2.3. Experiment Description

Temperature (both dry and wet bulb) and humidity measuring instruments were fixed at all the 27 micrometeorological stations as shown in Figure 1. Air temperature and relative humidity level data was measured and recorded by the wireless data logger units [T&D Corpora-

tion (Japan) Wireless Thermo Recorder Models: RTR-52 and RTR-72]. The instruments comprised of thermister sensor for air temperature, polymer humidity sensor for relative humidity and were protected by naturally ventilated radiation shields.

The instruments had been calibrated prior to the experimental phase by standardizing their readings with that of a standard psychrometer (Assmaan Psychrometer). The calibration process was carried out for a continuous stretch of 24 hours. Temperature correction coefficients were obtained by regression analysis of the measurements taken by the standard psychrometer and all the temperature data loggers used during the campaign and subsequently the measurements were corrected for the field campaign The instruments were installed at a height of about 1.5 m from the ground level facing the South direction such that the immediate surroundings at least upto 10 m from the instrument were open and not obstructed by any tree or building. Measurements were taken at interval of 1 minute from 25 May 2008 (morning)-28 May, 2008 (night extending into 29^th May) and averaged on hourly basis for analysis. The micrometeorological stations at these 27 sites provided the temperatures to obtain pockets of UHI in the city. In addition, 3 weather stations (Davis Wireless Vantage Pro2 Plus with Fan Aspirated Radiation Shield) installed at 15 m height on roof tops of the buildings gave information about various other meteorological parameters such as wind speed and direction, dry bulb temperature, atmospheric pressure, and solar radiation. Such experimental ground based measurements may not be sustained over a long time without infrastructural and security support. Hence the results of the in-situ measurements of the present study have also been compared with satellite data.

3. Weather Conditions during the Field Campaign

The experiment was carried out in the month of May 2008 in Delhi which is a typical summer month. As per the Climate-Normals of Delhi [25] the average maximum temperature during the month of May is about 39.6˚C while the average minimum temperature is about 25.9˚C. During the present study, the minimum temperatures ranged from about 20˚C to 22˚C while the maximum temperatures ranged from 33˚C to 37˚C. The major wind directions were observed to be North-West, West and South-West. The wind speeds during the experimental days ranged up to 3.6 m/s. The maximum solar radiation observed was 936 W·m^–2.

4. Urban Heat Island Intensity (UHI) across Delhi

With coming up of many satellite towns and city near the periphery of already established cities, definition of urban and rural areas have varied with different regions. As per classical definition of UHI [26], it is the difference in ambient air surface temperature between the city centre and a nearby rural area. However, Delhi is surrounded by peripheral satellite towns. Thus, it is difficult to find a rural site in within or near Delhi, free from influence of urban canopies. Nonetheless, Delhi with its variable LULC also offers many undeveloped/vegetated zones which can substitute for a rural area. Hence, the green areas in the city which were observed to have lowest temperatures have been considered as rural/non-urban temperature.

Stewart and Oke [22] noted that classifying field sites as conventionally “urban” or “rural” is especially difficult in regions where both cities and countrysides are densely populated and land uses intensely mixed. However, the core of the problem of urban heat island is that wherever high population and built up area exists, temperatures are higher and thus urban heat island effect is felt. Thus to estimate urban heat island intensity, a “rural” site may not be required for a region which has mixed land use. In the present study, mostly the green areas (parks and medium dense forests) in the city were observed to have the lowest temperatures. These green areas have thus been considered as baseline to asses UHI at other stations. Urban heat island intensity (UHI) is calculated as difference between highest temperature and lowest temperature observed among all micrometeorological stations across the study area at a given time.

These stations include all possible categories of LULC that may lead to both lowest (e.g., forest and other green areas, open areas and riverside) as well as highest temperatures (e.g., dense built up structures, commercial areas etc.).

To study the distribution of UHI over the entire monitoring network grid, UHI isopleths were constructed for all four days of the experimental phase for the specific (local) times of 0300, 0900, 1500 and 2100 hours. These hours were chosen based on approximate time of occurrence of minimum and maximum temperatures at all the stations and transitory conditions between morning-afternoon and evening-night respectively. UHI at each station for a given hour was calculated as difference of the temperature at that station and the lowest observed temperature at that hour among all stations of the micrometeorological network. The UHI values of all stations were then used to generate spatial distribution of UHI contours. Thus, each station (as marked by station number in the figure) is represented by its individual UHI value and UHI isopleths have been constructed by spatially interpolating the individual UHI values of 27 stations using Kriging technique across the study area of 32 × 32 km. Table 2 summarizes the maximum heat island intensity

for all days along with corresponding meteorological parameters. Overall, UHI was observed to range from 4.1˚C to 5.6˚C at 0300 hours, 4.6˚C to 6.4˚C at 0900 hours, 3.8˚C to 7.6˚C at 1500 hours, and 2.8˚C to 8.3˚C at 2100 hours. Thus, daytime hours 1500, 0900 also showed significant UHI along with night hours (0300 and 2100 hours). The trends of the current study are in conformity with some studies conducted in China [27] and Israel [28] where significant UHI is observed during day hours and evening-night transition. Further, significant daytime UHI has been observed in some other studies such as in Hong Kong [29,30], London [31] and Salamanca, Spain [32]. The major UHI zones consist of commercial centers and densely populated residential areas such as Sitaram Bazar, Connaught Place (CP), Bhikaji Cama and Noida [33].

Amongst all the days of the field campaign, May 28, 2008 showed the maximum insolation, ambient air temperature and UHI and hence the UHI isopleths for this particular day is shown in Figures 2(a)-(d) at 3:00 am, 3:00 pm, 9:00 am and 9:00 pm respectively and discussed as a representative case.

Lowest temperatures were always observed in green areas of Sanjay Van and Hauz Khas District Park. The other consistently cool region is around Buddha Jayanti Park. Prominent UHI pockets were Sitaram Bazar (station No. 30), Connaught Place (station No. 14) and Bhikaji Cama Place (station No. 7). The dense urban canopy of Sitaram Bazar and surrounding localities with major commercial complexes, densely populated residences and some of the busiest traffic intersections is the major cause behind building up of this heat island. The lower

sky view factor in dense built up areas emitting less longwave radiation and trapping of heat at late night hours may result into higher temperatures in these areas. Connaught Place is a major commercial complex with shops as well as office premises. Bhikaji Cama is also a commercial complex with a major traffic intersection in the vicinity.

5. Association of Urban Heat Islands with Meteorological Parameters

Urban heat island intensity is closely related to the prevailing meteorological conditions. Table 2 shows range of wind speed and direction for all three weather stations, PG-stability class [34,35], urban heat island intensity and hotspots for each day of the experimental duration. Winds of speed up to 3.8 m·s^–1 were prevalent during the experimental phase, but the night hours were marked by dominantly calm conditions (less than 0.3 m·s^–1). As seen in Table 2, in hours like 3:00 am and 9.00 pm, highest magnitudes of UHI are observed when calm conditions prevailed. In addition the heat island zones were observed in the zone encompassing the downwind direction. For example, the major UHI spots lie in the zone of downwind direction during hours such as 9:00 pm on 25^th May, 9:00 am on 26^th May and 3:00 pm on 27^th May.

Incidentally, the campaign period also covered an episode of rainfall thus facilitating a comparison between rainfall period and normal summer condition. The morning hours of 26^th May experienced rainfall for duration of roughly two hours from 9:00 am to 11:00 am. This time period was thus marked by zero incident direct solar radiation and on an average the wind speeds ranged up to 5 m/s. Table 2 shows UHI magnitudes for 8:00 am, 10:00 am and 12:00 pm respectively. These hours correspond to “before rainfall”, “during rainfall” and “after rainfall” hours respectively. The maximum heat island intensity decreases from 8:00 am (3.7˚C) to 10:00 am (2.2˚C) and again increases at 12:00 pm (2.5˚C) and later at 3:00 pm (3.8˚C). Thus contrast in UHI between rainfall period and rest of the period is evident where the UHI is 2.2˚C during rainfall against overall maximum of 8.3˚C.

In the present study, significant magnitudes of UHI are observed at unstable daytime hours also. Buildings decrease sky view factor and thus heat loss by building material due to radiation to cooler sky is contained. Though this phenomenon is prominent during night, it seems that for the present study the built up infrastructure plays greater role in trapping of heat and dominates over the convections in the atmosphere. Hence even though unstable and convective atmosphere during daytime is expected to facilitate temperature uniformity, the canopy structure of intense human activity with localized anthropogenic heat emissions and varying land use land cover in urban areas dominates the convective mixing thereby resulting in high magnitudes of many pockets of urban heat island intensity even during day time. Commercial areas like CP (site No. 14) are influenced by heat emissions from anthropogenic activities especially in daytime. Thus even in unstable conditions in maximum temperature period, higher UHI magnitudes have been observed. This could be a reason for an unclear relation between stability and UHI magnitude.

Absorption of solar radiation is the major factor of heat island formation in dense infrastructure area. Urban absorption of solar radiation is a function of many factors such as direct solar radiation, diffuse sky radiation, diffuse radiation reflected by buildings and streets, and the shortwave energy absorbed by urban structure [36]. Although all these variables have not been analysed in the present study, the daily maximum UHI has been found to increase with daily maximum solar radiation with the exception of 26 May which witnessed rainfall. For example, the maximum solar insolation on 25 and 26 May are 827 and 836 W·m^–2 with maximum UHI of 6.1˚C and 5.6˚C respectively against maximum solar insolation of 851 and 878 W·m^–2 on 27 and 28 May with maximum UHI of 7.8˚C and 8.3˚C respectively.

6. Comparison with Remote Sensing Data

The MODIS Terra spacecraft passes over the region of Delhi during 1000 hours and 1130 hours approximately during daytime and similarly from 2200 hours to 2330 hours during nighttime (by local time). Thus, the average data of this time period in the present field campaign has been used for comparison with the satellite data.

Figures 3(a) and (c) display nighttime and daytime UHI contours derived through land surface temperature obtained from MODIS-Terra data of the Monsoon Asia Integrated Regional Study (MAIRS) [20]. UHI contours based on in-situ measurements of the present campaign are display in Figures 3(b) and (d). It can be seen that many UHI hotspots of the present measurements have been captured well by satellite derived temperature fields especially during nighttime. The top UHI hotspots that are common in both MODIS contours and in-situ contours during nighttime are C.P. (14), Sitaram Bazar (30), Janakpuri (12), Bhikaji Cama (7) and Dwarka (6). During daytime however, only top two hotspots are common which are Sitaram Bazar and C.P. There is a satisfactory comparison of in-situ and MODIS observations in the identification of hotspots especially during night time. Although present study measurements are those of air temperature (~1.5 m above ground level), LU/LC have been shown to have relationship with both surface and air temperature [37]. Comparison of absolute temperatures and UHI at actual experimental locations is given in terms of bar diagram in Figure 4.

Night temperatures are correlated well with MODIS data in comparison to daytime temperatures. The magnitude of difference between absolute temperatures of MODIS and DELHI-I datasets ranges from 0.01˚C to 3.57˚C during nightime and from 1.8˚C to 7.9˚C during daytime. The difference in UHIs in both datasets ranges from 0.06˚C to 2.74˚C during night-time and from 0.02˚C to 3.34˚C during daytime. The differences in the above two observations could be attributed to different heat capacities of land and air. Thus on relative basis, the UHI and temperatures at the measurement locations as estimated from both satellite and current field campaign are comparable for night-time. Voogt and Oke, [38] and Mirzaei and Haghighat [39] explain that temperature measured by satellite sensors only relates to the spatial patterns of upward thermal radiance received by the remote sensor. However, the UHI phenomenon is majorly impacted by turbulence and velocity activities which are quite dominant at a larger scale in daytime in comparison to nighttime. In the present study also, the comparison of in-situ measurements with satellite data during daytime is poorer than night-time possibly for these reasons. Though,

quantitative comparison with satellite data is beyond the scope of this study, this qualitative comparison has validated the presence of temperature hotspots in the city.

7. Comparison with Other Cities

Table 3 lists some reported UHIs in major cities of the world based on field observations. Delhi, like Paris and London, is an inland city located at the bank of a river system. Although its population is lower than Tokyo and Beijing, the population density is higher than both these cities and similar to New York. With the exception of Beijing, all these cities belong to sub-tropical/temperate climate zone. However, climatologically Delhi is warmer than all these cities being quite close to tropical region.

New York, Paris London and Tokyo don’t have a significant precipitation in a particular season. On the other hand, Delhi, like Beijing, has a monsoon influenced climate and comparatively dry winters. Thus Delhi shares some similarities as well as dissimilarities with all these megacities. The heat island intensities stated against these cities in the Table 3 are from studies which were conducted using ground based measurements for a limited period. These studies were conducted in different time periods. The mean UHI (4.7˚C) observed in the present study is higher than those observed in New York (3˚C reported in year 2002) and Paris (2.6˚C reported in year 2006). The maximum UHI (8.3˚C) observed in present measurements falls in the range of those reported in London (8˚C, year 2003), Beijing (7.9˚C, year 2002) and Tokyo (8.1˚C, year 1992). However, it is important to note that in terms of built up infrastructure; all these cities are more urbanized than Delhi which is still developing.

Roth [52] analysed UHI and population of various continents and cities across the world. Population size and associated anthropogenic activities are expected to influence UHI. Based on observations, Oke [53] analysed relationship between UHI_max and populations of various North American and European cities and deduced that under clear and calm conditions, urban heat island magnitude is proportional to the city population. Accordingly, the following relationships were proposed 1) for North American cities and 2) European cities:

ΔT_u–r(max) = 2.96logP – 6.41 For North America (1)

ΔT_u–r(max) = 2.01logP – 4.06 For Europe    (2)

Delhi, with a population of 13.8 million [14] yields ΔT_u–r(max) as 14.7˚C and 10.2˚C following Equations (1) and (2) respectively. Thus, the observed ΔT_u–r(max) for Delhi (8.3˚C) is significantly lower than the results based on North American and European cities. The constants and slopes of Equations 1 and 2 are functions of artificial energy flux densities, heat capacity of the urban fabric, and evapo-transpiration in cities and therefore may have wide variations for different climates and demography. Further, Roth [54] reviewed several urban heat island studies in (sub) tropical regions and observed that in these UHI-population relationships, population is used as a convenient and widely available surrogate for urban geometry which has a more important control on urban heat island intensity. Delhi is still a developing megacity with mixed LULC having many interspersed green areas. In addition it is likely to have lower anthropogenic heat fluxes than the developed economies. Hence, the equations of Oke [53] yielded lower results than expected for Delhi.

Although, urban heat island intensity is expected to be related to population, Oke [53] also says that after a certain level, it is impossible to add more urban infrastructure to the city. For example, beyond a certain level of urbanization creation of new structures often results in destruction of old ones rather than the takeover of already scarce parkland. However for cities like Delhi, which are in phase of increasing urbanization, there is still scope of addition of more urban fabric. Built-up area of Delhi has witnessed an overall increment from 540.7 km² to 791.96 km² or 16.86% of the total city area (1,490 km²) during 1997 to 2008 which mainly came from waste land, scrub-land, sandy areas and water bodies [24]. Thus, with rising population and consequently increasing built up areas and associated anthropogenic activities, it is imperative to implement heat island mitigation strategies in the city in present time so that further intensification of heat island effect in the city can be prevented.

8. Conclusions

The main results and conclusions of the study are as follows.

•     The urban heat island intensity in Delhi is found to be significantly high during summer days for both afternoon (3.8˚C to 7.6˚C at 3 PM) and night hours (2.8˚C to 8.3˚C at 9 PM) in comparison to early morning hours at the time of occurrence of minimum temperature (4.1˚C to 5.6˚C).

•     The 3 top ranking UHI locations amongst the entire measurement network in the city are commercial areas namely CP, Sitaram Bazar and Bhikaji Cama Place. Other pockets with reasonably high UHI are residential or mixed use namely Noida, Dwarka, Janakpuri, Kaushambi, Adarsh Nagar etc.

•     The impact of meteorological parameters such as wind speed and direction as well as atmospheric stability was observed on the intensity and location of the urban heat island hotspots.

•     Comparison of in-situ measurements with satellite data showed that major hotspots in the city are well captured during night-time and can be helpful in assessing the major hotspots for other sites and times when in-situ measurements are inadequately available. The comparison of in-situ measurements with satellite data during daytime is poorer than night-time.

•     Comparison with maximum and average UHI of other cities of the world revealed that UHI in Delhi is comparable to other major cities of the world such as London, Tokyo and Beijing.

•     A relationship of maximum UHI and population showed that UHI in Delhi are lower than expected from the empirical relationships obtained for NorthAmerican and European cities with lower population than Delhi. Lower anthropogenic heat fluxes of Delhi due to a developing economy coupled with the climatic and demographic differences can be attributed for lower UHI in Delhi.

•     However, rising population and change in LULC and associated anthropogenic activities call for strategic mitigation measures in the city to prevent further strengthening of heat island effect.

9. Acknowledgements

The present study is a part of a research project “Implementation and Validation of Numerical Models for Heat Island Studies in Mega-city Delhi” conducted jointly by Indian Institute of Technology (Delhi and Roorkee) and Meisei University, Tokyo. We thank The Ministry of Education, Culture, Sports, Science and Technology (MEXT), Japan for the partial financial support. We express our gratitude to Head, Centre for Atmospheric Sciences, Indian Institute of Technology, Delhi for extending all possible support to this activity. We thank the participants of the field campaign from IIT Delhi and Roorkee. We would also like to thank the anonymous reviewers for their constructive comments which helped in improving the overall quality of the manuscript.

REFERENCES