Development of cities and human interventions replace natural vegetation with artificial urban materials, such as bricks, concrete, asphalt and steel. These significant anthropogenic modifications which lead to surface and atmospheric characteristics consequently result in inadvertent local weather and climate changes (Oke, 1988; Changnon, 1992; Cotton and Pielke, 1996). The urban heat island (UHI) is a well-known phenomenon which produces urban–rural temperature differences, making urban areas warmer than suburban and rural environments surrounding them (Marsh and Grossa, 1996; Brown, 1999; Robinson and Henderson-Sellers, 1999). Moreover, inter-annual variability of temperature tends to be lower in urban environments than in its rural neighbourhood (Camilloni and Barros, 1997). Knowledge of the effect of excessive UHI is needed for town planning, prevention of high concentration of air pollution and creation of optimum bioclimatic conditions (Kĺysik and Fortuniak, 1999).
There are many research papers in the scientific literature that document comparisons of urban and rural temperatures to detect UHI phenomenon. The most commonly used index of the urban heat island intensity (UHII) is the difference in temperature between representative urban and rural stations. This literature abounds with different conclusions. While urban warming was most conspicuous in winter, e.g. at Kew (Moffitt, 1972), in Prague (Brázdil and Budíková, 1999), in Fairbanks, Alaska (Magee et al., 1999) and in Seoul (Kim and Baik, 2002), other urban cities revealed the greatest heat island in summer (e.g. London (Lee, 1992), ĹódY, Poland (Kĺysik and Fortuniak, 1999) and Athens (Philandras et al., 1999)). Moreover, the UHII in some cities reaches its maximum at night, i.e. the minimum temperature being more affected. Examples of these are London (Lee, 1992), ĹódY (Kĺysik and Fortuniak, 1999), Fairbanks (Magee et al., 1999) and Seoul (Kim and Baik, 2002). Urbanization in Athens mainly affects the daytime temperature (Philandras et al., 1999).
Apart from the index of UHII discussed above, Karl et al. (1993) argue that urban development, among other local effects, may account for a decrease in diurnal temperature range (DTR) due to rising minimum temperature at a faster rate than maximum temperature. Sala et al. (2000), for instance, examined the urban and rural DTR for Spanish Mediterranean and showed that the decreasing trend is much more pronounced at the urban stations, and thus, is related to urban effect.
Although the results presented by Jones et al. (1990) show that
‘the urbanization influence in two of the most widely used hemispheric data sets is, at most, an order of magnitude less than the warming seen on a century timescale’
they concluded that
‘Indeed, careful selection, inspection and monitoring for urbanization influence in the climate record will be required’.
It can be noticed that the majority of the studies that dealt with urban influence on temperature were undertaken in the developed world. It seems that the developing countries lack such studies. Thus, the present study investigates the growth of the Khartoum heat island during the period 1941–2005.
2 Aspects of urban growth in Khartoum
Khartoum complex is comprised of three towns (Khartoum, Khartoum North and Omdurman) situated at the confluence of the White Nile and the Blue Nile. Like many other capital regions around the world, it has become increasingly urbanized during the past three decades as the national capital of Sudan. The population censuses carried out in 1983 and 1993 (Babiker, 1982; Davies, 1988; Central Statistics Bureau, 1993) marked Khartoum as the most outstanding population agglomeration throughout the country, with an annual growth rate of 6.29% since 1983. The population densities in the years 1956, 1983 and 1993 were estimated as 24.1, 85.9, 162.8 persons per square kilometre, respectively (Ministry of Culture and Information, 1994). According to the 2008 Sudan census committee, the State of Khartoum is now accommodating about 5.2 millions (Sudan Watch, 2009), i.e. ≅13% of the total population of the country. This elevates the population density to ≅248 persons per square kilometre. Several reasons can be held responsible for this enormous population growth: expansion of squatter settlements (El-Bushra and Hijazi, 1995; Yath, 1995), widespread rural–urban migration (Hänsel, 1991; Ruppert, 1991; Yath, 1991), large influx of southern Sudanese because of civil war (Ibrahim, 1991), Sudanese migration from western Sudan because of drought, desertification and ongoing conflicts in Darfur, return of expatriates from the Arabian Gulf region and so on.
Accompanying this dramatic population explosion has been the expansion of other human activities within the city, including the extensive use of cars and vehicles, leading to air pollution (Hassan, 1981, 1984), the spread of buildings, factories, paved roads, widespread use of air conditioning and so on. Integrating remote sensing and geographic information systems (GIS), Adam (2007) assessed the urban sprawl of the Greater Khartoum area during the years 1972, 1987 and 2000. Table I demonstrates the expansion of the residential, industrial and farm areas at the expense of other land uses and land cover.
Table I. Change in land use and land cover in the capital of Sudan
Land use and land cover type
Area in 1972 (km2)
Change (%) with respect to the area of the class in 1972
Two meteorological stations (first class stations) representing urban and rural sites are selected. These are respectively Khartoum International Airport in Khartoum (latitude 15°36′; longitude 32°33′; altitude 380 m) and Shambat Agrometeorological Station in Khartoum North (latitude 15°40′; longitude 32°32′; altitude 380 m), as shown in Figure 1. As can be noticed, there is no altitudinal difference between the two stations, thus making them ideal for the present investigation. The distance between the two stations is 15.5 km. Storied buildings, artificial surfaces and heavy traffic dominate around the airport. Using the Urban Climate Zone classification (Oke, 2006; Stewart and Oke, 2009), the urban development surrounding Khartoum Airport in the year 2000 (Figure 1) is a combination of compact housing and house and garden. Conversely, suburban housing, institutions such as universities and schools, ground-floor houses, natural vegetation, farms, vacant areas and scattered asphalt roadways characterize the vicinity of the station of Shambat.
Mean maximum and minimum temperatures, Tx and Tn, respectively, collected by Sudan Meteorological Authority (SMA) during the period 1941–2005 at both stations, are used. No inconsistency due to instrumentation is present in the data. Both stations utilize maximum and minimum thermometers for measurements. It should be mentioned that the data are checked after they have been received from the stations through at least two stages by the staff of SMA. Additional check was made by the author for any suspect outliers due to mistakes during data entry. This represented only one single case for Shambat and was replaced by the correct observation. More details on the homogeneity of the data for Shambat are given in Elagib and Mansell (2000a). Data sets of Tx and Tn are preferred than mean observations as mean daily UHII may conceal patterns of change in daily maximum and minimum UHII (Lee, 1992).
The time series of DTR, the temperature differences between Khartoum and Shambat (UHII) and the variance (square of standard deviation) of the within-year Tx, Tn and DTR values have been calculated. Two values are given for UHII, i.e. for Tx and Tn. The variance has been used herein as a measure of intra-annual variability of the 12-monthly values for the year as adopted by Michaels et al. (1998). Attention is also given to the comparison of inter-annual variability (variance) for two sub-periods, early (1941–1970) and late (1971–2005), in Tx, Tn and DTR between the two stations in addition to the frequency distribution of the values of UHII. Moreover, the UHII values have been compared between three averaging sub-periods: 1941–1970, 1971–2000 and 2001–2005. The analysis of UHI is performed in the present study on the monthly, seasonal and annual series. Elagib and Mansell (2000a, 2000b) define the seasons in central Sudan as dry (January, February, November and December), hot (March to May) and wet (June to October). It should be mentioned that rain in the capital of Sudan falls during the summer.
Kim and Baik (2002) defined the maximum UHII as the maximum temperature difference between urban and rural stations. However, this term is used in this study in different ways of analysis as follows:
1.The maximum of the within-season and the within-year UHIIs for Tx and Tn. For example, the maximum UHII for the wet season is the highest urban–rural temperature difference among June to October values, while the maximum for the year is obtained as the highest value among the 12-monthly values.
2.The highest UHII recorded for the monthly, seasonal or annual time series within the periods 1941–1970, 1971–2000 and 2001–2005.
This paper examines the linear regression lines of the temporal series of Tx, Tn, DTR, variance and UHII to deduce the rate of occurrence of UHI. All directions of trends and their significance have been authenticated by a non-parametric test, namely Spearman rank correlation (SRC) statistic (Kanji, 1997).
4 Results and discussion
4.1 Maximum, minimum and range of temperatures
The comparison between the time series of seasonal and annual mean temperatures for the urban and rural sites is given in Figures 2 and 3, for Tn and Tx, respectively. Overheating of Khartoum is quite evident although it had particularly dampened in the daytime of the wet season during the last third of the 20th century, more likely because of excess rainfall in the urban site (Marsh and Grossa, 1996), as revealed by rainfall data for the two stations (Figure 4). Contrary to this observation, Shambat exhibits higher range between its maximum and minimum temperatures DTR (Figure 5) compared to Khartoum. The trend rates of these time series, together with those for the months, are given in Table II. A gradual and statistically significant rising trend is noticeable for the overall hot and wet seasons as well as the annual series of Tn at both stations. The greater rates of warming are found at the station of Khartoum, except in the wet season as a result of excess rainfall in the urban site as shown in Figure 4. Urban rainfall enhancement has been reported by many investigators, including Huff and Changnon (1973), Jauregui and Romales (1996) and Shepherd (2006). Accordingly, the highest seasonal warming rate is registered for Khartoum in the hot season (0.29 °C century−1) and is more than twice that for Shambat (0.13 °C century−1). The warming in Shambat in the wet season (0.25 °C century−1) has been nearly twice that in the hot season (0.13 °C century−1). On monthly basis, the hot and wet seasons show significant warming trends all through, with the exception of July (Khartoum) and March (Shambat) for which cooling trends, being significant only for the urban station, are indicated. The month registering the highest warming trend is also a characteristic of the station: April (0.38 °C century−1) for Khartoum and September (0.33 °C century−1) for Shambat. The dry season exhibits overall insignificant increasing and decreasing trends, for Khartoum and Shambat, respectively. On monthly basis, significant cooling is indicated for Shambat in January at a rate of −0.25 °C century−1.
Table II. Linear trend rates ( °C century−1) of temperature and UHII during 1941–2005
Urban station (Khartoum)
Rural station (Shambat)
Significant values are given in boldface. Significance level α—
The time series of maximum temperature shows significant overall trends in Shambat for the hot and wet seasons as well as the whole year (warming) and in Khartoum for the dry season (cooling). It is to be mentioned that the SRC test does not confirm the significance of urban cooling of the dry season for Khartoum. In Shambat, the rate is most rapid for the wet season (0.24 °C century−1). January (dry season) and August (wet season) are the only months that reveal significantly decreasing and increasing trends, respectively, in Khartoum although the SRC test indicates a significance level (α) of the increasing trend for April of 0.036. Conversely, April and May (hot season) and July to October (wet season) all demonstrate warming of the climate in Shambat, with the highest warming rate registered for August (wettest month) as 0.43 °C century−1. This rate is ∼2.1 times the highest rate recorded during the hot season in May, thus emphasizing the associated temperature–rainfall behaviour discussed earlier by Elagib and Mansell (2000a) that the warmth during the past few decades appears to have coincided with the rainfall depletion.
All of the discussed Tx and Tn trends above have induced striking trends in DTR. The significant falling trends of the seasonal and annual time series are characteristic of the urban site of Khartoum International Airport. It can be noticed that the seasonal falling rate is greatest for the hot season and least for the wet season; the former is ∼2.4 times the latter. Nine months of the year (October to June) reveal a significant drop in the DTR at rates ranging from − 0.18 (November) to −0.32 °C century−1 (May) due to:
1.the general cooling daytime and warming night-time (dry season);
2.both warming and cooling daytime and warming night-time (hot and wet seasons).
These results indicate significance of the trends for the wet season only in the beginning and end of the season months. A rise in Tn at a rate of up to 2.89 times that in Tx can be seen in April for Khartoum compared to 1.34 for Shambat. Shambat has only two significant trends in its DTR series: a falling trend in the month of June (−0.15 °C century−1) and a rising trend in the month of August (0.20 °C century−1).
4.2 Variability in maximum, minimum and range of temperatures
The analysis of inter-annual variability abounds with remarkable mixed results in terms of both temperature variable and period of analysis (Figure 6). Although the difference in the variance between the two stations is small, the Tx shows higher values for Shambat than for Khartoum in April–September during 1941–1970 and in August and October–February during 1971–2005. Similarly, the Tn indicates higher inter-annual variance for January, April, May and August–October during 1941–1970 in comparison to only the hot season months of April and May during 1971–2005. Herein, the difference in the variance between the two stations is larger than in the case of Tx. During the normal period of 1941–1970, the DTR reveals higher inter-annual variability for the rural station in 9 months (January, February and April–October). For the same station, 9 months (December–August) show higher inter-annual variability in DTR during the second period of 1971–2005.
The salient characteristic of Figure 7 is the higher intra-annual variability of Tx, Tn and DTR for the rural station of Shambat compared to the urban station of Khartoum, with smaller differences in the variance pertinent to Tx than those obtained for Tn and DTR. Table III gives the respective trend rates from the linear regression for the three temperature variables. Upward trends are shown for the intra-annual variability in Tx and Tn and downward ones are clear for those relevant to DTR. The rural site has higher trend rates for the variance relating to Tx and Tn, while the urban site has higher rate for the variance pertinent to DTR. The significance of the trend is evident for all the time series except that for the DTR of the rural site.
Table III. Linear trend rates ( °C2 century−1) of intra-annual variance of temperature and their significance levels during 1941–2005
4.3 Urban heat island intensity
Figure 2 also depicts the evolution of the UHI for the seasons and the year in Khartoum, the intensity of which is more perceptible in the temperature minima than in the temperature maxima. Negative UHII (not shown in figures) can be registered in all months in the case of Tx, but it is confined to the months of May to February in the case of Tn. The mean seasonal and annual series of Tx exhibit cases of negative intensities of UHI. The negative urban effect is seen in the mean Tn series only for the wet season, possibly due to the effect of excess rainfall in the urban location. Seasonwise, no negative values appear in the series of maximum UHII (Figure 8) except in the wet season and only in the daytime records. Further investigation is needed to unravel the meteorological conditions that control these negative values.
Table II documents a clear deceleration of the temporal growth of UHI in Tx throughout the year because of three situations:
1.increasing trends for Shambat and decreasing trends for Khartoum (March, May, June, October, November and annual mean);
2.increasing trends for Shambat at higher rates than those for Khartoum (April, July, August, September, hot season mean and wet season mean);
3.decreasing trends for Shambat at slower rates than those for Khartoum (January, February, December and dry season mean).
This urban cooling is strongest in the hot season and weakest in the dry season. The hastiest monthly rates of urban cooling for the seasons in order of decreasing magnitude are those for May, October, February, November and December. Regarding the seasonal and annual maxima, urban cooling through time is only significant for the hot season series with a rate of −0.16 °C century−1. In contrast to the linear regression analysis, the SRC test gave significant results for the deceleration of UHII in the following additional cases: January (α = 0.042) and the maximum series for the year (α = 0.014), dry season (α = 0.012) and wet season (α = 0.004).
It can also be seen from Table II that an acceleration of the intensification of UHI in Tn is dominant in the hot and dry seasons. The cases of this urban warming are a result of the following trend combinations (refer to Table II):
1.More rapidly decreasing trends for Shambat compared to those for Khartoum (January).
2.More rapidly increasing trends for Khartoum than those for Shambat (April to June, October, December, hot season mean and annual mean).
3.Upward trends for Khartoum and downward trends for Shambat (February, November, March and dry season mean).
The most prominent occurrence of urban warming in Tn has a peak rate of 0.19 °C century−1 in both February (during the dry season) and March (during the hot season). Equal rates of intensifying seasonal UHI are shown for the dry and hot seasons. In terms of seasonal maximum UHII, there is a tendency for the UHI to increase significantly at a faster rate during the hot season compared with the dry season.
On the other hand, the period of the most remarkable decrease in urban–rural temperature differences in Tn coincides with the wettest period of the year (July–September), with the significant downward rate (−0.11 °C century−1) observed for September. This urban cooling appears to have happened in response to (see Table II):
1.downward trend for Khartoum and upward trend for Shambat (July);
2.increasing trends for Shambat at higher rates than those pertaining to Khartoum (August, September and wet season mean).
A comparison of the mean monthly UHII during the periods 1941–1970, 1971–2000 and 2001–2005 has also been made. During December to May of 1971–2000, the influence of urbanization on Tn was greater than during 1941–1970. This is a contrast to the wet season (June to October). The highest difference in the mean UHII values between the two periods is 0.6 °C and is noted in September. The last 5 years of the study period were markedly affected by urban warming from October to July. On the other hand, the monthly UHII in Tx is always less than 1 °C on average. The urban effect here happened to be weaker throughout the year after the first normal period (1941–1970). Normally, urban cooling is noticeable in the wettest months of July and August during 1971–2000, registering 17 cases of negative UHII in the two months compared to three and seven cases, respectively, during 1941–1970. In addition, the wet months of September and October also recorded negative values of UHII, but with less frequency of occurrence. The respective numbers of cases during 1971–2000 were nine and eight in comparison to three cases for each month within 1941–1970. The period 2001–2005 experienced the weakest urban effect on average. Several of these results can be authenticated by Figures 2 and 3.
Further comparison of the results for the three periods is presented in Figure 9, which reveals that the maximum values of UHII in Tx recorded during 1941–1970 were higher than those recorded during 1971–2000 in 9 months. The difference between the intensities reached a value of up to 2.6 °C, as in the case of May. In contrast, in 8 out of the 12 months the highest maximum UHII in Tn occurred during the period 1971–2000 rather than during 1941–1970, with the highest difference reaching up to 2.6 °C in February. The maxima of UHII for the entire 65-year period were as follows: 6.4 °C (May; hot season), 5.5 °C (February; dry season) and 5.5 °C (October; wet season) with respect to Tn and 4.1 °C (November; dry season), 3.6 °C (May; hot season) and 3.4 °C (July; wet season) in terms of Tx.
While such results demonstrate the fast changes of UHI (urban daytime cooling and urban night-time warming) in the recent period(s) by comparing UHII data for different periods, the inter-annual variability of UHII can play an important role. This could be seen in Figures 2, 3 and 6. For instance, decadal running analysis of the UHII (not shown) revealed stronger urban cooling in annual Tx during 1988–1997 (mean UHII = 0.11 °C) than during the last decade (mean UHII = 0.25 °C). As for the annual Tn, the mean UHII was highest during 1955–1964 (1.8 °C) and lowest during 1971–1980 (1.1 °C), while the last decade of the study period had an UHII of 1.7 °C. However, the same analysis tested for 30-year running periods accentuates the tendencies of UHIIs indicated by the results discussed earlier and in Table II.
Analysis of the seasonal and annual mean and maximum values of UHII demonstrated the following. Firstly, the Tn has the greatest difference between urban and rural temperatures in the hot season and the least difference in the wet season. Secondly, the urban warming affects the Tx series mostly in the dry season and least in the wet season. Thirdly, on average, Tn in the dry and hot seasons as well as the whole year was affected more by urban warming during 1971–2000 compared to 1941–1971. The contrary is true for Tx, a characteristic extending also to the wet season.
Figure 10 displays the histogram of the frequency distribution of the monthly UHIIs for the seasons and the whole year. Regarding the Tx records, a peak class of UHII of 0 to less than 1 °C is common for the seasonal and annual series, accounting for 78, 76, 61 and 71% of the observations, respectively, for the dry, hot, and wet seasons and the year. In the case of Tn records, the peak class is 1 to less than 2 °C for the dry season, hot season and annual records, with respective frequencies of 50, 45 and 43%, and 0 to less than 1 °C for the wet season records. It has also been found that around 93, 94, 86 and 87% of the cases lie, respectively, in the ranges of 0 to less than 3 °C (dry season UHIIs), 1 to less than 4 °C (hot season UHIIs), 0 to less than 3 °C (wet season UHIIs) and 0 to less than 3 °C (yearly UHIIs). In terms of maximum UHII, the night-time temperatures have 86% of the cases in the range of 1 to less than 4 °C, whereas the daytime temperatures comprise 88% of the cases within 0 to less than 2 °C. The peak class of the frequency histogram of nocturnal maximum UHII is 2 to less than 3 °C, accounting for 43% of the cases, while that of daytime conditions is 1 to less than 2 °C, accounting for 49% of the cases. The total number of monthly negative values of UHII during the entire period of study is 111 (∼14%) in the case of Tx. These cases are distributed among the seasons as 10 (dry), 14 (hot) and 76% (wet). Only seven negative values, representing ∼6%, occurred in the case of Tn during 1941–2005. These cases are distributed as 3, 1 and 3 cases for the dry, hot and wet seasons, respectively. The results of Figure 10 give additional evidence of greater urban influence on Tn over Tx.
5 Summary and conclusions
A gradual development of the capital of Sudan has taken place during the past few decades. Series of synchronic maximum and minimum temperature measurements at urban and rural stations over 1941–2005 enabled the analysis of the temporal structure of the UHI of Khartoum, using DTR and UHII as indices. Significant urban influence on DTR, i.e. rising night-time temperature at a higher rate than daytime temperature leading to significant reduction in DTR, dominates in Khartoum. This dominance characterizes the hot and wet seasons, with greater effect observed during the former season. There seems to be a slight upward tendency in the seasonal and annual DTR in Shambat. The second index (UHII) has values in Tx considerably smaller than those in Tn. Both nocturnal urban warming and daytime urban cooling through time were reported. The former is intensifying at significant rates in the series for the dry and hot seasons as well as for the whole year; the intensification reaches its maximum in the hot season. Significant urban cooling, on the other hand, is strongest in the hot season and is weakest in the dry season. The effect of probable excess rainfall conditions of the urban site reflects in the values of UHII for the wet season, giving significant downward trend in the Tn time series for September. Cases of negative UHII have been noted for the monthly Tx and Tn data, accounting for around 14 and 0.9% of the cases, respectively. On seasonal basis, the daytime cases of monthly negative UHII distribute among the season as follows: 76% for the wet season, 14% for the hot season and 10% for the dry season. The determination of the controlling climatic conditions and ranges regarding these negative values of UHII needs further investigation.
On one hand, the variance shows higher intra-annual variability in Tx, Tn and DTR for the rural site, and on the other hand, it displays mixed results with respect to inter-annual temperature variability.
As stated by several investigators (Moffitt, 1972; Brázdil, 1993; Brázdil and Budíková, 1999; Philandras et al., 1999), correction should be made in the temperature time series for sites exposed to urban influence to ensure reliable analysis of their temporal fluctuation and assessment of climate change.
The author is grateful to Mr Babiker F. E. E. Adam for assistance in producing Figure 1.