Estimation and Validation of Land Surface Temperature by using Remote Sensing & GIS for Chittoor District, Andhra Pradesh

Land Surface Temperature (LST) quantification is needed in various applications like temporal analysis, identification of global warming, land use or land cover, water management, soil moisture estimation and natural disasters. The objective of this study is estimation as well as validation of temperature data at 14 Automatic Weather Stations (AWS) in Chittoor District of Andhra Pradesh with LST extracted by using remote sensing as well as Geographic Information System (GIS). Satellite data considered for estimation purpose is LANDSAT 8. Sensor data used for assessment of LST are OLI (Operational Land Imager) and TIR (Thermal Infrared). Thermal band contains spectral bands of 10 and 11 were considered for evaluating LST independently by using algorithm called Mono Window Algorithm (MWA). Land Surface Emissivity (LSE) is the vital parameter for calculating LST. The LSE estimation requires NDVI (Normalized Difference Vegetation Index) which is computed by using Band 4 (visible Red band) and band 5 (Near-Infra Red band) spectral radiance bands. Thermal band images having wavelength 11.2 μm and 12.5 μm of 30th May, 2015 and 21st October, 2015 were processed for the analysis of LST. Later on validation of estimated LST through in-suite temperature data obtained from 14 AWS stations in Chittoor district was carried out. The end results showed that, the LST retrieved by using proposed method achieved 5 per cent greater correlation coefficient (r) compared to LST retrieved by using existing method which is based on band 10.


1.
Introduction Surface Temperature of the land could be obtained by using remote sensing from on board sensor in satellite. Extracted temperature is required for many applications like atmospheric models for the calculation of functional heat flux by assessing the change between the land surface temperature and the air temperature near the surface. It's assessment done by using brightness temperature of the Top-of-Atmosphere. Its evaluation additionally depends on aldedo, vegetation cover and soil moisture.
(Source: https://land.copernicus.eu/global/products/lst). Satellites with on board sensors of thermal infrared (TIR) instruments are the mainly available operational systems for collecting the LST data that has been used usually in agriculture drought assessment (M.S. Malik et al., 2018), land-atmosphere exchange simulation model, radiation budget estimation and evapo-transpiration etc.. During summer season farmers utilize land surface temperature specifying maps for evaluating water necessities for their crops when they are more prone to high temperature conditions. On the other hand, these maps assist citrus farmers in determining where and when orange orchards have been open to the elements to damaging chill in the winter season. Like this number of real-time applications make use of surface temperature parameter for different atmospheric models. (Source: https://earthobservatory.nasa.gov/). The observation of diurnal characteristics of land surface temperature provides the opportunity for exploration of the climate change, agricultural drought and estimation of crop yield (Ying Sun et al., 2020; Limin Yang, 2000 ). A number of LST assessment algorithms had been developed for LANDSAT 7 or 8 satellite data using thermal band data (F. Sattari et al., 2014). Mono-window algorithm is one of the extensively used methods for surface temperature estimation (Fei Wang et al., 2015;Jinqu Zhang et al., 2006). It is used for LST evaluation from single TIR band data of LANDSAT 7 or 8 satellite images using GIS. Other parameters necessary for LST assessment are atmospheric transmittance, ground emissivity and average effective atmospheric temperature. Land Surface Temperature is used for quantitative analysis of different land use or land cover categories by using Thermal Infra-Red (TIR) sensor of LANDSAT 8. Some of the thermal infrared sensors used for the study of LST with 1km resolution are NOAA (National Oceanic and Atmospheric Administration) Satellite -AVHRR sensor (Advanced Very High Resolution Radiometer), Terra and Aqua Satellites -MODIS (Moderate Resolution Imaging Spectroradiometer). Some of other sensors having high resolution are Terra Satellite -Advanced Space borne Thermal Emission and Reflection Radiometer (ASTER) with resolution of 90 m and LANDSAT 7 satellite -Enhanced Thematic Mapper (ETM+) and LANDSAT 8 Satellite -Thermal Infra-Red (TIR) sensors with 100 m resolution. The estimation of LST is useful for various applications like, Land use /  Rajani et al., 2020 ) The NDVI (Normalized Difference Vegetation Index) characterizes a number of vegetative properties of land. The LANDSAT 8 having sensor OLI captures images in 9 spectral bands, in that Visible Red band (Band4) and Near-Infrared band (Band5) used for NDVI estimation. NDVI index varies between -1.0 and +1.0. NDVI is used for land use or land cover classification. Based on the index values of NDVI Land cover classification was done into dense forest, water bodies, Baren land, built-up area and spare vegetation (A. Rajani et al., 2020). The present study focus is on feasibility of finding LST by using band11 data from TIR sensor of LANDSAT 8. The algorithm used for the proposed method is Mono Window Algorithm. After finding LST using band11 then it is compared with the AWS data i.e. error is estimated. And also considering data obtained by D.Jeevalakshmi et al., 2017 for finding LST using band10 and then its difference with AWS data. Finally correlation coefficient is generated using statistical tool for the proposed method and the method used by D. Jeevalakshmi et al., 2017. Land surface emissivity is one of essential parameter for the estimation of LST. LSE estimation requires proportion of vegetation. NDVI is used to estimate proportion of vegetation. The GIS software used for processing LANDSAT 8 satellite image is ArcGIS 10.3.

2.
Study Area Study area selected is Chittoor district which is extreme south of the Andhra Pradesh state. It lies between 12.616667 -14.133333 N latitudes and 78.05 -79.916667 E longitudes to the south of Andhra Pradesh state. The temperature is lower in Punganur, Madanapalle and Horsley Hills i.e western portions of the district when it is related to the eastern portions of the district. As western portions are at higher altitude compared to eastern portions so that lower temperatures are observed. In summer temperature ranges between 36.0° to 38.0 °C in the western portions and touches 46.0 °C in the eastern portions of the chittoor district. The winter temperatures of the eastern portion is around 16.0°C to 18.0 °C whereas western portions temperatures are at low ranging around 12.0 °C to 14.0 °C (source: https://en.wikipedia.org/wiki/Chittoor,_Andhra_Pradesh). Figure 1 illustrates Geographical site of the area of interest chosen for this particular study.  Table 1.

Methodology
The proposed work is to estimate LST of study area, i.e. Chittoor district by using band 10 and band 11 independently using Mono Window Algorithm. Mono Window algorithm comprises 6 steps process in the estimation of the LST of the study area. These steps are specifically used for processing the LANDSAT 8 band data only. LST assessment requires brightness temperature, Land surface emissivity and NDVI values. Land surface temperature is estimated by using band 10 and 11 separately for comparison of which one performs better. After retrieving land surface temperature from multiple spectral bands and multiple sensors, it is used to validate with in-suite data obtained from AWS stations. The LST estimated by using remote sensing and GIS is compared with in-suite AWS stations temperature data and difference of two temperatures is calculated. Eventually, the correlation coefficient of LST and AWS data is computed and analyzed. Flowchart for LST retrieval process and validation method is specified in figure 2. Step 1 Translation of satellite image Digital Number (DN) into spectral radiance called Top of Atmosphere (TOA) is done by applying the equation number (1) and the band specific parameters are presented in the Table  2. These parameters obtained from metadata file.
(1) Where, Lλ -Spectral radiance of TOA (mW /sr mm 2 ) ML -Multiplicative rescaling value AL -Additive rescaling factor of specific band Qcal -Digital Number (i.e. Quantized and calibrated pixel values) Oi -Adjustment factor Conversion of Top of Atmosphere (TOA) radiance into brightness temperature (BT) by using Lλ and band specific thermal conversion constants K1 and K2 specified in metadata file of satellite image. The resultant temperature is obtained in Kelvin. It is converted into Celsius by adding the absolute zero (-273.15 0 C). Brightness temperature in Celsius is estimated by the equation (2). The thermal conversion constants are shown in the Table 3. ( Where BT -Brightness Temperature in 0 Celsius Lλ -Top of Atmospheric spectral radiance (Band 10 or Band 11) K1 and K2 -Thermal Sensor constants used for conversion (Source: metadata file) Calculation of NDVI is necessary for estimation of vegetation proportion (Pv) and land surface emissivity (LSE i.e. ԑ) parameter which are needed for estimating the Land Surface Temperature.
Step 4 Vegetation proportion ( Pv ) is calculated by using equation number 4, where NDVI is obtained from step 3.
After the estimation of NDVI values of study area, then consider from that NDVI image lowest and highest values of the NDVI image.
Step 5 Land Surface Emissivity (ε) anticipated through the use of the NDVI threshold method. LSE is essential parameter for calculating the LST, because it is a proportionality thing that is used to scale blackbody radiance (Planck's law) so that emitted radiance can be predicted. Moreover Where, 0.004 -Standard deviation of soil bands, 0.989 -Average Emissivity (i.e average of soil and vegetation emissivity factors) Step 6 The ultimate step of estimating the LST is by using the equation (7) Where, λ = 10.8 µm i.e. Emitted radiance wavelength ελ -land surface emissivity and (8) Where

Results and Discussions
Land surface temperature obtained using remote sensing and GIS technique. Each pixel in the image represents the surface temperature of each object that may be group of numerous land cover types. By using above mentioned processing steps especially for LANDSAT 8 data LST maps are generated independently for both the thermal bands 10 Table 4 and 5 show the LST retrieved by using Band 10 and Band 11 by using Mono Window Algorithm and AWS data for 30-05-2015 and 21-10-2015. Table 4 and 5 also show various land cover types along with difference of temperatures between in-suite measurements from AWS stations and LST of band 10 and band 11, i.e. error by using Mono Window Algorithm estimation method. The correlation coefficient of AWS temperature data with LST retrieved by using band 10 is r= 0.801613419.
The correlation coefficient of AWS temperature data with LST retrieved by using band 11(Proposed method) is r= 0.855623779. From the results analysis, it is observed that LST retrieved by using band 11 is having 0.0487 greater correlation than LST retrieved by using band 10. The correlation coefficient can be estimated by using equation 9

(9)
From the results, it can be concluded that the land surface temperature retrieved by using band 11 is having more correlation coefficient compared to the LST retrieved by using band10. Figure 9 shows plotting of retrieved LST and AWS data for the study area by using band 10 and band 11 independent results graph for the day 30 th May 2015. From the plot it can be witnessed that the retrieved LST using band11 is having less deviation compared to AWS data from the study area for the chosen 14 weather stations.
Similar analysis is done for the LANDSAT 8 image for the day 21 st October, 2015 of the study area having 14 AWS stations.  Correlation coefficient of AWS temperature data with LST retrieved by using band 10 for 21 st October, 2015 is r= 0.81830048. Correlation coefficient of AWS temperature data with LST retrieved by using band 11 for 21 st October, 2015 is r= 0.86700979. From the analysis of results it is observed that the LST retrieved by using proposed method is having 0.05401 greater correlations than LST retrieved by using band 10. Figure 10 shows a plotting of AWS data and estimated LST for the chosen area using the independent results graph with band10 and band11 for the day 21 st October 2015. From the plot, it can be seen that the LST recovered using band11 has less deviation compared to AWS data for the selected 14 weather stations selected in the region studying Figure 10. Plot showing AWS data & LST retrieved using Band 10 & Band 11 for 21-10-2015 .

CONCLUSIONS
This research work considered 14 AWS stations as validation points from study area for the validation purpose. The LST was retrieved by using Mono Window Algorithm. Here both the thermal bands of TIR sensor of LANDSAT 8 are used. Most of the research work utilized band 10 for the extraction of LST. With the use of band 11 for LST extraction, more correlation coefficient is observed. The estimated LST using remote sensing and GIS were validated by in-suite measures taken from same geographical locations. And also both observations consider same date and time. Here for the proposed work estimation satellite data considered two dates 30-05-2015 and 21-10-2015. Time was local standard time 5.00 A.M. More deviation has been observed when compared to estimated LST using band 10 with AWS data. Whereas less deviation observed while comparing estimated LST using proposed method by using band 11 with AWS data. Therefore, retrieved LST based on band 11 measurements were having ~5 per cent more correlation coefficient compared to band10 LST of the in-suite AWS stations temperature of the study area.

Conflict of interest:
There is no conflict of interest.