Development of linear regression model to predict ground elevation from satellite elevation-statistical approach

Abstract

Digital Elevation Model is the versatile source for many applications. The elevation accuracy of DEM is the deciding factors for improving their applications in various fields. The accuracy of elevation differs based on the sources of data. The more economical and frequently obtained stereo capable Cartosat-1 is capable in generation of Digital Elevation Model. But the elevation accuracy is comparatively lower than Terrestrial, Aerial, LIDAR generated DEM's elevation. The idea of this study is to develop the prediction model from SPSS software to enhance the elevation accuracy of satellite based data by using ground elevation for different terrain conditions at Tirunelveli. Study covers a part of Tirunelveli, which comes under the Path and Row 550,354 of Cartosat-1 satellite data have been used in this study. The reliability check, correlation and Paired T tests were done before deriving the prediction model using SPSS statistical software in order to check the association between ground elevation and satellite elevation. It was observed from the study area that the minimum and maximum elevation differences between the satellite and ground was 6 m and 10 m respectively for same earth positions. In this study ground elevation used as dependent variable and satellite elevation used as independent variable for the prediction model development. The prediction model have been validated for the data have been collected in the field for the different terrain conditions, it was revealed that the differences between the predicted and measured values maximum was 0.62 meter and the minimum was 0.0009 meter. It was observed from the study that the prediction model identified from the different terrain conditions worked only for the same terrain types.