By Janet Nichol
Pui Hang To
The Hong Kong Polytechnic University
I. INTRODUCTION
A. Hong Kong’s urban heat island
Rising global temperatures bring into question the sustainability of tropical and sub-tropical cities, especially those such as Hong Kong, where dense and high-rise buildings are accompanied by an intense urban heat island (UHI) effect. At 21å¡N, Hong Kong lies at the boundary between the tropics and sub-tropics, experiencing hot humid summers and warm dry winters.
Air temperature typically reaches 37å¡C on the hottest summer days, cooling to 29å¡C in urban, and 24å¡C in rural areas at night. Kowloon, with a population of more than 2 million, is the largest continuous urban area and is separated from the rural New Territories by mountain ranges. Parts of Kowloon contain the highest urban population densities in the world, with more than 60 persons per km2 [1]. Reduced ventilation, high temperatures and the blocking of sea breezes by tall buildings on newly reclaimed land, the so-called Û÷wall effect,’ are contentious issues. Planners urgently require empirical and spatially comprehensive climatic data to support structure plans and policies but traditional methods of UHI analysis, the use of fixed stations and/or vehicle traverse have been unable to provide such data thus far.
In the light of the urgent need for information in Hong Kong discussed above, we examined a nighttime ASTER thermal satellite image of Hong Kong’s urban areas taken at 11:42 p.m. on Aug. 13, 2008, accompanied by Û÷in situ’ field measurements of surface and air temperatures. The aim was to assess the extent to which satellite-derived land surface temperatures (Ts) can represent the spatial pattern of air temperatures commonly prevailing across the city, over extended periods beyond the image time and which thus typify the character of the city’s urban heat island.
II. SATELLITE SENSORS FOR URBAN HEAT ISLAND ANALYSIS
Spatially complete and time-synchronous coverage of an urban area can be obtained from thermal wavelength satellite sensors which measure Surface Temperature (Ts), thus unlike fixed stations and vehicle traverse the actual maximum and minimum temperatures over a city region can be obtained [2] and intra-urban thermal patterns can be observed. However, because such images suffer from low temporal resolution, planners are reluctant to use the image data as representative of the UHI pattern of a city. For example, the ASTER sensor on board NASA’s TERRA satellite has approximately two potential acquisitions per month over Hong Kong, which may be during either day or night. It is unknown to what extent the UHI situation of a city depicted on a single thermal satellite image may be used to represent its typical, or commonly occurring heat island situation.
Figure 1. Air temperature (Ta) for the Hong Kong Observatory (urban) and Ta Kwu Ling (rural) climate stations for the 48 hours surrounding the image time.
A. Timing of image acquisition
The urban heat island is considered to be a night-time phenomenon, since it is best developed a few hours after sunset when the rural surface cools rapidly and heat is extracted from the adjacent air. (Figure 1). The urban atmosphere on the other hand continues to be warmed by the man-made urban surfaces which have higher heat capacity and thermal inertia than rural surfaces. In high-rise areas, these radiatively active surfaces are extensive, and potential energy loss by outgoing long wave radiation is blocked by the urban canyon geometry, with low sky view factors.
The relationship between surface and air temperatures increases closer to the surface [3], with lower wind speeds [4], and also at night when microscale advection is weaker [5]. On the night of the image in this study, the atmosphere was stable with an inversion height of 700m. Wind speeds were low, at less than 1ms, and relative humidity moderate, at 72%. By the image time 11:42 p.m., mean surface temperatures in urban areas had fallen to approximate air temperatures at 28-30 degrees, although the UHI was not fully developed until the time of minimum air temperature, around 4 a.m. This suggests that ASTER thermal images are not optimally timed for UHI analysis.
B. Image processing
The blackbody temperature image (TBBT) from ASTER, with a spatial resolution of 90m, was converted to an image representing Surface Temperature (Ts) at 10m resolution using the emissivity modulation method [6] (Equation 1). This method corrects the blackbody image for differences in emissivity (ε) between different surface materials by ratioing it in the equation with a land cover map at 10m resolution.
(1)
The image surface temperatures (Ts) were converted to air temperature (Ta) (Figures 3 and 4) by regressing 11 air temperature points measured in the field at the image time against the corresponding image Ts (Equation 2). The corresponding R2 between Ts and Ta is 0.82 (n = 11), suggesting that the field air temperature is suitable for converting the Ts to Ta
(2)
Accuracy was tested using 11 validation points from readings at automatic weather stations at the time of imaging. The high R2 value of 0.82 and Mean Actual Difference (MAD) of 1 suggest that the image-derived Ta values are suitable for studying patterns of air temperature distribution over the study area.
C. Hourly and daily analysis
The images are deemed more relevant if they can represent temperature patterns at other times of the 24-hour cycle and on days other than the image date. Therefore, to determine the hourly relevance, the air temperatures recorded at 11 automatic weather stations (AWS) several hours before and after the imaging time were regressed against the image-derived air temperature. Additionally, days in the same imaging season fulfilling all the following requirements were selected for comparison with the image values: (1) having the same wind direction of eight directions; (2) having wind speed differing by only 10% from that on the day of imaging; and (3) no rainfall recorded at any AWS.
Figure 2. Hourly time series correlation analysis between image-derived and AWS recorded air temperature on the nighttime image.
III. RESULTS
A. Hourly analysis
Figure 2 shows the correlation between the image-derived and AWS air temperatures surrounding the imaging time. This is high (R2 = 0.84) at the image time of 11:42 p.m.local time. The correlations remain relatively steady (R2 around 0.7) between 3.5 hours before, and 2 hours following the image time, but the results are significant at the 5% level for a 10-hour period, from 6 p.m. to 4 a.m. the next day. The spatial pattern of the UHI depicted on Figures 3 and 4 can therefore be considered representative of Hong Kong’s UHI, since air temperatures derived from the image and AWS are significantly correlated for this 10-hour period which includes the time of air temperature minimum (4 a.m.) when the UHI is most fully developed (Figure 1).
B. Daily analysis
Since summer corresponds to the rainy season in Hong Kong, only seven dates were able to satisfy the criteria of having no rainfall at any AWS station. For all of these, the R2 values were significant at the 5% level, ranging from 0.55 and 0.91. But two of the days other than the image date had a very high R2 value of 0.91. The moderate to very high correlations for nights other than the imaging date suggest that images obtained on one occasion can be used to retrieve
Figure 3. Summer nighttime air temperature image of Hong Kong derived from ASTER. The urbanized Kowloon Peninsula is at lower center, and shows the highest temperatures.
temperature patterns on other dates with similar atmospheric conditions. Of the seven dates which qualified as having similar weather conditions to the image date, one of the two lowest R2 values occurred on Aug. 20 (R2 = 0.56) and Sept. 13 (R2 = 0.24) (Table 1) and may be explained by the proximity of a typhoon on this day. Atmospheric conditions become complicated if a typhoon is approaching as it affects air pressure in adjacent offshore areas.
C. Daytime image
A daytime image obtained at 11:09 a.m. on the Aug. 20, 2009, and subjected to the same analysis as the nighttime image, was found to be less representative of other times, both on the same day and on different dates, than the nighttime image. At the image time, an R2 value of 0.79 was observed, and the values were significant for periods of 1.5 and 2.5 hours before and after the image time, respectively. The R2 values for the daily comparisons ranged from 0.24 to 0.79 and were considerably lower than the nighttime values.
IV. DISCUSSION
The data suggest that the nighttime image is fairly representative of air temperatures over Hong Kong for a considerable period before and after the image time, as well as on other days having similar weather conditions. A possible explanation for the period of longer validity at night is that convection and advection subside at night [5], producing a stable atmospheric environment [7], and sea breezes are minimal due to similar temperatures of land and sea (Figures 3 and 4). Additionally, the very high R2 obtained for nights other than the image date indicates that the nighttime image is better able to account for air temperature patterns observed on other dates than is the daytime image, again because the atmosphere is more stable at night.
Figure 4. Air temperature image of the Kowloon Peninsula with street network.
The long period of significance (6 p.m.-4 a.m.), a total of 10 hours, suggests that after sunset on this hot summer night, thermal patterns were quite stable over the urban area, probably due to the low wind speed (less than 1ms) and prevailing inversion, coupled with poor ventilation in the dense urban environment and blocking of sea breezes by the Û÷wall effectÛ buildings along the coast, a well known phenomenon in Hong Kong. This suggests that thermal satellite images can be used to represent the typical spatial distribution of the urban heat island in Hong Kong on hot summer nights, even though the time of image acquisition does not correspond to the time of maximum UHI development. Since the areas of maximum and minimum temperature can be located on the image, unlike other methods such as vehicle traverse and fixed stations, which are unable to locate the hottest and coolest areas, the magnitude and distribution of Hong Kong’s UHI can be evaluated.
ACKNOWLEDGEMENT
The authors would like to acknowledge Public Policy research grant 5006-PPR-09 from the Hong Kong government.
Author bios
Janet Nichol has a background in physical geography, with specialisms in remote sensing and ecology. She obtained her B.Sc. at London University, her M.A. at the University of Colorado, and her Ph.D. at the University of Aston in Birmingham, and has subsequently worked in the U.K., Nigeria, Singapore and the Republic of Ireland as a university lecturer. Since 2001, she has taught and undertaken research at the Department of Land Surveying and GeoInformatics of The Hong Kong Polytechnic University. Her research interests are in the application of remote sensing techniques to environmental assessment and monitoring, including the urban heat island, vegetation mapping, landslide hazard assessment and air quality monitoring.
Pui Hang To obtained his BSc(Hons) degree from The Hong Kong Polytechnic University, specializing in Geo-Information Technology. He is now an MPhil candidate at the same university. His research interests include thermal remote sensing, the application of Geo-IT to environmental management and sustainable urbanization.
REFERENCES
[1] E. Ng, ÛÏWind and heat environment on densely built urban areas in Hong KongÛ. Global Env. Res., vol. l3, pp.169-178, 2009. [2] G. L. Johnson, et al, ÛÏSimulation of surface urban heat islands under Û÷ideal’ conditions at night Part 1 : theory and tests against field dataÛ. Bound. Layer Meteor., vol. 56, pp.275-294, 1991. [3] J. E. Nichol, ÛÏHigh resolution surface temperature patterns related to urban morphology in a tropical city: a satellite-based studyÛ. J. App. Met., U.S.A, vol. 35, pp.135-146, 1996. [4] M. J. Stoll, and A.J, Brazel, ÛÏSurface-air temperature relationships in the urban environment of Phoenix, ArizonaÛ, Phys. Geogr., 13, pp.160-179, 1992. [5] J. A. Voogt, and T. R., Oke, ÛÏThermal remote sensing of urban climatesÛ, Remote Sens. Environ., vol. 86, pp.370-384, 2003. [6] J. E. Nichol, ÛÏAn emissivity modulation method for spatial enhancement of thermal satellite images in urban heat island analysisÛ. Phot. Eng. Rem. Sens., vol. 75, pp.547-556, 2009. [7] T. R., Oke, ÛÏBoundary Layer ClimatesÛ, 2nd ed., Great British: TJ International Ltd, 1978.