Temporal analysis of surface temperature obtained by remote sensing

Abstract

There is a growing concern with respect to the temperature increases that has caused several impacts on fauna and flora. In this sense, remote sensing (SR) data has been widely used for the acquisition of information on the Earth's surface temperature. Thus, this study aimed to correlate the temperature data of an automatic station of the Companhia Ambiental do Estado de São Paulo (Cetesb), located in Sorocaba, with data obtained remotely through digital image processing of the Landsat 8 satellite between the years 2013 to 2017. ArcGIS 10.4.1 software was used for image processing, using the fixed parameters of conversion of gray image levels to radiance and subsequent conversion of data into temperature values. To verify the variation of the data obtained by RS compared to the monitoring station, the Root Mean Square Error (RMSE), Pearson correlation coefficient (r) and index of agreement (d-index) were verified. The obtained RMSE values indicate an average temperature variation of 2.74 ºC, corroborating the r² values of 0.664, which indicates that 66.4% of the values obtained for the temperature remotely are directly related to the values obtained at the Cetesb monitoring, presenting positive agreement d-index values of 0.88. Therefore, the results were satisfactory for a better acquisition of temperature values obtained remotely indicating a good correlation with the data obtained by the Cetesb monitoring station.