Developing an algorithm for automatic estimation of Land Surface Temperature(LST) using landsat 8 images - Case study: Tabriz City

Land surface temperature (LST) is one of the key parameters in environmental studies on local to global scales. Considering the limitations of local meteorological stations, remote sensing has opened a new horizon in collection of suchinformation. Recently, successful launch of Landsat 8 with two thermal bands has provided a good opportunity for retrieving land surface temperature usingthermal remote sensing technology. Many studies had been performedwith the aim of retrieving land surface temperature, but available evidencesshow a significant calibration uncertainty inThermal Infrared Sensor (TIRS) of Landsat 8 band 11 and thus development of new studies based on onethermal band seems to be necessary. However, calibration documents issued by the United States Geological Survey (USGS) indicated uncertainty ofdata received from Band 11 Thermal Infrared Sensor (TIRS) of Landsat 8 and suggested using Band 10 data as a single spectral band for LST estimation.

In this study, mono-window algorithm with its three essential parameters (ground emissivity, atmospheric transmittance and effective mean atmospheric temperature)has been developedunderan automated algorithmin MATLABand was used for Landsat 8 data.Thermal band 10 was used to estimate brightness temperature. Bands 4 and 5 were also used to calculate the NDVI. Retrieval of LST from Landsat 8 TIRS data is performed based on the premise that brightness temperature (Ti)can be computed for any pixel of Band 10 using the mono-window algorithm.Since the observed thermal radiance for Band 10 of Landsat 8 TIRS is stored and transferredasa digital number (DNs) with 16 digits between 0 and 65,535, it is possible toconvertthe DN value into thermal radiance and then convert radiance into brightness temperature.Ground emissivity is calculatedusing land cover patterns received from other bands of Landsat 8, and the other two parameters are estimated based on the local meteorologicaldata. Usually, obtaining an accurate estimate of ground emissivity is very difficult, and the atmospheric water vapor content is considered to be a sensitive parameter in traditional LST retrieval methods.

The algorithm has been successfully applied to Tabriz city in north west of Iran with the aim of analyzing spatial distribution of LST. After running the algorithm on the satellite images of the study area on July 18,2016, a lower land surface temperature was observed in green spaces with 1.2°C accuracy as compared to urban areas and wastelands. The lowest temperature in the study area was 20°C and the highest temperature was 53°C and mean temperature was 38.78°C.Results indicate that the algorithm candiscover natural urban heat islands accurately. Moreover, spatial distribution of LST in the region is quite well matched with the land covers. Successful application of the algorithm proves the efficiency of improved mono-window algorithm as a method used for retrieving LST from Landsat 8 data.

Compared to common methods,the proposed algorithm estimates land surface temperature with minimum requirement for user intervention, least possible time and an acceptable accuracy. Itgives researches an opportunity to easily compute LST and apply it in other studies, and thus it is a significant tool.