|
|
|
Geoscience / DTM
|
Abstract
Digital Elevation Model (DEM) is indicpensable for many analysis such as topographic feature extraction, runoff analysis, slope stability analysis, landscape analysis and so on. Such analysis require a high accurate DEM. The accuracy of DEM is usually represented by spatial resolution and heigh accuracy. The height accuracy according to DEM generation emthod was evauated lays year in this s tudy, DEM accuracy according to spatial resolution was evaluated.
DEMs of various resolution(100m, 150m, 200m grid size) were prepared for an evaluation. Each analysis was carried out using previous DEMs. A differences of the results according to spatial resolution were concluded. The results showed spatial resolution influenced an inclination very well. Almost terrain analyses use comination of slope aspect and inclination. So, we must take care to use low resolution DEM. On the other hand, nearest neighbor showed the best method as resampling method. Mean value resampling orminimum value resampling made flat complicated terrain surface. So, nearest neighbor resapling should be used when low resolution DEM must be gerated.
1.Introduction
There are many kinds of DEM (Digital Elevation Model) generation methods such as a sereo matching from aerial photgraphor satellite image, an interferemetry from SAR and an interpolation of topographic maps. Some commercial computer software support such DEM generation fuction. On the other hand, we can use an accomplished DEM. For example, NGDC, NOAA offers global land one-km base elevation (GLOBE). And USGS Digital Chart of the World that has elevation information also. Nowadays, we have many analyses such as topographic feature extraction, runoff analysis such as topographic feature extraction , runoff analysis, slope stability analysis, landscape analysis and son on. We must consider which accuracy is appropriate for any analyses. Therefore, A study of DEM accuracy is very important.
By the way, we developed a new controur line interpolation method for DEM generation last year [1]. The method which is based on buffering showed higher performance rather than existing interpolation methods. However an accuracy of DEM was depending upon contour line interval. A high density controur line information was required for a high accurate DEM. If a percentage of total pixel number which consisted contour line indicates less 30%, a correct percentage by comparions with original DEM indicates less than 50%.
Therefore, an important of contour line interval was reconginzed in DEM generation.
In previous study, an accuracy of DEM incase of fixed spatial resolution was evaluated. In this study, an accuracy of DEM according to spatial resolution will be evaluated. We prepared a various grid size of DEMs. The DEMs will be used for topographical analysis, slope stability analysis and runoff analysis. After, that, results of each analysis will be compared with result from original DEM. A relationship between spatial resolution and DEM accuracy will be concluded.
2.2 Materials
An original DEM was generated by interpolation from 1:25000 contour maps. Its grid size is 50 m. Figure 1 shows a shaded image of the original DEM. The test area is mountainous area. So, slope stability analysis or runoff analysis can be carried out.
Figure 1 Shaded Image of Original DEM
For evaluation of spatial resolution, a various size DEM was prepared from the original DEM. In this study, 100m, 150m, 200m and 250m grid size DEMs were generated by resampling. There are some kinds of resapling method such as nearest neighbor, bi-linear, cubic convolution and so on. In case of DEM generation, nearest neighbor, mean and minimum value are used as resampling usually. For example, GLOBE supports maximum value, mean value and minimum value. We must consider which sampling method is the best for any analysis. So, all resampling methods were carried out to prepared materials.
3.Evaluations of DEM accuracy
3.1Slope Aspect Accuracy
A slope aspect can be expressed from DEM, which is one of the most important items for topographical analysis. In this study, the slope aspect means a direction along the maximum slope inclination at one target pixel. Figure 2 shows histogram of difference between original aspect data and resampled aspect data of each grid size. The histograms show symmetrical form. Figure 3 shows a relatiohship between grid size and percentage of correct pixels. The correct percentage in every resampling method has tendency to drop with grid size increase. Maximum value sampling showed the highest accuracy, minimum value sampling showed the worst accuracy.
Figure 2 Accuracy of Slope Aspect according to Spatial Resolution
Figure 3 Accuracy of Slope Aspect according to Spatial Resolution
3.2. Slope inclination Accuracy
A slope inclination can be expressed from DEM, which is also one of the most important items for topographical analysis. In this study, the inclination means maximum slope inclination at one target pixel. Figure 4 shows histogram of difference between original inclination data and resampled inclination data of each grid size. The histograms show asymmetrical form that is shifted to right. It means the resampled inclination data became gentle slope. Because, detailed terrain is gnarred by grid size incrase. Figure 5 shows a relationship between grid size and percentages of correct pixels. In case of slope in clinat in, correct pixel means difference with original data indicates inside of 20 degree. The nearest neighbor sampling almost showed the highest accuracy, mean value sampling showed the worst accuracy.
Figure 4 Accuracy of Slope Inclination according to Spatial Resolution
Figure 5 Accuracy of Slope Inclination according to Spatial Resolution
|
|
|
|
|
|
|
