Objectives Effective correction of ASTER GDEM (advanced spaceborne thermal emission and reflection radiometer global digital elevation model) elevation error is of great significance to the quality and application of ASTER GDEM data. Terrain parameters such as slope and aspect derived from digital elevation model (DEM) data have a significant impact on the accuracy of the data. Therefore, the relationship model between the ASTER GDEM elevation error and each influencing factor can be constructed to effectively correct its elevation accuracy.
Methods Four geomorphological types of Changwu, Yijun, Ganquan and Yanchuan in the Loess Plateau of northern Shaanxi are selected. With 1∶50 000 DEM as reference data, after data preprocessing, the elevation error values and related topographic factors and the surface coverage index of each point are calculated, by extracting a certain number of sampling points and check points, a random forest regression algorithm is introduced to establish an elevation error prediction model to correct the elevation accuracy, and compared with multiple regression model.
Results ASTER GDEM elevation error characteristics are closely related to the terrain conditions; the contribution of each impact factor to the model is different in different areas; the random forest regression prediction model is better than the multiple regression model overall, and has good applicability and error correction effect. It can reduce the error mean of Changwu, Yijun, Ganquan and Yanchuan by 3.08 m, 3.00 m, 3.61 m and 6.95 m.
Conclusions This research is helpful to understand the ASTER GDEM elevation error characteristics and its influencing factors in different geomorphological areas of the Loess Plateau in northern Shaanxi. At the same time, suitable regression models can be selected for areas with different terrain conditions to effectively correct the error values, further improve the application accuracy of ASTER GDEM data.
DownLoad: