Identification of the ‘Best’ DEM via Analysis of Morphological Parameters Derived From Different DEMS, and Applying Linear Regression for Accuracy Enhancement

In recent years, numerous Digital Elevation Models (DEMs) derived from satellite data have become available, offering varying resolutions and accuracies. These datasets are crucial for analyzing the topographic characteristics of river basins through the assessment of morphometric parameters. Researchers have extensively utilized these DEMs to explore basin characteristics and their drainage networks. However, the impact of DEM resolution, accuracy, and sources on these analyses remains a significant question. This study aims to evaluate the influence of DEM accuracy and sources on morphometric analysis. Additionally, it investigates a method to enhance DEM accuracy through the application of linear regression. The study initially employs DEMs from ASTER GDEM and SRTM 1-arc second to analyze the relief, surface, size, shape, and texture properties of a mountainous drainage basin. Elevations extracted from both DEMs for the same locations of available reference data were compared. The linear regression model was applied to both DEMs to enhance accuracy. The model demonstrated greater efficacy with the ASTER DEM, yielding a more accurate representation of terrain elevations compared to the SRTM DEM. However, the presence of outliers influenced the results, highlighting the importance for careful consideration of these anomalies. The study concludes by recommending further exploration of alternative models such as decision tree regression, random forest regression, support vector regression, and neural network regression, to potentially identify a more universally applicable approach for improving DEM accuracy.