Statistical Evaluation of Fitting Accuracy of Global and Local Digital Elevation Models in Iran

Abstract

ABSTRACT:Digital Elevation Models (DEMs) are one of the most important data for various applications such as hydrological studies, topography mapping and ortho image generation. There are well-known DEMs of the whole world that represent the terrain's surface at variable resolution and they are also freely available for 99% of the globe. However, it is necessary to assess the quality of the global DEMs for the regional scale applications.These models are evaluated by differencing with other reference DEMs or ground control points (GCPs) in order to estimate the quality and accuracy parameters over different land cover types. In this paper, a comparison of ASTER GDEM ver2, SRTM DEM with more than 800 reference GCPs and also with a local elevation model over the area of Iran is presented. This study investigates DEM's characteristics such as systematic error (bias), vertical accuracy and outliers for DEMs using both the usual (Mean error, Root Mean Square Error, Standard Deviation) and the robust (Median, Normalized Median Absolute Deviation, Sample Quantiles) descriptors. Also, the visual assessment tools are used to illustrate the quality of DEMs, such as normalized histograms and Q-Q plots. The results of the study confirmed that there is a negative elevation bias of approximately 5 meters of GDEM ver2. The measured RMSE and NMAD for elevation differences of GDEM-GCPs are 7.1 m and 3.2 m, respectively, while these values for SRTM and GCPs are 9.0 m and 4.4 m. On the other hand, in comparison with the local DEM, GDEM ver2 exhibits the RMSE of about 6.7 m, a little higher than the RMSE of SRTM (5.1 m).The results of height difference classification and other statistical analysis of GDEM ver2-local DEM and SRTM-local DEM reveal that SRTM is slightly more accurate than GDEM ver2. Accordingly, SRTM has no noticeable bias and shift from Local DEM and they have more consistency to each other, while GDEM ver2 has always a negative bias.

DEM Generation from Fixed-Wing UAV Imaging and LiDAR-Derived Ground Control Points for Flood Estimations

“…The first method is based on the root mean square error, RMSE, which is commonly used under the assumption that the set of {ΔZ} is normally distributed and is located over open areas not prone to outlier influence [42]. The second method is a measure of accuracy based on robust estimators, suggested in Höhle et al [43,66], the Normalized Median Absolute Deviation (NMAD): NMAD =1.4826 · mediani(|ΔZi− mΔZ|), where m ΔZ is the median of the errors, and ΔZ i are the individual errors. NMAD is thus proportional to the median of the absolute differences between errors and the median error.…”

Global Digital Elevation Model Comparison Criteria: An Evident Need to Consider Their Application

“…Although there is some controversy about the normality of elevation errors, this aspect is only analyzed in very few cases (e.g., [74,84,85] among others). In these cases, analysis based on Q-Q plot graphs is often applied (e.g., [86,87]) dismissing more formal statistical tests. Other possible distributions are almost ignored.…”