Models for Temperature Estimation from Remotely Sensed IR Data

Abstract

Thermal patterns of the terrain surface are very often used to know the state of the environment. They are usually obtained by remotely sensed IR data. Accurracy and reliability are essential for the practical applications of the remotely sensed data. They are examined by measurements obtained on the ground. As far as the authors know, the evaluation of them has been made merely through comparison (that is, only through good or bad coincidence) between the estimated and the observed temperatures. There is no established relation between the temperatures, on which the evaluation should be based. In the present paper, a temperature estimated from remotely sensed data is related to the observed temperature at the same place through a simle physical model. In the model the atmosphere is represented by a thin layer described by using a transmittance and an effective temperature. We consider four kinds of models slightly modified from each other. By using the extended Tⁿ relation of radiance to a temperature T which we already established, it is shown for all the models the regression coefficient given by regression analysis is not larger than 1. In the analysis the estimated temperature is taken as a dependent variable and the measured one as an independent variable. This gives us a reference relation between the both data. It is shown that this relation holds for most of the actual data.