Accepted Manuscript

Climate change projections for the Middle East - North Africa domain with COSMO-
CLM at different spatial resolutions

Edoardo Bucchignani, Paola Mercogliano, Hans-Jürgen Panitz, Myriam Montesarchio

PII: S1674-9278(17)30055-2
DOI: 10.1016/j.accre.2018.01.004
Reference: ACCRE 121

To appear in: Advances in Climate Change Research

Received Date: 8 May 2017

Revised Date: 17 January 2018
Accepted Date: 22 January 2018

Please cite this article as: Bucchignani, E., Mercogliano, P., Panitz, H.-J., Montesarchio, M., Climate
change projections for the Middle East - North Africa domain with COSMO-CLM at different spatial
resolutions, Advances in Climate Change Research (2018), doi: 10.1016/j.accre.2018.01.004.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.
 ACCEPTED MANUSCRIPT

Climate change projections for the Middle East–North

Africa domain with COSMO-CLM at different spatial
resolutions
Edoardo BUCCHIGNANI1,2*, Paola MERCOGLIANO1,2, Hans-Jürgen PANITZ3, Myriam
MONTESARCHIO1,2
1 Euro Mediterranean Center on Climate Change (CMCC) Via Maiorise, Capua (CE) 81043, Italy

PT
2 Italian Aerospace Research Center (CIRA) Via Maiorise, Capua (CE)81043, Italy.

3 Institute of Meteorology and Climate Research-Department Troposphere Research (IMK-TRO), Karlsruhe Institute of

RI
Technology (KIT), Karlsruhe 76131, Germany.

* Corresponding Author: Bucchignani, E., e.bucchignani@cira.it

SC
Abstract
In this study, projected changes in the future climate conditions for the Middle East-North Africa
domain over the 21st century have been investigated. Two simulations with the COSMO-CLM

U
model have been conducted respectively, at a spatial resolution of 0.44° and 0.22°, over the period
1979–2100, employing the IPCC RCP4.5 scenario. Initial and boundary conditions have been
AN
derived by the global model CMCC-CM. The main aim of the work is to provide regional climate
projections for this area, widening the range of the data already available, yet with higher
resolution, useful for many applications, especially as an input for the impact models. The two
M

different resolutions have been employed in order to quantify the differences due to the spatial scale
effects, comparing the results also with the driving global model. Climate projections show a
significant warming expected over the whole area considered at the end of the 21st century, along
D

with a reduction in precipitation, which is particularly evident over the western part of the domain.
Precipitation projections depend on the horizontal resolution, suggesting the need for additional
TE

simulations at higher resolution, especially for a proper representation of extreme weather events.
EP

1 Introduction

In recent years, the anthropogenic global warming of the arid regions received particular attention,
C

considering that water scarcity is a major challenge for food security. The Middle East and the
AC

adjacent Mediterranean have been identified as a hot spot for climate change (Giorgi and Lionello,
2008), with strong precipitation decrease and temperature increase projected for the near future. For
example, in 1989–2009, the northwestern Sahara experienced 40–50 heat wave days per year, while
there is a projected increase in this number of heat wave days over the 21st century (Russo et al.,
2016). The importance of assessing high-resolution climate projections over the 21st century is
universally recognized. In particular, climate projections over the CORDEX–MENA domain are
needed for the calculation of impacts on water resources in the region, and with regard to the
development of adaptation strategies. As CORDEX-MENA is one of the last domains that have
been defined in the frame of the CORDEX initiative (Giorgi et al., 2009), the number of literature
works available is still limited. Most of climate projections over this region are based on General
Circulation Models (GCMs), which are able to simulate large-scale climate features, but do not
reproduce some important aspects of local variability. As reported in the IPCC AR4 (Christensen et
 ACCEPTED MANUSCRIPT
al., 2007), according with the AR4 multi model ensemble of GCM projections under SRES A1B
scenario (IPCC, 2000) over Saharan sub-region, the median temperature increase lies between 3
and 4 , roughly 1.5 times the global mean response, while annual rainfall is likely to decrease in
much of Mediterranean Africa. According to the IPCC AR5 (Niang et al., 2014), the average air
temperature in Africa is projected to be beyond the band of variability of the 20th century by 2069
under Representative Concentration Pathway 4.5 (RCP4.5) (Moss et al., 2010) and already by 2047
under RCP8.5. Precipitation projections are more uncertain than temperature projections and exhibit
higher spatial and seasonal dependence than temperature projections. The CMIP5 ensemble (Taylor

PT
et al., 2012) projects very likely decreases in mean annual precipitation over the Mediterranean
region of northern Africa in the mid- and late 21st century periods for RCP8.5. Lelieveld et al.
(2016) investigated climate change and temperature extremes over MENA applying the ensemble

RI
results of CMIP5 models under RCP4.5 and RCP8.5 scenarios. They found out that the climate
warming is stronger in summer (associated with a thermal low), which is perceived as an extension
of the Persian trough, extending from South Asia to the eastern Mediterranean, and is projected to

SC
expand westwards.
To our knowledge, no regional climate projections over CORDEX-MENA domain have been
presented in literature works. Terink et al. (2013) performed a statistical downscaling over

U
CORDEX-MENA of an ensemble of nine GCMs from the CMIP3 under the SRES A1B emission
scenario, considering the future periods 2020–2030 and 2040–2050. They analyzed the changes in
AN
precipitation and evapotranspiration, finding that the annual precipitation sum will decrease for the
majority of countries, along with an average increase in evapotranspiration. Temperature increases,
combined with substantial decreases in precipitation and the resulting higher evapotranspiration
M

demand, will severely stress the water resources.

Existing regional projections over the regions included in this domain are only available through
D

other concentrated model efforts (e.g., Europe, Africa, and Asia). For example, Giorgi and Lionello
(2008) provided projections for the Mediterranean region based on the PRUDENCE simulations
TE

ensemble for the A1B emission scenario, highlighting a general increase of temperature in this area,
particularly in summer. Almazroui (2013) performed climate projections for Saudi Arabia using the
Regional Climate Model (RCM) PRECIS at 0.44° resolution, forced by the ECHAM5 GCM under
EP

the A1B emission scenario. He found out a warming tendency during the period 2041–2070 larger
than 0.4 per decade in the northern and central part of the country and a slow rising tendency in
the south. In the same period, rainfall is expected to increase mostly in the eastern and southeastern
C

areas, in good agreement with the IPCC precipitation projections for the Arabian Peninsula (IPCC,
2007). Laprise et al. (2013) performed climate projections for CORDEX-Africa domain, using the
AC

CRCM5 (Canadian Regional Climate Model), forced with two different GCMs. Both simulations
project a warming over the 21st century, while precipitation changes projected by the two
simulations differ significantly from each other. Dosio and Panitz (2016) analyzed the climate
projections obtained with COSMO-CLM at 0.44° resolution for the CORDEX-Africa domain,
according with RCP4.5 and RCP8.5 scenarios, comparing the results of the RCM with those of the
driving GCMs. They discovered that both COSMO-CLM and the driving GCMs project a large
temperature increase, while the projections are affected by a large inter-model and intra-model
variability. Buontempo et al. (2015) performed climate simulations over CORDEX-Africa using the
regional configuration of the Met Office Hadley Centre Climate model at 50 km resolution, using
the A1B emission scenario. They found that the spread in the climate anomaly provided by regional
simulations is generally narrower than in the driving GCM, highlighting the importance of local
effects in the analysis of the African climate. Lelieveld et al. (2012) analyzed climate changes over
 ACCEPTED MANUSCRIPT
the Eastern Mediterranean and the Middle East (EMME) domain, which is smaller than MENA
region and is actually completely included into it. They used both CRU data and projections over
the 21st century obtained with the RCM PRECIS, under the A1B emission scenario, indicating a
continual and gradual future warming.
Currently, several modelling groups are performing regional simulations over CORDEX-MENA
domain. In Bucchignani et al. (2016a; 2016b), COSMO-CLM has been proved to be able to
reproduce the general features of the climate of this complex area when adopting an optimized
model configuration. The effects of increasing spatial resolution were also investigated, comparing

PT
the results obtained respectively at 0.44° and 0.22° resolution. In the present work, climate
projections over the 21st century with COSMO-CLM for CORDEX-MENA have been analyzed,
according to RCP4.5 scenario. The main aim of this work is to enlarge the shortlist of projections

RI
already available in literature that is currently based only on GCM simulations, on RCM
simulations focused on different domains, or on statistical downscaling of GCM data. Results of the
two different spatial resolutions (0.44° and 0.22°) allow assessing the effects of increasing

SC
resolution on the quality of results. In fact, the added value of the higher-resolution is not always
obvious (Chan et al., 2013). Moreover, since climate extremes have a significant role and impact on
the environment and on economic activities, the present simulations have been analyzed as well in

U
order to provide climate projections of precipitation extreme events indicators. It is also important
to highlight that the scarcity of long term daily observational data over this area represents a strong
AN
limitation for the enhancement of climate studies on this area and, in particular for the correct
assessment of the model bias on the climate extreme and, consequently, for providing their
projections. Of course, the reliability of the projections can be improved by using multi-model
M

ensembles and weighting single elements with some measurements of model skills. This paper is
structured in 5 sections: in Section 2 a short description of the regional model and its setup is
provided, in Section 3 the validation of the GCM driven simulations is discussed, while in Section 4
D

future projections of temperature and precipitation are presented and discussed. Finally, the
conclusions are presented in Section 5.
TE

2 The regional model and setup

EP

The RCM COSMO-CLM model (Rockel et al., 2008) has been used adopting the same
configuration already described in Bucchignani et al. (2016a, 2016b). A third order Runge-Kutta
C

scheme has been used for time integration. The number of vertical levels in the atmosphere is equal
to 40 (top height of the model domain at 25 km) and the number of soil levels is equal to 7. The
AC

Tiedtke’s scheme (Tiedtke, 1989) is used for the convection parameterization. Moreover, the
parameterization of albedo derived from MODIS data (Lawrence and Chase, 2007) has been
adopted, in order to describe the Earth's surface reflectivity, and the NASA-GISS AOD
distributions (Tegen et al., 1996), since they provide a more accurate spatial and especially temporal
distribution of aerosols than the AOD distributions of Tanrè et al. (1984).
The model domain and its orography are shown in Fig. 1 (a). The physical domain, on which
analyses are performed, is identified with a black box. It was defined in the frame of the CORDEX
initiative, and it extends from 27°W to 76°E and from 7°S to 45°N. Two simulations have been
performed respectively at 0.44° and 0.22° spatial resolution; hereafter, these simulations are
referred to as MNA-44 and MNA-22. The high-resolution simulation MNA-22 is nested in the
lower resolution one. The computational domain for the 0.44° resolution (identified with a red box)
 ACCEPTED MANUSCRIPT
was discretized using a grid of 313×189 points while the computational domain for the 0.22°
resolution (identified with a blue box) was discretized using a grid with 524×296 points. Both
grids include a sponge zone of 15 grid points at each side, where the Davies boundary relaxation
scheme is used (Davies, 1983). The width of relaxation zone has been chosen through an analysis
aimed to exclude problems of reduced physical consistency in the zones close to the boundaries.
The time step has been set equal to 240 s for the 0.44° resolution and equal to 120 s for the 0.22°
resolution.
Simulations at both resolutions were performed over the period 1979–2100, but data related to 1979

PT
were removed from the analyses in order to reduce spin-up effects influenced by initial conditions.
The historical period 1979–2005 has been simulated according with the IPCC 20C3M protocol,
while the period 2006–2100 has been forced by the RCP4.5 scenario. Initial and boundary

RI
conditions are provided by the GCM CMCC-CM (Scoccimarro et al., 2011). CMCC-CM is a
coupled atmosphere–ocean general circulation model. Its atmospheric component is ECHAM5
(Roeckner et al., 2003) with a T159 horizontal resolution, while OPA 8.2 is the global ocean model

SC
with a 2° horizontal resolution, in the ORCA2 configuration. A model for the thermodynamics of
sea ice (Lovain-La-Neuve) is also included.

U
Analyses have been performed for average values of two-meter temperature (T2m) and total
precipitation, considering the same sub-regions already adopted in Bucchignani et al. (2016b) (Fig.
AN
1 b), which correspond to the main climates of the updated world map Köppen-Geiger climate
classification: equatorial (A), arid (B), warm temperate (C), snow (D) and polar climates (E).
Moreover, a subset of the standard ETCCDI indices (WMO, 2009) has been selected, in order to
M

evaluate the skill of COSMO-CLM to simulate extreme events and to assess future changes (Zollo
et al., 2016). The Extreme Weather Indices (EWIs) considered are: sdii, mean precipitation on wet
days (>1 mm); cdd, maximum number of consecutive dry days (< 1 mm); r10, number of days with
D

precipitation ≥ 10 mm; 90p, 90th percentile of daily precipitation. The sdii and 90p are measured in
mm d-1, while cdd and r10 in days per year.
TE
C EP
AC

Fig.1. (a) CORDEX-MENA domain (27°W–76° E, 7°S–45°N) with surface height (m). The red box
indicates the computational domain used to perform the simulation at 0.44°. The blue box indicates
the computational domain used to perform the simulation at 0.22°. The black box indicates the
 ACCEPTED MANUSCRIPT
physical domain (on which analysis are performed); (b) the five sub-regions for model evaluation,
chosen according to Köppen -Geiger evaluation.

3 Model evaluation

The capabilities of COSMO-CLM in reproducing the main climate features of the MENA domain
have already been assessed in Bucchignani et al. (2016b). Two ERA-Interim driven simulations

PT
performed at 0.44° and 0.22° (for the period 1980–2011) were analyzed and compared with respect
to a combination of available ground observations, satellite products and reanalysis. In the present
work, the GCM driven simulations at both resolutions (MNA-44 and MNA-22) have been

RI
evaluated over the same period, in terms of average properties, against CRU dataset (Mitchell and
Jones, 2005) in order to have the same reference dataset as many studies found elsewhere. Another
reason is that CRU values over this area are generally included in the range defined considering

SC
maximum and minimum values of other observational datasets (Bucchignani et al., 2016b). The
results of the CMCC-CM simulation are also included in the present analysis, in order to highlight
the effects of the increasing resolution over the quality of results. Moreover, precipitation EWIs

U
were also compared with Tropical Rainfall Measuring Mission (TRMM) dataset. TRMM provides
global daily precipitation estimates from a wide variety of meteorological satellites at 0.25° spatial
AN
resolution for the period 1998–2011 (Huffman et al., 2010).

Fig. 2 shows the seasonal mean T2m bias distribution ( ) against CRU of CMCC-CM, MNA-44
and MNA-22. In DJF, both COSMO-CLM simulations show a cold bias in the northern part of the
M

domain, which is clearly inherited by the driving GCM, while a warm bias is recorded over Congo
and Central African Republic. In MAM, for all the simulations the domain shows a dipole,
characterized by a cold bias in the northern part of the domain and a warm one in the southern part.
D

In JJA, the GCM exhibits a strong warm bias over central Africa, while for the RCM almost the
TE

whole domain is characterized by a positive bias, particularly evident over the Rub Al-Khali, which
is located in the south of Arabian Peninsula and has a hot desert climate. This finding confirms the
general tendency of climate models to show systematic biases in warm dry climates (Boberg and
Christensen, 2012). In SON, the GCM is affected by a large cold bias over North Africa and
EP

Europe, which is significantly reduced in the RCM simulations. Vice versa, COSMO-CLM exhibits
wide areas affected by a warm bias, while a small negative bias is recorded over North Africa,
Spain and the Gulf of Guinea. Table 1 contains the Root Mean Square Error (RMSE) and the spatial
C

correlation coefficient (CORR) of T2m compared to the CRU dataset for the sub-regions A, B, C,
D, E, related to CMCC-CM, MNA-44 and MNA-22. The high spatial correlation values confirm
AC

that both simulations (GCM and RCM) generally quite well reproduce the main characteristics of
T2m. Even if it is difficult to draw a general conclusion, a comparison between GCM, MNA-44 and
MNA-22 shows that, apart from few exceptions, the RCM is able to reduce the model bias with
respect to GCM. The resolution increase from 0.44° to 0.22° produces slight improvements in DJF
and MAM, while no significant differences are recorded in JJA and SON. This is probably related
to a not effective representation of the vegetation parameters (Hochman et al., 2017). It must be
taken into account that part of the model bias is related to observation inaccuracy especially over
Sahara and Rub Al-Khali deserts, which are areas with few CRU observation stations.
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
M
D
TE
C EP
AC

Fig.2. Seasonal bias of 2-m temperature ( ) of the GCM CMCC-CM and of COSMO-CLM
simulations (MNA-44 and MNA-22) over the period 1980–2011 with respect to CRU dataset.
Table 1. Seasonal values of Root Mean Square Error (RMSE) and spatial correlation coefficient
(CORR) for 2 m temperature with respect to CRU dataset over the period 1980–2011, averaged
over each of the five sub-domains considered. RMSE is expressed in .
Area CMCC-CM MNA-44
DJF MAM JJA SON DJF MAM JJA
RMSE CORR RMSE CORR RMSE CORR RMSE CORR RMSE CORR RMSE CORR RMSE CORR RMSE
A 1.4 0.8 1.2 0.9 1.8 0.9 1.0 0.9 1.5 0.7 1.8 0.9 1.9 0.8 1.1
B 3.5 1.0 2.4 0.9 2.4 0.8 2.3 0.9 3.1 1.0 1.5 1.0 2.7 0.9 1.6
 ACCEPTED MANUSCRIPT
C 2.4 1.0 2.5 0.9 3.0 0.7 2.5 0.9 1.9 1.0 1.9 0.9 2.2 0.8 1.5
D 4.1 0.6 3.2 0.7 3.0 0.8 3.1 0.7 3.1 0.7 3.1 0.8 2.6 0.8 2.4
E 8.0 -0.1 6.5 0.1 4.7 0.3 6.1 0.2 5.4 0.1 4.9 0.2 3.9 0.3 4.7

Fig. 3 shows the precipitation bias distribution against CRU for the four seasons. In DJF and MAM,
CMCC-CM provides a good agreement over almost the whole domain, while COSMO-CLM is
affected by underestimation over central Africa. In JJA, CMCC-CM precipitation is underestimated
over a wide area located in the central part of the domain. COSMO-CLM is not able to reduce this

PT
bias, but in any case, performances are better if compared with the ERA-Interim driven simulations,
especially over Central Africa. Here the strong overestimation documented in Bucchignani et al.
(2016b) was partially inherited by the forcing data. In SON, CMCC-CM provides good

RI
performances over the whole area, while RCM simulations are characterized by underestimation
over western Africa and the southern part of the domain. Table 2 contains the Root Mean Square
Error (RMSE) and the spatial correlation coefficient (CORR) of precipitation compared to the CRU

SC
dataset, for the sub-regions A, B, C, D, E defined in Section 2, related to CMCC-CM, MNA-44 and
MNA-22. Model results are better correlated to observations in DJF and JJA. Overall, it is evident
that COSMO-CLM (both resolutions) is able to reduce the bias of the driving GCM only in few

U
cases. Otherwise, the resolution increase from 0.44° to 0.22° does not produce improvements the
reason being that, as already explained in Bucchignani et al. (2016b), the mesoscale phenomena in
AN
this area are characterized by a higher resolution than those employed for the current simulations.
Moreover, the high resolution is not exploited by the moist convection scheme used here, which
probably works well with a cloud-resolving model. In fact, as documented in Cavicchia et al.
(2016), the enhancement in simulating precipitation is not only connected with the resolution, but
M

also with the model configuration, implying that MNA-22 needs a further tuning. The geographical
location of the domain, close to the mid-latitude and to the sub-tropical zones create additional
D

limitations to the usage of an RCM approach.

TE
C EP
AC
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
M
D
TE
C EP
AC

Fig.3. Seasonal bias of precipitation (mm per month) of the GCM CMCC-CM and of COSMO-
CLM simulations (MNA-44 and MNA-22) over the period 1980–2011 with respect to CRU dataset.
Table 2. Seasonal values of Root Mean Square Error (RMSE) and spatial correlation coefficient
(CORR) for precipitation with respect to CRU dataset over the period 1980–2011, averaged over
each of the five sub-domains considered. RMSE is expressed in mm per month for precipitation.

Area CMCC-CM MNA-44

DJF MAM JJA SON DJF MAM JJA
RMSE CORR RMSE CORR RMSE CORR RMSE CORR RMSE CORR RMSE CORR RMSE CORR RM
 ACCEPTED MANUSCRIPT
A 31.8 0.9 40.5 0.9 79.7 0.7 37.6 0.8 32.2 0.9 52.0 0.8 92.5 0.8 65.3
B 8.1 0.8 10.0 0.8 35.4 0.9 11.3 0.7 9.9 0.7 12.5 0.7 22.6 0.9 11.5
C 27.9 0.7 22.7 0.6 30.9 0.9 25.2 0.8 33.2 0.6 24.6 0.6 29.2 0.9 34.9
D 36.4 0.6 42.6 0.5 20.8 0.8 25.6 0.5 33.6 0.5 44.3 0.5 24.8 0.7 28.1
E 20.4 0.7 46.2 0.8 27.6 0.2 21.7 0.5 20.1 0.5 63.0 0.5 22.1 0.3 35.1

The high-resolution model is probably penalized in these verification measures because averaged
properties smooth out the extremes. In all cases, as already documented in similar works (e.g. Zollo

PT
et al., 2016; Bucchignani et al., 2017) the added value of the RCM simulations, especially at very
high resolutions, lies in their capabilities in simulating extreme weather events, especially those
connected with intense (or null) precipitation. Fig. 4 shows the spatial pattern of the bias (%)

RI
(MNA-44 and MNA-22) against TRMM dataset for the selected extreme precipitation indices,
averaged over the period 1998–2011. The cdd indicator is generally well reproduced over the
eastern part of the domain. It is worth noting that MNA-22 is generally able to reduce the bias

SC
affecting cdd over wide areas. For what concerns r10, it is generally underestimated in the same
areas where a precipitation negative bias has already been documented, but for this indicator MNA-
22 is able to provide a better representation than MNA-44. The sdii is generally overestimated on

U
wide areas by MNA-44, while 90p shows a dipole, with a positive bias over North Africa and a
negative one over Central Africa. The difficulties of COSMO-CLM in reproducing EWI over the
AN
eastern part of the domain could be related to the anomalous intense transport of dry air from the
Indian Ocean (Krichak et al., 2015), which plays an important role in extreme precipitation and is
not properly captured by RCMs. The overestimation in coastal areas is probably related to the small
M

value of the scaling factor of the laminar boundary layer for heat employed (Bucchignani et al.,
2016a), which increases the surface evaporation and as a consequence, the total precipitation. All
indices appear quite well reproduced over southern Europe. Overall, MNA-22 provides a quite
D

better representation, confirming the importance of high resolution in the simulation of extreme
events.
TE
C EP
AC
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
M
D
TE
C EP
AC

Fig.4. Spatial pattern of the bias (%) of the COSMO-CLM simulations (MNA-44 and MNA-22)
against TRMM dataset for the four extreme precipitation indices (1998–2011).

4 Climate projections

Climate projections have been performed according to the RCP4.5 scenario (Moss et al., 2010),
developed in the framework of the Coupled Model Intercomparison Project phase 5 (CMIP5). This
scenario raises the radiative forcing pathway to 4.5 W m-2 in 2100. The period 2071–2100 has been
selected for the analysis, as representative of the end of the 21st century. Following the approach
 ACCEPTED MANUSCRIPT
used in many literature works (Dosio and Panitz, 2016; Bucchignani et al., 2014), the mean
seasonal values of T2m and precipitation over the future period have been compared with the mean
seasonal values over the reference one (1981–2010). This period has been defined combining the
results of the historical simulation (1981–2005) with the first five years of the scenario period.
Projections obtained with the GCM CMCC-CM and with both COSMO-CLM simulations are
shown and discussed.

PT
4.1 Temperature

Fig. 5 shows the seasonal T2m change projections ( ) for the period 2071–2100 with respect to

RI
1981–2010, provided by CMCC-CM, MNA-44 and MNA-22. Both CMCC-CM and COSMO-
CLM simulations perform in a similar way, suggesting a general increase of temperature in the four
seasons. A comparison between MNA-44 and MNA-22 reveals that the coarser resolution projects a

SC
slight larger warming. These differences can be related to local processes linked to land processes
and parameterization, a better representation of topography and the location of land and sea in the
MNA-22 grid. Overall, a strong warming is projected, which is characterized by a marked

U
seasonality, ranging from a minimum of 2.5 in DJF to a maximum of 4 in JJA. In DJF and
MAM, the largest increases are projected over high altitude areas of Turkey, northern Iran and
AN
Afghanistan, in particular over the Hindu Kush, an 800 km-long mountain range that separates
climate zones B and D. Here the strong warming may be related to the snow-albedo feedback
(Giorgi et al., 1997). Large increases are also projected on the boundary between Iran and Iraq. JJA
M

is the season in which the largest warming is projected, as already found under A1B scenario for the
EMME domain (Lelieveld et al., 2012). In particular, northwestern Africa will be affected by
strong increases (up to 5 ). In SON, two distinct areas are clearly visible: a belt including North
D

Africa and Arabian Peninsula in which warming is more substantial, and the remaining part of the
domain where the warming is less intense.
TE
C EP
AC
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
M
D
TE
C EP
AC

Fig.5. T2m climate projections: seasonal differences ( ) between the average value over 2071–
2100 and 1981–2010 for the four seasons, provided respectively by GCM, MNA-44 and MNA-22.

The annual and seasonal drying/warming signal over the northern African region (Morocco,
Algeria, Libya, Egypt, and Tunisia) is a consistent feature, both in the global and the regional
climate change projections under the A1B and A2 scenarios (IPCC, 2014). A good agreement is
found also with the results by Dosio and Panitz (2016). They provided seasonal temperature
anomaly values for several sub-regions defined inside CORDEX-Africa, derived by an ensemble of
climate projections performed with COSMO-CLM at 0.44° under RCP4.5. In particular, they
 ACCEPTED MANUSCRIPT
projected at the end of the century an increase of about 2 over the Atlas Mountains (in JFM
season), of about 2.5 over West Africa North (in JAS season) and of about 2 over West Africa
South (in JAS season). Even if a direct comparison is not possible due to the different choice of
seasons, these values are very close to the findings of the present work. Projections over Saudi
Arabia are in quite good agreement with the values obtained by Almazroui (2013) with PRECIS,
even if some discrepancies concerning the spatial anomaly distribution is recorded, in particular he
found a strong warming over central Saudi Arabia, which is not recorded in the present results. An
excellent agreement is found with Lelieveld et al. (2016), where the ensemble results of CMIP5

PT
models are analyzed. In particular, their results fit well with the strong temperature increase
projected in JJA and with the peaks over North Africa, Spain and Turkey.
Fig. 6 shows the mean T2m time series (5-year running mean) and trend lines for the sub-regions A,

RI
B, C, D, E, provided by MNA-44 and MNA-22. The numerical values of the trends (for both
resolutions) are also reported. The significance of the projected changes has been verified by using
the Mann–Kendall nonparametric test with a confidence level of 95%, resulting statistically

SC
significant in all the sub-regions and for both resolutions. The plots highlight hardly a difference
between MNA-44 and MNA-22, both revealing an almost regular and gradual warming. Trends
differ from one climate region to another, the largest value occurring for the polar climate E. The

U
projected increases of temperature by 2100 with respect to the present climate range from 3.2 in
AN
A to 4.2 in E.
M
D
TE
C EP
AC
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
M
D
TE
C EP
AC

Fig.6. Time series of 2-m temperature ( ) over 1981–2100 and trend lines provided by MNA-44
and MNA-22 for the five subdomains considered. Numerical values of the trends ( per decade)
are reported in the title (* indicates that the trend is statistically significant).

4.2 Precipitation

Fig. 7 shows the seasonal precipitation climate projections (%) for the period 2071–2100 with
respect to 1981-2010 provided by CMCC-CM, MNA-44 and MNA-22. Both the GCM and the
RCM suggest significant percentage decreases in DJF in the western part of domain. However, this
 ACCEPTED MANUSCRIPT
area is characterized by very low precipitation values in the reference period, leading to high
percentage variations even if absolute changes are small. In JJA, the GCM projects reduction over
the Arabian peninsula, as confirmed also by MNA-22. A band of increase on the coast along the
Gulf of Guinea is visible in CMCC-CM and not projected by COSMO-CLM. This structure could
be related to a change in the West African Monsoon system, which is very difficult to be modeled.
In fact, a strong positive precipitation bias was observed here (Bucchignani et al. 2016b), which
needs to be better understood before interpreting the precipitation projection in this location. In
SON, according with COSMO-CLM, a wide area, including Egypt, Saudi Arabia and Iran
experiences a general increase, with absolute peaks of 50–100 mm per month. In MAM, according

PT
to CMCC-CM a general increase over the western half of the Sahel region is projected, but a
different behavior is projected by COSMO-CLM. Precipitation reductions are projected in Spain,
except in SON (but with limited absolute values). Differences between the two resolutions are

RI
smaller in SON and become more distinct in the other seasons. MNA-22 projects a strong decrease
in precipitation in JJA over the Arabian Peninsula, which is not fully detected in MNA-44. The

SC
precipitation variability increases with resolution since physical feedback processes are better
simulated with increased resolution and for a better-resolved orography (Hohenegger et al., 2009).
In fact, rainy events are often associated with penetration of cold air masses from North Europe and
warm humid masses from tropical Atlantic, so topography and coastal lines strongly influence the

U
spatial rainfall distribution (Krichak et al., 2010).
AN
M
D
TE
C EP
AC
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
M
D
TE
C EP
AC

Fig.7. Precipitation climate projections: seasonal differences (%) between the average value over
2071–2100 and 1981–2010 for the four seasons, provided respectively by GCM, MNA-44 and
MNA-22.

Dosio and Panitz (2016) projected at the end of 21st century a variation close to 0 over the Atlas
Mountains (in JFM) and modest reductions over West Africa North (-0.5 mm d-1) and South (-0.2
mm d-1) in JAS season, in qualitatively good agreement with the findings of the present work. Over
Senegal, Tall et al. (2017) found a precipitation decrease of more than 50% at the end of 21st
century, which is in quite good agreement with the present results for JJA. Over the northern basin
 ACCEPTED MANUSCRIPT
of Tunisia, global and regional models under the A1B scenario project a significant decrease in the
median, 10th and 90th percentile values of precipitation in winter and spring seasons (IPCC, 2014).
This finding is partially confirmed by the present simulations. Over West Africa, a good agreement
is found also with Buontempo et al. (2015), who projected precipitation reductions under A1B
emission scenario.
Fig. 8 shows the mean precipitation time series (5-year running mean) and trend lines for the five
sub-regions provided by MNA-44 and MNA-22, highlighting different behaviors in the five areas
and relevant differences between the two resolutions. The numerical values of the trends at both

PT
resolutions are also reported. In sub-regions A, B and D, trend values exhibit opposite sign, but it is
worth noting that values in B and D are not significant, while in A only the MNA-44 value is
statistically significant. In C, the trend values are negative and statistically significant, while in E

RI
they are positive and significant only for MNA-22.

U SC
AN
M
D
TE
C EP
AC
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
M
D
TE
C EP
AC

Fig.8. Time series of mean precipitation over 1981–2100 and trend lines for MNA-44 and MNA-22
for the five subdomains considered. Numerical values of the trends (mm per month per decade) are
reported in the title (* indicates that the trend is statistically significant).

4.3 Extreme weather events indicators

Fig. 9 shows the climate projections for the selected extreme precipitation indices; more
specifically, the percentage changes (%) between the average value over 2071–2100 and 1981–
2010 are shown. The cdd will increase largely over the southern part of the domain, and this is an
 ACCEPTED MANUSCRIPT
alarming finding, since zones already exhausted (e.g. central Africa) may suffer more drought in the
future. For r10, large percentage increases are projected at both resolutions over North Africa and
Arabian Peninsula. The sdii is projected to increase by MNA-44 over wide areas in the central part
of the domain, while MNA-22 projects low increases. Concerning 90p, MNA-44 projects
significant increases over Egypt and Iran, which are not projected by MNA-22, so a further
investigation is required over these areas. The growth of heavy precipitation over Saudi Arabia may
be related to the projected growth of temperature, since a hot atmosphere is characterized by a
higher moisture-holding capacity (Lenderink and van Meijgaard, 2010). The r10 will decrease over
wide areas of North Africa, due to the severe soil drying that reduces the moisture availability and

PT
hence the occurrence of precipitation extremes (Zollo et al., 2016). The general increase of sdii
projected over the Mediterranean coast may be related to the warmer climate and the consequent
enhanced hydrological cycle.

RI
U SC
AN
M
D
TE
C EP
AC
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
M
D
TE
C EP
AC

Fig.9. Climate projections for the four extreme precipitation indices: percentage changes between
the average value over 2071–2100 and 1981–2010.

5 Conclusions and discussion

This work provided a picture of the future climate conditions expected in the MENA region, as
projected by COSMO-CLM assuming the RCP4.5 scenario. High-resolution climate projections
make it possible to bridge the gap between climate and impact models, especially in regions
characterized by complex topography.
 ACCEPTED MANUSCRIPT
Climate projections in 2071–2100 highlight a general warming that can result in economic and
environmental consequences, especially in those regions in which a decline of precipitation is
projected. Temperature projections are statistically significant and generally highlight a strong
warming, especially in summer, which could be associated with a thermal low. It has been shown
that precipitation projections on this area, both in terms of average values and of extreme events
indicators, largely depend on the horizontal resolution, suggesting the need for additional
simulations at higher resolution.
A systematic and quantitative comparison with other projections was not possible, since regional

PT
climate simulations over CORDEX-MENA for the 21st century are not yet available in literature.
Comparisons with global projections and with data extracted by other computational domains (e.g.
CORDEX-Africa) revealed a good qualitative consistency for what concerns temperature, while

RI
projected precipitation changes are beset by larger uncertainties. Even if it is not possible to define
an optimal resolution for every geographical domain, (depending on the climate variability of the
area considered), it is evident that, at least in some areas, the high-resolution might provide good

SC
improvements. The reproduction of extreme temperature indices is more complicated and has not
been considered in the present analysis, due to the lack of a reliable daily observational dataset over
MENA domain. A critical question, already discussed in Bucchignani et al. (2014), is whether

U
climate change signals may be influenced by the presence of biases in the model simulations. In our
opinion, biases in the base and projected years are consistent, suggesting that projection changes
AN
provide a reasonable bias removal. Regarding the uncertainty induced by three different sources
(global model, regional model and emission scenario), as assessed in the IPCC report (Knutti et al.,
2010), a multi-model ensemble could improve the reliability of projections, especially if single
M

elements are weighted with a measure of the model skill. A possible approach is the Reliability
Ensemble Averaging (REA) proposed by Giorgi and Mearns (2002, 2003). The possibility of
comparing the present results with those provided by other GCMs or the possibility of forcing
D

COSMO-CLM with different driving data (e.g. those used in the work by Dosio and Panitz, 2016)
would effectively improve the data quality. It will be considered in future a work.
TE
EP

Acknowledgments
C

This work was developed within the framework of the GEMINA project, funded by the Italian
Ministry of Education, University and Research and the Italian Ministry of the Environment, Land
AC

and Sea.

References
Almazroui, M., 2013. Simulation of present and future climate of Saudi Arabia using a regional climate model
(PRECIS), Int. J. Climatol. 33, 2247-2259. doi: 10.1002/joc.3721

Boberg, F., Christensen, J.H., 2012. Overestimation of Mediterranean summer temperature projections due to model
deficiencies. Nat. Clim. Change 2, 433–436. doi: 10.1038/ntclimate1454

Bucchignani, E., Montesarchio, M., Cattaneo, L., et al., 2014. Regional Climate modeling over China with COSMO-
CLM: performance assessment and climate projections. J. Geophys. Res. 119(21), 12151-12170. doi:
10.1002/2014JD022219
 ACCEPTED MANUSCRIPT
Bucchignani, E., Mercogliano, P., Rianna, G., et al., 2016a. Analysis of ERA-Interim driven COSMO-CLM simulations
over Middle East - North Africa domain at different spatial resolutions. Int. J. Climatol. 36(9), 3346-3369. doi:
10.1002/joc.4559

Bucchignani, E., Cattaneo, L., Panitz, H.-J, Mercogliano, P., 2016b. Sensitivity analysis with the regional climate
model COSMO-CLM over the CORDEX-MENA domain. Meteorology and Atmospheric Physics 128(1), 73-95. doi:
10.1007/s00703-015-0403-3

Bucchignani, E., Zollo, A.L., Cattaneo, L.,et al., 2017. Extreme weather events over China: assessment of COSMO-
CLM simulations and future scenarios. Int. J. Climatol. 37(3), 1578-1594. doi: 10.1002/joc.4798

PT
Buontempo, C., Mathison, C., Jones, R., et al., 2015. An ensemble climate projection for Africa. Clim. Dyn. 44, 2097-
2118. doi: 10.1007/s00382-014-2286-2

Cavicchia, L., Scoccimarro, E., Gualdi, S., et al., 2016. Mediterranean extreme precipitation: a multi-model assessment.

RI
Clim. Dyn. doi:10.1007/s00382-016-3245-x.

Chan, S.C., Kendon, E.J., Fowler, H.J., et al., 2013. Does increasing the spatial resolution of a regional climate model
improve the simulated daily precipitation? Clim. Dyn. 41(5–6), 1475–1495. doi:10.1007/s00382-012-1568-9.

SC
Christensen, J.H., Hewison, B., Busuioc, A., et al., 2007. Regional climate projections, in: Solomon, S., Qin, D.
Manning, M. et al. (Eds.) Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the
Fourth Assessment Report of the Intergovernmental Panel on Climate Change []. Cambridge University Press,

U
Cambridge and New York.
AN
Davies, H.C., 1983. Limitations of some common lateral boundary schemes used in regional NWP models. Monthly
Weather Review 111, 1002–1012. doi:10.1175/1520-0493(1983)111<1002: LOSCLB>2.0.CO;2

Dosio, A., Panitz, H.J., 2016. Climate change projections for CORDEX‑Africa with COSMO‑CLM regional climate
M

model and differences with the driving global climate models. Clim. Dyn. 46 (5-6), 1599-1625. doi: 10.1007/s00382-
015-2664-4

Giorgi, F., Hurrell, J.W., Marinucci, et al., 1997. Elevation signal in surface climate change: a model study. Journal of
D

Climate 10, 288-296. doi: 10.1175/1520-0442(1997)010<0288:EDOTSC>2.0.CO;2

Giorgi, F., Mearns, L.O., 2002. Calculation of average, uncertainty range, and reliability of regional climate changes
TE

from AOGCM simulations via the “reliability ensemble averaging” (REA) method. J. Climate 15(10), 1141–1158.

Giorgi, F., Mearns, L.O., 2003. Probability of regional climate change based on the reliability ensemble averaging
(REA) method. Geophys. Res. Lett. 30(12), 1629. doi:10.1029/2003GL017130
EP

Giorgi, F., Lionello, P., 2008. Climate change projections for the Mediterranean region. Global and Planetary Change
63, 90–104. doi: 10.1016/j.gloplacha.2007.09.005
C

Giorgi, F., Jones, C., Asrar, G., 2009. Addressing climate information needs at the regional level: the CORDEX
framework. WMO Bulletin 58 (3), 175–183.
AC

Hochman, A., Bucchignani, E., Gershtein, G., et al., 2017. Evaluation of regional COSMO-CLM climate simulations
over the Eastern Mediterranean for the period 1979 – 2011. Int. J. Climatol. doi: 10.1002/joc.5232

Huffman, G., Adler, R., Bolvin, D., et al., 2010. The TRMM multi-satellite precipitation analysis (TMPA), in:
Gebremichael M., Hossain, F. (Eds.), Satellite Rainfall Applications for Surface Hydrology. Springer, Dordrecht, pp. 3–
22.

Hohenegger, C., Brockhaus, P., Bretherton, S., et al., 2009. The soil moisture-precipitation feedback 595 in simulations
with explicit and parameterized convection. J. Climate 22(19), 5003–5020.

IPCC, 2000. Special Report on Emission Scenarios. A Special Report of Working Group III of the Intergovernmental
Panel on Climate Change. Cambridge University Press, New York.

IPCC, 2007. Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth
Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge.
 ACCEPTED MANUSCRIPT
IPCC, 2014. Climate Change 2014: Impacts, Adaptation, and Vulnerability. Part B: Regional Aspects. Contribution of
Working Group II to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change Cambridge
University Press, Cambridge and New York.

Klingmüller, K., Pozzer, A., Metzger, S., et al., 2016. Aerosol optical depth trend over the Middle East. Atmos. Chem.
Physics 16, 5063–5073. doi:10.5194/acp-16-5063-2016

Knutti, R., Abramowitz, G., Collins, M., et al., 2010. Good practice guidance paper on assessing and combining multi
model climate projections. in: Stocker, T.F., Qin, D. Plattner, G.K., (Eds.) Meeting Report of the Intergovernmental
Panel on Climate Change Expert Meeting on Assessing and Combining Multi Model Climate Projections. IPCC

PT
Working Group I Technical Support Unit, University of Bern, Bern, Switzerland.

Krichak, S.O., Alpert, P., Kunin, P., 2010. Numerical simulation of seasonal distribution of precipitation over the
eastern Mediterranean with a RCM. Clim. Dyn. 34(1), 47:59.

RI
Krichak, S.O., Barkan, J., Breitgard, et al., 2015. The role of the export of tropical moisture into midlatitudes for
extreme precipitation events in the Eastern Mediterranean region. Theor. Appl. Climatol. 121, 499–515.

SC
doi:10.1007/s00704-014-1244-6.

Laprise, R., Hernandez-Diaz, L., Tete, K., et al., 2013. Climate projections over CORDEX Africa domain using the
fifth-generation Canadian Regional Climate Model (CRCM5). Clim. Dyn. 41, 3219-3246. doi:10.1007/s00382-012-

U
1651-2
AN
Lawrence, P.J., Chase, TN., 2007. Representing a new MODIS consistent land surface in the community land model
(CLM 3.0). J. Geophys. Res. 112(G1), G01023. doi: 10.1029/2006JG000168

Lelieveld, J., Hadjinicolaou, P., Kostopoulou, E., et al., 2012. Climate change and impacts in the Eastern Mediterranean
and the Middle East. Climatic Change 114, 667–687. doi:10.1007/s10584-012-0418-4
M

Lelieveld, J., Proestos, Y., Hadjinicolaou, P., et al., 2016. Strongly increasing heat extremes in the Middle East and
North Africa (MENA) in the 21st century. Climatic Change 137, 245-260. doi: 10.1007/s10584-016-1665-6
D

Lenderink, G., van Meijgaard, E., 2010. Linking increases in hourly precipitation extremes to atmospheric temperature
and moisture changes. Environ. Res. Lett. 5(2), 025208. doi: 10.1088/1748-9326/5/ 2/025208.
TE

Mitchell, T.D., Jones, P.D., 2005. An improved method of constructing a database of monthly climate observations and
associated high-resolution grids. Int. J. Climatol. 25(6), 693-712. doi: 10.1002/joc.1181

Moss, R., Edmonds, J., Hibbard, K., et al., 2010. The next generation of scenarios for climate change research and
EP

assessment. Nature 463, 747- 756. doi:10.1038/nature08823

Niang, I., Ruppel, O.C., Abdrabo, M.A., et al., 2014. Africa, in: Barros, V.R., Field, C.B., Dokken, D.J., et al. (Eds.)
Climate Change 2014: Impacts, Adaptation, and Vulnerability. Part B: Regional Aspects. Contribution of Working
C

Group II to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University
Press, Cambridge and New York.
AC

Rockel, B., Will, A., Hense, A., 2008. The regional climate model COSMO-CLM (CCLM). Meteorologische Zeitschrift
17(4), 347–348. doi: 10.1127/0941-2948/2008/0309

Roeckner, E., Bauml, G., Bonaventura, L., et al., 2003. The atmospheric general circulation model ECHAM5. Part I:
Model description, Max-Planck-Institut fur Meteorologie, Hamburg, Germany, (Rep. No. 349).

Russo, S., Marchese, A., Sillmann, J., et al., 2016. When will unusual heat waves become normal in a warming Africa?
Environ. Res. Lett. 11, 054016. doi:10.1088/1748-9326/11/5/054016

Scoccimarro, E., Gualdi, S., Bellucci, A., et al., 2011. Effects of tropical cyclones on ocean heat transport in a high
resolution coupled general circulation model. J. Climate 24, 4368–4384. doi: 10.1175/2011JCLI4104.1
 ACCEPTED MANUSCRIPT
Tall, M., Sylla, M.B., Diallo, I., Pal, J.S., et al., 2017. Projected impact of climate change in the hydroclimatology of
Senegal with a focus over the Lake of Guiers for the twenty-first century. Theor. Appl. Climatol. 129 (1-2), 655-665.
doi: 10.1007/s00704-016-1805-y

Taylor, K.E., Stouffer, R.J., Meehl, G.A., 2012. An overview of CMIP5 and the experiment design. Bull. Am.
Meteorol. Soc. 93, 485–498. doi: 10.1175/BAMS-D-11-00094.1

Tanrè, D., Geleyn, J.F., Slingo, J.M., 1984. First results of the introduction of an advanced aerosol-radiation interaction
in the ECMWF low resolution global model, in: Gerber, H., Deepak, A. (Eds.), Aerosols and Their Climatic Effects.
Deepak Publ, Hampton, VA, pp. 133-177

PT
Tegen, I., Lacis, A.A., Fung, I., 1996. The influence of mineral aerosols from disturbed soils on the global radiation
budget. Nature 380, 419-422. doi:10.1038/380419a0

RI
Terink, W., Immerzeel, W.W., Droogers, P., 2013. Climate change projections of precipitation and reference
evapotranspiration for the Middle East and Northern Africa until 2050. Int. J. Climatol. 33, 3055-3072. doi:
10.1002/joc.3650

SC
Tiedtke, M., 1989. A comprehensive mass flux scheme for cumulus parameterization in large-scale models. Monthly
Weather Review 117(8), 1779–1800. doi:10.1175/1520-0493(1989)117<1779: ACMFSF>2.0.CO;2

U
WMO, 2009. Guidelines on Analysis of extremes in a changing climate in support of informed decisions for adaptation.
Tech. Rep. WCDMP No. 72, WMO/TD-No. 1500, WMO, Geneva, Switzerland.
AN
Zollo, A.L., Rillo, V., Bucchignani, E., et al., 2016. Extreme temperature and precipitation events over Italy: assessment
of high resolution simulations with COSMO-CLM and future scenarios. Int. J. Climatol. 36(2), 987-1005. doi:
10.1002/joc.4401
M
D
TE
C EP
AC