1. Introduction
The Middle East–North Africa (MENA) region is considered to be one of the most prominent climate change hot spots, with significant temperature increases and rainfall reductions since the middle of the twentieth century (Lelieveld et al. 2012; Tanarhte et al. 2012; Zittis 2018; Ntoumos et al. 2020). Studies have shown that the MENA region is warming faster than the global average, along with a stronger summer warming (Lelieveld et al. 2016; Lionello and Scarascia 2018; Zittis et al. 2019, 2022). In the already environmentally stressed MENA, climate projections indicate further intensification of heat extremes (Zittis et al. 2016; El-Samra et al. 2018; Legasa et al. 2020; Ozturk et al. 2021). Maximum temperatures are expected to rise well above 50°C in many areas within MENA by the end of the twenty-first century (Lelieveld et al. 2016; Zittis et al. 2021; Ntoumos et al. 2022). The projected rapid warming and intensification of extreme heat conditions is expected to have numerous negative impacts on the societies in the MENA (Ahmadalipour and Moradkhani 2018; Dunne et al. 2013; Constantinidou et al. 2016) and will threaten the habitability of some regions (Pal and Eltahir 2016; Almazroui 2020).
Climate models are essential tools for climate extreme studies on both global and regional levels. In a recent study, Ntoumos et al. (2020) assessed the performance of several global climate models (GCMs) in simulating temperature extremes in the MENA region and found that GCMs were generally able to reproduce the historical climate trends and patterns. However, due to GCM relatively coarse horizontal resolution, systematic biases occur, especially in areas with complex orography. Therefore, despite the ability of GCMs to simulate essential large-scale climate features, they do not resolve adequately the regional climate responses to forcings, and their climate information output is too coarse to offer regional insight. The MENA region is such an area, with diverse physical characteristics such as steep orographic and climate gradients, many and diverse coastlines, variable land cover, and the presence of islands and major urban centers. High-spatial-resolution climate projections can be achieved through the use of regional climate models (RCMs), which simulate climate over limited areas of the globe by applying the dynamical downscaling technique (Giorgi and Gutowski 2015).
The representation of physical processes is considered to be one of the most challenging problems in numerical modeling of the atmosphere and climate (Dudhia 2014; Giorgi and Gutowski 2015; Donahue and Caldwell 2018). The performance of RCM simulations strongly depends on the choice of different physics parameterizations (e.g., of the land surface, boundary layer, convection, cloud microphysics, radiation). It has been shown that the effect of different physics schemes is often greater than the sensitivity induced by RCMs’ internal variability (Lavin-Gullon et al. 2021). The importance of RCMs’ physics is also highlighted in Oettli et al. (2011), who show that two configurations of the same RCM are sometimes more distinct than the two different RCMs themselves. The selection of the most appropriate combination of physics parameterizations varies according to the location of interest, type of application, season, or climate type since different climate variables have been proved to be sensitive to different model configurations (Katragkou et al. 2015; Kala et al. 2015).
For the MENA region specifically, Zittis et al. (2014) investigated the Weather Research and Forecasting (WRF) Model performance with combinations of three boundary layer, two cumulus, and two microphysics schemes and concluded that the best-performing ones for climate simulations over the MENA region at a 50-km horizontal resolution included the nonlocal Yonsei University (YSU) planetary boundary layer, the Kain–Fritsch (KF) cumulus, and the WRF single-moment 6-class microphysics scheme (WSM6). Zittis and Hadjinicolaou (2017) tested the WRF Model with different radiation schemes and found that simulations with an RRTMG scheme improve model performance, particularly over the desert areas of the region. Furthermore, Constantinidou et al. (2020b,a) focused on land surface schemes and concluded that Noah and its augmented version, Noah-MP, were the best-performing ones. In the current study, we assess RCM sensitivity under a combination of planetary boundary layer (PBL) schemes, which has not yet been extensively studied for the MENA region.
PBL schemes are used to parameterize the vertical subgrid-scale fluxes within the boundary layer that are associated with the exchanges of heat, moisture, and momentum between the surface and the free atmosphere (Hu et al. 2010; Cohen et al. 2015). Physical processes within the PBL have been shown to affect surface temperatures. Miralles et al. (2014) showed that during multiday heat events in subsidence regions, drying soils and high heat advection enhance the PBL growth, leading to further entrainment of warm air from its top. These processes could ultimately cause a further increase in surface temperatures. Several studies indicate that biases in vertical mixing within the PBL could very likely contribute to temperature biases (Hu et al. 2010; Wei et al. 2017; Zhao et al. 2017). As a result of climate change, increasing temperatures and decreasing relative humidity would lead to deeper PBL height in the future (Zhang et al. 2013; Zhou et al. 2020). Garcia-Díez et al. (2013) found, from a WRF simulation for 2001, that the YSU scheme produces warmer temperatures, and a different best-performing scheme was found for different seasons. According to this study, results from short-term studies should be carefully analyzed in order to derive the appropriate parameterizations to be used in long-term simulations.
The objective of this study is to evaluate the performance of the WRF Model in reproducing the climatic conditions over a 10-yr period, focusing on hot summer conditions. Our work is motivated by (and serves) the goals of the WCRP’s Coordinated Regional Climate Downscaling Experiment (CORDEX) (Giorgi and Gutowski 2015), by applying simulation techniques and assessing model output at long-term, climatic time scales, different from those of short-term, typical weather forecasting. Hence, we adopt the most common method in regional climate simulations (Xu et al. 2019), which is a long-term continuous simulation utilizing a single initialization of large-scale fields and continuous updating of lateral boundary conditions (LBCs). Long-term simulations with a single initialization are considered a common practice in similar regional climate modeling studies, for example, within the CORDEX program (Zittis et al. 2014; Katragkou et al. 2015; Zittis and Hadjinicolaou 2017), in order to capture interannual and decadal variability of climate. Our objective is to determine the best-performing physics configuration, with a particular focus on PBL schemes, for future multidecadal RCM studies.
Thus, we apply WRF in the MENA region in long-term (2001–10) simulations driven by ERA-Interim reanalyses, in order to explore the model sensitivity to three frequently used PBL schemes: YSU, Mellor–Yamada–Janjić (MYJ), and asymmetric convective model, version 2 (ACM2). The focus is on evaluating the performance of the WRF Model in simulating summer temperatures across the MENA region, with emphasis given to heat extremes. Differences among the three PBL schemes are obtained by comparing the model’s output with surface observation and reanalysis data and by deriving several statistical metrics. Additionally, we examine the physical causes of those differences via the analysis of relevant variables (such as boundary layer height and specific humidity).
2. Model configuration
a. Model setup and domain
In this study, the Advanced Research version of the WRF (WRF-ARW) Model, version 4.2.1, is used for simulations over the MENA region with a horizontal resolution of 0.22 (∼24 km) and 35 vertical levels, according to the CORDEX guidelines for the MENA-CORDEX domain (https://cordex.org/domains/cordexregion-mena-cordex). Our choice to use 35 vertical levels in the WRF Model simulations is consistent with recent regional climate modeling studies that use 30–40 vertical levels to downscale ERA-Interim data (Katragkou et al. 2015; Zittis and Hadjinicolaou 2017; Constantinidou et al. 2020b). The computational domain was discretized using a grid of 463 × 236 points. The model top was set at 30 hPa. The initial and boundary data are taken 6 hourly from the ERA-Interim reanalysis (Dee et al. 2011). The simulations were performed for an 11-yr period (2000–10), with the 2000 spinup year excluded from the analysis.
Following Zittis et al. (2014) and their suggested combination of physics schemes for the MENA domain, the common set of parameterizations used in all simulations includes the Betts–Miller–Janjić (BMJ) cumulus scheme (Janjić 1994) and the WSM6 (Hong and Lim 2006). For the radiation parameterizations, the RRTMG is used for short- and longwave radiation (Iacono et al. 2008) as it was shown to perform better for the MENA region than the other commonly used Community Atmosphere Model (CAM) radiation scheme (Zittis and Hadjinicolaou 2017). The Noah-MP (Niu et al. 2011) land surface scheme was selected and has already been evaluated for simulations over the MENA region (Constantinidou et al. 2020b). The Monin–Obukhov surface layer scheme was employed in the simulations with the YSU and ACM2 schemes, while in the MYJ scheme simulations, we used the Janjić Eta Monin–Obukhov surface layer scheme.1
The simulation domain covers the whole MENA CORDEX domain, as presented in Fig. 1, and the area that was included in the analysis is indicated by the shading. We have excluded the southern part of the domain since we focus on the Mediterranean area and the arid, hot areas of northern Africa and the Middle East.
Map of model domains. The outer rectangle is the domain of the dynamical downscaling, and the inner rectangle (gray shaded) is the MENA-CORDEX domain of the analysis.
Citation: Journal of Applied Meteorology and Climatology 62, 9; 10.1175/JAMC-D-22-0108.1
b. Model PBL parameterizations
One way to classify the different PBL parameterizations is on how they approach the turbulence closure problem (Cohen et al. 2015). A closure scheme is necessary in order to derive the turbulent fluxes (Stull 1988). One type of closure scheme is the local closure, in which variables and parameters are defined at each model level and therefore depend solely on local values and gradients. On the other hand, nonlocal closure schemes account for the whole vertical profile, which often leads to an improved representation of vertical mixing throughout the depth of the PBL. In this study, we chose three different PBL schemes, a nonlocal (YSU), a local (MYJ), and a nonlocal one that turns to local under stable conditions (ACM2). The schemes used are described as follows.
1) MYJ
The Mellor–Yamada–Janjić PBL scheme is a local, 1.5-order closure scheme (Janjić 1994). MYJ is a scheme based on turbulent kinetic energy prediction, and the PBL height is derived using the TKE profile. It has been found that, in case of a less stable PBL, MYJ underestimates vertical mixing, leading to shallower, more humid, and colder PBLs (e.g., Hu et al. 2010).
2) YSU
The YSU PBL scheme (Hong et al. 2006) is a nonlocal, first-order closure scheme very similar to the former Medium-Range Forecast Model (MRF) PBL scheme. In comparison with the MRF, the YSU scheme represents with higher accuracy the deeper vertical mixing in buoyancy-driven PBLs (Hong et al. 2006). The YSU scheme uses the bulk Richardson number to diagnose the PBL height.
3) ACM2
The ACM2 PBL scheme (Pleim 2007) is a hybrid (local nonlocal), first-order closure scheme. According to Pleim (2007), the inclusion of local transport under stable conditions improves the representation of potential temperature and velocity vertical profiles. Similar to the YSU scheme, the bulk Richardson number is used in order to define the top of the PBL.
3. Data and methods
a. Initial and boundary conditions
Initial conditions and lateral boundary conditions for the WRF simulations are provided by the ERA-Interim reanalysis dataset and were updated every 6 h. The ERA-Interim was developed by the European Centre for Medium-Range Weather Forecasts (ECMWF), and data are available from 1979 to August 2019. The data assimilation method used to produce ERA-Interim is based on a four-dimensional variational data assimilation (4D-Var) with a 12-h analysis window (Dee et al. 2011).
b. Datasets for verification
For the purpose of evaluating the model output, we used the ERA5 dataset (Hersbach et al. 2020). ERA5 is the fifth major global reanalysis produced by ECMWF from 1979 onward, updated monthly with a delay of approximately 2 months, and provides hourly estimates of a large number of atmospheric, land, and oceanic climate variables. The ERA5 is model-driven output of elaborated data assimilation applied on various observational sources and is commonly used for climate change assessment and regional climate model evaluation (Odnoletkova and Patzek 2021; Vautard et al. 2021). The horizontal resolution of the ERA5 reanalysis (30 km), the most recent long-term meteorological data product available, is close to the one used in these model runs (24 km), which makes it a more suitable reference dataset for assessing the current climate conditions and evaluating the WRF Model’s performance. Thus, the ERA5 data are a valuable tool for evaluating the current climate in hindcast mode, which provides the fidelity in the model to be applied for climate change projections. We test its ability to be used as a reference, quasi-observed dataset for temperature and related indices, by comparing 10-yr climatologies of maximum temperature (TX), minimum temperature (TN), and specific humidity with station measurements from the Global Summary of the Day (GSOD) derived from the Integrated Surface Hourly (ISH) dataset (both produced by the National Centers for Environmental Information; https://www.ncei.noaa.gov/access/metadata/landing-page/bin/iso?id=gov.noaa.ncdc:C00516). Inspection of Fig. 2 reveals that, of the 577 locations compared, the ERA5 temperatures are in reasonable agreement overall with the GSOD observations, with a modest underestimation of TX (median = −1.7°C) and a small overestimation of TN (median = +0.5°C). With regard to specific humidity, 420 locations are considered due to data availability. As shown in Fig. 2, the ERA5 demonstrates very good agreement with observed data (median bias = −0.6 g kg−1), thus justifying their subsequent use for model evaluation. The spatial distribution of ERA5 biases can be found in the maps of Fig. S1 in the online supplemental material.
Distribution (histogram with blue-line density curve) of ERA5 minus GSOD summer climatology (June–August 2001–10 average) for (top left) maximum temperature, (top right) minimum temperature, and (bottom) specific humidity. The vertical dashed red line defines the median value.
Citation: Journal of Applied Meteorology and Climatology 62, 9; 10.1175/JAMC-D-22-0108.1
Furthermore, the Earth System Research Laboratory (ESRL) radiosonde database (https://ruc.noaa.gov/raobs/) was utilized to evaluate ERA5 performance at levels beyond the surface. Daily soundings taken at 1200 UTC were selected for consistency, as the model results were being investigated at the same time in sections 4e and 4f. To provide a comprehensive evaluation, 10 stations were selected (Fig. S2 in the online supplemental material) based on their representation of different regions within the MENA area and on data availability. Only stations with 80% data availability were considered in the selection process. Figure 3 shows a comparison of the mean summer (2001–10) vertical profiles of potential temperature for 10 atmospheric soundings stations and their closest ERA5 grid points. Potential temperature vertical profiles are used in this study for the estimation of the PBL height of ERA5 and model simulations. Despite some deviations, ERA5 profiles show a high level of agreement with the observed soundings. At most locations, it was found that ERA5 accurately captures the height of the shift in potential temperature, which is usually indicative of the PBL height (PBLH) (Nielsen-Gammon et al. 2008). Following this method, the PBLH is estimated for both stations and ERA5 (Table S1 in the online supplemental material), and an overall good agreement is found.
Mean vertical profiles of potential temperature for 10 atmospheric soundings stations and the closest ERA5 grid points for the 2001–10 summer climatology.
Citation: Journal of Applied Meteorology and Climatology 62, 9; 10.1175/JAMC-D-22-0108.1
c. Bias analysis and indices
We assess the performance of the WRF Model employing the three different PBL schemes, with emphasis on the heat extremes. Since we focus on the simulated hot conditions across MENA, the June–August climatology of TX and TN have been analyzed. For the heat extremes, we consider the warmest day (TXx) and warmest night (TNx), which are among the indicators proposed by the Expert Team on Climate Change Detection and Indices (ETCCDI) (Karl et al. 1996). We also test the model’s ability to capture humid heat conditions (a combination of high temperature and humidity) through the calculation of the heat index (HI). Following the method used in a previous work (Ntoumos et al. 2022), the HI is derived via a multiple linear regression based on temperature and relative humidity (Rothfus 1990; Steadman 1979). The simulated climate deviations from the observed conditions were studied through the analysis of bias maps calculated from the average 2001–10 values. The model validation is not limited to the analysis of common surface variables (such as the 2 m temperature) but is enhanced with additional variables such as specific humidity and PBLH in an effort to explain the temperature biases. We also derive vertical profiles of potential temperature and specific humidity for selected areas of interest in order to explore in detail the differences among the three PBL schemes.
d. Evaluation metrics
To quantitatively assess the performance of each simulation, we used a number of statistical metrics using monthly values for TX and TN and annual values for the indices of extremes from the model output and the gridded observational data. A short description of each metric is presented below.
1) MAE
2) MIA
3) STDE
4. Results
a. Summer maximum and minimum temperature
Figure 4 shows the WRF biases for each PBL scheme for the summer maximum temperature relative to the ERA5 dataset. The model biases strongly vary according to geographic location. All schemes exhibit warm biases in most areas of the MENA region, particularly over the Middle East and northeast Africa where they overpredict maximum temperature by 2°–4°C, depending on the scheme used. This was also found in other RCM studies over the MENA domain (Bucchignani et al. 2016; Zittis and Hadjinicolaou 2017; Fonseca et al. 2020; Branch et al. 2021), which have shown similar warm biases for TX in these areas. They suggest that these biases are associated with a combination of model deficiencies such as overestimation of downwelling surface shortwave (SW) radiation, underestimation of albedo, and unrealistic cloud cover. In our case, we found an underestimation of the albedo value (Fig. S3 in the online supplemental material) in most of North Africa and the Middle East, which is around 0.2, while in ERA5 reanalysis, it is around 0.4. The underestimation of albedo can be linked to inaccurate representation of land surface characteristics, as surface reflectivity is strongly influenced by the type, color, and roughness of the surface, as well as estimations of radiation fluxes (Fonseca et al. 2020; Nelli et al. 2020).
(top left) ERA5 summer mean maximum temperature, and absolute biases of the three simulations with different PBL schemes relative to the ERA5 dataset averaged over the period 2001–10.
Citation: Journal of Applied Meteorology and Climatology 62, 9; 10.1175/JAMC-D-22-0108.1
It is clear that ACM2 is the warmest scheme, mainly in sub-Saharan Africa and the Mediterranean, where the other two schemes show opposite sign biases. MYJ predicts significantly lower temperatures in parts of sub-Saharan Africa and, on the contrary, has strong warm biases (>5°C) in northeast Africa and the Middle East. The same pattern is obtained for the YSU scheme, although it shows weaker biases than MYJ. Overall, the YSU scheme seems to be in better agreement with ERA5, which is also confirmed with the following quantitative assessment (section 4c).
In a similar way, the ERA5 mean minimum temperature and the respective WRF PBL scheme biases are presented in Fig. 5. Like before, the sign and magnitude of model biases strongly vary according to geographic location. All schemes predict colder temperatures in most of the Middle East, while differences are observed in the North Africa region. In contrast with maximum temperatures in which warm biases prevail, the minimum temperatures for the WRF runs tend to be colder in many areas, especially in the Arabian Peninsula, where cold biases exceed 5°C in the ACM2 scheme. Such nocturnal biases in these areas have been noted in previous analyses (Branch et al. 2014; Zittis et al. 2014; Fekih and Mohamed 2019; Weston et al. 2019). Fonseca et al. (2020) found that cold biases in arid regions are associated with a lower downward flux of longwave radiation, in a comparison with radiation measurements. Analysis of nighttime downward longwave (LW) radiation (Fig. S4 in the online supplemental material) shows that the ACM2 scheme underestimates the LW radiation, particularly in the arid areas, consistent with the observed cold bias. Regarding the other two schemes, both predict a weaker cold bias, showing a better performance overall.
As in Fig. 4, but for minimum temperature.
Citation: Journal of Applied Meteorology and Climatology 62, 9; 10.1175/JAMC-D-22-0108.1
b. Heat extremes
In this part of the study, we analyze the sensitivity of the different PBL schemes in simulating heat extremes.
Figure 6 depicts the ERA5 mean values and the absolute biases of the different schemes for the TXx index. A first look at model results indicates a warm bias in all simulations that has already been seen in the summer mean TX (Fig. 4). Here, the biases tend to be more uniform, with WRF predicting higher values for most of the MENA region and with magnitudes exceeding 8°C in a few locations. Larger biases are found in areas with more complex orography or different land surface characteristics (Nile Delta) and could be related to the performance of the land surface scheme (Constantinidou et al. 2020a). Spots with cold biases can be found mainly in mountainous, temperate areas in the northern part of the domain. These cold biases are not present in the ACM2 simulation, which is much warmer in the Mediterranean region, especially in the coastal areas. The MYJ scheme, on the other hand, exhibits substantial warm biases in the whole eastern part of the domain. In the southern part of the region, a general agreement is found for the three PBL schemes in contrast to the average summer maximum temperatures. This is attributed to the computation of the TXx index that accounts for temperature all year round. The southern part of our study domain is influenced by the seasonal march of temperature in the tropics, with the annual maximum not necessarily recorded in the summer months.
As in Fig. 4, but for TXx.
Citation: Journal of Applied Meteorology and Climatology 62, 9; 10.1175/JAMC-D-22-0108.1
For the TNx index (Fig. 7), the model biases differ notably among the PBL schemes and geographic locations. The WRF Model produces cold biases in many areas (in contrast with TXx), which is expected from the analysis of TN. Especially in the eastern part of the domain (Middle East and Anatolia), all schemes underestimate the TNx index. For ACM2, the largest cold biases are obtained in the arid areas (Sahara and Middle East) exceeding 6°C (e.g., in Saudi Arabia). On the other hand, in southern Europe, the ACM2 is again the warmest scheme simulating a higher TNx index mainly in the Balkans. YSU and MYJ perform better in the Middle East showing weaker biases, while they are much warmer in North Africa, with the YSU being the warmest. In southern Europe, both schemes are in good agreement with ERA5.
As in Fig. 4, but for TNx.
Citation: Journal of Applied Meteorology and Climatology 62, 9; 10.1175/JAMC-D-22-0108.1
Apart from the temperature-based indices, the HI is also examined in order to account for the effect of humidity. For our calculations, we used hourly data of temperature and relative humidity (RH) from the ERA5 reanalysis and the WRF output, and then, the maximum values of HI were derived. The main results for the annual maximum of HI (HIx) are illustrated in Fig. 8.
As in Fig. 4, but for HIx.
Citation: Journal of Applied Meteorology and Climatology 62, 9; 10.1175/JAMC-D-22-0108.1
A different spatial distribution of biases is evident when compared with the analysis of indices that depend solely on temperature. These differences can be attributed to the humidity patterns as shown in section 4d, with the MYJ scheme being the wettest and the ACM2 being the driest scheme. Therefore, we note that MYJ simulates much higher HIx in the arid areas of the region (Sahara and Middle East) with warm biases approaching 10°C in locations over northeast Africa. On the other hand, the ACM2 simulation predicts colder HIx in many arid areas (e.g., Gulf region), showing better performance over the Sahara Desert, with small biases from the ERA5. In the YSU scheme, the spatial patterns of biases, albeit weaker, are in agreement with the ones in MYJ. For the more temperate Mediterranean areas, MYJ and YSU show slightly lower HIx on average, while the ACM2 is warmer, resembling the patterns apparent in TX and TXx.
c. Statistical evaluation
The different statistical metrics described in the methodology section were applied to all grid points of the analysis domain for the model output and the ERA5, and their MENA domain average values are presented in Table 1. For the summer TX and TN, the YSU scheme has the best performance, mainly according to MAE and MIA, and the ACM2 is the worst performing, confirming the main points of the previous visual analysis. A similar performance is evident in the TNx index, while for the TXx index, the differences in the metrics among the three schemes are smaller. This can be attributed to the larger errors of the YSU in the Middle East relative to ACM2, which performs better. Regarding the HIx index, the ACM2 performs best in MAE and MIA, in contrast to its lowest scores for those metrics in all other variables. The ACM2 includes the lowest errors in standard deviation for all variables (apart from HIx), indicating a better representation of the observed interannual variability relative to the other two schemes.
Statistical metrics for TX, TN, TXx, TNx, and HIx for the WRF simulations with the three PBL schemes (YSU, MYJ, and ACM2) averaged over the MENA domain. The ones indicating best performance (with smallest MAE and STDE and largest MIA) are in boldface type.
Table 2 shows the percentage of grid points of the analysis domain with a good agreement (MAE < 1) between ERA5 and WRF simulations for each scheme. Regarding the TX and TN, the YSU scheme appears to be the best performing, while ACM2 performs the worst. For the YSU scheme, the percentage of low-bias grid points reaches 30% and 40% for TX and TN, respectively. The same pattern is observed for the TNx index, while for the TXx index, all schemes are presenting low values of good-performing grid points. Similar to Table 1, the ACM2 simulation seems to be in better agreement with ERA5 for the HIx index. However, since the heat index is a combination of two variables (temperature and relative humidity), the model may provide the right average value for the wrong reason, as a result of a compensation of errors.
Percentage of grid points with MAE < 1 relative to the ERA5. The best-performing scheme for each variable is in boldface type.
d. Specific humidity
We next expand our analysis with additional variables to support interpretation of the previous results. Figure 9 shows the summer climatology of surface-specific humidity (top left), as well as the biases of each scheme relative to ERA5.
As in Fig. 4, but for specific humidity.
Citation: Journal of Applied Meteorology and Climatology 62, 9; 10.1175/JAMC-D-22-0108.1
The ERA5 map shows an expected spatial distribution of specific humidity with very low values (<6 g kg−1) over the Sahara Desert and the Middle East and higher values (8–10 g kg−1) around the Mediterranean and particularly in sub-Saharan Africa (>10 g kg−1). The largest discrepancies among the three schemes appear around sub-Saharan Africa, where MYJ overestimates specific humidity, while the opposite is true for the ACM2 scheme. The YSU biases are moderate and in better agreement with ERA5. MYJ is also the moistest scheme in other areas like southern Europe, where it exhibits weak wet biases and, on the other hand, ACM2 seems to be the driest scheme.
These wet biases of the MYJ scheme are in line with previous studies that indicate that the MYJ simulations overestimated surface humidity (Hu et al. 2010; Garcia-Díez et al. 2013; Avolio et al. 2017). As a local closure scheme, MYJ produces weaker vertical mixing (Brown 1996) than do the nonlocal YSU and ACM2 schemes, and, therefore, it entrains less warm and dry air into the PBL. Furthermore, the large differences in surface moisture derived for sub-Saharan Africa can be linked to the model representation of the West African monsoon (WAM) patterns. Klein et al. (2015) showed that the choice of the PBL scheme strongly affects the position of the WAM rainband with a southward shift in the ACM2 scheme and a northward shift in the MYJ scheme. Figure S7 in the online supplemental material illustrates that it is obvious that the MYJ scheme overestimates the amount of precipitation along sub-Saharan Africa, while ACM2 is significantly drier. This is also consistent with the moisture and maximum temperature–simulated patterns. In the following sections, we will further investigate the entrainment processes within the PBL via the estimation of PBL thickness and vertical profiles.
e. PBL height
One way to further investigate the physical mechanisms related to the temperature biases is via the estimation of the PBLH. According to relevant studies, PBLH could contribute to model temperature biases (Hu et al. 2010; Garcia-Díez et al. 2013; Wei et al. 2017), as it is considered a good indicator of the degree of the thermally induced turbulent mixing and entrainment (Stull 1988). In general, weaker vertical mixing hinders PBL growth and produces insufficient entrainment of warmer and drier free-tropospheric air into the PBL, which leads to cold and moist PBLs. The opposite situation holds true for the regions with warmer and drier PBLs.
Usually, the transition from the PBL to the free atmosphere can be distinguished by a significant change in temperature and water vapor. Therefore, the simplest approach for estimating PBLH is to identify where the greatest change occurs in the vertical profile. Following recent studies (Hu et al. 2010; Garcia-Díez et al. 2013), the 1.5-theta-increase method was applied in order to derive the PBLH. For this, we use potential temperature vertical profiles at each grid point, and subsequently, the PBL top height can be defined as the point where the potential temperature first exceeds the minimum potential temperature within the boundary layer by 1.5 K (Nielsen-Gammon et al. 2008). This method is favored above others, especially for well-developed mixed layers, which is the case in our domain during summer days (Garcia-Carreras et al. 2015; Zhou et al. 2020).
Figure 10 shows PBLH maps for the three summer months at 1200 UTC, approximately at the formation of the deep daytime mixed layer. The ERA5 climatology map (top left) shows the deepest PBLs (up to 5 km) in the Middle East and parts of the Sahara Desert. This is in agreement with relevant studies showing that PBLH can stretch over 4 km in daytime in these areas (Gamo 1996; Marsham et al. 2008). All schemes are simulating deeper PBLs in comparison with ERA5 in the northern parts of Africa and in the Middle East, which is consistent with the warm biases that are observed in these areas (Fig. 4). Large discrepancies can be found in the southern parts of the Sahara and sub-Saharan Africa. In these areas, the ACM2 scheme overestimates the PBLH, while the other two schemes predict lower PBLH. Also in southern Europe, ACM2 is predicting the highest, and MYJ the lowest, PBLH. This supports the hypothesis that temperature biases are related to differences in vertical mixing strength and entrainment. However, in parts of the Middle East, ACM2 is predicting deeper PBL while showing weaker warm biases when compared with the other two. An analysis of surface sensible heat flux values (Fig. S5 in the online supplemental material) does not show any difference between the three schemes in these areas and could in part explain this inconsistency.
As in Fig. 4, but for PBLH at 1200 UTC.
Citation: Journal of Applied Meteorology and Climatology 62, 9; 10.1175/JAMC-D-22-0108.1
f. Vertical profiles
To obtain a more detailed view of the behavior of the different PBL schemes and further investigate entrainment processes, the vertical profiles of temperature and moisture have been analyzed.
Figures 11 and 12 show the 2001–10 climatology profiles of potential temperature, which has already been used for the PBLH estimation, and specific humidity for four different locations with different climate types and land surface characteristics. All profiles show averaged values for the summer months during daytime at 1200 UTC. Color lines represent the three different WRF simulations and dots represent the ERA5 values.
Mean profiles of (top) potential temperature and (bottom) moisture at 1200 UTC for the summer months (JJA) and for land locations in the central Mediterranean and sub-Sahara.
Citation: Journal of Applied Meteorology and Climatology 62, 9; 10.1175/JAMC-D-22-0108.1
As in Fig. 11, but for two locations in the Middle East.
Citation: Journal of Applied Meteorology and Climatology 62, 9; 10.1175/JAMC-D-22-0108.1
More specifically, vertical profiles in Fig. 11 refer to locations in central Italy (left) and in sub-Saharan Africa (right). Profiles with colder conditions near the surface predict higher moisture as well. Among the three schemes, the MYJ, a local closure scheme, predicts the lower temperature and higher moisture at the lowest levels. Local closure schemes are reported to underestimate vertical mixing in the convective boundary layer (Brown 1996). This leads to a less efficient transport of water vapor to the higher levels, along with a weaker entrainment of warm and dry air from the free troposphere. In the case of the central Italy profiles, both the ERA5 and MYJ scheme show larger values of moisture near the surface, while they exhibit less moisture aloft. On the other hand, the YSU and ACM2 schemes are much drier and warmer in the lower troposphere. This is in line with other studies (Hu et al. 2010; Garcia-Díez et al. 2013), which found that nonlocal schemes tend to entrain warmer and drier air into the PBL. Profiles in sub-Saharan Africa show larger discrepancies in the three schemes from the ERA5 reanalysis. The ACM2 scheme simulates much warmer and drier conditions than the ERA5, indicative of much stronger vertical mixing, as also demonstrated with the overestimation of PBL thickness in those areas (Fig. 10).
Figure 12 depicts the mean profiles for locations in Saudi Arabia (left) and the Cairo area (right). A pattern different from the one in Fig. 11 is apparent here, with the colder profiles not being the wettest ones and vice versa. As shown in the potential temperature profiles for the Cairo area, MYJ is the warmest among the three schemes, while it predicts more moisture than the other two schemes. Below 2000 m, all three parameterizations predict warmer temperatures and more moisture than the ERA5, implying a common cause for this bias. ERA5 shows more moisture near the surface, while being drier in the higher levels, which indicates weaker heat and moisture vertical transport relative to the WRF simulations. This could be related to land surface characteristics around the Nile Delta. This is also evident from the latent heat flux maps (Fig. S6 in the online supplemental material). In these maps, WRF simulations show lower values of latent heat flux around the Nile Delta relative to ERA5, implying reduced evaporation from the surface. In Saudi Arabian profiles, ACM2 is the driest, despite being the coldest among the three schemes. As illustrated in Fig. 10, ACM2 also predicts the highest PBL during daytime, suggesting more entrainment of warmer and drier air into the PBL.
5. Summary and conclusions
We have performed and analyzed decadal RCM simulations (2001–10) for the MENA-CORDEX domain to assess the performance of three PBL schemes employed (MYJ, YSU, and ACM2) for climate applications focusing on surface air temperature and temperature extremes. For this, the WRF Model was applied at a 24-km resolution and driven by the ERA Interim reanalysis. For the evaluation of the WRF runs, we used related meteorological variables from the ERA5 reanalysis, which was shown to be a reliable, near-observed alternative for surface temperature while it also provides additional variables, not available in other types of datasets (e.g., gridded observations or satellite data).
The model output tends to overpredict maximum (TX) and underpredict minimum (TN) summer-average temperatures. Comparison of the ERA5 dataset with station measurements reveals that ERA5 underestimates TX (−1.7°C) and slightly overestimates TN (+0.5°C) and, therefore, that model biases are possibly lower. Maps of biases reveal notable differences among the three PBL schemes with regard to the magnitude and the sign. For TX, the ACM2 scheme shows warm biases throughout the simulation domain, while YSU and MYJ have cold biases in sub-Saharan Africa and southern Europe. For TN, ACM2 simulates the lowest TN, with MYJ and YSU being warmer.
The performance of the three PBL schemes in simulating temperature extremes was also scrutinized. Model biases for the warmest-day (TXx) index follow the spatial distribution that was observed in summer TX. Large positive biases (>6°C) were found in areas with complex orography and coastlines, with the ACM2 scheme showing the warmest biases overall. Colder biases were found for the warmest-night (TNx) index, with the sign and magnitude of model biases in each scheme strongly dependent on the geographic location. ACM2 is the coldest scheme mostly in the arid areas (biases exceeding 5°C), while it is warmer in southern Europe, where YSU and MYJ are in good agreement with ERA5. Very different spatial patterns of biases were found for the annual maximum of heat index (Hix), owing to the effect of the humidity, with MYJ being the warmest, while ACM2 was found to be the coldest scheme.
Further interpretation of the model biases was achieved via the analysis of additional variables. Analysis of surface-specific humidity showed the MYJ to be the wettest scheme and ACM2 the driest one. The MYJ wet biases can be attributed to weaker vertical mixing and entrainment, common in local closure schemes (Brown 1996; Hu et al. 2010; Garcia-Díez et al. 2013). The YSU scheme is drier than MYJ, wetter than ACM2, and in better agreement with ERA5. These large discrepancies among the three schemes in sub-Saharan Africa could be linked to the model representation of the West African monsoon patterns (Klein et al. 2015). This could also help explain the large differences regarding the simulation of TX that were noted before.
A qualitative estimation of the model entrainment was obtained by inspecting spatial variations of the simulated PBL height. WRF overestimates the PBLH in most of North Africa and the Middle East, suggesting stronger entrainment of warm and dry free-tropospheric air into the PBL. This is connected to the model’s warm bias in these areas. On average, ACM2 produces thicker PBLs than the other two schemes, consistent with the warmer temperature simulated by this scheme. On the other hand, MYJ, which is the coldest scheme, on average simulates lower PBLH as well.
Vertical profiles of potential temperature and specific humidity show that colder schemes predict more surface moisture while they are drier at higher levels and, thus, provide reduced transport of water vapor and weaker entrainment. This pattern is not applicable, though, to the profile over Saudi Arabia, in which ACM2 is the driest despite being the coldest scheme. On the other hand, all profiles over the Nile Delta are warmer than ERA5 due to differences in the representation of land surface moist processes.
The present study showed that model errors are dependent on the geographic location and on the variable being studied, as different patterns were observed for maximum and minimum temperatures. Therefore, the performance of the different PBL schemes strongly varies according to the atmospheric conditions that prevail at the different regions within MENA and times of the day. Overall, the YSU scheme can be identified as the best-performing scheme for the MENA-CORDEX domain. This is also evident from the quantitative analysis performed, in which the YSU scheme shows higher scores of statistical metrics for most variables. Given the large extent of the simulation domain, different schemes can perform better in specific locations, and this must be considered when selecting a suitable model configuration for subregional climate modeling studies.
Acknowledgments.
This work was cofunded by the European Regional Development Fund and the Republic of Cyprus through the Research Innovation Foundation CELSIUS Project EXCELLENCE/1216/0039. It was also supported by the EMME-CARE project that has received funding from the European Union’s Horizon 2020 Research and Innovation Programme, under Grant Agreement 856612, as well as matching cofunding by the government of the Republic of Cyprus. The authors declare no conflict of interest.
Data availability statement.
All observational and reanalyses data can be assessed from the internet sources mentioned in the text. The model data can be obtained by request to the corresponding author.
Footnotes
It is not possible to use a single common surface layer scheme in the WRF Model because certain PBL schemes are tied to specific surface layer schemes (Skamarock and Klemp 2008).
REFERENCES
Ahmadalipour, A., and H. Moradkhani, 2018: Escalating heat-stress mortality risk due to global warming in the Middle East and North Africa (MENA). Environ. Int., 117, 215–225, https://doi.org/10.1016/j.envint.2018.05.014.
Almazroui, M., 2020: Summer maximum temperature over the Gulf Cooperation Council states in the twenty-first century: Multimodel simulations overview. Arabian J. Geosci., 13, 477, https://doi.org/10.1007/s12517-020-05537-x.
Avolio, E., S. Federico, M. M. Miglietta, T. L. Feudo, C. R. Calidonna, and A. M. Sempreviva, 2017: Sensitivity analysis of WRF Model PBL schemes in simulating boundary-layer variables in southern Italy: An experimental campaign. Atmos. Res., 192, 58–71, https://doi.org/10.1016/j.atmosres.2017.04.003.
Branch, O., K. Warrach-Sagi, V. Wulfmeyer, and S. Cohen, 2014: Simulation of semi-arid biomass plantations and irrigation using the WRF-NOAH Model—A comparison with observations from Israel. Hydrol. Earth Syst. Sci., 18, 1761–1783, https://doi.org/10.5194/hess-18-1761-2014.
Branch, O., T. Schwitalla, M. Temimi, R. Fonseca, N. Nelli, M. Weston, J. Milovac, and V. Wulfmeyer, 2021: Seasonal and diurnal performance of daily forecasts with WRF v3.8.1 over the United Arab Emirates. Geosci. Model Dev., 14, 1615–1637, https://doi.org/10.5194/gmd-14-1615-2021.
Brown, A. R., 1996: Evaluation of parametrization schemes for the convective boundary layer using large-eddy simulation results. Bound.-Layer Meteor., 81, 167–200, https://doi.org/10.1007/BF00119064.
Bucchignani, E., L. Cattaneo, H.-J. Panitz, and P. Mercogliano, 2016: Sensitivity analysis with the regional climate model COSMO-CLM over the CORDEX-MENA domain. Meteor. Atmos. Phys., 128, 73–95, https://doi.org/10.1007/s00703-015-0403-3.
Cohen, A. E., S. M. Cavallo, M. C. Coniglio, and H. E. Brooks, 2015: A review of planetary boundary layer parameterization schemes and their sensitivity in simulating southeastern U.S. cold season severe weather environments. Wea. Forecasting, 30, 591–612, https://doi.org/10.1175/WAF-D-14-00105.1.
Constantinidou, K., P. Hadjinicolaou, G. Zittis, and J. Lelieveld, 2016: Effects of climate change on the yield of winter wheat in the eastern Mediterranean and Middle East. Climate Res., 69, 129–141, https://doi.org/10.3354/cr01395.
Constantinidou, K., P. Hadjinicolaou, G. Zittis, and J. Lelieveld, 2020a: Performance of land surface schemes in the WRF Model for climate simulations over the MENA-CORDEX domain. Earth Syst. Environ., 4, 647–665, https://doi.org/10.1007/s41748-020-00187-1.
Constantinidou, K., P. Hadjinicolaou, G. Zittis, and J. Lelieveld, 2020b: Sensitivity of simulated climate over the MENA region related to different land surface schemes in the WRF Model. Theor. Appl. Climatol., 141, 1431–1449, https://doi.org/10.1007/s00704-020-03258-5.
Dee, D. P., and Coauthors, 2011: The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Quart. J. Roy. Meteor. Soc., 137, 553–597, https://doi.org/10.1002/qj.828.
Donahue, A. S., and P. M. Caldwell, 2018: Impact of physics parameterization ordering in a global atmosphere model. J. Adv. Model. Earth Syst., 10, 481–499, https://doi.org/10.1002/2017MS001067.
Dudhia, J., 2014: A history of mesoscale model development. Asia-Pac. J. Atmos. Sci., 50, 121–131, https://doi.org/10.1007/s13143-014-0031-8.
Dunne, J. P., R. J. Stouffer, and J. G. John, 2013: Reductions in labour capacity from heat stress under climate warming. Nat. Climate Change, 3, 563–566, https://doi.org/10.1038/nclimate1827.
El-Samra, R., E. Bou-Zeid, H. K. Bangalath, G. Stenchikov, and M. El-Fadel, 2018: Seasonal and regional patterns of future temperature extremes: High-resolution dynamic downscaling over a complex terrain. J. Geophys. Res. Atmos., 123, 6669–6689, https://doi.org/10.1029/2017JD027500.
Fekih, A., and A. Mohamed, 2019: Evaluation of the WRF Model on simulating the vertical structure and diurnal cycle of the atmospheric boundary layer over Bordj Badji Mokhtar (southwestern Algeria). J. King Saud Univ. Sci., 31, 602–611, https://doi.org/10.1016/j.jksus.2017.12.004.
Fonseca, R., and Coauthors, 2020: On the analysis of the performance of WRF and NICAM in a hyperarid environment. Wea. Forecasting, 35, 891–919, https://doi.org/10.1175/WAF-D-19-0210.1.
Gamo, M., 1996: Thickness of the dry convection and large-scale subsidence above deserts. Bound.-Layer Meteor., 79, 265–278, https://doi.org/10.1007/BF00119441.
Garcia-Carreras, L., and Coauthors, 2015: The turbulent structure and diurnal growth of the Saharan atmospheric boundary layer. J. Atmos. Sci., 72, 693–713, https://doi.org/10.1175/JAS-D-13-0384.1.
Garcia-Díez, M., L. Fita, and C. Yag, 2013: Seasonal dependence of WRF Model biases and sensitivity to PBL schemes over Europe. Quart. J. Roy. Meteor. Soc., 139, 501–514, https://doi.org/10.1002/qj.1976.
Giorgi, F., and W. J. Gutowski Jr., 2015: Regional dynamical downscaling and the CORDEX initiative. Annu. Rev. Environ. Resour., 40, 467–490, https://doi.org/10.1146/annurev-environ-102014-021217.
Hersbach, H., and Coauthors, 2020: The ERA5 global reanalysis. Quart. J. Roy. Meteor. Soc., 146, 1999–2049, https://doi.org/10.1002/qj.3803.
Hong, S.-Y., and J.-O. J. Lim, 2006: The WRF single-moment 6-class microphysics scheme (WSM6). Asia-Pac. J. Atmos. Sci., 42, 129–151.
Hong, S.-Y., Y. Noh, and J. Dudhia, 2006: A new vertical diffusion package with an explicit treatment of entrainment processes. Mon. Wea. Rev., 134, 2318–2341, https://doi.org/10.1175/MWR3199.1.
Hu, X.-M., J. W. Nielsen-Gammon, and F. Zhang, 2010: Evaluation of three planetary boundary layer schemes in the WRF Model. J. Appl. Meteor. Climatol., 49, 1831–1844, https://doi.org/10.1175/2010JAMC2432.1.
Iacono, M. J., J. S. Delamere, E. J. Mlawer, M. W. Shephard, S. A. Clough, and W. D. Collins, 2008: Radiative forcing by long-lived greenhouse gases: Calculations with the AER radiative transfer models. J. Geophys. Res., 113, D13103, https://doi.org/10.1029/2008JD009944.
Janjić, Z. I., 1994: The step-mountain eta coordinate model: Further developments of the convection, viscous sublayer, and turbulence closure schemes. Mon. Wea. Rev., 122, 927–945, https://doi.org/10.1175/1520-0493(1994)122<0927:TSMECM>2.0.CO;2.
Kala, J., J. Andrys, T. J. Lyons, I. J. Foster, and B. J. Evans, 2015: Sensitivity of WRF to driving data and physics options on a seasonal time-scale for the southwest of Western Australia. Climate Dyn., 44, 633–659, https://doi.org/10.1007/s00382-014-2160-2.
Karl, T. R., R. W. Knight, D. R. Easterling, and R. G. Quayle, 1996: Indices of climate change for the United States. Bull. Amer. Meteor. Soc., 77, 279–292, https://doi.org/10.1175/1520-0477(1996)077<0279:IOCCFT>2.0.CO;2.
Katragkou, E., and Coauthors, 2015: Regional climate hindcast simulations within EURO-CORDEX: Evaluation of a WRF multi-physics ensemble. Geosci. Model Dev., 8, 603–618, https://doi.org/10.5194/gmd-8-603-2015.
Klein, C., D. Heinzeller, J. Bliefernicht, and H. Kunstmann, 2015: Variability of West African monsoon patterns generated by a WRF multi-physics ensemble. Climate Dyn., 45, 2733–2755, https://doi.org/10.1007/s00382-015-2505-5.
Lavin-Gullon, A., and Coauthors, 2021: Internal variability versus multi-physics uncertainty in a regional climate model. Int. J. Climatol., 41, E656–E671, https://doi.org/10.1002/joc.6717.
Legasa, M. N., and Coauthors, 2020: Assessing multidomain overlaps and grand ensemble generation in CORDEX regional projections. Geophys. Res. Lett., 47, e2019GL086799, https://doi.org/10.1029/2019GL086799.
Legates, D. R., and G. J. McCabe Jr., 1999: Evaluating the use of “goodness-of-fit” measures in hydrologic and hydroclimatic model validation. Water Resour. Res., 35, 233–241, https://doi.org/10.1029/1998WR900018.
Lelieveld, J., and Coauthors, 2012: Climate change and impacts in the eastern Mediterranean and the Middle East. Climatic Change, 114, 667–687, https://doi.org/10.1007/s10584-012-0418-4.
Lelieveld, J., Y. Proestos, P. Hadjinicolaou, M. Tanarhte, E. Tyrlis, and G. Zittis, 2016: Strongly increasing heat extremes in the Middle East and North Africa (MENA) in the 21st century. Climatic Change, 137, 245–260, https://doi.org/10.1007/s10584-016-1665-6.
Lionello, P., and L. Scarascia, 2018: The relation between climate change in the Mediterranean region and global warming. Reg. Environ. Change, 18, 1481–1493, https://doi.org/10.1007/s10113-018-1290-1.
Marsham, J. H., D. J. Parker, C. M. Grams, C. M. Taylor, and J. M. Haywood, 2008: Uplift of Saharan dust south of the intertropical discontinuity. J. Geophys. Res., 113, D21102, https://doi.org/10.1029/2008JD009844.
Miralles, D. G., A. J. Teuling, C. C. Van Heerwaarden, and J. V.-G. de Arellano, 2014: Mega-heatwave temperatures due to combined soil desiccation and atmospheric heat accumulation. Nat. Geosci., 7, 345–349, https://doi.org/10.1038/ngeo2141.
Nelli, N. R., and Coauthors, 2020: Impact of roughness length on WRF simulated land-atmosphere interactions over a hyper-arid region. Earth Space Sci., 7, e2020EA001165, https://doi.org/10.1029/2020EA001165.
Nielsen-Gammon, J. W., and Coauthors, 2008: Multisensor estimation of mixing heights over a coastal city. J. Appl. Meteor. Climatol., 47, 27–43, https://doi.org/10.1175/2007JAMC1503.1.
Niu, G.-Y., and Coauthors, 2011: The community Noah land surface model with multiparameterization options (Noah-MP): 1. Model description and evaluation with local-scale measurements. J. Geophys. Res., 116, D12109, https://doi.org/10.1029/2010JD015139.
Ntoumos, A., P. Hadjinicolaou, G. Zittis, and J. Lelieveld, 2020: Updated assessment of temperature extremes over the Middle East–North Africa (MENA) region from observational and CMIP5 data. Atmosphere, 11, 813, https://doi.org/10.3390/atmos11080813.
Ntoumos, A., P. Hadjinicolaou, G. Zittis, Y. Proestos, and J. Lelieveld, 2022: Projected air temperature extremes and maximum heat conditions over the Middle-East-North Africa (MENA) region. Earth Syst. Environ., 6, 343–359, https://doi.org/10.1007/s41748-022-00297-y.
Odnoletkova, N., and T. W. Patzek, 2021: Data-driven analysis of climate change in Saudi Arabia: Trends in temperature extremes and human comfort indicators. J. Appl. Meteor. Climatol., 60, 1055–1070, https://doi.org/10.1175/JAMC-D-20-0273.1.
Oettli, P., B. Sultan, C. Baron, and M. Vrac, 2011: Are regional climate models relevant for crop yield prediction in West Africa? Environ. Res. Lett., 6, 014008, https://doi.org/10.1088/1748-9326/6/1/014008.
Ozturk, T., F. S. Saygili-Araci, and M. L. Kurnaz, 2021: Projected changes in extreme temperature and precipitation indices over CORDEX-MENA domain. Atmosphere, 12, 622, https://doi.org/10.3390/atmos12050622.
Pal, J. S., and E. A. B. Eltahir, 2016: Future temperature in southwest Asia projected to exceed a threshold for human adaptability. Nat. Climate Change, 6, 197–200, https://doi.org/10.1038/nclimate2833.
Pleim, J. E., 2007: A combined local and nonlocal closure model for the atmospheric boundary layer. Part I: Model description and testing. J. Appl. Meteor. Climatol., 46, 1383–1395, https://doi.org/10.1175/JAM2539.1.
Rothfus, L. P., 1990: The heat index “equation” (or, more than you ever wanted to know about heat index). National Weather Service Tech. Attachment SR 90-23, 2 pp., https://www.weather.gov/media/ffc/ta_htindx.PDF.
Skamarock, W. C., and J. B. Klemp, 2008: A time-split nonhydrostatic atmospheric model for weather research and forecasting applications. J. Comput. Phys., 227, 3465–3485, https://doi.org/10.1016/j.jcp.2007.01.037.
Steadman, R. G., 1979: The assessment of sultriness. Part I: A temperature-humidity index based on human physiology and clothing science. J. Appl. Meteor., 7, 861–873, https://doi.org/10.1175/1520-0450(1979)018<0861:TAOSPI>2.0.CO;2.
Stull, R. B., 1988: An Introduction to Boundary Layer Meteorology. Kluwer Academic, 666 pp.
Tanarhte, M., P. Hadjinicolaou, and J. Lelieveld, 2012: Intercomparison of temperature and precipitation data sets based on observations in the Mediterranean and the Middle East. J. Geophys. Res., 117, D12102, https://doi.org/10.1029/2011JD017293.
Vautard, R., and Coauthors, 2021: Evaluation of the large EURO-CORDEX Regional Climate Model ensemble. J. Geophys. Res. Atmos., 126, e2019JD032344, https://doi.org/10.1029/2019JD032344.
Wei, N., L. Zhou, and Y. Dai, 2017: Evaluation of simulated climatological diurnal temperature range in CMIP5 models from the perspective of planetary boundary layer turbulent mixing. Climate Dyn., 49, 1–22, https://doi.org/10.1007/s00382-016-3323-0.
Weston, M., N. Chaouch, V. Valappil, M. Temimi, M. Ek, and W. Zheng, 2019: Assessment of the sensitivity to the thermal roughness length in Noah and Noah-MP land surface model using WRF in an arid region. Pure Appl. Geophys., 176, 2121–2137, https://doi.org/10.1007/s00024-018-1901-2.
Willmott, C. J., 1981: On the validation of models. Phys. Geogr., 2, 184–194, https://doi.org/10.1080/02723646.1981.10642213.
Willmott, C. J., and K. Matsuura, 2005: Advantages of the mean absolute error (MAE) over the root mean square error (RMSE) in assessing average model performance. Climate Res., 30, 79–82, https://doi.org/10.3354/cr030079.
Xu, Z., Y. Han, and Z. Yang, 2019: Dynamical downscaling of regional climate: A review of methods and limitations. Sci. China Earth Sci., 62, 365–375, https://doi.org/10.1007/s11430-018-9261-5.
Zhang, Y., D. J. Seidel, and S. Zhang, 2013: Trends in planetary boundary layer height over Europe. J. Climate, 26, 10 071–10 076, https://doi.org/10.1175/JCLI-D-13-00108.1.
Zhao, Y., W. Mao, K. Zhang, Y. Ma, H. Li, and W. Zhang, 2017: Climatic variations in the boundary layer height of arid and semiarid areas in East Asia and North Africa. J. Meteor. Soc. Japan, 95, 181–197, https://doi.org/10.2151/jmsj.2017-010.
Zhou, L., Y. Tian, N. Wei, S.-p. Ho, and J. Li, 2020: Rising planetary boundary layer height over the Sahara Desert and Arabian Peninsula in a warming climate. J. Climate, 34, 4043–4068, https://doi.org/10.1175/JCLI-D-20-0645.1.
Zittis, G., 2018: Observed rainfall trends and precipitation uncertainty in the vicinity of the Mediterranean, Middle East and North Africa. Theor. Appl. Climatol., 134, 1207–1230, https://doi.org/10.1007/s00704-017-2333-0.
Zittis, G., and P. Hadjinicolaou, 2017: The effect of radiation parameterization schemes on surface temperature in regional climate simulations over the MENA-CORDEX domain. Int. J. Climatol., 37, 3847–3862, https://doi.org/10.1002/joc.4959.
Zittis, G., P. Hadjinicolaou, and J. Lelieveld, 2014: Comparison of WRF Model physics parameterizations over the MENA-CORDEX domain. Amer. J. Climate Change, 3, 490–511, https://doi.org/10.4236/ajcc.2014.35042.
Zittis, G., P. Hadjinicolaou, M. Fnais, and J. Lelieveld, 2016: Projected changes in heat wave characteristics in the eastern Mediterranean and the Middle East. Reg. Environ. Change, 16, 1863–1876, https://doi.org/10.1007/s10113-014-0753-2.
Zittis, G., P. Hadjinicolaou, M. Klangidou, Y. Proestos, and J. Lelieveld, 2019: A multi-model, multi-scenario, and multi-domain analysis of regional climate projections for the Mediterranean. Reg. Environ. Change, 19, 2621–2635, https://doi.org/10.1007/s10113-019-01565-w.
Zittis, G., and Coauthors, 2021: Business-as-usual will lead to super and ultra-extreme heatwaves in the Middle East and North Africa. npj Climate Atmos. Sci., 4, 20, https://doi.org/10.1038/s41612-021-00178-7.
Zittis, G., and Coauthors, 2022: Climate change and weather extremes in the eastern Mediterranean and Middle East. Rev. Geophys., 60, e2021RG000762, https://doi.org/10.1029/2021RG000762.