Comparison of the resampling methods for gridded dem downscaling
In this paper, a comparison and evaluation of three resampling methods for gridded DEM is implemented. The evaluation was based on the results of bilinear, bi-cubic and Kriging resampling methods for an experiment using both degraded and sampled datasets at 20 m, 60 m and 90 m spatial resolutions. The evaluation of the algorithms was accomplished comprehensively with visual and quantitative assessments. The visual assessment process was based on direct comparison of the same topographic features in different resampled images, scatter plots and profiles. The quantitative assessment was based on the most commonly used parameters for DEM accuracy assessment such as root mean square errors (RMSEs), linear regression parameters m and b, and correlation coefficient R of the resulted DEMs versus the reference DEM. Both visual and quantitative assessment revealed a greater accuracy of the Kriging resampling over the other two conventional methods with the RMSE of the Kriging interpolated DEMs decreased by approximately 58%, 23%, 50%, and 58% for the 20 m and 30 m degraded DEMs in Nghe An province, 5 m sampled DEM in Lang son province, and 30 m sampled DEM in Dac Ha, Vietnam, respectively.