycle. An older version of SECHIBA was
included previously in PROMES. Several improvements
have been introduced with respect to that version. Among
them, 13 different plant functional types are now used
instead of 7 vegetation types in the previous version, and
the water uptake and transpiration are related now to the
Fig. 1 Domain and orography
(m) of PROMES simulations
M. Domı ´nguez et al.: Regional climate model simulation251
123

Page 4

distribution of the roots, allowing a more realistic repre-
sentation of the vegetation behaviour. The standard
hydrology of ORCHIDEE is described with the so-called
Choisnel model. It accounts for two soil moisture layers
(Ducoudre ´ et al. 1993, de Rosnay and Polcher 1998). A
physically-based soil hydrology has been developed, in
which the soil is multi-layered and the vertical soil mois-
ture dynamics is governed by the Richards equation (de
Rosnay et al. 2002; d’Orgeval et al. 2008). LPJ is the
parameterization that simulates the processes which define
the competition processes, covering a time scale of one
year. STOMATE describes the processes of phenology,
terrestrial carbon cycle and photosynthesis. Its time scale is
several days and it acts as a link between the relatively fast
hydrological processes represented in SECHIBA and the
slow vegetation processes described in LPJ. In the simu-
lations described here, only the SECHIBA module of
ORCHIDEE has been applied.
Fig. 2 Distribution of precipitation (first column mm day-1) and 2-m
temperature (second column ?C) average (June, July, August and
September 2004–2006): ALMIP reference fields (first row) and bias
of PROMES parameterization tests (second row NOSHAL-NOFR
run, third row SHAL-NOFR run, fourth row SHAL-FR run)
252M. Domı ´nguez et al.: Regional climate model simulation
123

Page 5

2.2 Simulations set-up
The simulation domain covers the area from 318W to 218E
and from 48S to 308N, approximately (Fig. 1). The domain
has been designed taking into account the ALMIP domain,
and is similar to the domains used in other regional climate
model studies over West Africa like Galle ´e et al. (2004) or
Messager et al. (2004). The topography of the domain has
been interpolated from GTOPO30 data (Verdin and
Greenlee, 1996). The horizontal domain includes 137 9 97
gridpoints with 40 km resolution. 36 levels are used in the
vertical, with higher resolution near the surface.
The boundary and initial conditions have been taken
from the analysis of European Center for Medium-Range
Fig. 3 Latitudinal cross section of atmospheric moisture, averaged over June, July, August and September 2004–2006: SHAL-NOFR simulation
(kg kg-1) and percentage change in NOSHAL-NOFR simulation (NOSHAL-NOFR–SHAL-NOFR, % difference)
Fig. 4 Distribution of net short-wave radiation (W m-2) during August 2004–2006: ALMIP reference fields and bias of NOSHAL-NOFR
simulation, SHAL-NOFR simulation and SHAL-FR simulation
M. Domı ´nguez et al.: Regional climate model simulation 253
123

Page 6

Weather Forecast (ECMWF) with a spatial resolution of
18 9 18.
The simulated period is from December 1999 to
November 2006. Two complete iterations of the year 1999
have been used as spin-up period to initialize the soil
moisture and temperature.
Several of the parameterization changes previously
described have been tested over this domain. Here we
present only results of the two changes affecting more the
simulation of the West African Monsoon: the activation of
shallow convection in the moist convection parameteriza-
tion and the introduction of the new parameterizations of
radiation and fractional cloud cover. The following three
simulations are analysed:
•
NOSHAL-NOFR simulation: shallow convection is
switched off and the older radiation parameterization
(with no fractional cloud cover) is used
SHAL-NOFR simulation: shallow convection is acti-
vated, but the older radiation parameterization (with no
fractional cloud cover) is used
•
•
SHAL-FR: both shallow convection and the new
radiation and fractional cloud cover parameterizations
are used
2.3 The AMMA land-surface model intercomparison
project
The main source of data to compare the simulation results
is ALMIP. In the recently completed phase-1 of ALMIP,
an ensemble of state-of-the-art LSMs (11 participants)
have been run offline (i.e. decoupled from an atmospheric
model) at a regional scale over West Africa for five annual
cycles (2002–2006) (Boone et al. 2009). All participating
LSMs used the following input forcing fields: precipitation,
short-wave and long-wave radiative fluxes, wind speed and
direction, 2-m air humidity and temperature and surface
pressure. Vegetation and soil parameters were provided by
the ECOCLIMAP data base (Masson et al. 2003). For each
LSM, two ALMIP experiments were conducted with dif-
ferent precipitation and radiative-flux forcing. In the
Fig. 5 Latitudinal cross section of zonal wind (m s-1), averaged over August 2004–2006: ECMWF analysis, NOSHAL-NOFR simulation,
SHAL-NOFR simulation and SHAL-FR simulation
254M. Domı ´nguez et al.: Regional climate model simulation
123

Page 7

control experiment (EXP1), the LSMs were forced with the
European Centre for Medium-Range Weather Forecast
(ECMWF) forecasts for 2002–2006. In the second experi-
ment (EXP2), ECMWF fields were hybridised with the
satellite based precipitation products obtained in estimation
des pluies par satellite-seconde ge ´ne ´ration (EPSAT-SG)
(Chopin et al. 2004) and the ocean and sea-ice-satellite
application facility (OSI-SAF) radiative fluxes (Boone and
de Rosnay 2007; Geiger et al. 2008). Boone and de Rosnay
(2007) have shown that the hybridised forcing data set used
in EXP2 is more realistic. In particular, the extension of the
African Monsoon to the north is better represented than in
the ECMWF model precipitation which underestimates
rainfall occurrence and intensity over the Sahel. The
higher quality EXP2 forcing data are available only for the
period 2004-2006, which is the reason for the use of this
3-year period for most of the comparisons with PROMES
results.
Both EXP1 and EXP2 were performed at a 0.5 degrees
resolution over the West African domain (from 58S to
208N and from 208W to 308E). ALMIP outputs have been
provided for each ALMIP LSM at a 3 h time step. They
include soil moisture and soil temperature profiles, runoff,
sensible and latent heat fluxes. More information on the
Fig. 6 Time–latitude diagrams
of precipitation (first column
mm day-1) and 2-m
temperature (second column ?C)
for 2004, averaged over
108W–108E: ALMIP reference
values (first row) and PROMES
parameterization tests
(NOSHAL-NOFR,
SHAL-NOFR and SHAL-FR
simulations)
M. Domı ´nguez et al.: Regional climate model simulation255
123

Page 8

ALMIP experiments design and input output can be found
on http://www.cnrm.meteo.fr/amma-moana/amma_surf/
almip/index.html.
ALMIP EXP2 forcings and outputs are considered as the
reference fields for comparison with PROMES output in
the present study. ALMIP fields are either satellite based
observations (like EPSAT-SG precipitation) or simulated
fields (model output) that are the best available approxi-
mation for observations (like the surface fluxes and soil
fields) as they have been generated using observed forc-
ings. The use of the same LSM (ORCHIDEE) in ALMIP
offline simulations and in PROMES coupled simulations
facilitates the comparison. In order to compare the
simulated vertical cross sections of zonal wind component,
the ECMWF analysis has been used.
3 Results
3.1 Parameterization tests: overall analysis
In this subsection an analysis of the overall impact of the
two parameterization changes on the simulation of the
West African Monsoon will be done. This analysis will be
done for the 3-year period 2004–2006, when the higher
quality forcing data used in ALMIP are available.
Fig. 7 Time–latitude diagrams
of precipitation (first column
mm day-1) and 2-m
temperature (second column ?C)
for 2005, averaged over 108W–
108E: ALMIP reference values
(first row) and PROMES
parameterization tests
(NOSHAL-NOFR, SHAL-
NOFR and SHAL-FR
simulations)
256 M. Domı ´nguez et al.: Regional climate model simulation
123

Page 9

In order to give a general idea of the quality of the
simulations for the average of the wet seasons of 2004–
2006, Fig. 2 shows the distribution of precipitation and the
2-m temperature from ALMIP forcings (average for June,
July, August and September 2004–2006), as well as the
biases of the different simulations with regard to these
reference fields.
NOSHAL-NOFR simulation underestimates precipita-
tion over most of the domain, including the Sahel. NO-
SHAL-NOFR shows a particularly strong precipitation
deficit over the two maxima seen in ALMIP precipitation:
one over the ocean, along the southwestern coast approx-
imately between 108W–158W and 68N–108N, and another
one at 88E, 58N (over the Cameroon highlands). Addi-
tionally, an extended negative bias of about 2 mm/day is
found over the northern Sahel area. This is linked to an
insufficient northward extension of the monsoon, as will be
shown more clearly later through an analysis of time-lati-
tude diagrams.
The activation of shallow convection (SHAL-NOFR
simulation) generates a precipitation increase particularly
south of 108N. In this case, the intensity of the south-
western maximum is overestimated. In contrast, the maxi-
mum over the Cameroon highlands is captured rather well
in location and intensity in this simulation. This is a
positive aspect of this simulation, as other regional climate
Fig. 8 Time–latitude diagrams
of precipitation (first column
mm day-1) and 2-m
temperature (second column ?C)
for 2006, averaged over 108W–
108E: ALMIP reference values
(first row) and PROMES
parameterization tests
(NOSHAL-NOFR, SHAL-
NOFR and SHAL-FR
simulations)
M. Domı ´nguez et al.: Regional climate model simulation257
123

Page 10

simulations (Ramel et al. 2006; Afiesimama et al. 2006)
tend to overestimate this eastern maximum. However,
SHAL-NOFR simulation doesn’t extend this maximum to
the ocean as observed. Other RCM simulations have shown
a similar deficit of precipitation over sea (Messager et al.
2004; Ramel et al. 2006).
The precipitation increase associated to the activation of
shallow convection indicates that this process plays an
important role in reproducing the intensity of monsoon
precipitation. Shallow convection increases the atmo-
spheric moisture, as can be seen in Fig. 3. The increased
atmospheric moisture can be explained by higher evapo-
ration in SHAL-NOFR simulation over the ocean, parti-
cularly south of about 98N, compared with NOSHAL-NOFR
simulation (figure not shown). A plausible mechanism for
this evaporation increase is known (Neggers et al. 2007):
shallow convection tends to transport moisture from the
PBL to the free troposphere, which increases surface
evaporation over the ocean due to the reduction of PBL
moisture. In the present case, this seems to increase the
total atmospheric moisture transported to the continent by
the monsoon flow, generating more precipitation over land.
The introduction of a fractional cloud cover para-
meterization and a new radiation module (SHAL-FR) has a
strong impact on precipitation. Precipitation over the Sahel
increases strongly, generating in this case a large positive
bias over this area. This is associated to a larger northward
extension of the monsoon. The overestimation of precipi-
tation over the southwestern coast is much larger than in
SHAL-NOFR simulation. Overall, NOSHAL-NOFR pre-
cipitation shows less spatial correlation with ALMIP pre-
cipitation values (0.54) than SHAL-NOFR (0.77) and
SHAL-FR (0.65).
Regarding the 2-m temperature (Fig. 2), NOSHAL-
NOFR and SHAL-NOFR simulations have a clear positive
bias, which is lower in the latter simulation. The maximum
positive bias is found over a band between 128N and 188N.
A comparison of the simulated net shortwave radiation at
Fig. 9 Monthly ‘monsoon parameters’ for SHAL-NOFR simulation
(black line), SHAL-FR simulation (red line) and ALMIP reference
data (green line) from March to October 2004, averaged over 20?W–
15?E and 0–20?N: precipitation intensity (mm/day), latitude of the
monsoon center (degree), longitude of the monsoon center (degree)
and width of the monsoon band (degree)
258 M. Domı ´nguez et al.: Regional climate model simulation
123

Page 11

the surface with ALMIP reference values (Fig. 4) reveals a
positive bias in these simulations over most of the domain,
which explains at least partly the extended temperature
bias. The positive temperature bias disappears when the
new parameterizations of solar radiation and fractional
cloud cover are used. A negative temperature bias appears
in SHAL-FR simulation over most of the Sahel, but it’s
clearly smaller in absolute value (about 1?C at most) than
the positive bias of more than 4?C found in the other two
simulations. This is associated to a clear improvement in
solar radiation at the surface in SHAL-FR simulation, as
can be seen in Fig. 4. A relatively small shortwave radia-
tion deficit is obtained in this simulation over the Sahel.
The value of the spatial correlation with ALMIP reference
values is high in all three runs, with some improvement
from NOSHAL-NOFR (0.94) to the other two simulations
(0.97).
Trying to analyse how well the different runs simulate
important dynamical features of the monsoon like the
African easterly jet (AEJ) and the Tropical easterly jet
(TEJ), the latitudinal cross section of zonal wind averaged
over August 2004–2006 is displayed in Fig. 5. The inten-
sity of the AEJ is captured well by all simulations. The
latitudinal location of the AEJ improves with the para-
meterization changes: from 108N in NOSHAL-NOFR to
128N in SHAL-NOFR and 148N in SHAL-FR. The latter
value agrees very well with the ECMWF analysis. The
improvement in the location of the AEJ in SHAL-FR is
associated to a better representation of the meridional
temperature gradient, which is displaced to the south in the
other two simulations (Figs. 6, 7, 8). Both the location and
the intensity of the TEJ improve with the new parameter-
izations, as shown by the remarkably good agreement of
the TEJ in SHAL-FR simulation compared with ECMWF
analysis. Regarding the low level westerly monsoon winds,
its northward extension is also improved in SHAL-FR, but
the maximum intensity of the westerly winds is too large.
These results indicate that a better representation of
Fig. 10 Monthly ‘monsoon parameters’ for SHAL-NOFR simulation
(black line), SHAL-FR simulation (red line) and ALMIP reference
data (green line) from March to October 2005, averaged over 20?W–
15?E and 0–20?N: precipitation intensity (mm/day), latitude of the
monsoon center (degree), longitude of the monsoon center (degree)
and width of the monsoon band (degree)
M. Domı ´nguez et al.: Regional climate model simulation259
123

Page 12

convective and radiative processes not only affects directly
the precipitation and temperature, but can also improve the
representation of monsoon dynamics.
Figures 6, 7, 8 show time-latitude diagrams of precipi-
tation and temperature for the years 2004, 2005 and 2006,
respectively. The time-latitude diagrams of precipitation
show that precipitation increases with the inclusion of the
new parameterizations. The maximum intensity values are
better represented in SHAL-NOFR, with the other two
simulations showing large biases of opposite sign. The
amplitude of the latitudinal displacement of the precipita-
tion band is underestimated in the simulations with the
previous radiation parameterization and no fractional cloud
cover. This aspect has improved in SHAL-FR run, which
represents better the southern location of the precipitation
band in spring and the northward shift of the monsoon
during the summer, though it extends the precipitation
excessively to the north.
In order to quantify how well the simulations reproduce
the precipitation band and its displacement, the diagnostics
proposed by d’Orgeval et al. (2006) have been calculated
for SHAL-NOFR and SHAL-FR simulations and compared
to ALMIP values. The average precipitation intensity, the
latitude and longitude of the monsoon center (calculated
trough a spatial average with a weight dependent on pre-
cipitation intensity) and the width of the precipitation band
(the latitudinal standard deviation of precipitation values)
have been calculated with a monthly frequency for the
period 2004–2006. For these calculations, precipitation
values above a threshold of 2 mm/day in the area between
20?W–15?E and 0–20?N have been used. These diagnostics
are shown in Figs. 9, 10, 11. The precipitation intensity is
overestimated for both simulations compared to ALMIP
data. The precipitation intensity values in SHAL-FR simu-
lation are clearly higher for all years than in SHAL-NOFR
and ALMIP. The latitude of the monsoon center is
Fig. 11 Monthly ‘monsoon parameters’ for SHAL-NOFR simulation
(black line), SHAL-FR simulation (red line) and ALMIP reference
data (green line) from March to October 2006, average over 20?W–
15?E and 0–20?N: precipitation intensity (mm/day), latitude of the
monsoon center (degree), longitude of the monsoon center (degree)
and width of the monsoon band (degree)
260 M. Domı ´nguez et al.: Regional climate model simulation
123

Page 13

overestimated by SHAL-FR run but the amplitude of the
latitudinal displacement of the precipitation band is better
captured than in SHAL-NOFR. The evolution of the lon-
gitude of the monsoon center in the simulations is quite
different to ALMIP. Both simulations locate the monsoon
center excessively to the west, which is probably due to the
fact that the positive bias at the western coast is much
larger than for other areas. In general, the width of the
precipitation band is better represented in SHAL-FR sim-
ulation than in SHAL-NOFR simulation.
Regarding 2-m temperature, Figs. 6, 7, 8 illustrate that
the simulations without the new radiation scheme (NO-
SHAL-NOFR and SHAL-NOFR) show an overall positive
bias during spring-summer, and a stronger meridional
temperature gradient, located more to the south, than in
ALMIP data. In these simulations, the temperature bias
develops during the spring and diminishes at the end of the
monsoon. In contrast, SHAL-FR simulation shows a posi-
tive bias over the northern Sahel in spring but represents
much better the temperature field in summer. The
improvement in the location of the summer temperature
gradient is associated to the already commented improve-
ments in the representation of the AEJ.
Some aspects of the observed intraseasonal precipitation
variability are reproduced in the simulations. When the
simulated contrast between rainy and break periods is
large, the timing of these periods is well captured in all
simulations. This can be better seen in the 2-m temperature
evolution, as rainy periods correspond to peaks of cooler
monsoon air penetrating into the Sahel. Good examples are
the cooler peaks at the end of July and at the end of August
in 2004, as well as the cooler peaks at the beginning of
May and at the beginning of June in 2005. SHAL-FR
simulates much better the amplitude of these cooler peaks
than the other two runs.
The coincidence in the timing of the main active and
break periods among the simulations is a remarkable result,
and suggests that this timing depends basically from factors
different than convective or radiative processes within the
model domain. A likely reason for this behaviour can be
the dependence of the timing of active convection periods
on larger scale features, which are transmitted through the
Fig. 12 Distribution of the biases (PROMES simulations–ALMIP) of SHAL-NOFR and SHAL-FR mean monthly latent heat flux (W m-2) (left)
and mean monthly sensible heat flux (W m-2) (right) in August (averaged over 2004–2006)
M. Domı ´nguez et al.: Regional climate model simulation261
123

Page 14

lateral boundary conditions and are therefore the same in
all three runs.
3.2 Parameterization tests: land-surface fluxes and soil
fields
We analyse now land-surface fluxes and soil fields in
SHAL-NOFR and SHAL-FR simulations, as the inclusion
of the fractional cloud cover and new radiation para-
meterizations produces the largest changes in these fields.
Taking the output from ALMIP offline simulation as
reference, we have analysed the errors in land-surface
fluxes and soil variables in our simulations. Figure 12
represents the mean monthly bias of latent and sensible
heat flux for August (2004–2006 average). In SHAL-
NOFR, the bias in latent heat flux shows a dipolar structure
(positive bias to the south, negative bias to the north). The
bias in sensible heat flux is mostly positive. Despite this
difference, a rather clear spatial correlation between both
fluxes can be observed. The behaviour over the southern
part of the domain (positive biases for both fluxes) can be
explained to a large extent by excessively high downward
solar radiation over that area. The improved representation
of solar radiation in SHAL-FR simulation is associated
with a strong change in the surface heat fluxes. The biases
of latent and sensible heat flux are much smaller in this
simulation, and their sign is now opposite over most of the
domain, as expected when the radiative fluxes are better
represented. The latent heat flux in SHAL-FR shows a
positive bias over most of the domain, in accordance with
the sign of the precipitation bias.
In order to analyse better the relationship of simulated
precipitation and 2-m temperature with the land-surface
fluxes, the variation with latitude of the respective biases is
shown in Figs. 13, 14. In Fig. 13, the precipitation bias is
plotted together with the evapotranspiration and soil
moisture biases, for the months of May and August
(averaged over 2004–2006 and longitudinally over 10?W–
10?E). In SHAL-NOFR simulation, there is a positive
correlation between evapotranspiration bias and precipita-
tion bias, but soil moisture and precipitation biases show an
opposite sign for the latitudes with more rain. This rela-
tionship changes completely with the new radiation
parameterization. In SHAL-FR simulation, the soil mois-
ture bias follows closely the precipitation bias, whereas the
evapotranspiration bias in August does not change with
Fig. 13 Latitudinal variation of monthly precipitation bias (mm)
(black line), monthly evapotranspiration bias (W m-2) (red line) and
monthly soil moisture bias (kg m-2) (blue line) for SHAL-NOFR
simulation (upper panels) and SHAL-FR simulation (lower panels),
for May and August (averaged over 2004–2006 and over 108W–108E)
262 M. Domı ´nguez et al.: Regional climate model simulation
123

Page 15

latitude. This strong change in the relationships may be
linked to the change in radiation biases. In SHAL-NOFR
the downward solar radiation bias is positive, which should
enhance evapotranspiration at the expense of soil moisture
storage. In contrast, the negative downward solar radiation
bias in SHAL-FR will cause the opposite effect. It seems
that the soil moisture biases depend on other biases and
don’t play a major role in explaining precipitation biases.
Figure 14 shows the 2-m temperature bias together with
downward solar radiation and sensible heat flux biases.
While the positive sign of the solar radiation bias can
explain in part the positive sign of the 2-m temperature bias
in SHAL-NOFR, the magnitude of both is clearly uncor-
related. The relationship between both biases is even less
clear in SHAL-FR, particularly in August when an almost
zero temperature bias coincides with a negative solar
radiation bias. In contrast, the 2-m temperature and sensi-
ble heat flux biases follow rather closely each other.
4 Conclusions
In this paper we have analysed the effect of different
parameterizations on the simulation of the West African
Monsoon with a regional climate model and the devia-
tions of modelled land-surface fluxes and soil fields from
reference fields taken from corresponding offline simu-
lations in ALMIP experiment. Results have been pre-
sented for two parameterization changes (activation of
shallow convection and change of radiation parameteri-
zation including a fractional cloud cover parameteriza-
tion) which have a clear impact on the monsoon
simulation. Three simulations have been analysed: NO-
SHAL-NOFR (without shallow convection and with the
older radiation parameterization), SHAL-NOFR (where
shallow convection is activated) and SHAL-FR (where
additionally the new radiation and fractional cloud cover
parameterizations are used).
Both NOSHAL-NOFR and SHAL-NOFR simulations
have a common behaviour in some aspects of the monsoon
simulation: they don’t simulate adequately the northward
extension of the monsoon, as the temperature gradient is
displaced to the south (linked to a southward displacement
of the AEJ) and the westerly monsoon flux doesn’t extend
enough to the north. There is a positive temperature bias,
associated to an excess of downward solar radiation. The
amplitude of the latitudinal displacement of the precipita-
tion band is underestimated in these simulations.
Fig. 14 Latitudinal variations of monthly down solar radiation bias
(W m-2) (black line), sensible heat flux bias (W m-2) (red line) and
2-m mean temperature bias (8C) (blue line) for SHAL-NOFR
simulation (upper panels) and SHAL-FR simulation (lower panels),
for May and August (averaged over 2004–2006 and over 108W–108E)
M. Domı ´nguez et al.: Regional climate model simulation263
123