Simulation of present and future climate of Saudi Arabia using a regional climate model (PRECIS)

Authors

  • Mansour Almazroui

    Corresponding author
    • Center of Excellence for Climate Change Research/Department of Meteorology, King Abdulaziz University, Jeddah, Saudi Arabia
    Search for more papers by this author

Correspondence to: M. Almazroui, Center of Excellence for Climate Change Research/Department of Meteorology, King Abdulaziz University, P. O. Box 80234, Jeddah 21589, Saudi Arabia. E-mail: mansour@kau.edu.sa

ABSTRACT

In this article, climate parameters (rainfall and temperature) are simulated for Saudi Arabia by downscaling the European Centre for Medium-Range Weather Forecasts (ECMWF) 40-year reanalysis (ERA40) and the European Community − Hamburg Atmospheric Model (ECHAM5) data using the UK Met Office Regional Climate Model PRECIS (Providing REgional Climates for Impacts Studies). Simulations are performed for a present climate of 30 years (1971–2000) using ERA40 and ECHAM5, and future climate is predicted for a period of 50 years (2021–2070) using ECHAM5 A1B emissions scenario. The results show that the spatial distribution of the present-day rainfall and temperature simulated by PRECIS are consistent with the observed dataset. In addition, their annual cycle and interannual variability are reasonably well reproduced. The dry precipitation and warm temperature biases exhibited in the driving fields (compared with the observations) are reduced because of the improvements in specific humidity and in the wind field within the PRECIS simulations compared with the driving fields. The projection using the calibrated national average temperature exhibits a positive trend in mean temperature of around 0.65 °C per decade. For rainfall projection, the results show that the coastal areas along the central parts of the Red Sea and the south-southwestern areas of Saudi Arabia may experience more extreme rainfall events, whereas the northern and central parts of the country may undergo a drying trend. © 2013 The Authors. International Journal of Climatology published by John Wiley & Sons Ltd on behalf of the Royal Meteorological Society.

1. Introduction

There is unequivocal evidence from air and sea temperature data that the climate is changing at both global and regional scales. Variations in rainfall and temperature are considered to be the key measures of climate change in any part of the world. The Arabian Peninsula is one of the driest and most water-scarce regions in the world. Although rainfall is considered to be the main source of fresh and renewable water in most regions, the per capita water share of renewable water resources in the Arabian Peninsula is less than 10% of the global average. According to the climate projections of the Intergovernmental Panel on Climate Change (IPCC), in the Arabian Peninsula the annual rainfall may decrease, and the frequency and intensity of extreme events are likely to increase in the future (IPCC, 2007). Recently, the Arabian Peninsula has experienced an increase in extreme events such as warm spells, droughts, flash floods and storm surges (Al-Sarmi and Washington, 2011). Thus, sustainable management of water resources is an obligatory requirement as water scarcity is increasingly becoming a constraint, impeding the socio-economic development of many countries in the Peninsula.

The climate of Saudi Arabia, which is the largest country in the Arabian Peninsula, is characterized by dry and semi-arid areas. The meagre water resources of Saudi Arabia are greatly affected by the adverse impacts of climate change, affecting the country's various socio-economic sectors. Recently, two extreme rainfall events were recorded in the country, both causing localized flash flooding. The first event occurred in the city of Jeddah on 25 November 2009, when 74 mm of rainfall was recorded by the Presidency of Meteorology and Environment (PME) at the city's airport vicinity in the north. The second event occurred on 26 January 2011, when 111 mm of rainfall was recorded in just 4 h by the weather station located in the Department of Meteorology, King Abdulaziz University in the southern part of Jeddah. Both events resulted from cloud bursts and, because of the intensity and rapidity of their occurrence, were responsible for the flash floods in the area. As such events are rare, unpredictable, fast and intense, their anticipation (with sufficient time to broadcast warnings) is a primary subject of concern. Similarly, 2010 was the warmest year on record in Saudi Arabia, where the temperature in Jeddah broke all previous records and reached 52 °C on 22 June 2010, resulting in widespread disruption. It is important to assess the impact of climate change (and the level of socio-economic vulnerability) for the adaptation and mitigation strategies of any such events. Nowadays, climate models are the main tools available for generating projections of climate change scenarios (Houghton et al., 2001) for impact assessment and adaptation studies. Alongside climate projection studies, these models are also very useful in understanding the behaviour of the recent past or base period climate of any region under consideration.

In the past, general circulation models (GCMs) produced climate variables in a coarse horizontal resolution (˜300 km) (Murphy and Mitchell, 1995), while regional climate models (RCMs) were considered the best tool for the dynamic downscaling of climate features in order to obtain detailed information of a particular region (Giorgi and Hewitson, 2001; Jones et al., 2004). Nowadays, GCM resolution is significantly improved and many centres around the world are using GCM resolution comparable to that of RCM (e.g. Cozzetto et al., 2011; Joyce et al., 2011). The use of an RCM is vital because of its low computational cost, data handling and ease of regional analysis. Thus, adopting the RCM approach for the dynamic downscaling of GCM simulations has become an accepted strategy (Pal et al., 2007). Recently, Almazroui (2011a) used an RCM for the case study of a heavy rainfall event in Jeddah, Saudi Arabia. PRECIS (Providing REgional Climates for Impacts Studies) is an RCM for performing the dynamic downscaling of GCM outputs, and thereby generating high-resolution climate variables for a region (Jones et al., 2004).

PRECIS has also been used in different parts of the world, including Bangladesh (Islam, 2009), China (Yinlong et al., 2006), Eastern Mediterranean (Kotroni et al., 2008), India (Kolli et al., 2006), Niger (Beraki, 2005), Pakistan (Islam et al., 2009), South Africa (Hudson and Jones, 2002) and South America (Alves and Marengo, 2010). To date, no state-of-the-art climate model free of uncertainties has yet been developed; i.e. the model output inherently involves some biases compared with the observational datasets. Therefore, it is important that the model-simulated near-term climate predictions should be corrected for the biases (WCRP, 2011). In addition, the uncertainty in model output should be taken into consideration when using model data for application-oriented tasks.

In this study, an attempt is made to project bias-corrected PRECIS-simulated rainfall and temperatures (maximum and minimum) over Saudi Arabia. The work is done mainly to assist impact studies that require dynamic-downscaling outputs with low biases. Thus, the main objective of this work is to construct low-bias RCM scenarios for near-future climate change predictions. It is envisaged that this study would be very useful for policy makers, assisting them in taking the necessary measures to mitigate the possible effects of climate change and to cope with weather-related disasters in the future.

2. Data and methods

2.1. Data used

The surface observed rainfall and air temperature (minimum and maximum) data are obtained from PME in Saudi Arabia for the period 1978–2009. The gridded observed data for precipitation and temperature are retrieved from the Climate Prediction Center (CPC) Merged Analysis of Precipitation (CMAP; Xie and Arkin, 1997) and the Climate Research Unit (CRU; New et al., 2000), respectively. The CMAP (2.5° × 2.5° gridded—available from 1979) and the CRU (0.5° × 0.5° gridded) datasets are used for verification purposes.

2.2. PRECIS model description and design of experiment

PRECIS is a regional modelling system developed by the UK Hadley Centre to provide regional climate information for impact studies (Jones et al., 2004). It is a hydrostatic, primitive equation grid-point model, containing 19 levels described by a hybrid vertical coordinate (Simon et al., 2009). The main variables comprising the lateral boundary forcings for PRECIS are atmospheric pressure, horizontal wind components, temperature and humidity. In this work, the boundary forcings are taken from the European Centre for Medium-Range Weather Forecasts (ECMWF) 40-year reanalysis datasets (1957–2001; 2.5° × 2.5° grid), hereafter referred to as ERA40, and from the ECHAM5 Global Climate Model (Max Planck Institute for Meteorology, Germany) using A1B emission scenario experiments. For the model runs, the experiment set-up is as follows:

  1. The model run simulation is completed for the present climate using both ERA40 and ECHAM5 boundary forcings with 0.44° (˜50 km) resolution. The domain extends from 3°N–38°N and 20°E–62°E, encompassing the Arabian Peninsula and large surrounding areas (see Figure 1). The model run period is from 1 December 1969 to 31 December 2000. The first year of simulation is allowed for spin-up time, and therefore the base period used is from 1971 to 2000.
  2. The future climate model run simulation is done using the ECHAM5 A1B scenarios for the period 1 December 2020 to 31 December 2070, where the first year of simulation is not used in the analysis (spin-up year).
Figure 1.

The analysis domain, which covers Saudi Arabia and its surroundings. The open triangles represent the observed meteorological stations throughout the country.

2.3. Methods of analysis

The simulated rainfall and temperature data are extracted at the coordinates of the observed meteorological stations for the study area (see Figure 1). The extracted data are processed on daily, monthly and annual scales, and objectively compared with the observed data at the same scale. The regression coefficients are obtained for rainfall and for maximum and minimum temperatures of both observed and simulated datasets. The obtained regression coefficients are used in the calibration of the PRECIS output as described by Islam (2009) and Almazroui (2011b). The expressions for the calibration of rainfall and temperatures are as follows:

display math(1)
display math(2)
display math(3)

where Calibrated is the variable to be obtained after calibration, α is the slope and β is the intercept. The corresponding subscripts are used for the rainfall, and the minimum and maximum temperatures. The calibration using the regression equations (Equations (1)-(3)) is termed PRECIS-REGN. The bias calibration method (Equation (4) below) of the World Climate Research Programme (WCRP) is also used in this study (WCRP, 2011). The WCRP calibration method further strengthens the performance of the PRECIS model.

display math(4)

where <X> and <Y> are the long-term climatological averages derived from the observed and modelled variables (for the observed period 1978–2000), j identifies the initial times and r the forecast range. The calibration using Equation (4) is termed as PRECIS-WCRP.

The simulated climatic parameters are compared with the observed data for the calibration period of 1978–2000. Similarly, the projected parameters at station level obtained from the calibrated PRECIS-ECHAM5 A1B output are averaged across the country for period of 2021–2070. Thus, the climate change scenario is obtained for the near future (2021–2070) with respect to the base period.

3. Results and Discussion

3.1. Mean present-day climatology

The distribution of mean annual rainfall of the driving fields (ERA40 and ECHAM5; Figure 2(a) and (b)), PRECIS simulations (ERA40-PRECIS and ECHAM5-PRECIS; Figure 2(c) and (d)) and observations (CMAP and CRU; Figure 2(e) and (f)) shows that the simulation patterns are comparable to the observations. For instance, the dry zone over Egypt, modest rain over the Empty Quarter (Rub Al-Khali: 44°30′E–56°30′E, 16°30′N–23°00′N) in the Arabian Peninsula and the large rainfall belt over Sudan–Ethiopia are similar for both the simulations and observations. Similarly, the light rain in the southeastern and northwestern regions, and the heavy rain in the southwestern region of Saudi Arabia are all visible in both the simulations (ERA40-PRECIS and ECHAM5-PRECIS) and in the CRU. The driving fields ERA40 and ECHAM5 confine the rainfall to the southwest of the Arabian Peninsula, with very light rain over Saudi Arabia and the Arabian Gulf countries; however, these values are much lower than the observations (Figure 2(a), (b), (e) and (f)). The unique characteristic of the Peninsula, the large rain area along the southern Red Sea coast observed in CRU data, is missing in the driving fields, but well captured in ERA40-PRECIS and ECHAM5-PRECIS (Figure 2(c) and (d)). Thus, the simulations show a great improvement in the downscaling of the driving fields for the rainfall climatology over the whole domain. The improvements achieved through PRECIS are clearly visible over Saudi Arabia, where the driving fields show little rain (below 20 mm except for the southwestern region), yet both the simulations and the observations show heavier rain (at least above 40 mm). This demonstrates the capability of PRECIS in downscaling the driving large-scale coarse datasets (ERA40 and ECHAM5) to reproduce the present-day climatological features for rainfall in and around Saudi Arabia. However, the simulated amounts are slightly higher/lower than the gridded CMAP and CRU data for both the ERA40-PRECIS and ECHAM5-PRECIS simulations. It is important to note that the rainfall in the southwestern region is a unique feature of the Arabian Peninsula in that rain falls in both the wet and the dry season (Almazroui, 2011b, Almazroui et al., 2012a); this is well simulated by PRECIS compared with the driving fields.

Figure 2.

The distribution of annual mean rainfall (mm): (a) ERA40 boundary forcing, (b) ECHAM5 boundary forcing, (c) simulated by PRECIS using ERA40 boundary forcing, (d) simulated by PRECIS using ECHAM5 boundary forcing, (e) obtained from CMAP and (f) obtained from CRU. The rainfalls are averaged over 1979–2000. Rainfall amounts below 10 mm are not shown in the legend.

Figure 3 displays the spatial distributions of mean temperatures (°C) averaged over 1971–2000, obtained from the driving fields (ERA40 and ECHAM5), the simulations (ERA40-PRECIS and ECHAM5-PRECIS) and the observation (CRU) data. In general, the observed CRU data show that temperatures are low in the southwestern and northwestern areas of the Arabian Peninsula, but high in the eastern and southeastern parts as well as along the Red Sea coast. It is evident that the simulation results agree well with the CRU data, particularly the low temperatures in the southwestern mountainous regions, which is absent in the driving fields, although the simulation slightly overestimates the temperatures over the sand desert of the Rub Al-Khali. Overall, the simulation tends to produce warmer temperatures over Saudi Arabia compared with the gridded CRU data.

Figure 3.

The distribution of annual mean temperature (°C): (a) ERA40 boundary forcing, (b) ECHAM5 boundary forcing, (c) simulated by PRECIS using ERA40 boundary forcing, (d) simulated by PRECIS using ECHAM5 boundary forcing and (e) obtained from CRU. The temperatures are averaged over 1979–2000.

The reduction of dry precipitation bias and warm temperature in the driving fields by downscaling the simulations is investigated with lower level specific humidity and superimposed wind vectors, as shown in Figure 4. Both the driving fields (ERA40 and ECHAM5) show dryness over the Arabian Peninsula (Figure 4(a) and (b)). Northwesterlies are coming from the Mediterranean region and southwesterlies are entering the land from the Arabian Sea. Two main wind vector flows are creating convergence zones over the middle of the Red Sea, which in turn is favourable for the development of convective rainfall systems. In both simulations, the wind vectors (humidity) became stronger (wetter) compared with the driving fields for the entire domain. Convergence zones in the simulated wind vectors are also observed over the Red Sea, which triggered convection, enhancing precipitation over the region.

Figure 4.

The distribution of 10 m wind field (vector in m s−1) and specific humidity (shade in g kg−1) for: (a) ERA40 boundary forcing, (b) ECHAM5 boundary forcing, (c) simulated by PRECIS using ERA40 boundary forcing, and (d) simulated by PRECIS using ECHAM5 boundary forcing. The wind and specific humidity data are averaged over 1979–2000.

Among the simulations, humidity is slightly higher for ERA40-PRECIS than ECHAM5-PRECIS over the Red Sea, Arabian Sea and Arabian Gulf. The precipitation dry bias and temperature warm bias are reduced in the PRECIS simulations because of the feed of strong wind and moist air from the Arabian Sea and Mediterranean region. Because ERA40-PRECIS is more humid over the bodies of water than ECHAM5-PRECIS, the former overestimated precipitation (see Figure 2(c)) in the Peninsula, whilst the latter has a better result in the reduction of dry bias (see Figure 2(d)). It is concluded that by PRECIS is useful in the reconstruction of climatic variables (rainfall and temperature), relative to the driving fields; the biases are significantly reduced.

Figure 5 displays the annual cycles of rainfall and temperature obtained from surface observations, ERA40 driving forcing and ERA40-PRECIS averaged over Saudi Arabia for the period 1978–2000. It is evident that the ERA40 driving forcing underestimates rainfalls in all months except for August and September, which it overestimates (Figure 5(a)). Although the simulated rainfalls follow the annual cycle obtained from the observed data, the model overestimates rainfall for February through October and slightly underestimates the same for the months of November, December and January (Figure 5(a)). The simulated rainfall amounts are closer to the surface observed amounts for the first 4 months (November to February) of the wet season and only for the first month (June) of the dry season. Overall, the simulated, driving forcing and observed rainfalls are about 12.64, 3.78 and 8.27 mm month−1, respectively. Hence, large uncertainties are evident in the simulated and driving rainfall amounts on the annual scale, seasonal scale and even larger at the monthly scale. The model data, therefore, cannot be utilized directly in impact assessment studies without bias correction. The biases are calculated for each month of the period 1978–2000, and are presented in Table 1, which shows that the ERA40-PRECIS rainfall biases (in %) are small and negative for January, November and December and large and positive for the remaining months. The simulated rainfall bias is about 52.83% on the annual scale with maximum positive bias (331.16%) for the month of July. The ERA40 driving field has negative bias in 8 months; however, the maximum positive bias (328.37% in September) reduced the annual bias to −54.32%. Hence, the simulation's reduced rainfall dry bias of the ERA40 driving field introduced a wet bias over Saudi Arabia. Thus, even though the annual rainfall from the simulated data can be calculated with the same patterns as those of the observed data (see Figure 2), they differ in terms of magnitude at the monthly, seasonal and annual scales. Therefore, irrespective of month and/or season, the ERA40-PRECIS amounts are not free from uncertainties and need further investigation.

Figure 5.

The monthly (a) rainfall (mm/month) and (b) mean temperature (°C) obtained from the observed data, ERA40 driving forcing and simulated by PRECIS using ERA40 boundary forcing. The rainfalls and temperatures are averaged throughout the country over 1978–2000.

Table 1. Rainfall and temperature with biases when PRECIS is driven using ERA40. The bias is calculated for both ERA40 driving and ERA40-PRECIS with respect to the observed data for the calibration period (1978–2000)
 Rainfall (mm month−1)Rainfall bias (%)Temperature (°C)Temperature bias (°C)
ObservedERA40PRECISERA40PRECISObservedERA40PRECISERA40PRECIS
January12.153.9711.07−67.30−8.8415.1514.5114.85−0.64−0.30
February7.523.149.09−58.2620.8516.8116.4616.36−0.35−0.45
March17.785.2921.76−70.2622.3620.1320.1419.600.01−0.53
April15.174.5823.17−69.8152.7224.6724.8524.060.18−0.60
May8.761.0917.47−87.5599.4228.7729.2328.320.46−0.45
June1.680.493.16−70.6288.2331.1531.8530.820.70−0.33
July3.062.5913.18−15.34331.1632.2332.6932.140.46−0.09
August5.078.7016.6671.54228.7632.2032.4231.890.23−0.31
September1.536.556.52328.37325.6030.0830.6330.020.55−0.06
October5.452.609.50−52.3474.3525.7226.0925.670.37−0.05
November11.373.0310.65−73.37−6.3420.4720.3520.16−0.12−0.31
December9.713.319.45−65.96−2.6916.6716.0116.32−0.65−0.34
Annual8.273.7812.64−54.3252.8324.5024.6024.180.10−0.32

As shown in Figure 5(b), the ERA40 driving forcing overestimates Saudi Arabian mean temperature from April to October and underestimates for other months. The ERA40-PRECIS temperatures closely follow the annual cycle with slight underestimations for all months, except for September and October. For ERA40-PRECIS, the lowest/highest cold bias is about 0.05 °C/0.60 °C in October/April (see Table 1). This is because the month of October covers the transition period from dry season to wet season in Saudi Arabia, whilst April is the last month of wet season (Almazroui, 2011b). However, the average cold (warm) bias is about −0.32 °C (0.10 °C) at the annual scale for ERA40-PRECIS (ERA40 driving forcing).

Figure 6 displays the interannual variability of mean annual rainfall (mm) and temperature (°C) obtained from the ERA40 driving field, ERA40-PRECIS simulation and observed datasets, averaged throughout the country for 1978–2000. For interannual variability of mean annual rainfall, the PRECIS simulation significantly improves the rainfall amounts in all years compared with the too-low rainfall values of the ERA40 driving field (Figure 6(a)). From Figure 6(a), it is evident that ERA40-PRECIS (observed) dataset shows an increasing trend of 14.0 (8.3) mm per decade, with highly significant correlation (of 0.84). The PRECIS model shows an excellent capability in simulating the interannual variability of rainfall although the values tend to be larger than the observed datasets; thus, there are still some biases between the two data sources.

Figure 6.

The time sequences of yearly (a) rainfall (mm) and (b) mean temperature (°C) obtained from the observed data, ERA40 driving forcing and simulated by PRECIS using ERA40 boundary forcing. The rainfalls and temperatures are averaged throughout the country.

The time sequences of mean temperature for the period 1978–2000 show that the ERA40 driving field closely follows the observed temperature pattern, with warm bias in all years (Figure 6(b)). The PRECIS simulation reduces the warm bias found in the ERA40 driving field; however, it shows a slightly cold bias relative to the observation. The evident increase in temperature (Figure 6(b)) is found to be at a rate of 0.57 (0.27) °C per decade for the observed (ERA40-PRECIS) data. On average, the model underestimates temperature by 0.30 °C, ranging from 0.15 to −0.71 °C for 1978–2000, with the exception of slight overestimations for 5 years (1979, 1980, 1983, 1984 and 1990), ranging from 0.07 to 0.15 °C. However, the correlation coefficient between simulated and observed data for temperature is 0.88. Furthermore, PRECIS accurately detects the interannual variability of temperature, which is important for climate change impact studies. The mean values of temperature for the period 1978–2000, for the observed, ERA40 and ERA40-PRECIS are 24.33, 24.74 and 24.04 °C, respectively.

The time sequence of the daily rainfall obtained from PRECIS-ERA40 averaged over all the studied stations for 1978–2000 is displayed in Figure 7. The country's average simulated rainfall (dashed line) closely follows the pattern of the observed data (continuous line). The simulated amounts are particularly close to the observed amounts for November through April (wet season), whilst generally they are overestimated for May through October (dry season), with very large overestimations for the months of July and August. The overestimations mainly occur at the coastal and southern stations (not shown), and therefore, regional as well as seasonal considerations are recommended in utilizing model-simulated rainfall in application-oriented tasks.

Figure 7.

The time sequences of the daily rainfall (mm d−1) averaged from all studied stations for the period 1978–2000.

For future climate change variability, it is important to calibrate any RCM-based projected rainfall and temperature scenarios. In this study, PRECIS-REGN (Equations (1)-(3)) and PRECIS-WCRP (Equation (4)) are used to calibrate the PRECIS-ECHAM5 A1B emissions scenario (Figure 8). The ECHAM5-PRECIS rainfall without calibration shows large positive bias from June through October compared with the observed data (Figure 8(a)). The calibrated amounts, either for PRECIS-REGN or for PRECIS-WCRP, agree well with the annual cycle of the observational dataset. This emphasizes the need to calibrate PRECIS outputs for future scenarios for utilization in application-oriented tasks. Similarly, the calibrated temperatures fit well with the annual cycle of the observations (Figure 8(b)).

Figure 8.

The annual cycle of calibrated ECHAM5-PRECIS: (a) rainfall (mm d−1) and (b) mean temperature (°C) averaged from all the studied stations over 1978–2000.

3.2. Future changes in climate

The future changes (2021–2070) in climate parameters (rainfall and temperature) are performed using PRECIS with respect to the base period ECHAM5 A1B data (1971–2000). A time length of 30 years is selected to divide the future period into two groups (2021–2050 and 2041–2070, with 2041–2050 being common to both groups).

Figure 9 displays the spatial distribution of changes in the mean annual rainfall (in %) obtained from the driving field and simulated by PRECIS. The driving field shows an increasing tendency for rainfall in Saudi Arabia, except for pockets of a decreasing tendency in the northern and eastern border regions for the period 2021–2050 (Figure 9(a)). During the period 2041–2070, the rate of increasing tendency over most parts of Saudi Arabia is high (above 70%, reaching 100%); however, over the northern and eastern border regions, the decreasing pockets are again evident but more intense (Figure 9(b)). The changes in the simulated rainfall are generally positive over most parts of Saudi Arabia for the 2021–2050 period, except for a pocket of decreasing values (below 10%) in the northwestern region (Figure 9(c)). In contrast, during the period 2041–2070, the rainfall tends to increase mostly in the eastern and southeastern areas (and in southwestern areas including the coastal belt along the Red Sea to a lesser extent), and in some places the increase is above 70% (Figure 9(d)). It is important to note that during this latter period (2041–2070), a prominent decreasing tendency is observable over a large area in the northwestern region of the country, which extends up to the northern border of the analysis domain. Overall, the simulated results differ from the driving fields, based on the base period downscaling performance of the model. Nevertheless, the PRECIS results are in good agreement with the IPCC precipitation projection for the Arabian Peninsula (IPCC, 2007); the results show that the northern and central areas tend to be drier, and they also show that the southeastern parts are likely to be considerably wetter, which may lead to an increase in the number of extreme events. Further, PRECIS-ECHAM5 (for both 2021–2050 and 2041–2070) shows increases in rainfall for the Red Sea coast, where Jeddah is located, exposing this city also to an increase in the number of extreme events in the future.

Figure 9.

The spatial distribution of the change in mean annual rainfall for ECHAM5 boundary forcings (a and b) and generated by PRECIS (c and d) for the projection period 2021–2050 (a and c) and 2041–2070 (b and d), with respect to the base period (1971–2000).

The change in mean temperature for the driving field and simulation w.r.t. base period is shown in Figure 10. The driving field shows that the rising tendency in temperature for 2021–2050 in Saudi Arabia is apparent in most locations, with a general increase of >0.4 °C in the south and >0.3 °C in the north of the country (Figure 10(a)). During the period 2041–2070, the warming tendency for temperature is >0.40 °C in the northern and central-to-Arabian Gulf regions (Figure 10(b)). In this latter period, a slow rising tendency in temperature (<0.20 °C) is evident in the southwestern regions. Importantly, the rising tendency in simulated temperature is evidently about six times higher than the driving field across the country for both periods. In the period 2021–2050, the rising tendency is >1.6 °C for most parts of Saudi Arabia; it is >1.4 °C in the border areas (Figure 10(c)). However, for 2041–2070, the rising tendency for temperature is >2.5 °C (>3.0 °C) in the southern (northern) region of the country (Figure 10(d)). Overall, the rising tendency in simulated temperature is more rapid than that for the driving field, which is a typical signature of the improvement of climate projection through dynamical downscaling. These analyses clearly show the tendency for greater increase in temperature during the period 2041–2070 (compared with 2021–2050) over Saudi Arabia, which would be of particular significance in application-oriented tasks vis-à-vis climate change.

Figure 10.

The spatial distribution of the change in mean temperature for ECHAM5 boundary forcings (a and b) and generated by PRECIS (c and d) for the projection period 2021–2050 (a and c) and 2041–2070 (b and d), with respect to the base period (1971–2000).

The possible changes in rainfall and temperature obtained from the PRECIS-ECHAM5 simulated scenarios, which are extracted at the observed coordinates (averaged over the country) and calibrated using the WCRP suggested method, are presented in Figure 11. The tendency for rainfall change indicates a high degree of variability during the period 2021–2070, implying the possibility of extreme events occurring during the study period. The results show a possibility of surplus rainfalls for less than 10 years, while the deficit in rainfall is for more than 40 years (Figure 11(a)). This implies that generally the future climate of Saudi Arabia will tend to be drier compared with the present-day climate.

Figure 11.

The trends of change for (a) rainfall and (b) mean temperature obtained from PRECIS with the ECHAM5 A1B emission scenarios during 2021–2070, with respect to the calibration period (1978–2000), averaged over the country.

The calibrated PRECIS-WCRP extracted mean temperature shows a tendency of sharply increasing trend for the scenario during 2021–2070 (Figure 11(b)). The largest possible change for temperature is 4.46 °C in 2063. Overall, the rate of possible increase is about 0.65 °C per decade for temperature during the period 2021–2070. This possible rate of temperature increase is closer to the current mean temperature trend of 0.60 °C per decade in Saudi Arabia (Almazroui et al., 2012b).

This study shows that the model usually underestimates or overestimates climate parameters and involves varying degrees of uncertainty and bias. However, the utilization of calibrated PRECIS-ERA40 outputs could be useful for assessing vulnerability vis-à-vis adaptation or mitigation strategies (or risk management studies) against any future climate change. ERA40 is also useful in understanding the model performance for the base period climatic conditions in the region. To achieve the goal of future projection, ECHAM5 A1B, emissions scenario is calibrated with observed data (see Figure 8). The calibrated parameters can then be utilized in assessing possible changes in the near-future climate with PRECIS-ECHAM5 data (see Figure 11). These analyses offer the opportunity to utilize PRECIS-simulated calibrated scenarios at the local level for application-oriented tasks. It facilitates utilization of information at the local level for devising solutions in application-oriented responsibilities, and thereby eliminating the possibilities of erroneous outcomes.

4. Conclusions

The analyses of the PRECIS-simulated rainfall and temperature for the base period (1971–2000) and near future (2021–2070) are conducted using ERA40 and ECHAM5 datasets. The patterns of simulated rainfalls and temperatures match well with the gridded CMAP and CRU datasets. Overall, the dynamical downscaling of large-scale forcings by the PRECIS simulation improved the projected climate of Saudi Arabia relative to the observations. The simulated rainfall and temperatures precisely follow the annual cycle and interannual variability obtained from the observed data.

In the case of the PRECIS simulation driven by ERA40, the correlation coefficient between the simulated rainfalls and the observed data is approximately 0.84, whereas for temperature the correlation is approximately 0.88. For the case of PRECIS driven by ECHAM5, the calibrated rainfall and temperature agreed well with the annual cycle of the observed data. Thus, the utilization of calibrated PRECIS-generated scenarios for 2021–2070 is found to be more suitable in climate change impact studies. The projected rain tends to be in surplus over the Red Sea (including the coastal areas) and over the southeastern parts of the Peninsula (also to a lesser extent over the southwestern parts) and in deficit in the northern and central regions of the study area. The PRECIS ECHAM5 A1B emissions scenario results confirm the IPCC-reported precipitation characteristics over the Arabian Peninsula (IPCC, 2007). This work discloses that the calibrated model data demonstrate better agreement with the observations; it may be relied on for future projections with higher level of confidence. As many areas of Saudi Arabia are likely to be more vulnerable to extreme events in the future climate scenario, the outcomes of this study may be useful in enhancing the reliability and confidence of impact analysis and formulation of adaption strategies to cope with the effects of climate change.

Acknowledgements

The Presidency of Meteorology and Environment in Saudi Arabia is acknowledged for providing the observational datasets. The Department of Metrology at King Abdulaziz University (KAU) is also acknowledged for providing the 2011 rainfall records, and the UK Met Office is appreciated for providing PRECIS with driving forcings. The CMAP and CRU data were obtained from their websites at http://www.esrl.noaa.gov/psd/data/gridded/ and http://www.cru.uea.ac.uk, respectively.

Ancillary