A Study of Canopy Urban Heat Island of Baghdad, Iraq

Based on different two sets of data such as Landsat satellite imageries and daily air temperatures at 2 m high, canopy urban heat island (CUHI) phenomenon was investigated over Baghdad city for time span of three years: 2008, 2013 and 2019. The detection of changes in built-up areas was carried out to extract built-up index using ArcGIS 10.4.1 software with an overall accuracy of 94%. The results show that the built-up areas in the surface of the city have been increased from 621.7 km2 in 2008 to 727.3 km2 in 2019. This increase with high population density produced adverse thermal effects on atmospheric environment represented by CUHI. This was verified by analyzing daily, season and annual means of air temperature at the same years above measured at two different places: 1) urban site located at city center (Mustansiriyah University) and 2) rural site at the countryside of the city (International Baghdad Airport). Using the difference in daily temperatures between the two sites, the intensity of CUHI is computed. The results showed that the maximum daily CUHI is found in summer month (July) in 2019 with value 3.4°C, while the annual means are ranged from 1.5 (in both 2008 and 2013) to 2.4°C (in 2019). Finally, CUHI intensity is related with population density and built-up index in nonlinear variation.


INTRODUCTION
The urbanization process associated with the surface changings in the landscape creates a modification in temperature for both land surface and air layer called as canopy layer.In this layer where people live lies from ground to below the tops of tree and roof levels, which is the lowest part of atmospheric boundary layer (ABL) (Oke, 1976).In this layer, air properties are affected by the individual structures (houses, trees, paved roads, lawns, bridges, etc.) of the urban system.This leads to a modification in the atmosphere, creating a new local climate (Roth et al., 1989;Landsberg, 1981), which is a local disturbance of the regional climate.Almost, the modified climate over urban area is thermally and dynamically warmer than their surroundings rural regions, giving rise to the canopy urban island.This phenomena have long been thermally studied (e.g., Mukhelif et al., 2016;Montavez et al., 2000) in terms of air temperature difference between the cities and their surrounding regions, which call the canopy urban heat island (CUHI) (Bhargava et al., 2017).The study of CUHI is of interest attracted the attention in increased energy consumption, adverse effects of human health, urban planners and urban investors, as well as air/water pollution aspects, humidity (Golroudbary et al., 2018;Zhang et al., 2017).
Many remote sensing techniques using satellite images could detect the rapid alterations in the landscape of land use and land cover through several studies, which confirmed the surface UHI (Achmad et al., 2019;Rousta et al., 2018).Naturally, the physical surface at urban environment consists of different built material such as asphalt, gravel, stones, pebbles, concrete, etc., which characterizes with high heat capacities to store the sun's energy.The conversion of natural green areas into residential and commercial areas significantly increases land surface temperature.These surfaces directly or indirectly influence the local thermal climate of the city and even economic perspectives (Wang et al., 2018).When the surface UHI occurs, upward heat convection will arise especially with weakness wind, and consequently CUHI will set up over urban space with higher temperature than their peripheral or rural areas.
In the present study, two different sources of data for a period of three census years: 2008, 2013 and 2019 were used.(1) Satellite sensing images for Landsat-7 Enhanced Thematic Mapper plus (ETM+) for 2008 and Landsat-8 Operational Land Imager (OLI) for 2013 and 2019.These data analyzed to detect the changes in land use in Baghdad Metropolitan and to map built-up index area.(2) Time series for daily minimum and maximum air temperature measured by two weather stations with different surface features.Using the differences in these meteorological data, the CUHI intensity and its temporal distribution were obtained during four seasons of above years.

METHODOLOGY
Modification in the urban surface and the morphology of the city leads to increasing the urban warmth whereas air passing over this surface gets warmer than the adjacent rural surrounding air temperature.This defines as CUHI which forms from the ground to below the roof and tree levels.Thus CUHI intensity is the difference between air temperature in urban site (Tu) and the ambient air temperature in rural site (Tr), as given The formation of CUHI phenomenon depends mainly on weather conditions (i.e.cloudless skies, and absent wind), feature of urban area, length of night and day time, and weakly the anthropogenic activities (i.e. heat, pollution, etc.).In order to assess the randomly continuous development and expansion of cities, towns or villages with reducing vegetation, we adopted the built-up index (BUI) proposed by Zha, Gao and Ni (2003), to extract the changes in urban built up area from other land cover types using the following Equation: where NDBI is the Normalized Difference Built up Index and NDVI is the Normalized Difference Vegetation Index.This method has advantage in unique spectral response to build-up area and provide automatically the spatial pattern.The output binary image shows positive pixels for built-up and barren areas, and the rest of pixel for all other covers (Zha et al., 2003).NDBI could calculate through the Equation belowMIR-NIR NDBI = ----------------- (3) MIR + NIR where MIR is the Middle Infrared = Band 5 for Landat-7 EMT+ (2008) and = Band 6 for Landsat-8 OLI (2013 and 2019), while NIR is the Near Infrared = Band 4 in Landsat-7 ETM+ and = Band 5 for Landsat-8 OLI.Finally, NDVI retrieves using the following equation where RED is Red which reports Band 3 in Landsat-7 ETM+ and Band 4 in Landsat-8 OLI.For further explanation, a flowchart of the used algorithm is illustrated in Fig. 1.

1 Site Description
Baghdad, the capital governorate city, is located in central Iraq including surrounding metropolitan area.It extends along both banks of the Tigris River, see Fig. 1 and occupies an area of 894.3 km 2 and is in the heart of ancient Mesopotamia.The terrain of Baghdad and its sur rounding areas is a roughly flat and low-lying land adjacent to a stream alluvial plain (Al-Jiboori et al., 2020).Geographical coordinates are of latitude 33°12 -33°29 N, longitude 44°10 -44°30 E and 34 m above mean sea level.Baghdad had a population of a 6.974 million in 2019 and forms one of the populous urban agglomerations of the Middle East.
From build-up area of about of about 10 km 2 at the beginning of the 20th century, Baghdad has expanded into a bustling metropolis with suburbs spreading north and south along the river and east and west onto the surrounding plains (Britannica.com/place/Baghdad,2020).Baghdad has about 12 bridges spanning the river-joining the east and west of the city and has multiple small Mesopotamian Marshes.It has a prosperous and sophisticated city whose rich cultural life can be measured by its many museum universities and institutes and by myriad scholars and literature.The architecture of the city ranges from traditional two-or three-story brick houses to modern steel, glass and concrete structures.The Köppen climate classification for Baghdad features is BWH, semiarid, extremely hot, dry summer and damp winter (Roth, 2007).Baghdad has an average temperature 25°C and average rainfall per year is approximately 140 mm with no rain at summertime.The readers can refer to the reference (Al-Jiboori et al., 2020) for more details.

2. 1 Meteorological Data
The climate data used in this paper are daily mean air temperature (T) recorded by two automatic weather sta-tions in the city of Baghdad.The horizontal distance between them about 20 km ( Jaber and Al-Jiboori, 2018) as shown in Fig. 1 denoted by 1 and 2 and their photographs are displayed in Fig. 2a and 2b.Classifying the stations into different types is a key issue in UHI research.
First station is set up on the roof of building with 4 floors (13 m high over the ground) following Atmospheric Sciences Department, College of Science, Mustansiriyah University (see its photograph in Fig. 2a), which ideally represents the air of canopy layer of the city.This station has been found in the last three months of 2007 that records many weather elements at every hour.Although its height is relatively high compared with the other station, its records of outside air temperature do not appear serious differences with a personal amateur station.This agrees with a constant heat flux theory in the surface layer of ABL (Stull, 1989).The site of this station is ideally considered as urban site, because of the location of the university approximately at the center of Baghdad.Another station is located at International Baghdad Airport belonging to Iraqi Meteorological Organization and Seismology that is in accordance with World Meteorology Organization standards.This station has been installed in very open area without any obstacles around the device (see its photograph in Fig. 2b), which has the longest continuous climate data.Owing to its nature this station is considered as a rural one.
According to the discussion above, the condition to mea sure CUHI becomes available as urban versus rural.Data collected from both sites were analyzed for only    quality from both sources was carefully examined.In addition, these pair stations are located at flat land with no distinguish topography.

2. 2 Satellite Images
The detection of urban development for Baghdad city was performed using Landsat satellite images for the three different time-points, i.e. 2008, 2013 and 2019.These images were in UTM projection system (Zone 38) and acquired from the United States Geological Survey portal (www.earthexplorer.usgs.gov)for summer season to avoid the cloud cover (<15%) and interference.The study area is downloaded by sub setting the image with a square area of interest to include the changes in the surface features clearly.The images of Landsat-7 Enhanced and Thematic Mapper plus (ETM+) for the year 2008 and Landsat-8 Operation Land Imager (OLI) for the two years 2013 and 2019 were selected.Table 1 shows the details of Landsat datasets of satellite imageries collected in this work.Images of Landsat-8 OLI were already corrected for radiometric, while that of Landsat-7 ETM+ was corrected using squline method to remove lines.

Spatial Analysis of Built-up Index Extraction
Based on Equation (2) adapted in this paper, the resultant binary images were extracted with positive values for the built-up and barren pixel and negative values for non-urban-up (or other land cover types) these pixels are mapped automatically for each year of 2008, 2013 and 2019, as shown in Fig. 4. A comparative analysis of BUI extraction in these years is made to monitor and detect the land change in Baghdad city.Total number of pixels is 993743 (each pixel is 30 m 30 m in dimension) the changes in land use/land cover in terms of BUI are ultimately increased during the period 2008-2019.The occupied areas by built-up and nonbuilt-up with their  ratios respect to the total area are reported in Table 2.
From Table 2, built-up areas and their ratios are increased from 621.7 to 727.3 km 2 with adding 105.6 km 2 with ratio 11.9% during the period study.The great addition in built-up area (84.6 km 2 with 9.5%) occurred from 2008 to 2013.Inversely, there is reduction in nonbuiltup area from 272.6 to 167 km 2 from 2008 to 2019 with large ratio in 2013.In general, eastern side of Baghdad has more resident houses with high populous density than that for the opposite site.In 2019, the extension in eastern side of Baghdad as shown in Fig. 4c is approximately limited because of oldest area with high populous density.The evolution patterns in urban agglomerations for western side in 2019 can reveal the urban footprints since 2008.This can visualize the spatial expansion of this side by increasing the traffic infrastructure, paved roads, new settlements, vertical resident groups, gardens inside the houses and large supermarket malls.In both sides, so many small barren lands have been transformed to building such as settlement, small stadiums for footballs especially in northern part of Baghdad where so many vegetation have been altered to built-up since 2013.This occurred due to the continuous increasing in populous growth rates as will be showed later in Table 3 subsection 4. 3.
The accuracy of each BUI map was spatially assessed by selecting many point samples and comparing them with the ground truth using a field knowledge and Google Earth at main places, and road intersections.For the assessment of accuracy at the entire area, nearly 100 pixels of the objects were identified by visual interpreta-tion from the mapped results.These pixels were distributed as 70 pixels for the built-up land and 32 pixels for the other land cover (nonbuilt-up).The result of these verification points was 63 and 28 were correctly mapped as urban and nonurban area with overall accuracy of 94%.

2. 1 Monthly Variation
The impact of the increased changes in BUI during the period from 2008 to 2019 on the thermal climate of Baghdad, has been investigated through examining the seasonal behavior of CUHI (ΔT u-r ) intensity.This was performed by studying the daily variation of ΔT u-r at each year by choosing the months: January, April, July and October, which represent winter, spring, summer and autumn, respectively.Figs.4a-4d display these seasonal variations respectively, plotted together for the years 2008, 2013 and 2019 for comparing among them, and hence, to detecting the warming.The intensity of CUHI has roughly positive values across all seasons.This means that mean air temperatures at urban city of Baghdad are larger than those at rural site.The large difference occurred during summer month (Fig. 4c), especially at the year of 2019, while less difference occurs during winter month with slightly increasing in 2019.The similar behavior is also found in spring month which have less values of ΔT u-r in 2008.Fig. 4d shows relatively large value of ΔT u-r , especially in 2019 comparing to other years.

2. 2 Seasonal Variation
Daily variation for ΔT u-r across all seasonal months are found to be highly fluctuated such as in summer and autumn.This result is expected because the weather in Baghdad is characterized by instability along the year days.This belongs to that Baghdad and even Iraq is one of regions affected by extreme atmospheres being lied in mid-latitude zones.To reduce high fluctuation in daily variation for ΔT u-r and noting which year showing more warming.Monthly means were taken with calculating standard deviation (SD), which represents the spreading around the mean value.They are displayed in Fig. 5 as bars and vertical lines for SD.Looking at Fig. 5, one can see that all differences, ΔT u-r , are absolutely positive for all years.This means that mean air temperatures in Baghdad is usually larger than the rural site during any year.Across all the seasons the intensity of CUHI in 2019 is largest with a value of 3.6±0.9°Ccomparing to the two years 2008 and 2013 and also to other seasons.In autumn month, the 2019 year has ΔT u-r = 2°C and also SD is large (±1.2°C), as shown in Fig. 5.In winter month, there was not seriously temperature difference among the years.The reason is that so many synoptic pressure systems are influenced Baghdad, which causes the surface cooling and even the atmospheric and canopy urban heat island.This is different in summer month where these systems have no effect and never falling rain  ( Al-Jiboori et al., 2020).In addition, the length of day in this season is largest that almost reaches 14 hours.

3 Annual Variation of CUHI
In this subsection, it discusses the impact of expanded and increased area of built-up index over time for the years 2008, 2013 and 2019 on the intensity of CHUI as described in subsection 4.1.Before explaining this, it should notice that BUI areas strongly depended upon land uses produced by numerous number of individuals in any urban city.Therefore, statistics of the population and population density (in simple terms, defined as the midyear number of population living in a land area per square kilometer (Clark, 1951)) were added to achieve this analysis, whereas the relationship between these statistics and urban heat island intensity has been fixedly documented by several researchers (Golroudbary et al., 2018;Oke, 1973).Population density (PD) of Baghdad were computed from the database of population available at the central statistical organization, Planning of Ministry.
Using seasonal means of CUHI (i.e.ΔT u-r ) reported in the previous subsection, annual means of ΔT u-r with their SD were calculated at each year.The results for CUHI intensity and its dependency on both BUI areas and PD were investigated for the years of 2008, 2013 and 2019.All of the values of these results are reported in Table 3.The linear increasing in population and PD was also associated with the corresponding increase in the areas of BUI.This means that the continuous increasing in PD and BUI with the years shown in Table 3 explores positive extreme signs of urbanization warmth, especially in the eastern part of the city which describes with high PD (Alsamaw and Zboon, 2009).
Also in the recent decade, most ground housing units with more than 100 m 2 have been divided to small ones even small gardens inside the houses which have been also transformed settlement units.The city's population growth was not only causing exceeding the BUI areas but also rising the level of anthropogenic heat emissions.All of the increases produced warmth in the air canopy layer of Baghdad with annual mean of ΔT u-r = 1.5°C for both years 2008 and 2013, but more seriously in 2019 with 2.5°C.Now we attempt to study the physical behaviors among the three parameters PD, BUI and CUHI (or ΔT u-r ) with falling out the time (years).The values of these parameters were drawn together in 3D graph pre-sented in Fig.

CONCLUSION
This study incorporated an integrated analysis of both remotely sensed imagery and air temperature data for the three years (2008, 2013 and 2019) recorded by automatic weather stations at two sites of Baghdad city, first is urban site approximately located at the center (Mustansiriyah university) and another is International Baghdad Airport located in the countryside as rural site.Satellite images of Landsat-7 ETM+ and Landsat-8 OLI were used to extract built-up index as useful method for monitoring and detecting distinguish between built-up and nonbuilt-up areas land change in the city during these years.This index was computed from NDVI and NDBI and used to make spatial maps for built-up area in the city, which shows the rapid expansion in these areas ranging from 621.7 (in 2008) to 727.3 km 2 (in 2019).The difference between these values ( = 105.6 km 2 ) associated with increasing in population density produced the local warm climate in city center which expressed by canopy urban heat island along the days of any studied year.
The existing of CUHI was noticed through daily difference in air temperature between rural and urban sites, ΔT u-r .This indicator was investigated at four months which represent the middle of each season: January (winter), April (spring), July (summer) and October (autumn).The results demonstrated that the summer month has the largest CUHI with 3.6°C and less value at winter month.The seasonal and annual means of ΔT u-r were also taken, which illustrate that the summer season at each year has relatively exceeds from the remaining seasons with noticing the great extreme in 2019.Annual means of ΔT u-r has the same positive magnitudes (1.5± 0.5°C) in both years (2008 and 2019) and largest (2.4± 0.9°C) in 2019.As an overall result, the nonlinear relation of CUHI with a given increment in density population and Built-up index area was enhanced in 3D graph.

Fig. 1 .
Fig. 1.Flowchart of methodology comprised of three parts: Satellite images, population and meteorological data.

Fig. 3 .
Fig. 3. Photographs of two automatic weather stations at (a) Mustansiriyah university (as urban site) and (b) International Baghdad Airport (as rural site).

Fig. 6 .
Fig. 6.Histogram bars of seasonal CUHI (ΔT u-r ) with their standard deviation for the studied years.
7. The relations (ΔT u-r vs. PD), (ΔT u-r vs. BUI), and (PD vs. BUI) show nonlinear increases of ΔT u-r dependent on the human modifications in PD and BUI.Dashed-solid line represents the values resulting from PD and BUI (i.e.ΔT u-r is function of both PD and BUI: CUHI = ƒ (PD,BUI)).The SD values denoted by the red vertical lines shows the largest diffusion in CUHI values (±0.9°C) from the seasonal means for the last 2019 year.In concluding, the continuous population growth came from various parts of Iraq (i.e.internal immigration) and even the numerous numbers of new births has been a mixed effect on local thermal climate of canopy Baghdad environment.

Fig. 7 .
Fig. 7. Variation of ΔT u-r with PD and BUI area.The lines with dots , and are ΔT u-r vs. BUI, ΔT u-r vs. PD and PD vs. BUI, respectively.

Table 1 .
Properties of datasets for Landsat scenes and their date of acquisition.

Table 2 .
Built-up and nonbuilt-up areas extracted from built-up index with their percent ratios.

Table 3 .
Comparison the results among PD, BUI and CUHI intensity during the years of this study.