• Open Access

Decadal variability of the observed daily temperature in Saudi Arabia during 1979–2008


  • H. Athar

    Corresponding author
    1. Center of Excellence for Climate Change Research, Department of Meteorology, King Abdulaziz University, P.O. Box 80208, Jeddah 21589, Saudi Arabia
    • Center of Excellence for Climate Change Research, Department of Meteorology, King Abdulaziz University, P.O. Box 80208, Jeddah 21589, Saudi Arabia.
    Search for more papers by this author


Variability in the observed daily temperature for the 30-year period (1979–2008) is studied from a total of 19 stations in Saudi Arabia (SA) by calculating the empirical anomaly probability distribution functions (PDFs) on annual basis. The 30-year period is divided into three decades. As compared with the first decade, the PDFs for the remaining decades display a relative frequency rise in warmer temperatures. The mean values of the PDFs depict an average decadal positive shift of 0.83, 0.66, and 0.49 °C, for the maximum, the mean, and the minimum temperature, respectively, relative to the 30-year base value. Copyright © 2012 Royal Meteorological Society

1. Introduction

A technique to quantify the relative changes in the future climate, as compared to the observed one, is to use probability distribution functions (PDFs) (see, for instance, Ballester et al., 2010). Therefore, obtaining PDFs based on observed datasets remained an important task for the validation and calibration of the climate projections. In particular, this is the case for several of the climatic variables on daily basis, using both global and regional model output datasets, after being downscaled through various downscaling techniques (e.g. Winkler et al., 1997; Mearns et al., 1999). Variability assessment of the observed daily temperature is of paramount importance and utility to perform climate projection studies, and to assist the policy makers to form policies of climate change impacts (IPCC, 2007), in particular for an arid country such as Saudi Arabia (SA) (Al-Jerash, 1984). SA alone covers almost 80% of the Arabian Peninsula area (CSD, 2010).

The observed temperature datasets for the stations in SA have been analyzed on annual and monthly basis, in the recent past. The mean monthly maximum and minimum temperatures for the stations at Riyadh and Dhahran (in addition to rainfall) in SA were analyzed for the 32-year period (1961–1992), in addition to other 21 stations in the Arabian Peninsula by Nasrallah and Balling (1996). No statistically significant trend in either the maximum or the minimum temperature was found. The observed daily temperature (in addition to rainfall) for one station at Dhahran in SA was studied for the 37-year period (1970–2006) by Rehman (2010). A warming trend of the local air was reported. The mean annual and monthly station-based and regional temperature trends (in addition to rainfall) for six stations (Tabuk, Riyadh, Jeddah, Khamis Mushait, Gizan, and Dammam) in SA for the 24 year period (1985–2008), except for Dammam (2000–2008), were recently studied by AlSarmi and Washington (2011). Depending upon the season, and the temperature variable, a general statistically significant warming trend was noticed. Updated seasonal and annual temperature and rainfall climatological analyses based on the monthly datasets from 27 stations in SA (in addition to the gridded datasets) for the 31 year period (1979–2009) were presented by Almazroui et al. (2012a, 2012b). Statistically significant linear trends in the temperature variables were found.

However, decadal variability analysis of the observed daily temperature in terms of PDFs has not been presented for stations in SA (to our knowledge). Change in the relative frequency of a particular temperature range may be identified through such analysis, and therefore the analysis presented here supplements, broadens, and updates the scope of findings in several of the previous studies (e.g. Alexander et al., 2006, and references cited therein).

2. Data and methodology

The stations-based daily temperature datasets were obtained from the Presidency of Meteorology and Environment of SA, Jeddah. From a total of 28 stations, 19 stations were selected for this analysis based on the availability of the longest dataset. Of these 19 stations, there is no missing data for the following 6 stations: Turaif, Arar, Tabuk, Al-Qaysumah, Gassim, and Madina. Cumulatively, for the remaining 13 stations, the missing daily data is less than 0.02%. The details of these 19 stations are provided in Table I (see Figure 1 also). These include the International Civil Aviation Organization (ICAO) and World Meteorological Organization (WMO) code for each station, the latitude (°N), the longitude (°E), the elevation (m) above the mean sea level, along with the station name. The stations are listed from north to south in the descending order. The dataset from each station is 30 years long. Figure 1 displays the geographical location of each station from which the temperature dataset is analyzed in this study.

Figure 1.

Geographical locations of the stations in Saudi Arabia (including the topography, in m), from which the daily temperature data are used

Table I. The detailed geographical description of the stations in SA used in this study
NoStation nameICAO codeWMO codeLat (°N)Lon (°E)Elevation (m)
10Riyadh Old40438OERY24.7146.73610
16Khamis Mushait41114OEKM18.2942.802047

Given the multifaceted utility of the daily maximum, the daily mean, and the daily minimum temperatures (see, for instance, Rehman et al., 2011), all three temperature variables are analyzed here. The daily mean temperature was calculated by averaging over the daily maximum and the daily minimum temperatures. Commonly used statistics (such as the mean, the standard deviation, the range, and the skewness) are employed in this analysis to quantify the relative changes in the shape of the PDFs (see, for instance, Wilks, 2006).

First, for each of the above three temperature variables, the 30-year climatology for all the stations was calculated, with respect to the base period of 30 years, on annual basis. The mean of each calendar day's maximum, minimum, and mean temperature constitutes the 30-year climatology. The anomaly-based estimates represent the relative change in the three variables with respect to the base period. The PDFs were calculated for each station and then a total PDF was calculated for the all SA, on annual basis, using a 2 °C bin, respectively. The PDFs for the three decades were then constructed. To quantify the change in the extremes of the PDFs, the 10th and 90th percentiles were also calculated on seasonal basis, for the first and the second decade.

Station-wise trends in the anomaly-based annual temperature datasets were computed using the linear regression. The daily temperature datasets were used to construct the annual temperature datasets. The statistical significance of the trend (α) was estimated using the Mann-Kendall's test (see, for instance, Wilks, 2006). The 5% level of statistical significance was used.

3. Results

Figures 2–4 display the PDFs for the observed daily maximum, mean, and the minimum temperatures for the all SA, on the decadal basis, respectively. Given the approximate latitude (10–32°N) and longitude (30–60°E) boundaries of the SA, the contributing factors in controlling the temperature distributions include the persistence of the Hadley cell-based subsidence during the summer season and the southward intrusions of the cold air masses in the winter season (Walters and Sjoberg, 1988). During the spring and the fall seasons, the Mediterranean and Monsoonal weather systems are retreating/intruding simultaneously.

Figure 2.

Top panel: All SA daily maximum temperature PDFs on decadal basis, relative to the 30-year base period, for the first decade (black bars), and the second decade (red face bars). The maximum temperature anomaly ( °C) is plotted along the horizontal axis, whereas on the vertical axis the relative frequency is plotted. The left vertical line indicates the 10th percentile limit, whereas the right vertical line indicates the 90th percentile limit, both for the first decade. Middle panel: same as for the top panel, except for the daily mean temperature. Bottom panel: same as for the top panel, except for the daily minimum temperature

Figure 3.

Same as Figure 1 except for the second decade (black bars) and the third decade (red face bars)

Figure 4.

Same as Figure 1 except for the first decade (black bars) and the third decade (red face bars)

On the individual station basis, the coastal region stations including Wejh (No. 9 in Table I), Yenbo (No. 12 in Table I), Jeddah (No. 13 in Table I), and Gizan (No. 19 in Table I) have narrower spread, in the PDFs, around the base value (approximately ± 10 °C), for all temperature variables, as compared with other inland stations (not shown).

A positive decadal shift in the mean value of the PDFs is noticeable; the maximal (minimal) shift occurs during the third decade as compared with the first decade (second decade as compared to the first decade) decade for the maximum (minimum) temperature (Table II). Relative to the 90th percentile for the first decade, for all the remaining decades and for all the temperature variables analyzed, the right end of the PDF tail consists of more frequent higher temperature occurrences. The relative frequency of occurrence of lower temperatures has decreased during the third decade as compared with the first decade. The relative maximal decadal shift in the mean value in the maximum temperature as compared to the mean and the minimum temperatures is broadly consistent with the maximal warming in the maximum temperature on the mean annual basis (AlSarmi and Washington, 2011).

Table II. The one sigma standard deviation ( °C), the range ( °C), and the skewness for the PDFs (displayed in Figures 2–4), on decadal basis, for all the three temperature variables during the first (1979–1988), second (1989–1998), and the third decade (1999–2008), relative to the 30-year base period (1979–2008)
 PeriodStandard deviation ( °C)Range ( °C)Skewness
Maximum temperature    
 1979–19887.8546.30− 0.35
 1989–19988.1349.80− 0.39
 1999–20088.0349.00− 0.39
Mean temperature    
 1979–19887.5042.45− 0.32
 1989–19987.6443.75− 0.36
 1999–20087.6443.80− 0.36
Minimum temperature    
 1979–19887.5343.00− 0.27
 1989–19987.5147.20− 0.30
 1999–20087.6647.30− 0.29

As indicated by Table II, another characteristic feature of the comparative analysis of the decadal PDFs is the observation that a maximal change occurs in the average range for the maximum temperature (defined as the difference between the absolute maximum and the absolute minimum values of the temperature variables averaged over the three considered decades), as compared to that for the mean and the minimum temperature. This is yet another signature indicative of changing climate in the SA as diagnosed in this study.

The dominating negative skewness indicates progressively relative lower frequency of occurrence for below 10th percentile temperatures (Table II). In general, given the small negative skewness values the distributions are asymmetrical, in all three decades.

The standard deviation of the decadal PDFs displayed a progressively increasing character. Except for the maximum temperature in the third decade, the standard deviation has increased for all other temperature variables in all decades, reflecting a smearing and broadening of the PDFs, relative to the base value (Table II).

Table III indicates that, in general, station-wise, the statistically significant anomaly-based annual linear trends are higher for the maximum temperature as compared to the mean and the minimum temperatures. The warmest anomaly-based annual trends (0.11 °C year−1) were shown by the stations at Turaif (in maximum temperature) and at Al-Qaysumah (in minimum temperature). Only the station at Gizan displayed a statistically significant cooling trend (−0.02 °C year−1) for the minimum temperature. Overall, country-wise, a statistically significant annual linear warming trend seems to exist. These results are consistent with the observations made through the PDF-based analysis presented earlier.

Table III. The anomaly-based station-wise yearly linear trend in units of °C year−1, and the corresponding statistical significance (α) relative to the 30-year period (1979–2008), for the maximum, the mean, and the minimum temperatures, respectively
 Maximum temperatureMean temperatureMinimum temperature
Station nameTrendαTrendαTrendα
  1. The geographic description of each listed station is provided in Table I.

Arar0.050.0010.05< 0.00010.040.000
Al-Jouf0.050.0010.03< 0.00010.02< 0.0001
Tabuk0.09 0.070.0140.060.007
Gassim0.03< 0.00010.030.0010.03 
Dhahran0.03< 0.00010.04< 0.00010.040.000
Riyadh Old0.08
Yenbo0.08< 0.00010.07< 0.00010.050.002
Bisha0.09< 0.00010.06< 0.00010.02< 0.0001
Khamis Mushait0.090.0000.07< 0.00010.05< 0.0001
Abha0.06< 0.00010.05< 0.00010.04< 0.0001
Najran0.06< 0.00010.05< 0.00010.05 
Gizan0.07 0.030.014− 0.020.001

4. Conclusions

A 30-year (1979–2008) variability analysis of the daily temperature from a total of 19 stations in SA is presented in an attempt to quantify the relative change in the frequency and intensity of the extremes (≤10th and ≥ 90th percentile). The results are presented on decadal basis. The anomaly-based empirical PDFs for the daily maximum, mean, and minimum temperatures are calculated. During the third decade (1999–2008), the extremes (≥90th percentile) occur more frequently for all the analyzed temperature variables, relative to first decade (1979–1988). The increase of the daily maximum temperature is more prominent than the increase of the daily mean and minimum temperatures.

During the second and third decade, progressively more negatively skewed PDFs characterize the changing climate for all the three temperature variables analyzed. The analysis presented in this paper is indicative of climate change in SA.

The shifts in the computed statistics (mean, range, skewness, and standard deviation) of the anomaly-based probability density functions for the observed daily temperatures may serve to provide some guidance not only for the climate models attempting to simulate the climatological features over SA, but also for the potential climate change impact studies. The noted rise in the extreme temperature in this study will have considerable socioeconomic impacts. These include, but not limited to, excessive heat stress on humans and ecosystems, in particular, during the heat waves.


The author thanks Presidency of Meteorology and Environment, Jeddah, Saudi Arabia and Dr Mansour Almazroui for providing the dataset.