Placing in situ instruments into an urban environment to measure energy fluxes raises the question of representativeness. Although instruments can be put into the constant flux layer, a measurement finally stands only for the actual source area and can be compared directly only to areas with similar surface characteristics (Oke, 2007). To a lesser extent, the same is true for agricultural or natural areas. The urban surface can be described by different roughness elements like buildings, paved surfaces, their thermal and optical properties, the density of optional vegetation, and bare soil/sand coverage. The common heterogeneity of an urban landscape puts a constraint on the representativeness of in situ measurements carried out at a single place. To overcome this restriction, a multitude of towers would be needed to cover a wider area. This was done by Kanda et al. (2006), using five towers in a densely built-up area in Tokyo, Japan. He found that there is a remarkable spatial variability, especially in the morning sensible heat flux, which was related to the areal fraction of vegetation of the immediate environment of the measurement: a 200-m radius circle around each tower. The highest difference of their ensemble-averaged sensible heat flux was found to be about 50 Wm−2 shortly after noon. As their area of interest was still more or less homogeneous, comparing fluxes of different quarters of a city (e.g. comparing a dense with a less dense built-up area) might show an even higher variability. To account for this problem, one might want to use a technique that senses wider areas at the same time, e.g. remote sensing techniques. Using measurements from space, the whole of a megacity can easily be captured at once. However, no direct measurements of most variables of the energy balance are possible by satellite; therefore, there is intensive research into the estimation of the energy budget from space. Most studies have aimed at the investigation of natural or agricultural surfaces (Roerink et al., 2000; Jia et al., 2003; French et al., 2005; Li et al., 2008). However, Rigo and Parlow (2007) have modelled the ground heat flux of an urban area using remote sensing data, and Xu et al. (2008) have derived the whole energy budget for an urban surface using data from a high-resolution sensor mounted on a helicopter. Such methods allow a spatial investigation of fluxes of different urban surfaces, even in comparison with the outer environment. A further advantage is the easy access to remote areas where it is hardly possible to conduct in situ measurements due to geographical, political, or social reasons.
Following this idea, we wondered whether it would be possible to derive the terms of the energy budget satisfactorily from remote sensing data over an urban surface of a city with non-optimal political and social conditions, preferably without using in situ instrumentation. To control this hypothesis results must be cross-checked with a set of in situ data and methods should probably be refined. Therefore, a control city featuring many different microclimates and interesting contrasts had to be selected. Existing algorithms for the derivation of heat fluxes can then be tested for their robustness and general practicability.
The city of Cairo, Egypt, was chosen because of its unique location: Situated in a hot and dry climate and nonetheless partly surrounded by agriculture, a variety of different rural and urban microclimates are evolving. This spatial heterogeneity asks for a process-oriented approach that accounts for the climatic differences in the spatial domain. Further, Cairo is one of the most heavily polluted megacities in the world. The pollution, originating from traffic and industries, is dangerous to human health and has a further impact on the radiation budget.
In the framework of Cairo Air Pollution and Climate (CAPAC), a measurement campaign was conducted from November 2007 to February 2008. At three different stations, all main components of the energy budget were measured continuously additionally to air temperature and humidity. A side aspect of the CAPAC campaign focused on the air pollution of the city. In situ measurements at different locations provided a first understanding of background and street-side concentrations of coarse and fine particulate matter (PM). Further, very high-resolution CO2 flux measurements complete the picture.
In this paper, the setup of the three stations and the main characteristics of the measured variables (except PM and CO2 data) will be presented. A short discussion of Cairo as an urban heat island (UHI) will be included. The analysis of PM and CO2 is beyond the scope of this paper. Inherent to the technique of data capture from remote sensing platforms, satellite images provide top of the atmosphere radiances from a single very short integration time per unit area of the surface. Potential problems arising from the connection of in situ measurements to this kind of data will be discussed in the conclusions section (Section 5.). The main conclusions of the whole energy balance study including the remote sensing analysis will appear in a forthcoming paper (‘Flux Measurements in Cairo, Part 2’). The results of the CO2 analysis will also be presented in a follow-up paper (‘Flux Measurements in Cairo, Part 3’).
2. Location and setup
Greater Cairo is the largest city in Egypt and on the African continent. Nearly one in five Egyptians lives in Greater Cairo. For centuries Cairo has been a leading city, dominating the social, economic, and political life of the region (Weeks et al., 2005). What is seen from space as one megacity is Greater Cairo, which consists of the governorate of Cairo on the east side of the River Nile, the governorate of Giza that lies along the west bank of the River Nile, and the southern tip of the governorate of Qalyubiyya (also known as the Northern City), which represents the northernmost fringe of Greater Cairo. However, the latter two governorates also incorporate many villages and other free-standing urban settlements (Sutton and Fahmi, 2001).
The city core consists of a densely built-up area, which was already partly developed in the 10th century AD. To the east and west, the city is surrounded by non-arable desert land. Partly to the south along the River Nile but especially to the north in the Nile Delta, prime agricultural land is found. Greater Cairo is a rapidly growing megacity. According to the Central Agency for Population, Mobilization, and Statistics (CAPMAS), Greater Cairo had an estimated population of 17 600 000 in 2006 (Fahmi and Sutton, 2007). Other sources estimate a population of about 20 million inhabitants (Schlink et al., 2007).
Around Cairo new towns and settlements are being constructed to relieve the population pressure. The 10th Ramadan on the road to Ismailia is an example of these new towns (Sutton and Fahmi, 2001). A major factor boosting this suburbanization process was the construction of the ring road connecting Cairo's fringes to its centre.
From November 2007 to February 2008, a micrometeorological campaign was conducted in Greater Cairo to measure in situ surface energy fluxes. The purpose of these measurements was to deepen our understanding of the energy budget in a megacity like Cairo, obtain local knowledge of the area, and measure ground-truth data for the remote sensing-based energy balance study of Greater Cairo using Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER, http://asterweb.jpl.nasa.gov/) data from NASA.
For the duration of the campaign, three stations were maintained: an urban station in the district of Giza, a suburban-agricultural station in the area of Bahteem in the north of Greater Cairo, and a suburban-desert station close to 10th Ramadan City, northeast of the agglomeration of the megacity. Figure 1 shows a LANDSAT RGB composite depicting the locations of the three stations.
The urban station (30°01′33.39″N, 31°12′27.81″E) was located on the roof of a 15-m building in the campus of Cairo University (http://www.cu.edu.eg/english/), which belongs to the governorate of Giza, west of the River Nile. A 12-m mast was mounted on the roof of the building, which was situated in the most southern part of the campus and was taller than most of the other buildings. This resulted in a total measuring height of 27 m. The campus is a secured area with massive three- to four-storey buildings, paved roads with cars, squares, footpaths, and planted greens. The campus is partly surrounded by roads with heavy traffic, but in the south and east some extended green areas are found. They consist of a sports ground, botanical testing fields of the university, a zoological garden, and a park. To the west and north of the campus, two areas of shabby blocks of flats separated by narrow alleys are found.
The suburban-agricultural station (30°08′38.58″N, 31°15′25.26″E) was placed on an alfalfa field inside the meteorological station of Egyptian Meteorological Authority (EMA, http://www.nwp.gov.eg) in Bahteem town, an outskirt in the North of Cairo. Planted fields extended to the east and north of the station. To the south and west of the station, residential areas are found. Bahteem is one of the poorer quarters of Cairo, only a few streets are asphalted, and rubbish scattered around is a major problem. Some small and primitive glass, fertilizer, and ironworks factories are found.
The suburban-desert station (30°14′44.04″N, 31°43′08.64″E) was situated outside 10th Ramadan City, which was built into the desert. Around the station, which was also a measuring site of EMA, some new asphalt roads, compounds (mostly under construction), newer factories, and a planted green area in the east were found. But mostly the environment consisted of sandy surfaces. Figure 2 shows Google Earth cutouts depicting the surroundings of the three stations.
The purpose of selecting these three stations was to cover the three most dominant landscape features of the region: urban areas, agricultural fields, and desert. Therefore, the sites should be as representative as possible of each feature. However, Cairo is a highly diverse megacity with many different quarters. The extremely dense housing in poor quarters cannot be compared to the planned new cities at the rim of Cairo with their even streets and well kept villas or the spacious quarters of public buildings like the campus of Cairo University. The urban station, therefore, does not represent the whole megacity, but only gives an idea of the selected place, and certainly it can be used for comparison to the remotely sensed data. The same considerations are true for the agricultural station. Farming is mostly done as a family business, so commonly the fields are very small and crops are diversified (El-Khattib et al., 1996).
Beside the scientific demands, the selection of the sites was dominated by security and networking considerations. Egypt is a third world country, and therefore permanent protection of the stations was required. Furthermore, it was necessary to get permission from the authorities for the mounting of the stations. Therefore, the stations were mounted only on restricted sites on governmental ground. Cooperation with the two organizations (Cairo University and EMA) ensured a protected zone for the measurements. Table I shows the characteristics of the three stations; Table II shows the instruments used and their descriptions.
Table I. Station characteristics
Measurement height h (m)
Zeroplane displacement height d (m)
Altitude a.s.l. (m)
27 (temperature: 20)
Table II. Instrument setup during CAPAC campaign
CNR1 by Kipp and Zonen
Four-component net radiometer
Shortwave radiation (incoming and outgoing)
Longwave radiation (incoming and outgoing)
CSAT3 by Campbell
Three-dimensional sonic anemometer
Wind direction and wind speed
Sensible heat flux
In combination with LI-7500: latent heat flux and CO2 flux
LI-7500 by Licor
Open path infrared gas analyser
CO2 and H2O concentration
In combination with LI-7500: latent heat flux and CO2 flux
H2O concentration In combination with LI-7500: latent heat flux
Ventilated temperature measurement
Wet bulb and dry bulb temperature
Barometric pressure sensor
Shortwave radiation (reference)
Longwave radiation (reference)
Soil heat plates
Soil heat flux (each station five pieces)
Soil temperatures (each station five pieces)
Deposition by sedimentation on an adhesive foil
Coarse particulate matter (PM) > 2.5 µm (urban three pieces)
Ventilated system sampling the PM2.5 fraction
The basic equation for all surface energy balance studies is well known and can be expressed as:
where Q* is the net radiation, Qs is the soil heat flux, QH the turbulent sensible heat flux, and QLE the turbulent latent heat flux. The anthropogenic heat flux is included in the measurement and not listed separately. More recent research has shown that horizontal advection QA may play a significant role in the energy budget (Feigenwinter et al., 2008). QA should be added to Equation (1), especially as our measurement height is less than twice the building height. Unfortunately, it is not possible to measure this term with a standard eddy covariance tower (Nemitz, 2002; Foken, 2008). Equation (1) therefore assumes zero advection, as was done by Oke et al. (1999), Grimmond and Oke (2002), and Spronken-Smith (2002). In the following, basic equations and correction methods of each of these terms will be explained.
3.1. Net radiation
Net radiation was determined using the four-component radiometer CNR1 (Table II). Four domes measure separately the shortwave downwelling radiance (K↓), the shortwave upwelling radiance (K↑), the longwave down- welling radiance (L↓), and the longwave upwelling radiance (L↑). The net radiation Q* is:
The measurement by the two pyrgeometers on the CNR1 includes the thermal emission of the instrument. Therefore, it has to be added to the registered radiance, using the pyrgeometer temperature, which is measured with a Pt-100 inside the CNR1. Calibration constants were determined at the end of the campaign in a comparison experiment. All CNR1 were mounted side by side next to references CM22 and CG4 (Table II), which were calibrated at the World Radiation Centre (WRC) in Davos, Switzerland. The radiation values were stored as 1-min averages.
3.2. Soil heat flux
The soil heat flux Qs was measured in two layers using three heat flux plates (HFPs) in the upper layer (5 cm) and two HFPs in a deeper layer (Table II). For the calculation of the energy balance, only the HFPs of the upper layer were used. In addition, thermocouples were buried at 5 cm to measure the soil temperature Tsoil. Soil heat flux was calculated as:
where Qs(z) is the soil heat flux measured at the depth z, f is the Philips correction factor, and Qs(0–z) is the soil heat storage of the soil layer above z.
Cv is the volumetric soil heat capacity (Jm−3K−1), which had to be estimated for the respective soils for this study. The used values were 2.5 × 1E6 for Bahteem and 1.5 × 1E6 for the 10th Ramadan station. The factor f of the Philip correction (Philip, 1961) was calculated using the geometrical measurements of the HFPs, as well as the thermal conductivity (Wm−1K−1), which also had to be estimated. The used values were 0.5 for Bahteem and 0.7 for the 10th Ramadan station. The constant value at Bahteem leads to a certain mistake, as its water content was not constant due to irrigation (Tsoar, 1990; Abu-Hamdeh and Reeder, 2000). There is some controversy about the Philip correction in the literature (Sauer et al., 2003), but for this analysis it was anticipated that the effect of omitting this correction would exceed the errors induced by the application of the Philip correction. Especially at the desert station, the thermal conductivity of the sand was assumed to differ considerably from the HFP's conductivity.
Using the soil temperatures of 5 cm depths misrepresents the storage heat flux substantially, as the temperature gradient is strongest in the top layers of the soil. In the morning, Qs therefore showed a time lag behind the solar irradiation, which could be explained by the unaccounted warming of the uppermost layer of the soil. In the afternoon, though, Qs was overestimated. To account for this, the radiative temperature from the CNR1 was introduced as an additional soil temperature. Therefore, dTsoil in Equation (4) is a weighted mean of the two temperature differences of both the soil and the radiative temperature. The soil heat flux was also stored as 1-min averages.
3.3. Turbulent heat fluxes
Turbulent heat fluxes were measured using an eddy covariance system coupling a sonic anemometer (CSAT3) with an open path infrared gas analyser (LI-7500) at both Cairo University and Bahteem (Table II). At 10th Ramadan, a fast hygrometer (KH2O) replaced the open path infrared gas analyser. The measurement rate was 20 Hz. Raw fluxes were then calculated online on a 30-min basis.
Sensible and latent heat fluxes were calculated from:
where ρa is the density of air, cp is the specific heat of air, λ is the latent heat of vaporization of water, and w′, T′, and q′ are the fluctuations in vertical velocity, temperature, and water vapour mixing ratio, respectively.
For several reasons, it is not possible to measure w′T′ and w′q′ directly. Therefore, a series of four different corrections need to be applied.
The Schotanus correction accounts for the influence of humidity on the sonic temperature T (Schotanus et al., 1983; Kaimal and Gaynor, 1991; Oncley et al., 2007). The Schotanus correction was applied automatically online by the CSAT3 for the sonic data of all three stations.
The krypton hygrometer KH2O is a highly sensitive hygrometer that measures rapid fluctuations in atmospheric water vapour using two absorption bands in the ultraviolet region. A correction for the absorption of ultraviolet light by oxygen was necessary (Van Dijk et al., 2003). This correction of the latent heat flux needed to be applied only at 10th Ramadan and was usually very small for the CAPAC campaign (see also Oncley et al., 2007).
Further, the Webb, Pearman and Leuning (WPL) correction (Webb et al., 1980), which compensates for the influence of the fluctuations in temperature and water vapour on the vertical flux of water vapour density, was applied for all three stations.
Eddy covariance systems attenuate the true turbulent signals at very high and low frequencies. Limitations in sensor response, path-length averaging, sensor separation, and signal processing lead to a loss of information. Moore (1986) has presented a set of spectral transfer functions, which are relatively comprehensive. However, they require a priori assumptions about the cospectral shape. In case actual cospectra resemble the assumed one, this approach provides useful correction factors (Massman and Lee, 2002). Although in the meantime some other methods have been presented (Massman and Lee, 2002; Spank and Bernhofer, 2008), in this study the traditional transfer functions of Moore for sensor separation and sensor line-averaging were used. All these corrections are conventional and no further research on new correction methodologies was pursued, as the main interest of this study was to provide background information on the energy balance of Cairo for the subsequent remote sensing investigation.
All measured data were further checked for erroneous values resulting from the regular maintenances, occasional rain events, and insects. A despiking algorithm took care of any other unknown events. Missing values were interpolated according to the following rule: In time-series of 1-minute averages, a maximum of up to 30 continuous missing values were replaced; in 30-min averages, the number of allowed continuous missing values was 2. This rule left a few periods without valid data; these values were set to a missing value. The calculation of the ensemble averages in this study includes these missing values. At Cairo University, missing values of the 30-min averages accounted for 10% of all values. At Bahteem, however, 52% of all values were missing. This high number is mainly because of a high proportion of missing values during the night. Between the hours of 06:00 and 16:30, the percentage is 19%. At 10th Ramadan, missing values accounted for 23%. Missing values are distributed evenly. It is anticipated that they do not have a significant influence on the findings (excluding nocturnal fluxes at Bahteem).
The following gives a brief overview discussing the measured terms of the energy balance one by one. The main focus is on the comparison of the three stations.
4.1. Wind speed and direction
Both wind speed and direction were recorded as 30-min averages. Wind speed showed a clear diurnal course at all three stations, with high values during the day and low values during the night. The 10th Ramadan showed higher wind speeds than Bahteem during day and night. The maximum ensemble-average wind speeds of the whole duration of the campaign were 4.2 ms−1 for 10th Ramadan and 2.6 ms−1 for Bahteem. As the measurement was at a higher level at Cairo University than at the other two stations, it is problematic to compare these values. However, the maximum ensemble-average wind speed at Cairo University was 3.6 ms−1. Two dominant directions were observed: a northern and a southern direction, whereas in 10th Ramadan the directions are rather northeast and southwest and in Bahteem they are north and southwest (Figure 3). The southwesterly winds are mainly apparent in winter and spring. They are called ‘Khamasin winds’ and are often associated with dust storms (Favez et al., 2008). Such a storm occurred from 29 to 30 January 2008, when a maximum wind speed of 10 ms−1 was measured. This storm subsequently corrupted the measurements, as the instruments got very dirty. Generally, wind direction did not follow any diurnal course when blowing from the northern sector. However, during ‘Khamasin’ days, the wind direction followed a clear diurnal pattern with a slight west shift in the later afternoon. Figure 3 shows wind direction and speed for the three stations.
4.2. Air temperature
Air temperatures showed clear diurnal cycles at all stations and decreased from November until the end of January to rise again in February. The mean temperature at Cairo University during the campaign was 14.9 °C, while the maximum ensemble-average temperature was 19.3 °C and the minimum ensemble-average temperature was 9.6 °C. The mean maximum temperature (mean of the daily maximum) at Cairo University was 19.2 °C, and the mean minimum temperature (mean of the daily minimum) was 10.9 °C. These values are characteristic for a hot desert climate (the annual mean temperature is over 18 °C and there is only a little precipitation, which mostly falls in the winter months).
Cairo features a typical nocturnal heat island, as was found by Robaa (2003). To cancel out the topographic effect on temperature, in the following temperatures were corrected for their height using simply the dry-adiabatic temperature gradient (0.00981 K m−1). This correction, however, did not account for the fact that air temperatures at Cairo University were measured at 20 m (5 m above the roof), while the other two measurements were made at only 1.9 m. According to Kanda et al. (2005), a distinct gradient exists in the street canyon, even if maximum temperatures do not always occur at the same height. For this study, an average deviation of 1 K between the tower measurement and the temperatures at ground was estimated (not included in the data). An unambiguous heat island study should also include topographic wind effects. However, this would reach beyond the scope of this paper. Figure 4(a) shows the ensemble-average air temperatures of the three stations.
Air temperatures were higher at Cairo University than in Bahteem and 10th of Ramadan during the whole night. The daily mean maximum difference between Cairo University and Bahteem was 5.0 °C, and that between Cairo University and 10th Ramadan was 3.2 °C (not shown in Figure 4(a)). These differences, together with Figure 4(a), document the nocturnal heat island of Cairo. In the morning, urban temperatures did not increase as much as the others; at noon the urban temperatures were lowest. Compared to 10th Ramadan, even a midday cool island seemed to evolve. The subsequent cooling in the evening was strongest in Bahteem, and lowest at Cairo University, which led to the re-evolving nocturnal heat island. The coolest temperatures were found in Bahteem during the night.
UHI studies are conducted using in situ measurements of air temperature (Chow and Roth, 2006; Garcia-Cueto et al., 2007; Parlow, 2007; Gaffin et al., 2008), but also using thermal remote sensing images (Hung et al., 2006; Jusuf et al., 2007; Stathopoulou and Cartalis, 2007; Frey et al., 2007). Figure 4(a) and (b) shows that the results of these studies should not be compared directly, as air and radiative temperatures vary in behaviour and magnitude. Voogt and Oke (2003) therefore used the terms UHI and surface urban heat island (SUHI) for the respective case. During the night, the difference between the two temperatures is not so obvious. Bahteem and 10th Ramadan cooled to the same radiative temperature, while Cairo University featured considerably higher radiative temperatures ( = nocturnal SUHI). However, during the day, Cairo University had higher radiative temperatures than Bahteem, but clearly cooler radiative temperatures than 10th Ramadan. The strong heating during the day and cooling during the night of the desert surface appears here in contrast to the urban surface, which seemed to store the heat more efficiently in the buildings. A reason for the lower daytime radiative temperatures in Bahteem compared to its relatively higher daytime air temperatures was the wetness of the soil. Before irrigation events, surface temperatures in Bahteem might even exceed those at Cairo University. Nevertheless, after the irrigation, Bahteem's daytime radiative temperatures dropped considerably. Overall, during the day radiative temperatures were higher than the air temperatures, while at night, they were slightly cooler. The conversion from radiative to brightness temperatures would slightly enhance the differences between Bahteem and the other two stations, due to the lower estimated emissivity of sand and urban construction materials.
To quantitatively characterize the UHI of Cairo, mean, mean maximum, and mean minimum differences (MD, MDmax, and MDmin) as well as standard deviations between Cairo University and the two suburban sites were calculated for day and night hours from air and radiative temperatures. Day was defined as the time between 06:00 and 17:00 (local time, approximate sunrise to sunset) and night as the time from 17:00 to 06:00. Table III shows that Cairo functioned as a heat island not only in the night but also during the day, compared to Bahteem station, when averaged over these time intervals. However, 10th Ramadan station was hotter during the day. This result was expected, as 10th Ramadan station is located in the desert, where almost no latent heat flux occurred and most available energy served to heat up the soil and the air. While both the air and radiative temperatures show the same patterns, the magnitudes of the radiative temperatures are much higher.
Table III. Daytime and nocturnal mean differences (MD), mean maximum differences (MDmax), and mean minimum differences (MDmin) and the standard deviations, respectively; air (Tair) and radiative temperatures (Trad)
Cairo University–10th Ramadan
The comparison shows further that the definition of an UHI/SUHI depends considerably on the reference station in the environment (see also Grimmond et al., 1993).
The Cairo nocturnal heat island was not a permanent feature, but depended on the wind situation. Bahteem and 10th Ramadan featured lower air temperatures than Cairo University, particularly when the main wind direction was north-northeast to north-northwest. On such occasions, wind speed was low. On the contrary, Bahteem and sometimes even 10th Ramadan air temperatures might equal Cairo University temperatures, when the strong south to southwest winds were blowing. It seemed that the whole UHI of Cairo was drifting in a north-northeast direction. Figure 5 shows the air temperatures from the three stations for 3 days along with the wind speeds. Obviously, on the first two nights, temperatures were very similar due to strong winds turning to the south, while on the third night temperatures were much cooler in Bahteem and 10th Ramadan compared with Cairo University, while light northern winds prevailed.
4.3. Radiation fluxes
The measurement of solar radiation K↓ indicated a positive gradient in atmospheric transmissivity from the city to the urban fringe and beyond, as was found by Shaltout et al. (2000) and Robaa (2009). At Cairo University, the smallest K↓ values were measured, which can be explained by the strong air pollution in Cairo. At Bahteem, K↓ was only slightly higher, which might be due to air pollution on one hand and high levels of dispersed dust from the commonly unpaved roads on the other hand. The 10th Ramadan station showed the highest K↓ (Figure 6). The ensemble-average difference between K↓ from 11:30 to 12:30 at Cairo University and 10th Ramadan station was − 62.0 Wm−2, including all weather situations. The difference between Cairo University and Bahteem at the same time of day was only − 16.6 Wm−2. Wind direction did not influence these differences. The lower slope of Cairo University and Bahteem in the morning is also noticeable, and might be due to the morning fog in Cairo, which sometimes occurs during the winter months.
Shortwave reflectance is dependent on the albedo α. The albedo was derived using α = K↑/K↓ from the times 11:00 to 13:00 only, as the ratios showed the typical bowl shape. The α value was not constant, but altered with time. A first rise in α at Cairo University was due to the clearing activities on the roof at the beginning of the campaign. They ended on 6 December 2007. Further, α was mainly influenced by rain events and by the rate of drying afterwards. The low runaway values of Cairo University and 10th Ramadan coincide with rain events, which were experienced by the first author herself. In the case of Bahteem station, important drops were due to the irrigation events that took place on 20 November 2007, 6 December 2007, 17 January 2008, and 11 February 2008. All of them clearly show up in the data (Figure 7). Apart from these short alterations of α, there was a decreasing trend at Cairo University and 10th Ramadan over the whole period. A reason might be the rising humidity of the ground, as rain events started only in November after the hot and dry summer season, so the ground was extremely dry at the beginning of the measurements. Generally, different sun elevation angles can change measured urban α values (Christen and Vogt, 2004). However, the decreasing trend after the winter solstice does not support this explanation. Another reason might be a changing phenology in the green areas of the campus. Further oscillations, especially at Cairo University, could not be explained.
The longwave downwards radiation L↓ showed a similar pattern to the air temperature, and the differences are mainly influenced by the UHI effect. During the day, the differences were very small (<2.5 Wm−2 from 11:30 to 12:30); in the night, mean differences were a little higher (<9.2 Wm−2 from 20:00 to 06:00). Longwave upward radiation L↑ showed similar behaviour to the radiative temperature.
Figure 8 shows that Q* was highest at Bahteem station, where the surface albedo was lowest. Cairo University had only slightly higher values than 10th Ramadan station, which might be due to the albedo. During the night, Cairo University had the lowest Q*, which was due to the high surface temperatures and consequential high longwave emission. Bahteem and 10th Ramadan were also slightly negative during the night.
4.4. Soil heat flux
The soil heat flux Qs of Cairo University was not measured directly. Owing to the non-closure of the energy balance at Bahteem and 10th Ramadan (vide infra), it would be error prone if only the residual was used, and therefore it is not shown here. Qs of 10th Ramadan took a considerable amount of Q* during the day. From 09:00 to 15:00, it took about 37% of Q* on average. The maximum of the ensemble average was 158 Wm−2 at 10:05. For Bahteem, this relation was smaller: during the same time span, Qs took about 13% of Q*. The maximum of the ensemble average was 69 Wm−2 at 11:11.
The storage term Qs(0–z) accounted for a considerable part of Qs. Its average ensemble values peaked at 68 Wm−2 in Bahteem at 09:55 and at 117 Wm−2 in 10th Ramadan at 10:05, which is 122 and 74% of Qs, respectively. Qs(0–z) exceeded the original heat flux considerably from morning until early afternoon. Similar magnitudes of the storage term were measured by Ochsner et al. (2007). Omitting the storage term would lead to an extreme underestimation of Qs at the two stations and further decrease the energy balance closure. In the afternoon, the storage term quickly became negative with a daily minimum and remained negative until the morning. The minimum ensemble-average value of the storage term was − 41 Wm−2 at 17:08 in Bahteem and − 70 Wm−2 at 16:29 in 10th Ramadan. Qs(0–z) showed a strong statistical spread, due to the high fluctuations of the surface radiative temperature.
Figure 8 shows the ensemble-average Q* at the three stations and the ensemble-average Qs at Bahteem and 10th Ramadan for the period 20 November 2007 to 20 February 2008.
4.5. Turbulent fluxes
During the day, unstable conditions (expressed by z/L < 0) were the normal case at all three stations, while z is the measurement height and L is the Monin-Obukhov length. On average, daytime z/L was significantly lower at Cairo University than in Bahteem or 10th Ramadan. During the night, stable conditions (z/L > 0) prevailed at all three stations. The turbulent fluxes showed a great variability during the whole campaign. But a common pattern reigns during the whole time: At Cairo University and 10th Ramadan, the available energy was mainly going into QH, whereas in Bahteem most energy was fed into QLE. At 10th Ramadan station, QLE was generally very low. Still, on some occasions a strong QLE was recorded. These values were very likely connected with rain events. Unfortunately, no rain data were available, but especially in the second half of the campaign, it was likely that short duration and small-scale rain events occurred, as experienced by the first author. Also at Cairo University, a low average QLE was found, despite the green areas neighbouring the campus. During the night, the ensemble-average QH at Cairo University and 10th Ramadan was slightly negative, while at Bahteem it became clearly negative down to − 32 Wm−2 shortly after sunset. This strong negative QH at Bahteem did not show up every night, but often did so in southwest wind situations with high wind speed. The magnitude was strongly dependent on the latter: the higher the wind speed, the more negative QH. However, even during nights with very low wind speeds, H remained mostly slightly negative. Owing to the high percentage of missing values in Bahteem during the night, these findings may not represent the whole duration of the campaign. Nevertheless, Spronken-Smith (2002) found a similar negative QH after sunset in winter. The nocturnal negative QH at Cairo University is not in agreement with the findings of Oke et al. (1999) and Christen and Vogt (2004), who found a positive nocturnal QH in Mexico City, and in Basel, Switzerland. Although heat release is strong during the night at the urban site, it was not enough to maintain a positive QH. Figure 9 shows ensemble-average QH and QLE at the three stations.
QLE in cities is known to be driven by the fraction of vegetation of the surface (Christen and Vogt, 2004; Moriwaki and Kanda, 2004). As all stations show a considerable heterogeneity in the closer environment, it is expected that fluxes show certain directionality. To control this hypothesis, the surroundings of each station were divided into two sectors. The limiting factor was whether there were vegetated areas or not.
The data of each station were then divided into two groups according to the wind direction. The limiting angles of attack for the vegetated sector of each station were as follows (clockwise rotation):
Cairo University: 40°–260°
10th Ramadan: 30°–190°
Mean daytime (06:00–16:30) fluxes were calculated over the whole period. The highest directionality showed QLE at Bahteem station. Average daytime QLE was found to be 100.5 Wm−2 in the vegetated sector, whereas the non-vegetated sector showed only an average flux of 50.0 Wm−2. The high fluxes came predominantly from the northeastern sector, where winds crossed agricultural fields. Southwesterly winds blew over the suburbs of Bahteem, the non-vegetated sector; consequentially QLE was low (Figure 10). QH did show only a small directional difference. At Cairo University, a small influence of the fields to the south or of the gardens to the east could be detected in QLE, with a difference of 11.9 Wm−2. But this vegetated sector produced not only a higher QLE but also a higher QH (difference of 16.8 Wm−2). At 10th Ramadan, the tower was located southwest of a small wooden garden belonging to a compound within a distance of 350 m. In the south, a roundabout with irrigated beets was found. Surprisingly, no influence of this planting was found at all in the data. The vegetated sector even showed a lower ensemble-average QLE. However, the vegetated areas at this location were small and species were mostly xerophilous (Table IV). QLE in relation to the wind direction is shown in Figure 10.
Table IV. Mean QH and QLE according to the vegetated and non-vegetated sectors
The Bowen ratio β = QH/QLE was positive for all three stations during the day. The β value started to rise in the morning and reached its peak in the afternoon at Cairo University and 10th Ramadan. The maximum ensemble-average β peaked at β = 11.0 at 15:00 in 10th Ramadan, while it reached β = 4.4 at 14:30 in Cairo University. In contrary, in Bahteem a weak peak occurred already in the morning (maximum ensemble-average β = 0.8) and β decreased during the day continuously. This disparity could be attributed to the humidity in the air and soil. At Cairo University and 10th Ramadan, only a small latent heat flux was measured. Most available water was evaporated until the afternoon, forcing β to rise. In Bahteem, the evaporation occurred from the crop during the whole day, so β did not rise more. Daytime mean Bowen ratios were β = 1.9 for Cairo University, β = 0.1 for Bahteem, and β = 6.7 for 10th Ramadan station (06:00–16:30). The β value showed the same dependence on the wind direction as QLE. The β values found in this study are comparable to values in the literature. Spronken-Smith (2002) found summer daytime β for Christchurch from 1.29 to 2.11. Moriwaki and Kanda (2004) found daytime β of a residential area in Tokio in the range of 1.29–5.0, depending on the season. Christen and Vogt (2004) found daytime urban β close to β = 2, and Grimmond and Oke (1995) reported daytime values of β = 0.78–1.83 in four US cities, with a clear gradient of the water availability in the city. Oke et al. (1999) found the Bowen ratios of Mexico city to be between β = 4 and 12, which is more similar to our desert conditions.
The Bowen ratios at Cairo University and 10th Ramadan showed a considerable statistical spread. The daytime mean standard deviations of the ensemble averages were 14.7 for Cairo University and 23.1 for 10th Ramadan. In Bahteem, this value was 4.9. Grimmond et al. (1993) found daytime β to be more constant at rural sites than at urban sites. This observation can be confirmed only for Bahteem and Cairo University. The 10th Ramadan, also a rural site (albeit desert), showed an even higher diversity than Cairo University.
4.6. Energy balance closure
At Cairo University, the soil heat flux was not measured, so no closure could be determined. At Bahteem and 10th Ramadan, the energy balance was not closed. While the residual was slightly negative during the night, it reached considerable dimensions during the day, especially in Bahteem, where the maximum of the ensemble-average residuum was 148 Wm−2, which is 45% of Q* at that time. At 10th Ramadan, the maximum ensemble-average residuum was 57 Wm−2, which is 20% of Q*. Figure 11 shows the ensemble-average residual at Bahteem and 10th Ramadan.
No final explanation for these high residuals could be found. According to Foken (2008), measuring errors of the terms of the energy balance, including storage terms, do not explain such big daytime residuals if measurements are done carefully. As a result of the EBEX campaign, Oncley et al. (2007) found advection to be a major reason for afternoon residuals, which are assumed to be zero in Equation (1). Foken (2008) goes further and assumes that the closure problem is generally a scale problem, as the eddy covariance measurements only include the turbulent exchange of smaller eddies. But larger eddies seem to have a major influence on exchange processes in a heterogeneous landscape. So a closure is only possible on a landscape scale. However, fluxes in a totally uniform landscape, like a desert or a uniform bush land, should not be influenced by larger eddies. This hypothesis is partly supported by the data of 10th Ramadan, where the residual is lower than that of Bahteem. But as Figure 2 shows, 10th Ramadan is also homogeneous, but in the microscale. The outer surroundings include several buildings, factories, some trees, and street lamps. Further, the residual shows directionality. When northeasterly winds prevail in Bahteem, the residual is considerably smaller than in southwestern wind situations. The daytime (06:00–16:30) mean ensemble-average residual for the vegetated sector (Section 4.5.) of Bahteem is 40.4 Wm−2; whereas for the non-vegetated sector, it is double: 91.3 Wm−2. This can be explained by the higher heterogeneity farther into the southwestern sector, where the surface is mainly urban. The 10th Ramadan station also shows a slight directionality, but here the difference between the two daytime means is only 10.5 Wm−2, with the non-vegetated sector (open desert) having the lower residual.
In the framework of CAPAC, a micrometeorological field campaign was conducted in Greater Cairo from November 2007 to February 2008. All key variables of the energy balance were measured using standard eddy covariance instrumentation in three typical microclimates: an urban, a suburban-agricultural, and a suburban-desert station. The terms of the energy balance will be compared to the output of a remote sensing study. Hence, following the results of this study will be related to some aspects of remote sensing.
Cairo features a nocturnal heat island with higher air and radiative surface temperatures in the city than in the suburban stations. In south to southwest wind situations when wind speed was high, the nocturnal differences disappeared; the UHI seemed to be shifted to north. The results also highlight the fact that findings from radiative temperatures, as used in remote sensing studies, differ significantly from findings from studies using air temperature. The urban setting better stored heat than the suburban-desert soil. This was expressed by the lower magnitude of the diurnal cycle of radiative temperature at the urban station compared with the suburban-desert station.
Net radiation of the suburban-agricultural station was highest, while the suburban-desert station had the lowest values. The main factor determining these differences was the surface albedo. The latter variable was not constant over time, but changed due to several reasons. Therefore, it is suggested that surface albedo should be updated regularly when doing remote sensing studies of the area. Daytime irradiation showed a gradient away from the city, most probably due to air pollution. It follows that the atmospheric properties change over space significantly. So, a single radiosonde ascent is not sufficient to accurately describe the composition of the atmosphere in such an environment. This fact must be considered when modelling, for example, solar irradiation in the spatial domain for remote sensing studies. Longwave incident radiation did show nocturnal differences due to the urban heat island. However, the magnitude of these differences was fairly small (<9.2 Wm−2) which falls into the measurement accuracy. All terms of the net radiation could be determined with sufficient accuracy and can be directly used for the comparison study with remote sensing or modelled data, whereas scaling issues of the respective pixels must be accounted for. The findings for the soil heat flux were similar. The soil heat flux was estimated using two terms: the soil heat flux at depth z and the heat storage above z. The storage term proved to be a major contributor and omitting it would lead to a considerable underestimation of the soil heat flux. However, the storage term also showed extreme short-term fluctuations originating from rapid radiative surface temperature changes. Thus, the radiative temperatures of a satellite image might depict a condition, which is not necessarily representative of a longer period (e.g. 30 min).
The analysis of the turbulent heat fluxes showed that at the urban and the suburban-desert station the main part of the available energy was going into the sensible heat flux, while at the suburban-agricultural station the latent heat flux was dominating. Several constraints were found in using the in situ measured turbulent flux data for the remote sensing analysis. The lack of a correct determination of the turbulent heat fluxes with a sufficient closure limits the applicability of these data to comparison studies with remote sensing data. The influence of large-scale eddies is as yet unsolved and puts a major restriction on the use of these data. As stated earlier, the purpose of these measurements was to measure ground-truth data for a remote sensing-based energy balance study. The non-closure of the energy balance and a very probable underestimation of the turbulent fluxes limit the direct usage of the in situ flux data for input in models later applied to remote sensing data. A simple redistribution of the residual to the sensible and latent heat flux according to the Bowen ratio (Twine et al., 2000) is a first approach, but remains problematic, as the ratio might be different for large eddies (Foken, 2008).
Important for the comparison with remote sensing data is the directionality found in the flux data. As turbulent fluxes were strongly influenced by their source area, especially in Bahteem, it is straightforward to define this source area in the remote sensing image for further flux comparison (Li et al., 2008). This of course assumes knowledge of the wind speed and direction. So, wind speed and direction are indispensable variables in heterogeneous surfaces.
The high daytime temporal variation of the partition of the turbulent fluxes does not allow a methodology to be applied that uses a constant Bowen ratio or equivalent. This is another major restriction, as the Bowen ratio would allow only one turbulent heat flux to be determined.
Finally, fluxes measured with the eddy covariance technique are always time averages, in our case 30-min averages, while remote sensing images are observations done in a very short time frame. The in situ measured fluxes are affected by changes in surface variables like surface temperature during the averaging period. However, the use of remote sensing images will not incorporate these changes.
A sound methodology for comparing fluxes retrieved from satellite images with in situ measured fluxes incorporating all these constraints is outstanding and the subject of ongoing research.
This work was supported by the Swiss National Science Foundation (grant number 200021-1094).