Home Articles A remote sensing-GIS evaluation of urban growth-land surface temperature relationships in Selangor,...

A remote sensing-GIS evaluation of urban growth-land surface temperature relationships in Selangor, Malaysia

Asmala Ahmad
Dept. of Science & Mathematics,
Centre for Academic Services (CAS),
Kolej Universiti Teknikal Kebangsaan Malaysia (KUTKM)
Locked Bag 1200, Ayer Keroh, 75450 Malacca, MALAYSIA
Phone: 606-233 3034 Fax: 606-233 3144
[email protected]

Noorazuan Md Hashim
School for Social, Development and Environmental Studies,
Faculty of Social Sciences and Humanities
Universiti Kebangsaan Malaysia (UKM), 43600 Bangi,
Selangor Darul Ehsan, MALAYSIA
Phone: 03-8926 6213
[email protected]

Abstract
The state of Selangor has experienced rapid development and high growth in urbanisation over the past two decades. Urban encroachment was significant especially in urbanised mukims such as Kajang, Cheras and Dengkil. This article investigates the potential application of remote sensing and GIS for detecting urban growth as well as determine the growth impact to land surface temperature (LST) and urban heat island (UHI) phenomenon. Urban heat island due to rapid growth in urbanisation can be determined by using the surface temperature information from thermal infrared band of Landsat image. However, an extra effort need to be done for LST retrieval from the original raw DN value of band 6 TIR. A subset of Landsat TM acquired on April 17th, 1988 and February 11th, 1999 (path 127/row58) that covers an urbanised mukims (Kajang-Cheras-Dengkil) was used in this study. Erdas Imagine 8.5 was the main software for image classification of urban growth 1988-1999, while GIS-Arcview packages were used to derive land surface temperature through GIS-Grid Calculator functions. An initial results showed an increase of urban cover within the period, from 69330ha in 1988 to 91420ha in 1999. Eventhough the urban cover increase within the period, the heat island phenomenon was less in the latter period. A further analysis showed that the antecedent precipitation prior to the scene acquisition for February 11th 1999 was very high comparatively and produced high value of soil moisture content. In overall, the integration of GIS and remote sensing has been found to be effective in analysing urban-land surface temperature relationships.

1. Introduction
The state of Selangor is experiencing rapid economic growth especially within the past two decades. In late 1990s, a large area of agricultural land has been converted into urban areas, hence changed the surface profile of the area. These surfaces absorb heat and increase the temperature comparatively to the surrounding area. The heat bubble i.e. urban heat islands not only reduced human comfortability, but it also could increase energy consumption in buildings. Latest research has also showed the effectiveness of urban heat island in changing the urban micro-climate e.g. changes of intensity and frequency of rainfall.

This article investigates the potential application of remote sensing and GIS for detecting urban growth as well as determine the growth impact to land surface temperature (LST) and urban heat island (UHI) phenomenon. Urban heat island due to rapid growth in urbanisation can be determined by using the surface temperature information from thermal infrared band of Landsat image. However, an extra works need to be done for LST retrieval from the original raw DN value of band 6 TIR.

Land-surface temperature (LST) can be defined as the thermal emission from the landscape “surface”, including the top of the canopy for vegetated surfaces as well as other surfaces (such as bare soils). LST is an important parameter in the field of atmospheric sciences as it combines the result of all surface-atmosphere interaction and energy fluxes between the ground and the atmosphere and is, therefore, a good indicator of the energy balance at the Earth’s surface (Wan and Snyder, 1996). LST controls the surface heat and water exchange with atmosphere. Estimation of LST from satellites infrared radiometers has been proven useful. Most studies have focused on the use of polar orbiting satellite systems because of their high spatial resolution (Sun et al., 2004).

2. Data
A subset of Landsat TM acquired on April 17, 1988 and February 11th, 1999 (path 127/row58) that covers an urbanised mukims (Kajang-Cheras-Dengkil) was used in this study. The TM data are scanned simultaneously in seven spectral bands. Band 6 scans thermal (heat) infrared radiation. For image classification band 4, 3 and 2 was used to analysed the urban changes within the period. Band 6 of Landsat TM was used to determine the LST in this project.


Figure 1. A subset of Landsat TM covers three main urban mukims, i.e. Dengkil, Kajang and Cheras

3. Methodology
The image processing of the Landsat TM and ETM normally involves a number of pre-processing procedures. Image pre-processing aimed to improve the quality of the classification input data and calibrate it to units of reflectance (Smith et al., 2001). The image pre-processing involved spatial sharpening or image enhancement of the Landsat data, with geometric and radiometric correction (Jensen, 1986). Several pre-processing steps were undertaken prior to image classification:

3.1 Radiometric and atmospheric calibration
Prior to image analysis, remotely sensed data has to be modified to improve the quality of interpretability. It also requires correction, radiometrically and geometrically. For image enhancement, the contrast adjustments (histogram equalisation method) of the subset images were done in ERDAS Imagine 8.5 (ERDAS, 1999). The radiometric and atmospheric corrections are essential as these two images (1988 and 1999) will be used in detecting the urban change.

Both images belong to different periods, seasons and the sun elevation angles. Correction of the atmospheric and radiometric errors is necessary as the study involves multi-sensors and multi-temporal images (Thamm et al., 2004; Madhavan et al., 2001). Radiometric normalisation is an important when two dates of different images are compared. This is due to the different in atmospheric condition between the image dates. Ibrahim et al. (2003) and Teillet (1986) have developed an algorithm of radiometric normalisation that undertakes several types of atmospheric correction.

Table 1: The total Julian days, solar angles and sun-earth distance for each scene
Scanner type & date scene Total julian days (days)Solar angle(degrees)Sun-earth distance (in astronomical units) TM5 (17/4/88)12169.811.0076 TM5 (11/2/99)6074.630.9909 Source: Markham and Barker (1986) and Chengquan et al. (2002)

The original sources of the imageries from MACRES were not atmospheric corrected. Thus, the atmospheric effect was corrected by using the real time of sun elevation angle and sun-earth distance for each scene by using the COST model. The spatial model, i.e. the COST model developed by Chavez (1996) has been downloaded from the public domain ERDAS Leica Geosystems official webpage and run into ERDAS Imagine 8.5.

The exo-atmospheric reflectance, at-satellite temperatures, total Julian days, sun elevation angle and sun-earth distance for each scene were derived from Markham and Barker (1986), Chengquan et al. (2002), the Measurement and Instrumental Data Center (MIDC) official website ) and NASA official webpage (https://lambda.gsfc.nasa.gov/toolbox/tb_converters_ov.cfm). In the model, a reduction in between-scene variability can be achieved through a normalization for solar irradiance by converting spectral radiance, as calculated above,

Where:
ρp= Unitless planetary reflectance

= Spectral radiance at the sensor’s aperture

d= Earth-Sun distance in astronomical units from nautical handbook

ESUNλ= Mean solar exoatmospheric irradiances from

θs= Solar zenith angle in degrees

3.2 LST Retrieval
LST retrieval was carried out through three phases.

  1. Conversion from Digital Number to Radiance
  2. All TM bands are quantized as 8 bit data thus, all information is stored in digital number (DN) with range between 0 to 255. The data was converted to radiance using a linear equation as shown below:

    CVR = G (CVDN )+ B (2)

    Where:

    CVR is the cell value as radiance
    CVDN is the cell value digital number
    G is the gain (0.005632156 for TM6 and 0.003705882 for ETM+6)
    B is the offset (0.1238 for TM6 and 0.3200 for ETM 6)

  3. Conversion from Radiance to Brightness Temperature
  4. By applying the inverse of the Planck function, thermal bands radiance values was converted to brightness temperature value.

    (3)

    Where:

    T is degrees Kelvin
    CVR is the cell value as radiance
    K1 is calibration constant 1 (607.76 for TM) and (666.09 for ETM+)
    K2 is calibration constant 2 (1260.56 for TM) and (1282.71 for ETM+)

  5. LST Retrieval
  6. LST was derived from TM6 using model developed by Sobrino,et al (2004) and Jackson, et al (2004) which use spectral surface emmissivity and NDVI values of the particular scenes.

    (4)

    Where,

    St = LST
    λ is wavelength of emitted radiance (λ = 11.5 μm),
    ρ = h

Where,

St = LST
λ is wavelength of emitted radiance (λ = 11.5 µm),
ρ = h×c/σ (1.438 ×10-2 m K),
σ = Boltzman constant (1.38 ×10-23 J/K),
h = Planck’s constant (6.626×10-34 J s), and
c = velocity of light (2.998 ×108 m/s)
Emissivity ε can be estimated through :

(5)

where εv and εs denote emissivity of vegetation and soil, while fv can be expressed as (Sobrino, et al (2004):

(6)

Where,

NDVI max = NDVI for complete vegetation cover,
NDVI min = NDVI for bare soil.
α = function of leaf orientation distribution with the canopy

After the atmospheric and radiometric calibartion were done, the imageries were transfered into grid system and exported in to GIS Arcview to gain LST retrieval using Grid Calculator.
3.3 Image classification
Land use/land cover maps published by the Department of Agriculture, Malaysia were used side by side with the latest (year 2004) Quickbird image of Southern Selangor to identify the basic land cover types in the study area. A preliminary analysis of the images based on an iteartive self-organising data analysis (ISODATA) unsupervised classification was performed to differentiate the spectral clusters corresponding to the basic land cover types.

3.4 Soil moisture analysis based on API
Another indicator was used to highlight the changes of urban heat island pattern through application of soil moisture information, i.e. Antecedent Precipitation Index (API). According to Ward (1978) and Chow et al. (1988), the basin’s antecedent condition is important in determining the moisture content. The calculation of API was made possible by using formula derived by Gregory & Walling (1973).

APId = (APId-1 ) K + P (7)

Where,
APId = API for d day
APId-1 = API for preceding day
K = API constant (K = < 1.0 and usually 0.85-0.98)
P = rainfall in preceding 24 hours
The value of K in this study has been determined as 0.9 (Gregory & Walling, 1973)

4. Result and Discussion

4.1 Landuse/Landcover Image


Figure 2. Image classification for 1988


Figure 3. Image classification for 1999. There was an increase in the urban area compared to 1988.

4.2 LST Image


Figure 3. 1988 temperature image – red region indicates area with temperature 240 C-260 C (146,236 pixels).


Figure 4. 1999 temperature image – red region indicates area with temperature 240 C-260 C (9274 pixels). There was a decrease in the number of hot pixels compared to 1988


Figure 5. Distribution of LST greater than 27.50 C (Average temperature for April 1988 – based on 4 stations) in the year 1988 within the urban area. 3557 pixels has LST greater than 27.50 C


Figure 6. Distribution of LST greater than 25.5 0 C (Average temperature for November 1999 – based on 4 stations) in the year 1999 within the urban area. There are only 2656 pixels greater than 25.5 0.


Figure 7. Temperature distribution based on LST analysis for 1988


Figure 8. Temperature distribution based on LST analysis for 1999

4.3 Soil Moisture Difference
The API index was based on 5-day record prior to the acquisition date of the related scenes. Only three main meteorological stations were used in this analysis, namely Stesen Subang Airport, Petaling Jaya (Mets. Off) and Universiti Malaya meterological station. The result show that the 1988 scene taken in dry season, comparativey to the other scene (1999) (Table 2). Low satellite radiance in 1999 was believed due to wet surface resulted from high rainfall.

Table 2 : Image acquisition date and total rain for 1988 and 1999

Acquisition Date Total rain 5 days before image acquisition date at UM station Total rain 5 days before image acquisition date at PJ station Total rain 5 days before image acquisition date at Subang station
April 17th, 1988 (dry) 12 mm 0 mm 7.2 mm
February 11th, 1999 (wet) 111.5 mm 72 mm 21.4 mm

Table 3: Selected points for accuracy assessment

Forest Cover (x,y Coor) Urban Cover (x,y Coor)
X Y Temp (Celcius) X Y Temp (Celcius)
403578 322929 22 419222 325640 28.2
405553 318073 20 416106 328342 28.2
408104 329431 23.4 424322 332548 26.9
426458 337580 22.4 418688 334923 26.9
404372 318508 19.7 419674 336668 27.7
416857 317640 23.4 416852 335250 27.3
415921 318816 23.8 412253 324849 23.4
422194 321346 24.2 423130 330148 22.9
418741 322396 23.4 419627 336915 25.6
322863 322863 23.8 416718 331653 24.2
Average 22.6 Average 26.1


Figure 9. Accuracy Assessment

Conclusion
The result shows that integration of GIS and remote sensing has been found to be effective in analysing urban-land surface temperature relationships. In order to produce more reliable and meaningful result future study will incorporate the using of other higher resolution platform such as SPOT and IKONOS satellite data.

References

  1. Asmala Ahmad and Noorazuan Hashim (2006). “Remote Sensing of Land surface Temperature Using Landsat TM Thermal Infrared Band”. Proceedings of The Asia GIS 2006 International Conference, UTM Johor, Malaysia, 9-10 March 2006.
  2. Asmala Ahmad, Noorazuan Hashim, Zolkepli Buang (2006). “Estimation Of Land Surface Temperature Using Landsat Tm Thermal Infrared In Selangor-Negeri Sembilan”. Proceedings of The National Seminar on Science and Its Applications in Industry (SSASI 2006), Malacca, Malaysia, 14th – 15th Februaty 2006. Section1 : 1.
  3. Li, F. Jackson, T. J., Kustas, W. P., Schmugge, J., French, A. N., Cosh, M. H. and Bindlish, R., 2004 Deriving land surface temperature from Landsat 5 and 7 during SMEX02/SMACEX, Remote Sensing of Environment, vol. 92, pp. 521-534.
  4. Sobrino, J.A, Jime´nez-Mun˜oza, Paolini. 2004.Land surface temperature retrieval from LANDSAT TM 5. Remote Sensing of Environment 90 (2004) 434–440