Recent observed climate change over the Arabian Peninsula



[1] We have examined trends in temperature and precipitation parameters for the Arabian Peninsula (AP) during the last 2 to 3 decades. The data set has been carefully quality controlled and checked for homogeneity. Although of low density (21 stations) and relatively short time period, a clear picture of climate change in the region has emerged. The general pattern of the AP mean annual temperature trend is one of warming, with 14 of 21 stations show statistically significant warming at 0.05 level and most at 0.001 level and only one (Seeb) showing statistically significant cooling. The highest statistically significant mean annual warming trends are found in Oman (Sur = 1.03°C decade−1) and Emirates (Dubai = 0.81°C decade−1). The season of maximum warming in mean temperature is March to April. The highest monthly mean temperature trend in the AP occurs in Sur in May (1.47°C decade−1). There is a broad statistically significant increase in mean annual maximum temperature in AP in 12 out of 21 stations, with the highest trends in central and eastern/southeastern AP. Only SW AP and the Gulf of Oman do not show warming. The highest monthly maximum temperature trend in the AP occurs in Bahrain in March (2.27°C decade−1). The second highest significant warming trends are reported in Doha in February (1.54°C decade−1). For minimum temperature, 16 out of 21 stations show statistically significant warming trends, with the highest annual trends observed in the Emirates (Dubai = 1.24°C decade−1), northwest Oman (Sohar = 1.17°C decade−1) and Qatar (Doha = 1.13°C decade−1). The highest monthly minimum temperature warming rate occurred in October. Both Dubai and Kuwait reported the highest significant rate of 2.00°C decade−1. The general mean annual diurnal temperature range trend is negative in the AP, with six out of 21 stations show statistically significant negative trends while three stations show statistically significant positive trends. Trends in mean annual precipitation are significant at only two stations which show a decrease in precipitation.

1. Introduction

[2] Many parts of the world and especially developed countries have benefited from assessments of observed climate changes, especially in relation to temperature and precipitation owing to the societal importance of these parameters [Goddard et al., 2001]. According to the IPCC fourth assessment report [Solomon et al., 2007], the surface temperatures over land regions have warmed at a faster rate than oceans in both hemispheres with land based warming trends of about 0.27°C versus 0.13°C per decade over the past two decades. Climate change studies have, in the past, been largely restricted to North America, Europe and Australia as a result of data availability [Freiwan and Kadioglu, 2008]. Less is known about trends in temperature and precipitation in the Arabian Peninsula (AP). To some extent, this is because the region covers a broad range of countries, some of which have poor data availability, quality and consistency [Zhang et al., 2005; Kwarteng et al., 2009; Nasrallah and Balling, 1993]. This is in addition to the fact that much of the area is desert and there are very few sites with instrumental data. The region, located in the southwestern Asia, is characterized by unique topography which varies between deserts and high mountains. It is a dry environment with very high temperature especially in summer. It is vast with several contrasting climate zones, the northern part of which is considered to be subtropical (north of 20°N) while the southern is a tropical (monsoonal) type.

[3] Zhang et al. [2005] were responsible for the first region-wide trend analysis of the Middle East extreme indices for 1950–2003 at 52 stations covering 15 countries. The study shows statistically significant, spatially coherent trends in temperature indices that indicate temperature increase in the region. In a broader study covering all dryland regions, Hulme [1996] and Jones and Reid [2001] demonstrated that warming may have contributed to a reduction in the P/PE ratio in many of these dryland regions. Over the Middle East, Zhang et al. [2005] concluded that trends in precipitation are weak in general and do not show spatial coherence while over Oman during 1977 to 2003, Kwarteng et al. [2009] reported that in general, the monthly and yearly patterns of precipitation are stable and no significant trends were observed over the study period.

[4] Economic and Social Commission for Western Asia [2010] urges Arab countries to demonstrate commitment and preparation in relation to emerging international climate change frameworks and that these countries should provide more inputs to the IPCC negotiations. Additionally, climate change projections for the AP shows little agreement among the models, particularly with regards to the sign of precipitation change. A concerted effort is therefore needed in the case of the AP to address the knowledge deficit in climate. The aim of this paper is to evaluate observed temperature parameters and precipitation trends over the relatively neglected AP region using an updated data series which includes several new stations compared with previous efforts. To our knowledge, this is the first trend analysis of climate parameters for several countries in the AP especially Oman, Qatar, Emirates and Yemen. Previous studies focus on one or a few stations per country using only one variable [Elagib, 2008; Nasrallah et al., 2004] or use coarse gridded data with few stations [Nasrallah and Balling, 1993]. This paper uses data for 21 stations covering AP's 7 countries to analyze trends of 5 climate variables.

2. Data

[5] Monthly mean temperature, maximum temperature, minimum temperature and precipitation data have been provided by the National Meteorological Services NMSs of the AP, namely Oman, UAE, Qatar, Bahrain, Saudi Arabia, Kuwait and Yemen for a total of 21 stations (Figure 1).

Figure 1.

The distribution of 21 stations used in this study in the Arabian Peninsula, where x axis is °E and y axis is °N.

[6] Seventeen stations were provided from Oman and 4 from UAE, 6 from Saudi Arabia, 2 from Yemen and 1 station from each of Qatar, Bahrain and Kuwait. Station data which is included have the following characteristics: (1) less than 5% missing values in the whole period. (2) They have passed the homogeneity and quality control tests. (3) Bahrain, Marisah and Salalah have the longest data record of more than 50 years (1943–2008) while most of the remaining stations have data from 1980 to 2008. To maximize the number of stations used in the analysis, stations which have at least 24 years of data during the period 1980–2008 were selected to form the bulk of the study. A small proportion of stations did not have data for the complete period (Table 1) but have been included because no data from these regions has been previously available. Approximately 70% of the stations have no missing values in the period 1980–2008 for the temperature and precipitation variables while the remaining stations have between 0.3% and 4.9% missing values.

Table 1. Stations List
CountryStationWMOLatitudeLongitudeElevation (m)Period TemperaturePeriod Precipitation
 Ras AlKhaimah4118425.3755.5631.01980–20081980–20081980–20081980–2008
Saudi ArabiaTabuk4037528.2236.36768.11985–20081985–20081985–20081985–2008
 Khamis Mushait4111418.1742.482093.41986–20081986–20081986–20081985–2008
 Dammam4041726.2749.4912. 02000–20082000–20082000–20082000–2008

[7] The seasonal definition will follow a modified version of work by Fisher and Membery [1998] over AP, namely winter (DJF), spring (MA, 1st transitional period), early summer (MJ, premonsoon), late summer (JAS) and autumn (ON, 2nd transitional period or postmonsoon).

2.1. Quality Control

[8] Data supplied by the region's NMSs were quality controlled in two stages: (1) a check for physically implausible data (e.g., negative rainfall or maximum temperature < minimum temperature) and (2) analysis of outliers which are identified by numerical and visual checks. These were evaluated by comparing their values to nearest stations in order to determine whether the flagged values were associated with real or anomalous weather events. A few physically implausible values have been set to missing.

2.2. Homogeneity Checks

[9] A homogeneous climate time series is defined as one where variations are caused only by variations in climate [Conrad and Pollak, 1950]. It is important, therefore, to remove the inhomogeneities. Data homogeneity is assessed using the RHtestV3 software which uses a two-phase regression model applied to monthly data in order to check for multiple stepchange points that could exist in a time series [Wang, 2008, 2003].

[10] Identified step changes are checked against the station history (if available). Significant change points (5% level) were identified in 8 stations for mean temperature, 10 for maximum temperature, 8 for minimum temperature and 3 for precipitation. Most of these change points were regionally widespread (detected in many countries) and were found to be physically real, which concurs with other similar studies [Alexander et al., 2006; Peterson et al., 1998] (Table 2). For example the highest number of significant temperature discontinuities occurred in 1997 and 1998. It is apparent from the spatial (not shown) and temporal signal that these change points are linked to heating associated with the 1997/8 El Nino. Most change points occurred in the southern AP stations earlier (mid 1997) than those detected in the north (Bahrain, Qatar, Emirates and Saudi Arabia) which followed several months later. The heating anomalies are evident in several reanalysis data sets (not shown).

Table 2. Details of Significant Detected Change Points Showing Years, Stations, Variables, and Dates
YearStationVariableDate (year.month)
Ras AlKhaimahprecipitation1995.11
1997Dohamax T1997.07
Masirahmax T1997.06
Sana'amean T1997.06
Adenmean T1997.07
Adenmin T1997.07
1998Bahrainmean T1998.01
Bahrainmax T1998.01
Bahrainmin T1998.05
Dohamean T1998.01
Dubaimax T1998.01
Dubaimean T1998.01
Ras AlKhaimahmax T1998.02
Riyadhmax T1998.05
Ras AlKhaimahprecipitation1998.03
1999Dubaimean T1999.06
Dubaimax T1999.08
Bahrainmin T1999.06

[11] Figure 2 shows an example from Bahrain monthly mean temperature. The station shows a large inhomogeneity in 1998, which corresponds to the exceptional region-wide warming, associated with the 1998 El Niño year. Historical explanations for the cause of the change points, such as station relocation, are found for only two stations namely Khasab and Sur. Therefore, we use the adjusted monthly time series of these stations where the change points are significant and supported by metadata.

Figure 2.

Bahrain station monthly mean temperature change point of 1998.

[12] After quality control and homogeneity testing, 8 stations were retained for Oman, 2 from UAE, 6 stations from Saudi, 2 from Yemen and 1 station from Qatar, 1 from Bahrain and 1 from Kuwait. The final list of stations is summarized in Table 1.

3. Methods

3.1. Trends

[13] Linear trends are computed for the following variables: mean temperature, maximum temperature, minimum temperature, diurnal temperature range (DTR) and precipitation series using the nonparametric approach by Sen [1968] modified by Wang and Swail [2001] to account for time series autocorrelation. This is a robust approach which is resistant to outliers and can be computed when data are missing. Annual missing values are excluded from the analysis when calculating the linear trend. The method has been widely used to compute trends in climate studies [Aguilar et al., 2009; Alexander et al., 2006; Butt et al., 2009; Zhang et al., 2005].

[14] The significance of the trend is determined using Mann-Kendall's test for which data need not conform to any particular distribution. A similar method was adapted by Manton et al. [2001]. The 5% level of statistical significance is used. Monthly, seasonal and annual trends are calculated for all the 21 stations based on the stations time period.

[15] Two trends analyses were carried out; the first used station data with a period equal or more than 24 years during 1980–2008 (18 stations) while the second analysis used stations data with a period more than 50 years during 1943–2008 (Bahrain, Marisah, Salalah).

3.2. Area Averaging

[16] The AP regional average time series for the 5 variables were calculated following Aguilar et al.'s. [2009] method which avoids the domination of stations with high mean values by averaging the stations time series anomalies. A base reference of 1985–2005 was used to calculate the station normals and anomalies following Jones and Moberg [2003] who used at least 20 years of data within the 30 year period to calculate the normals. Data for the base period 1985 through 2005 were used in the calculation of the anomalies; at least 18 years needed to be available in this base period for the station to be used. In addition, the regional standardized time series of all the variables have been calculated and regional trends obtained. Both the anomaly and standardized averaged time series lead to similar trend shape, sign and significance. The standardized time series are not used and the results are based on the anomaly time series. No adjustment for the varying number of stations during some years have been made but it is important to note that the first 5 years (1980–1984) did not include Saudi Arabia's stations. New et al. [1999] states that monthly anomalies tend to be more a function of large-scale circulation patterns and are relatively independent of physiographic control. We assume the network studied here is sufficient to describe the month-to-month departures from the mean climate.

[17] The area average time series have been computed for 3 areas: the all AP, the monsoonal area (stations south of 20°N under the direct influence of the monsoon which comprise 5 stations located over Oman and Saudi Arabia namely Masirah, Thumrait, Salalah, Gizan and Khamis Mushait) and the nonmonsoonal stations located at north of 20°N covering the remaining 13 stations excluding the short period stations.

4. Results

4.1. Stations

[18] Trends of mean, maximum and minimum temperature, diurnal temperature range and precipitation are presented in this section. Trends for individual stations precede regional trends.

4.1.1. Trends During the Period 1980–2008 Mean Temperature

[19] The general pattern of the Arabian Peninsula (AP) mean annual temperature trend is warming, with 14 of 21 stations show statistically significant warming at 0.05 level and most at the 0.001 level.

[20] One station (Seeb-Oman) shows significant cooling (Figure 3) while the remaining 6 station trends are not statistically significant. The highest statistically significant warming trends are found in Oman (Sur = 1.03°C decade−1) and UAE (Dubai = 0.81°C decade−1) (Figure 4). Both countries are located in the eastern/southeastern parts of the AP. Other stations showing statistically significant high rates of warming are Doha (0.65°C decade−1), Khamis Mushait (0.60°C decade−1) and Kuwait (0.57°C decade−1).

Figure 3.

Muscat monthly anomalies time series with linear regression line showing the cooling trends.

Figure 4.

Mean temperature trends in the Arabian Peninsula in degrees °C per decade for the period 1980–2008 (except Saudi Arabian stations 1985–2008). The slope of the warming trend is proportional to the diameter of the circles. Shaded circles are statistically significant at the 0.05 level. Warming is shown in red, and cooling is shown in blue. Stations enclosed by the rectangle have a much shorter record.

[21] The lowest trend values are observed on the south and southwest coast of AP (Masirah and Gizan, (0.22°C decade−1).Trends in annual temperature at the rest of the southern AP stations are not statistically significant.

[22] We start the seasonal analysis with the March-April trends which have three notable features: (1) the highest significant warming trend is in central AP (Riyadh = 1.22°C decade−1); (2) 8 stations in the Arabian Gulf (AG) have warming in excess of 0.55°C decade−1; and (3) there are no statistically significant trend along the southern AP coast (Figure 4).

[23] The western areas of AP have higher statistically significant warming in JAS season compared with MA season (Figure 4). The warming extends from Sur in east Oman to Jeddah in west Saudi Arabia (the line of largest warming) while Seeb shows significant cooling (−0.55°C decade−1), as is evident also in the mean annual trends. The warming pattern in MJ season is closer to MA season with Sur station at the eastern Oman reporting the highest statistically significant monthly mean temperature warming trend in all AP in May (1.47°C decade−1).

[24] In the DJF season, the areas of most rapid warming are generally the same as for the mean annual temperature (i.e., eastern and southeastern AP). In this season, 10 of 21 stations show statistically significant trends. Surprisingly, the highest rate of warming is observed over the mountains in northern Oman (Saiq = 0.85°C decade−1) at around 2 km above mean sea level and distant from large settlements. Sur, Dubai and Khamis Mushait have significant trends of more than 0.70°C decade−1). In the DJF season, warming trends in the northern stations are not as steep as those for the mean annual trends. In contrast, some of the eastern and southern trends are larger during DJF. It is also striking that Khamis Mushait (2 km above MSL) reported the highest statistically significant warming in January (0.80°C decade−1). Maximum Temperature

[25] There is a broad statistically significant increase (at 0.05 level) in mean annual maximum temperature in AP with 12 out of 21 stations showing statistically significant warming. The exception is southwest AP and Gulf of Oman (Figure 5). The highest mean annual trends are observed in the central and eastern/southeastern areas where Sur and Khasab (Oman) reported the values of 0.93 and 0.88°C decade−1, respectively. Bahrain shows a similarly high trend of 0.81°C decade−1. Both Saiq over the Northern Oman Mountains and Masirah over Oman's east coast reported the lowest statistically significant trend (0.27°C decade−1). The cooling in the southwest AP is not statistically significant except in Khamis Mushait (−0.40°C decade−1).

Figure 5.

Same as Figure 4 but for maximum temperature.

[26] The MA season shows the highest number of stations with high trend values. Riyadh and Kuwait reported the highest MA season significant warming of 1.50 and 1.25°C decade−1, respectively. The month of March witnessed the highest monthly maximum temperature trend value in the AP of 2.27°C decade−1 (at Bahrain; see Figure 5). The second highest significant warming trend is reported at Doha in February (1.54°C decade−1).

[27] In MJ a line of significant warming trends appeared from Dubai (UAE) to Jeddah (in west Saudi Arabia) with warming trends ranging from 0.52 to 0.76°C decade−1. This line also appeared in the mean temperature MJ and JAS seasons (Figure 5).

[28] A line of intense significant warming is seen in central AP in the JAS season (Bahrain = 1.10°C decade−1) but intense significant cooling occurred in the southwest areas (Khamis Mushait = −0.75°C decade−1) (Figure 5). In ON season most of the statistically significant trends are observed in Oman and UAE while significant cooling persists in the southwest AP. Minimum Temperature

[29] From the data in Figure 6, it is apparent that the warming trends of the mean annual minimum temperature in AP are statistically significant and these are more spatially coherent at more of the stations compared with those for the annual mean and annual maximum temperature.

Figure 6.

Same as Figure 4 but for minimum temperature.

[30] Sixteen out of 21 stations show statistically significant warming trends, the highest trend values are observed in the UAE (Dubai = 1.24°C decade−1), northwest Oman (Sohar = 1.17°C decade−1) and Qatar (Doha = 1.13°C decade−1). One single striking difference with the other variables is that the trend of the mean annual minimum temperature is statistically significant even over the south AP coast. There is only one statistically significant cooling trend (Seeb, −0.32°C decade−1).

[31] Widespread and significant warming trends are evident in MJ and JAS with many stations reporting high trend values in excess of (1.00°C decade−1). The line of intense warming from east to west AP apparent in the MJ season is even more marked in the JAS season (Figure 6). This line is seen from February through to September (not shown).

[32] ON season witnessed the highest significant trends which cover all the eastern AP. Dubai reported the highest trend of 1.61°C decade−1. Other stations reporting trends higher than 1.00°C decade−1 include: Sohar, Saiq, Thumrait and Doha (Figure 6). In October both Dubai and Kuwait reported the highest significant rate of 2.00°C decade−1.

[33] The statistical significant warming in DJF covers most of central and southern AP (Doha = 1.23°C decade−1, Saiq = 1.20°C decade−1) while significant cooling is observed over Seeb(−0.63°C decade−1). Interestingly in the MA season the statistically significant trends are limited to AG stations as Kuwait (1.22°C decade−1), Doha (1.19°C decade−1) and Dubai (1.00°C decade−1). Diurnal Temperature Range

[34] The general mean annual DTR trends pattern is negative in the AP (Figure 7) with 6 out of 21 stations show statistically significant negative trends. The highest negative rate occurred in the northwest Oman (Sohar = −1.45°C decade−1), Qatar (Doha = −0.79°C decade−1) and UAE (Dubai = −0.71°C decade−1). Sur, Khasab and Bahrain stations show statistically significant positive trends with values 0.58°C decade−1, 0.47°C decade−1 and 0.44°C decade−1, respectively. The results show few statistically significant trends in the first half of the year but the negative trends increase significantly in the second half of the year, i.e., MJ, JAS and ON seasons especially over eastern, southeastern and the southwest AP (e.g., ON Figure 7). October witnessed the highest monthly statistically significant trends (Figure 7) with values exceeding −2.00°C decade−1 as in Sohar (Oman) and Kuwait (Kuwait).

Figure 7.

Same as Figure 4 but for diurnal temperature range (DTR). Precipitation

[35] The total annual precipitation trends values show only 2 stations with statistically significant trends. Both stations show negative trends: Saiq = −67.71 mm decade−1 which is −59% of the base mean and Tabuk = −20.90 mm decade−1 (−163% of the base mean). Most of the remaining stations trends are negative and not statistically significant (Figure 8).

Figure 8.

Precipitation percentages from 1985 to 2005 mean in the Arabian Peninsula per decade for the period 1980–2008 (except Saudi Arabian stations 1985–2008). The slope of the drying/wetting trend is proportional to the diameter of the circles. Shaded circles are statistically significant at the 0.05 level. Drying is shown in blue, and wetting is shown in red. Stations enclosed by the rectangle have a much shorter record.

[36] Of the rainy seasons, the largest number of negative statistically significant trends occurs in MA (Figure 8), covering northern Oman (Saiq = −10.23 mm decade−1, −94% of the base mean), Qatar (Doha = −3.87 mm decade−1, −85% of the base mean) and Emirates (Dubai = −3.52 mm decade−1, −67% of the base mean).

[37] In the DJF season the trend values shows 1 negative statistically significant trend over Thumrait (south Oman) and 1 positive statistically significant trend over Kuwait (6.94 mm decade−1, 82% of the base mean). In January there is a widespread increase of precipitation over AG and west of AP but it is statistically insignificant.

[38] The AlKhareef summer rain over Salalah (south Oman) shows statistically significant negative trend (−3.72 mm decade−1, −53% of the base mean).

4.1.2. Trends During the Period 1943–2008

[39] Analysis of trends for stations with a record of more than 50 years (see Table 3) was done for Salalah, Masirah and Bahrain. The results show that the mean annual mean temperature is increasing significantly at 0.01 level over all the stations. Salalah and Bahrain warmed at similar rate of 0.15°C decade−1 while Masirah warmed at a lower rate of 0.08°C decade−1. Over Salalah all the seasons are warming significantly (0.001 level) except JAS. Masirah cooled in early summer (MJ) with the remaining seasons warming significantly, except JAS season. Most of the seasonal warming over Bahrain is in early and late summer (significant at 0.001 level). The highest significant warming month over Salalah, Masirah and Bahrain are October, November and June, respectively.

Table 3. Regional Trends and Significance (α) and Overall Change of Mean, Maximum, and Minimum Temperatures, DTR, and Precipitation Over the Arabian Peninsula for Long-Term Data for Salalah, Masirah and Bahraina
  • a

    Precipitation given as percentages of the base mean 1985–2005. DJF, December-February; MA, March-April; MJ, May-June; JAS, July-September, ON, October-November.

Mean Temperature (°C decade−1)
JAS0.04 −0.05 0.240.001
Minimum Temperature (°C decade−1)
DJF0.250.0010.10 −0.05 
Maximum Temperature (°C decade−1)
MJ0.200.050.15 0.710.001
JAS0.05 0.11 0.680.001
DTR (°C decade−1)
DJF− 0.10 
MA0.04 0.08 0.340.1
MJ−0.13 −0.02 0.460.05
JAS− 0.620.001
ON−0.07 0.05 0.440.01
Annual− 0.420.001
DJF0.00 5.76 29.26 
MA0.745 0.00 0.172 
MJ10.87 0.00 0.00 
JAS−22.11 0.00 0.00 
ON0.00 0.00 0.00 
Annual−4.33 −10.73 27.40 

[40] The mean annual maximum temperature is increasing significantly (at 0.001 level) with similar trends (0.20°C decade−1) over Salalah and Masirah. The warming is large by a factor of 2 at Bahrain (0.46°C decade−1) with similar significance. Most of the significant warming is in DJF, MA and ON in Salalah and Masirah while all the seasons are warming significantly over Bahrain except DJF season. May reported the highest warming month in Salalah, November in Masirah and July in Bahrain.

[41] Both Salalah and Masirah reported statistically significant (at 0.001 level) increase in mean annual minimum temperature with higher trends reported over Salalah (0.27°C decade−1). The mean annual DTR is increasing over Bahrain (0.42°C decade−1) but is decreasing significantly over Salalah station with rate of −0.11°C decade−1.

4.2. Arabian Peninsula Regional Trends

[42] An area-average time series of the AP was constructed based on monthly station anomalies. The contribution of the station data to the area average was weighted to remove the effect of the domination of the regional time series by the higher density of stations in some regions. The trends were calculated following the methods described in section 3.1. The resulting annual cycle of regional AP trends is shown in Figures 9 and 10. As was clear from the discussion of station trends, there are generally higher values in the north and lower values on the coastal margin in the south where smaller trends occur. As a result, the AP region was divided into 2 subregions: (1) stations south of 20°N under the direct influence of the monsoon and (2) nonmonsoon stations located north of 20°N. The monsoonal area consists of 5 stations located over Oman and Saudi Arabia (Masirah, Thumrait, Salalah, Gizan and Khamis Mushait). The nonmonsoonal area covers the remaining 13 stations (excluding the short period stations).

Figure 9.

Seasonal and annual cycle trends in degrees °C per decade for the period 1980–2008 (except Saudi stations 1985–2008) of mean, maximum and minimum temperatures for all Arabian Peninsula (blue), nonmonsoonal subregion (red), and monsoonal subregion (green).

Figure 10.

Same as Figure 9 but for precipitation in mm decade−1.

4.2.1. Mean Temperature

[43] The all AP annual mean temperature increased at a rate of 0.40°C decade−1 (0.001 significance level) during the period 1980–2008 while the nonmonsoonal annual mean temperature increased at a rate of 0.46°C decade−1 (0.001 significant level). In contrast to the all AP and the nonmonsoonal regions, the temperature increases over the monsoonal subregion is lower (0.20°C decade−1 which is almost half of the all AP).

[44] The warming is significant over all AP in almost all seasons and months except January (see Table 4). The annual cycle (Figure 9) of the mean temperature trends peaks in MA season with a minimum in JAS in both all and nonmonsoonal regions. The mean temperature trend in the nonmonsoonal subregion is 6 times higher than the monsoonal subregion in MJ season. However, in the monsoonal area, the mean temperature is increasing significantly in DJF and MA seasons with other peak in ON season.

Table 4. Regional Trends for the Period 1980–2008 and Significance (α) and Overall Change of Mean, Maximum, and Minimum Temperatures, DTR, and Precipitation Over the Arabian Peninsulaa
 Mean TemperatureMaximum TemperatureMinimum TemperatureDiurnal Temperature RangePrecipitation
  • a

    Regions: All, all Arabian Peninsula; NMon, nonmonsoonal subregion; Mon, monsoonal subregion. Precipitation given as percentages of the base mean 1985–2005.

Jan0.24 0.70.21 0.60.21 0.6−0.03 −0.1−0.07 −0.20.17 0.50.31 0.90.22 0.60.350.11.0−0.31 −0.9−0.340.1−1.0−0.25 −0.72.93 54.73.60 53.7−0.22 −12.0
Feb0.640.0011.90.700.0012.00.460.011.30.690.012.00.900.012.60.390.051.10.770.012.20.760.052.20.850.052.5−0.03 −0.10.09 0.2−0.61 −1.8−5.21 −106.2−6.58 −102.1−0.680.05−74.6
Mar0.530.011.50.630.011.80.21 0.60.55 1.60.54 1.60.420.11.20.560.051.60.650.011.90.34 1.00.05 0.10.06 0.20.01 0.0−4.35 −64.2−4.26 −55.3−1.01 −23.2
Apr0.500.051.50.590. 1.00.51 1.50.13 0.40.830.012.40.960.052.80.470.051.4−0.24 −0.7−0.27 −0.8−0.22 −0.6−1.39 −35.4−1.66 −51.9−0.31 −5.3
May0.520.0011.50.680.0012.00.16 0.50.500.0011.40.440.011.30.630.051.80.710.012.10.810.012.30.390.051.1−0.18 −0.5−0.33 −1.00.44 1.3−0.10 −6.3−0.24 −23.70.93 30.4
Jun0.370.0011.10.510.0011.50.00 0.00.17 0.50.310.10.9−0.09 −0.30.680.0012.00.760.0012.20.390.011.1−0.470.01−1.4−0.410.01−1.2−0.530.05−1.50.17 17.90.460.0594.3−0.20 −9.8
Jul0.330.0011.00.390.011.10.22 0.60.11 0.30.08 0.20.16 0.50.500.0011.40.600.0011.−0.390.05−1.1−0.500.01−1.4−0.07 −0.2−0.20 −7.60.25 13.6−0.55 −11.6
Aug0.380.0011.10.540.0011.6−0.12 −−0.21 −0.60.590.0011.70.670.0011.90.460.0011.3−0.370.05−1.1−0.22 −0.6−0.560.001−1.6−0.50 −18.3−0.52 −32.6−1.20 −20.8
Sep0.440.0011.30.480.0011.−0.420.05−1.2−0.440.1−1.3−0.350.1−1.0−0.11 −10.80.00 0.0−0.72 −38.6
Oct0.400.0011.20.510.0011. 0.40.10 0.30.23 0.70.850.0012.51.000.0012.90.580.011.7−0.780.001−2.3−0.950.001−2.8−0.18 −0.5−0.01 −0.5−0.04 −4.00.06 2.5
Nov0.440.011.30.440.011.30.310.010.90.25 0.10.590.011.70.530.051.50.740.0012.2−0.35 −1.0−0.19 −0.5−0.770.01−2.20.40 14.90.51 16.40.00 0.0
Dec0.470.051.30.490.051.40.450.011.30.12 0.30.02 1.20.610.051.8−0.30 −0.9−0.42 −1.2−0.20 −0.61.28 25.41.45 23.1−0.03 −1.5
DJF0.490.011.40.470.051.40.340.0011.−0.20 −0.6−0.09 −0.3−0.400.1−1.2−0.36 −7.20.29 4.5−0.63 −43.3
MA0.560.011.60.640. 0.70.530.011.50.660.011.90.390.11.1−0.12 −0.3−0.10 −0.3−0.15 −0.4−3.14 −58.7−2.96 −54.4−1.50 −29.4
MJ0.440.0011.30.600.0011.70.10 0.30.320.010.90.390.011.10.30 0.90.740.0012.10.830.0012.40.370.011.1−0.250.1−0.7−0.310.1−0.9−0.16 −0.5−0.01 −0.50.14 18.7−0.16 −6.4
JAS0.380.0011.10.470.0011.40.13 0.20.570.0011.60.630.0011.80.440.0011.3−0.360.001−1.0−0.350.01−1.0−0.370.01−1.1−0.11 −5.2−0.20 −14.5−0.60 −14.6
ON0.410.0011.20.480.0011. 0.30.15 0.40.11 0.30.700.0012.00.730.0012.10.640.0011.8−0.570.01−1.6−0.610.01−1.8−0.490.1−1.40.19 9.30.24 11.80.14 7.5
ANN0.400.0011.10.460.0011.30.200.0010.60.320.010.90.370.−0.260.01−0.7−0.210.05−0.6−0.250.05−0.7−6.59 −16.9−8.72 −21.7−1.91 −5.3

4.2.2. Maximum Temperature

[45] The all AP annual mean maximum temperature is increasing significantly at a rate 0.32°C decade−1 which is less than the mean temperature. The rate of the annual mean maximum temperature for the monsoonal and nonmonsoonal region is 0.21°C decade−1 and 0.37°C decade−1, respectively.

[46] Over all AP, the maximum temperature warming is significant in February, May and September and in MJ and JAS. In the JAS season, the nonmonsoonal area warming was 4 times higher than the monsoonal area. For the monsoonal area, the maximum temperature trends have the highest values in the first half of the year (only significant in DJF season) with a minimum in the second half of the year (Figure 9).

4.2.3. Minimum Temperature

[47] The warming rate of annual mean minimum temperature is higher than the mean and maximum temperatures over all AP and the 2 subregions. The annual mean minimum temperature trend is 0.55°C decade−1 over all AP. It is 0.61°C decade−1 over the nonmonsoonal subregion. Over the monsoonal subregion the annual warming rate is 0.44°C decade−1.

[48] Similar to the mean temperature, the warming over all AP is significant in all seasons and months except January and December. Minimum temperature peaks significantly (at 0.05 level) in MJ and ON at both all and nonmonsoonal regions. There is also a DJF peak in the monsoonal region (Figure 9).

4.2.4. DTR

[49] As a result of higher warming trends of annual mean minimum temperature relative to annual mean maximum temperature, there is a significant decrease of the annual mean DTR over all regions. The trend values reported over all AP, the nonmonsoonal and monsoonal subregions are as follow −0.26°C decade−1, −0.21°C decade−1 and −0.25°C decade−1, respectively.

[50] The DTR trend is insignificant in DJF and MA at all and the nonmonsoonal regions but the highest reduction in DTR occurs significantly in ON season (Figure 9). In the monsoonal subregion, the lowest significant reduction occurred in JAS season.

4.2.5. Precipitation

[51] The general trend for precipitation is drying over all the regions with overall decrease of annual total around 19.1 mm (−16.9% of the base mean) over all AP although this trend is not significant.

4.2.6. Sensitivity to Missing Data

[52] Trends for all AP and the two subregions (nonmonsoon and monsoon) have been recalculated without the Saudi Arabian stations to assess the impact of the missing data (1980–1984) for the Saudi Arabian stations. Excluding Saudi Arabian stations yields almost identical trends of the mean, maximum and minimum temperatures and DTR in all AP, nonmonsoon and monsoon subregions. Some differences do exist, notably the maximum temperature in the monsoon subregion where the trends are higher when excluding the Saudi Arabian stations relative to the trends when including of Saudi Arabian stations. The following values are the difference between the trends for Monsoon region without the Saudi Arabian stations and the monsoon region with all stations included: 0.24°C decade−1 in May, 0.38°C decade−1 in October, 0.23°C decade−1 in MJ season and 0.30°C decade−1 in ON season.

[53] The precipitation results are insensitive to the exclusion of the Saudi Arabian stations.

5. Discussion

5.1. For the Period 1980–2008

5.1.1. Precipitation

[54] Observed precipitation trends are generally less spatially coherent and of lower level of statistical significance than temperature changes [Alexander et al., 2006]. This is true also of the AP. Nasrallah and Balling [1996] suggested some possible reasons for the lack of clear trends in AP including: (1) the low rainfall totals that dominate the region and (2) the great temporal and spatial variability of precipitation in such an arid and topographically diverse environment.

[55] Most of the drying in all AP and the subregions occurred in MA season. Most of the AP rainfall mechanisms occur during winter and spring westerly upper air troughs [Fisher and Membery, 1998; Galvin, 2009].

[56] The AP precipitation anomalies time series was correlated with the North Atlantic Oscillation (NAO) and El Niño-Southern Oscillation (ENSO) indices (Table 5). The indices time series were extracted from the NCEP/NCAR website ( AP rainy seasons (DJF and MA) have higher and significant correlation with ENSO (Nino3.4 and SOI indices) than NAO. Correlation is insignificant with NAO.

Table 5. Arabian Peninsula Precipitation Anomalies Correlation and Significance With El Niño-Southern Oscillation (Nino3.4 and SOI) and North Atlantic Oscillation (NAO) Indices
 Correlationp Value

[57] A noticeable result of the precipitation analysis is that the annual total data from years 1995–1997 for many AP stations include record high levels of precipitation. Further analysis of daily rainfall totals and the mechanisms associated with these events is required before a fuller picture can emerge.

5.1.2. Mean Temperature

[58] During the past two decades the surface temperatures over land regions have warmed at about 0.27°C versus 0.13°C over oceans per decade [Solomon et al., 2007]. The rate over land is faster than over the oceans in both hemispheres. Klotzbach et al. [2009] found a closer global trend of 0.31°C decade−1 over the period 1979–2008 using surface temperature data sets from the National Climate Data Center (NCDC). Comparison of individual station mean annual temperature trends within AP with global trends (Figure 11) allows the distinction of four types of station trends. First Bahrain, Sohar, Tabuk and Jeddah are all similar to global land trends. Second, trends at Doha, Riyadh, Khamis Mushait, Dubai, Kuwait, Khasab, Saiq, Ras AlKhaimah and Sur are 1.5 to 3.5 times higher than the global land mean trends. Third stations which located further south in the AP (Masirah, Salalah, Thumrait and Gizan) report lower trends. Fourth, one station Seeb reported a significant cooling trend (Aden and Dammam are also negative but over a much reduced period).

Figure 11.

Arabian Peninsula stations mean annual temperature trends for the period 1980–2008 (except Saudi Arabian stations 1985–2008) in °C per decade relative to global mean annual temperature according to IPCC [Solomon et al., 2007] and Klotzbach et al. [2009].

[59] The recent significant warming over AP during the last few decades is in agreement with several other dryland or AP region studies. Hulme [1996] reported that all dryland regions have warmed. This study shows that the nonmonsoonal region located north of 20°N has experienced higher rates of warming. This is in line with Solomon et al. [2007] who reported greater increases in the interior compared with coastal regions. Temperature trends are specific to sampling periods [Manton et al., 2001; Haylock et al., 2006; Elagib and Mansell, 2000] and the area of study and this is true over AP. El Gindy [1994] found no regional warming for Doha (1962–1992), Seeb (1974–1990), Sur (1977–1990), or Masira (1956–1990). Analysis by Nasrallah and Balling [1996] over the period 1979 to 1990, found similar results to the earlier ground-based assessments, namely a slight warming trend that is not statistically significant. The additional years available to this study change this picture substantially, pointing to the importance of recent years in leading to a statistically significant warming trend.

[60] The AP mountainous stations are located at around 2000 m above mean sea level. Saiq station is located within a diffuse settlement of 2 km width while Khamis Mushait is larger. Over Khamis Mushait the rate of warming is 10 times higher than Gizan in MA season while the trend at Saiq is 5 times higher than Sohar in JAS season. These stations merit further attention.

[61] Dry lands have experienced a negative association between precipitation and temperatures, such that warmer (colder) years have been drier (wetter) [Hulme, 1996; Jones and Reid, 2001]. The temperature and precipitation regimes appear to be very closely coupled (r = −0.456, p = 0.01) over the AP region (Figure 12). Sustained negative precipitation anomalies are matched by sustained positive temperature anomalies since 1998.

Figure 12.

All Arabian Peninsula annual anomalies time series for both temperature and precipitation for the period 1980–2008.

[62] For all AP the highest significant warming occurs in MA season, then DJF and MJ. Some AP regions have experienced cooling, notably the monsoonal subregion in JAS season and the cooling at Seeb of −0.27°C decade−1 (with a July peak of −0.58°C decade−1). Possible reasons for this cooling include the enhancement of the SW monsoon and strengthened upwelling regimes over the western Arabian Sea since 1997 where average summertime phytoplankton biomass has increased more than 350% [Goes et al., 2005]. In addition, Sultan and Ahmad [1993] have demonstrated that during the southwest monsoon (summer) the upwelled cold water from the southern coast of Arabia appears to compensate the heat gain in the Gulf of Oman.

5.1.3. Maximum Temperature

[63] For all AP the maximum temperature is increasing significantly (0.032°C decade−1), but at a lower rate than the mean temperature. The AP mountain maximum temperature trends are striking in that the two stations Saiq and Khamis Mushait, show no significant warming (the latter shows significant cooling) even though the mean annual temperatures here are increasing (1.3–1.5°C) as a result of changes to minimum temperatures. Reasons for the behavior of maximum temperature are unclear.

[64] The warming of maximum temperature over AP is significant, largest and most widespread during late winter and spring. For example in February, 15 stations reported an overall increase more than 1.5°C and 5 stations more than 3°C. This might be due to the sensitivity of maximum temperature to the decrease of precipitation in the highest rainy months as in February (Doha reported overall increase of 4.5°C) and March (Bahrain reported 6.6°C). This is also true for the MA season. The large increase of maximum temperature in the monsoon area during DJF could be related to the weakening of the Sub-Tropical High (STH) over Asia and associated circulation changes such as a reduction in temperature advection by the northerly trade winds [deMenocal and Rind, 1993].

[65] Salalah is an example of how the maximum temperature trends may be sensitive to the start and retreat of the monsoon season. Over this monsoonal station the warming trend is significant and reaches its highest value on May (1.11°C decade−1) prior to the onset of the cooler monsoon season. After the retreat of the monsoon in October the warming trend changes to more positive (relative to summer months) over Salalah (1.00°C decade−1).

5.1.4. Minimum Temperature

[66] For the all AP and the subregions, the increasing trend in mean temperature is matched and derives partly from an increase in the minimum temperature. Minimum temperature is increasing faster than maximum temperature, which is a widespread finding [Aguilar et al., 2005; You et al., 2008; Zhang et al., 2005; Labraga and Villalba, 2009; Manton et al., 2001].

5.1.5. DTR

[67] Solomon et al. [2007] reported that the global average DTR has stopped decreasing. Widespread (but not ubiquitous) decreases in continental DTR have coincided with increases in cloud amounts and the trends are highly variable from one region to another. This is not the case in AP. During 1980–2008 the DTR shows a significant reducing trend at a rate of −0.26°C decade−1.

[68] Due to the uniform high and significant rate of warming of the mean annual minimum temperature over the north and south AP, the mean annual DTR trends do not show high variation between the subregions. Both AP Mountain's stations reported reduction in DTR. Khamis Mushait is reducing at −0.50°C decade−1 (significant at 0.01 level) while Saiq is reducing at −0.40°C decade−1 (significant at 0.1 level). There are some seasonal and monthly variations in the DTR trends. In winter, the monsoonal region witnessed high and significant (at 10% level) DTR reduction. This could be due to the expected RH increase in the monsoonal area and weakening of NE monsoon [deMenocal and Rind, 1993].

5.1.6. Urban Influence on Temperature

[69] In order to investigate urban influence on temperature, which has recently been studied by Parker [2010], we define two categories of stations: large cities and smaller towns/rural settings using a cutoff population of 100,000 inhabitants. A similar approach was adopted by Tayanç et al. [2009]. The population data were derived from the United Nations Demographic Yearbook 2007 ( The large cities category comprise 11 stations (Bahrain, Salalah, Sohar, Dubai, Ras AlKhaimah, Doha, Riyadh, Khamis Mushait, Tabuk, Jeddah and Gizan) while the small towns comprise 6 stations (Khasab, Masirah, Saiq, Sur, Thumrait and Kuwait). Time series for each category were computed for mean, minimum, maximum temperature and DTR.

[70] Mean annual mean temperature trends from large and small/rural cities both show statistically significant warming (at 0.001 level), the larger cities reported 0.43°C decade−1 while the smaller/rural towns reported 0.50°C decade−1. For the rest of the seasons and months the difference are small except in September and ON seasons where surprisingly the smaller/rural cities reported higher significant warming relative to large cities by ≥0.20°C decade−1. In the case of minimum temperature, the larger cities reported higher and significant trends in minimum temperature during late summer (0.77°C decade−1) relative to small cities (0.40°C decade−1). As the maximum temperature trend (0.22°C decade−1) is less than the minimum temperature over the large cities during late summer, this leads to sharp significant (at 0.001 level) decrease in DTR.

[71] Any urban warming signal should be most evident in summer, when urban heat islands are stronger owing to greater storage of solar heat in urban structures [Parker, 2004].

5.2. For the Period 1943–2008

[72] Mean annual warming for Bahrain (0.16°C decade−1) and Salalah (0.15°C decade−1) is similar though slightly lower than the global land temperature increases for the same period (0.18°C decade−1 based on National Climatic Data Centre). Masirah is warming at a still lower rate (0.08°C decade−1). Analyses of extended periods from earlier studies which ended in the 1990s show the importance of warming in more recent years. For example, Nasrallah and Balling [1993] found an insignificant mean annual temperature increase over the Arabian Gulf (1950–1990) although warming was significant in the summer months.

[73] Nasrallah and Balling [1996] also analyzed trends in minimum temperature (1951–1990) for 3 stations (Doha, Dhahran and Riyadh), finding no statistically significant trends. Furthermore they report a small increase in the DTR. The new results from this study show that both Salalah and Masira report significant warming. The annual DTR is decreasing significantly over Salalah but increasing significantly over Bahrain. Recently Elagib [2011] found significant urban influence on DTR over Khartoum. Using data during 1941–2005 he observed rising nighttime temperature at a higher rate than daytime temperature leading to significant reduction in DTR with greater effect on the hot season.

[74] The additional years (to 2008) therefore leads to a revised view of temperature trends. Additionally, the extended period study shows that Bahrain is warming significantly in summer months not in winter but it is the reverse over Salalah and Masirah, pointing to the importance of the monsoon in modulating the temperature response.

[75] Precipitation trends remain insignificant (as in work by Nasrallah and Balling [1996]) even with the additional years available in this study.

6. Conclusion

[76] During the last 2–3 decades 14 out of 21 stations in the AP reported significant warming in the mean annual temperatures. This warming is mainly derived from the minimum temperature (16 out of 21 stations reported statistically significant warming) where the warming rate was higher, more spatially coherent and more statistically significant than for maximum temperature (12 out of 21 stations reported statistically significant warming). This leads to a general decrease of the DTR (6 out of 21 stations reported statistically significantly negative trends with 3 stations reporting significant positive trends) while the precipitation decreases significantly only in two stations. In general, most of warming occurred over the eastern/southeastern areas. The trends show considerable spatial consistency within countries and across regions (subregions) even though the climate varies across the region. Some site specific stations reported significant mean temperatures warming at more than 1.5–3.5 times the global rate. Minimum temperature increases reached approximately 3–6°C in October in both urban and rural stations over the period 1980 to 2008. Overall, these observational data underscore the concerns about global climate change with the projections, as summarized by IPCC, having not exaggerated but may in some respects actually underestimated the change [Rahmstorf et al., 2007]. Regional climate change attribution experiments are much needed for the AP region. These changes need attention by the AP communities to enable planning and adaptation.


[77] We wish to express our gratitude to all the AP NMSs for providing the climate data. Special thanks go to the Ministry of Transport and Communications in the Sultanate of Oman for sponsoring the lead author's doctorate degree of the Arabian climate trends, variability, and change. Data used in this study will be made available for the research community after the permission has been gained by relevant AP NMSs.