Satellite Land Surface Temperature for Sarawak Area
Hadi Batatia1 and nabil Bessaih2
Managing Director
University Malaysia Sarawalk
94300 Kota Smarahan
Sarawak - Malaysia
1Tel: + 60 82671000 Fax: + 60 82672301
E-mail :bessaih@feng.unimas.my
2Tel: + 60 82671000 Fax: 60 82672317
E-mail :bessaih@feng.unimas.my
Abstract
Land Surface Temperature is a key parameter in many remote-sensing applications. This temperature can be estimated from rremotely sensed infrared radiance. Two major phenomena have direct effects on the infrared radiance received by the satellite, namely atmospheric effects and surface emissivity. Many algorithms have been developed to take into account these effects. This paper presents the results of a study that aimed to evaluate the importance of these effects in the region of Northern Borneo. The study consisted in the implementation of an algorithm that corrects water vapour and emssivity effects on the estimated land surface temperature. This algorithm uses the split window method, which consists in measuring the temperature using two different infrared channels, and eliminating the effect between these two measurements. Uncorrected and corrected land surface temperature maps were determined for the region under study. These maps show the importance of the surface emissivity and atmospheric effects. The study suggests an accurate estimation of water vapour profile and emissivity in order to estimate accurately land surface temperature.
Introduction
Land Surface Temperature is a key parameter in energy budget models, evaportanspiration models (Serafini, 1987; Bssieres, 1995), estimating soil moisture (Price, 1980), forest detection and forecasting, monitoring the state of the crops, studing land and sea breezes and nocturnal cooling.
Land Surface Temperature can be determined by measuring the radiation emitted by the earth's surface. This radiation will be converted into temperature brightness using the inverse of Planck's radiation equation. Unfortunately, land surface does not behave as a blackbody. The radiation measured by the satellite radiometer include also radiation emitted by the atmosphere and radiation reflected by the land surface. Gases and suspended particles in the atmosphere and radiation reflected by the land surface. Gases and suspended particles in the atmosphere may absorb radiation emitted from objects, resulting in a decrease in the energy reaching a thermal sensor. Scattering, in the presence of suspended particles, can also attenuate ground signals. Atmospheric absorption and scattering tend to make the signals from ground objects appear colder than they are. For this reason, measurement are usually made in two atmospheric absorption free regions (windows ) of the electromagnetic spectrum lying between 3.5mm to 3.9 mm and 10mm to 13mm. Unfortunately, in this spectrum regions, there is residual gaseous absorption, mainly by water vapour. Sobrino et al, (1991), reported that the difference between the real Land Surface Temperature and the satellite surface temperature is of 7oC for a tropical atmosphere. By using a second infrared measurement, either at a different wavelength or at different zenith angle, it is possible to improve the estimation of the surface radiant temperature.
Split window method
The method widely used for correcting the water vapour effects is the split window method. This method makes use of the observation that the transmission of a path trough a most atmosphere at one wavelength is closely correlated with the transmission through the same path at a second nearby wavelength. The advanced very high resolution radimeters (AVHRR) on board of the NOAA 14 satellite employs two channels in the infrared windows between 10mm and 13mm which can be used to exploit this differential effect.
Price (1980,1984), sobrino et al (1991), vidal (1991), Ottle and Vidal-Madjdar (1992), prata (1993) developed split window algorithms, which have global form with coefficients that depends upon local conditions. These algorithms have the following general form:
T0 = T4 + A(T4-T5) + B (1)
where,
T
0 : real Land Surface Temperature
T
4 : channel 4 brightness temperature
T
5: channel 5 brightness temperature
A and B are coefficients which depend on the surface emissivities, atmospheric absorption coefficients and total water vapour amount (Sobrino et al, 1991)
Data analysis
Sarawak occupies a large proportion of the northern part of Borneo island. It is located within the longitudes E110o to E114o and N1o to N5o. Over this area cloud cover is frequent. This makes selection of data sets for analysis a tedious task. Ten NOAA-14 AVHRR GAC data sets occurring during June and July 1996 near 5:00 UCT (12:00 LST), which are close to the "midday" conditions required by many applications, have been acquired for this investigation. The satellite data were mapped to UTM projection by using ERMAPPER(™) image processing software. Satellite counts were converted into measured radiance for the five NOAA-AVHRR channels using methods provided in the "NOAA Polar Orbiter Data User Guide".
The radiance measured by channel 3, 4 and 5 were converted to brightness temperature using the equations provided in Kidwell (1995).
For channels 1 and 2, the percent albedo measured by the sensor were converted to spectral radiance using the method described in Kidwell (1995).
Normalised Difference Vegetation Index (NDVI) values were calculated using visible channel (CH1) and near -infrared channel (CH2) reflectance values (Eq.2),
| NDVI = |
CH2 -CH1
-------------------
CH2+CH1 |
(2) |
Applying a threshold on the NDVI maps identified the cloudy regions. The cloudy pixels were then eliminated from the Temperature Brightness maps. It was notices that cloudy pixels had brightness temperature lower than 13oC, which is acceptable for the region.
The cloud-free maps for channel 4 and 5 wee overlaid with a map containing land regions. An emissifity of value 1 was to all non-land regions. An emissivity of 0.98 was assigned to all land pixels. The sobrino algorithm (eq.1) was then applied to correct the temperature.
