Materials and Methods

Survey of India (SOI) toposheets of 1:50000 and 1:250000 scales were used to generate base layers. Field data were collected with a handheld GPS. Remote sensing data used for the study are: Landsat MSS (1973), Landsat TM (1992), Landsat ETM+ (2000 and 2009) [Landsat data downloaded from http://glcf.umiacs.umd.edu/data/], IRS (Indian Remote Sensing) LISS (Linear Imaging Self Scanner)-III of (1999 and 2006), MODIS (Moderate Resolution Imaging Spectroradiometer) Surface Reflectance 7 bands product [downloaded from http://edcdaac. usgs.gov/main.asp] of 2002, MODIS Land Surface Temperature/Emissivity 8-Day L3 Global and Daily L3 Global (V004 product) [http://lpdaac.usgs.gov/modis/dataproducts.asp#mod11]. Google Earth data (http://earth.google.com) served in pre and post classification process and validation of the results. Latest data for 2010 (IRS – Indian remote Sensing) was procured from the National remote Sensing Centre (http://www.nrsc.gov.in), Hyderabad. The methods adopted in the analysis involved:

1. Georeferencing of acquired remote sensing data to latitude-longitude coordinate system with Evrst 56 datum: Landsat bands, IRS LISS-III MSS bands, MODIS bands 1 and 2 (spatial resolution 250 m) and bands 3 to 7 (spatial resolution 500 m) were geo-corrected with the known ground control points (GCP’s) and projected to Polyconic with Evrst 1956 as the datum, followed by masking and cropping of the study area.
   
   1. Band 1, 2, 3 and 4 of Landsat 1973 data to 79 m.
   2. Band 1, 2, 3 and 4 of Landsat TM of 1992 to 30 m.
   3. Band 1, 2, 3, 4, 5 and 7 of Lansat ETM+ to 30 m.
   4. MODIS bands 1 to 7 to 250 m.
   5. IRS LISS-III band 1, 2 and 3 to 23.5 m.
   6. Thermal band of TM (resampled to 120m), ETM+ (to 60m) and MODIS (to 1 km) and Panchromatic bands of ETM+ (resampled to 15 m).
2. Supervised Classification using Bayesian Classifier: In supervised classification, the pixel categorisation process is done by specifying the numerical descriptors of the various LC types present in a scene. It involves (i) training, (ii) classification and (iii) output.
3. Accuracy assessment: Accuracy assessments were done with field knowledge, visual interpretation and also referring Google Earth (http://earth.google.com).
4. Computation of Normalised Difference Vegetation Index (NDVI): It separates green vegetation from its background soil brightness and retains the ability to minimize topographic effects while producing a measurement scale ranging from –1 to +1 with NDVI-values < 0 representing no vegetation.

Derivation of Land Surface Temperature (LST)

LST from Landsat TM: The TIR band 6 of Landsat-5 TM was used to calculate the surface temperature of the area. The digital number (DN) was first converted into radiance LTM using

L_TM = 0.124 + 0.00563 * DN ….. (Equation 1)

The radiance was converted to equivalent blackbody temperature TTMSurface at the satellite using

T_TMSurface = K₂/(K₁ – lnLTM¬) – 273 ….. (Equation 2)

The coefficients K₁ and K₂ depend on the range of blackbody temperatures. In the blackbody temperature range 260-300K the default values (Singh, S. M., 1988) for Landsat TM are K₁ = 4.127 and K₂ = 1274.7. Brightness temperature is the temperature that a blackbody would obtain in order to produce the same radiance at the same wavelength (λ = 11.5 μm). Therefore, additional correction for spectral emissivity (ε) is required to account for the non-uniform emissivity of the land surface. Spectral emissivity for all objects are very close to 1, yet for more accurate temperature derivation emissivity of each LC class is considered separately. Emissivity correction is carried out using surface emissivities for the specified LC (table 1) derived from the methodology described in Snyder et al., (1998) and Stathopoulou et al. (2006).

Table 1: Surface emissivity values by LC type

The procedure involves combining surface emissivity maps obtained from the Normalized Difference Vegetation Index Thresholds Method (NDVITHM) (Sobrino and Raissouni, 2000) with LC information. The emmissivity corrected land surface temperature (Ts) were finally computed as follows (Artis and Carnhan, 1 982)

                ............(3)

where, λ is the wavelength of emitted radiance for which the peak response and the average of the limiting wavelengths (λ = 11.5 μm) were used, ρ = h x c/σ (1.438 x 10^-2 mK), σ = Stefan Bolzmann’s constant (5.67 x 10^-8 Wm^-2K^-4 = 1.38 x 10^-23 J/K), h = Planck’s constant (6.626 x 10^-34 Jsec), c = velocity of light (2.998 x 10⁸ m/sec), and ε is spectral emissivity.

LST from Landsat ETM+: The TIR image (band 6) was converted to a surface temperature map according to the following procedure (Weng et al., 2004). The DN of Landsat ETM+ was first converted into spectral radiance LETM using equation 4, and then converted to at-satellite brightness temperature (i.e., black body temperature, T_ETMSurface), under the assumption of uniform emissivity (ε ≈ 1) using equation 5 (Landsat Project Science Office, 2002):

L_ETM = 0.0370588 x DN + 3.2 ….. (Equation 4)

T_ETMSurface = K2/ln (K1/ LETM + 1) ….. (Equation 5)

where, TETMSurface is the effective at-satellite temperature in Kelvin, LETM is spectral radiance in watts/(meters squared x ster x μm); and K2 and K2 are pre-launch calibration constants. For Landsat-7 ETM+, K2 = 1282.71 K and K1 = 666.09 mWcm^-2sr-1μm^-1 were used(http://ltpwww.gsfc.nasa.gov/IAS/handbook/handbook_htmls/chapter11/chapter11. html). The emissivity corrected land surface temperatures Ts were finally computed by equation 3.