Modeling and estimation of the spatial distribution of elevation error in high resolution DEMs from stereo-image processing

The authors examine the spatial variability of elevation error in high-resolution DEMs derived from stereo-image processing. DEM error models are developed by examining the correlation between various parameters and the observed DEM error. The DEM vertical errors are obtained from a database containing more than 51,000 points of known elevation. The error models are shown to have strong correlation with the magnitude of the random DEM vertical error. In addition, the models capture the full dynamic range of the observed error and are able to predict the overall error in four different high-resolution DEMs to within ±5%. Moreover, the final model is used to estimate the vertical error at every point in a DEM containing ~20 million elevations