Evaluation on the accuracy of digital elevation models

The resolution of digital elevation model (DEM) used for watershed delineation and data extracted from the delineated watershed relevant to hydrography plays an important role in the calibration of hydrological models. The effect of DEM resolution on the drainage length, watershed area and average slope are studied in many earlier works. DEM resolution can have impact on hydrologic modeling parameters, water quality simulation results. The impacts of DEM resolution on results of the watershed response and water quality modeling are studied using HSPF (Hydrologic Simulation Program Fortran) model applied to a watershed in Kentucky in this study. Three major streams in this watershed are impaired due to pathogen pollution according to the current state of Kentucky water quality standard set for recreation. The effects of DEM resolution on water quality modeling and calibration for fecal coliform bacteria are studied using a HSPF model. Results from this study indicate that the DEM resolution affects hydrography parameters of the delineated watershed, influence the calibration process, and the non-point and point source loadings to the streams in the watershed.

Accurate and precise topographic information from Digital Elevation Models (DEM) is of fundamental requirement for many geoscience and engineering applications. With a goal of providing accurate global topographic products, German Aerospace Centre (DLR) and Airbus Defence and Space developed X-band TanDEM-X mission. The TanDEM-X Global DEM is generated with a pixel spacing of 12×12 m from interferometric data acquired in strip map mode with a resolution of 3.3 m [1]. In this study, the vertical accuracy of TanDEM-X Global DEM is assessed by comparing it with various elevation datasets. The elevation datasets include DGPS measurements, SRTM 30m, and 90 m DEM, CartoSat DEM, LiDAR data, photogrammetry data, intermediate and fused DEMs’ derived from TanDEM-X CoSSC [2], ICESat elevations and the newly released TanDEM-X 90 m DEM. The study areas include Bihar (flat terrain), Mumbai (moderate and urban terrain), Godavari (moderate and hilly), Koyna (hilly and vegetated), Manali (rugged Himalayan) and Kartarniaghat Wild Life Sanctuary-KWLS (flat and forest) with different terrain conditions. Widely used statistical parameters such as mean, standard deviation (SD) and root mean square error (RMSE) are used for vertical accuracy assessment [3]. The minimum and maximum differences between elevations are also noted, which are useful in exploiting quality of DEM under evaluation. The accuracy of the TanDEM-X Global DEM is found to be less than 1.6 m, 4 m, 5 m, 13 m, 6 m for plain, moderate relief, hilly, Himalayan, vegetated terrains respectively. In vegetated areas (KWLS), we observed an 8 m difference in tree height estimation using TanDEM-X Global DEM and GPS measurement. Elevation and difference profiles were drawn to understand the vertical accuracy of these DEMs’. In all cases, the elevations of TanDEM-X Global DEM closely matched with heights derived from DGPS, ICESAT, LiDAR and photogrammetry data.

Urban flood models that use Digital Elevation Models (DEMs) to simulate extent and depth of flood inundation rely on the accuracy of DEMs for predicting flood events. Despite recent advances in developing vegetation corrected DEMs, the effect of building height and density errors in global DEMs in urban areas are still poorly understood, and their correction remains a challenge. In this research we developed a methodology for building error correction that can be applied to any other case study, where building density data and a local reference DEM data are available. This methodology was applied to Nairobi, Kenya using six global DEMs (SRTM, MERIT, ALOS, NASADEM, TanDEM-X 12 m, and TanDEM-X 90 m DEM). Our results show building error at highest building density varying between 1.25 m and 5.07 m for the DEMs used, with the MERIT DEM showing the smallest vertical height error from the reference DEM. The six DEMs were corrected by deriving a linear relationship between building density and DEM error. Our findings show that the removal of building density error resulted in the improvement of the vertical height accuracy of the global DEMs of up to 45% for MERIT and 40% for ALOS. This methodology was also applied to the Central Business District (CBD) area of Nairobi, characterized by taller buildings and high building density. The error parameters in the CBD area resulted to be between 15 to 45% higher than those of the Nairobi city wide area for the six global DEMs, thus providing further insights into the contribution of building heights to errors in global DEMs. Building height data is still unavailable on a global scale and our results show that global DEMs can be usefully corrected for building density errors in urban areas, even where specific building height data are not available.

Digital Elevation Model (DEM) is main input for watershed modelling. Recently, (DEM) is available online for free in different accuracies, and spatial resolutions as a product of several remote sensing satellites. Hence, it is necessary to find out which one is the best for watershed modeling in the study area. In this study, the different accuracies 30m spatial resolution DEMs of (Copernicus, SRTM, and ASTER) can be examined by using Remote sensing (RS) and Geographic Information systems (GIS) techniques to delineate and calculate the topographic characteristics for five different size and topography watersheds (Swaidy, Garlond, Khuwayr Hirah, Naqab, and Kalak) located on both sides of the Mosul reservoir in the northeastern part of Iraq. The analysis results can be led to find that the Copernicus (GLO-30) 30 m resolution DEM is the optimum and most accurate DEM in the selected study area at the vertical accuracy (1.3521 m) and with 95% confidence level is (2.6502 m) represented by the minimum Root Mean Square Error (RMSE) of the elevations differences between check points (MOWR elevations and LIDAR DEM 1 m resolution) and the used DEMs. The watershed delineation and calculated topographic characteristics (watersheds boundary, elevations, area, perimeter and slope areas) are affected by DEM accuracy. Where the considerable accuracy of the differences is with the (Copernicus-SRTM) DEMS at the minimum RMSE of watersheds characteristics.

To accurately assess landslide susceptibility and model debris flow paths, acquisition of high-resolution elevation data is essential. These data enable precise topographical analysis, considering factors like slope and curvature. But most of the time such high-resolution data will have inaccuracies due to vegetation, especially in tropical region. This study proposes a comprehensive approach that integrates field data and Structure from Motion (SfM) technology to rectify such DEM inaccuracies caused by dense vegetation. The goal is to enhance the accuracy of landslide simulations and volumetric analysis for effective post-disaster management. The Western Ghats, running parallel to India's western coast, has experienced a surge in rainfall-induced landslides, resulting in significant casualties in recent years. Notably, the Pettimudi village landslide in 2020 caused 70 deaths. The unique geomorphological features of the Western Ghats, such as concave curvature, colluvium deposits, and evidence of paleo landslides, contribute to the area's susceptibility. The study emphasizes the need for detailed assessment and mitigation strategies. The proposed method aims to improve the post-event high-resolution DEM accuracy by integrating field-collected elevation values and utilizing Agisoft Metashape software with the SfM algorithm. The rectification process involves combining elevation differences measured during a field study with photogrammetric elevation data. The field-collected elevation differences are crucial for rectifying these points, and enhancing the Digital Elevation Model (DEM) accuracy. The inaccessible source region is improved using a SfM-created DEM from drone footage, resulting in a more accurate post-event DEM. Correcting the 1 m DEM using an SfM-generated DEM proves challenging but significantly improves detail and accuracy, especially over the landslide source area. The impact of the rectification on accuracy is assessed by comparing volumes calculated from the initial DEM and the newly corrected DEM. The difference in volumes of debris depletion and accumulation, computed using initial and corrected DEMs, highlights variations, with depleted volumes being significantly larger due to the extensive depth increase over the landslide source region. These volumes are then utilized in Rapid Mass Movement Simulation (RAMMS) to validate the rectification process and enhance landslide impact predictions. Debris flow simulations in RAMMS, utilizing the rectified 1 m DEM, show specific outcomes at the landslide source region and toe of the landslide. The study emphasizes the importance of integrating SfM technology with field data to improve DEM accuracy, acknowledging the significance of additional field data for further refinement. The potential adoption of highly precise aerial imagery from Unmanned Aerial Vehicle (UAV) surveys is suggested to further enhance the SfM DEM. Debris flow modeling with RAMMS serves as a vital step in validating the accuracy and reliability of the rectified elevation model. Keywords: Landslides, Western Ghats, Pettimudi, Digital elevation model (DEM), Structure-from-Motion (SfM) technology, Rapid Mass Movement Simulation (RAMMS)

Satellite-derived Digital Elevation Model (DEM) selection, preparation and correction for hydrodynamic modelling in large, low-gradient and data-sparse catchments

Digital Elevation Model (DEM) errors tend to be spatially correlated, inevitably affecting DEM-based topographic change detection. Traditional topographic change detection methods often ignore the spatial distribution of the DEM error. This paper aims to develop a workflow that considers the spatial autocorrelation of the error in topographic change detection. Firstly, the DEM of Difference (DoD) is obtained from two-period DEMs, and the Monte Carlo method is employed to evaluate the Spatially Distributed Errors (SDE) in DEMs. Secondly, DoD errors are calculated by propagation based on spatially distributed DEM errors. At the same time, its spatial distribution is quantified using the semi-variance function. Finally, topographic changes (erosion, deposition, and net changes) are calculated based on the spatial distribution analysis and significance detection. The results in two small catchments indicate that DEM errors are spatially correlated, increasing the volume calculation errors. However, using Standard Deviation of Errors (SDE) instead of Root Mean Square Error (RMSE) can effectively reduce the sensitivity of the detection results in the significance threshold. When the significance threshold increases from 68% to 95%, the observations loss using the spatially distributed error is 4.67% −6.92% lower than that using the RMSE. The level of detection has little impact on the net topographic change and significantly influences gross erosion and deposition. In particular, the use of level of detection can effectively reduce the misclassification of erosion or deposition in stable topography areas. The proposed method can be effectively utilized in various applications like surface deformation monitoring, erosion monitoring, and sediment transport assessment.

Abstract. The Antarctic Peninsula (AP) is one of the widely studied polar regions because of its sensitivity to climate change and potential contribution of its glaciers to global sea level rise. Precise digital elevation models (DEMs) at a high spatial resolution are much demanded for investigating the complex glacier system of the AP at fine scales. However, the two most recent circum-Antarctic DEMs, the 12 m TanDEM-X DEM (TDM DEM) from bistatic interferometric synthetic aperture radar (InSAR) data acquired between 2013 and 2014 and the Reference Elevation Model of Antarctica mosaic (REMA mosaic) at an 8 m spatial resolution derived from optical data acquired between 2011 and 2017 have specific individual limitations in this area. The TDM DEM has the advantage of good data consistency and few data voids (approx. 0.85 %), but there exist residual systematic elevation errors such as phase-unwrapping errors in the non-edited DEM version. The REMA mosaic has high absolute vertical accuracy, but on the AP it suffers from extended areas with data voids (approx. 8 %). To generate a consistent, gapless and high-resolution topography product of the AP, we fill the data voids in the TDM DEM with newly processed TDM raw DEM data acquired in austral winters of 2013 and 2014 and detect and correct the residual systematic elevation errors (i.e., elevation biases) in the TDM DEM with the support of the accurately calibrated REMA mosaic. Instead of a pixelwise replacement with REMA mosaic elevations, these provide reference values to correct the TDM elevation biases over entire regions detected through a path propagation algorithm. The procedure is applied iteratively to gradually correct the errors in the TDM DEM from a large to small scale. The proposed method maintains the characteristics of an InSAR-generated DEM and is minimally influenced by temporal or penetration differences between the TDM DEM and REMA mosaic. The performance of the correction is evaluated with laser altimetry data from Operation IceBridge and ICESat-2 missions. The overall root mean square error (RMSE) of the corrected TDM DEM has been reduced from more than 30 m to about 10 m which together with the improved absolute elevation accuracy indicates comparable values to the REMA mosaic. The generated high-resolution DEM depicts the up-to-date topography of the AP in detail and can be widely used for interferometric applications as well as for glaciological studies on individual glaciers or at regional scales.

Using the physical deterministic model to analyze landslide stability has become a hotspot of landslide disasters research all over the world. The Digital Elevation Model (DEM) resolution has a great influence on the simulation effect of 3 D physical models. However, few researchers have studied the prediction performance of the 3 D models under different DEM resolutions. Therefore, based on the 3 D model Scoops3D, the spatial distribution of landslides was simulated and predicted under five different DEMs resolutions (2.5 m, 5 m, 10 m, 20 m, and 30 m). The optimal parameters of the model were obtained through field investigations and laboratory experiments, and then, the simulation results were compared with the actual landslides distribution. Receiver operating characteristic (ROC) analysis and %LRclass index were used to quantitatively evaluate the prediction performance of the 3 D model under five different DEMs resolutions. The results show that Scoops3D has good performance in landslide spatial distribution prediction. In addition, we also found that the simulation results of high-resolution DEM were not ideal, while the prediction results of medium resolution DEMs (i.e., 5 m and 10 m) were more accurate. Therefore, this study provide a reference to select the most suitable DEM resolution for landslide stability analysis.

Many Digital Elevation Model (DEM)-based methods have been proposed to enable automated gully mapping. However, the accuracy of a gully map derived from a DEM is inevitably affected by the DEM resolution. Related literature evaluated the accuracy of DEMs generated from stereo imagery of different satellite sensors; however, limited work has been done on quantitative analysis concerning the capability and accuracy of satellite DEMs in Gully erosion mapping. Therefore, this study was centered to map gully erosion and to investigate the accuracy of measured heights of gully erosion using field acquired data (RTK GNSS) and Global Satellite DEMs obtained from ALOS World 3D 30m and 30m ASTER GDEM. The Etim Umana Gully Erosion site in Uyo Local Government Area of Akwa Ibom State, Nigeria was used as the test site for this study. These datasets were referenced to orthometric height, therefore were converted to ellipsoidal height using the Earth Gravitational Model (EGM) 2008. Digital Elevation Model of the field data were generated using various kriging method (simple, ordinary, universal) of which the ordinary method assumed the best with the least Root-Mean-Square Error(RMSE) of 1.326199. Ground Control Points (GCP) (32) values were used to validate the accuracy of the field acquired data and obtained a RMSE of 1.125463 and a Mean Absolute Error (MAE) of 0.8. The spot height values obtained from field were used to validate the satellite DEMs by calculating their RMSE from the difference between the measured values and the predicted values. The ALOS had a RMSE of 10.89154 when compared to the ASTER which had the least RMSE of 10.71999 which indicates the propagation of error. The Mean-Absolute-Error was calculated recoding MAE of 8.539147 for ALOS and ASTER 9.346781 and a Standard deviation of 9.176663 for the ALOS and ASTER 10.53712. ALOS had a mean error of -8.5174 and ASTER 7.706339. The result from this research indicates that they are similar levels of error propagation between the ALOS and ASTER and this can further be ascertained by carrying out validation on different terrain surfaces. These results indicate that it can be map to scale with contours based on the specifications of the National Mapping Accuracy Standards (NMAS)

Interferometric synthetic aperture radar (InSAR) processing, if lacking high quality ground control points (GCPs), may produce large errors in the final DEM. Some kind of alignment or registration can reduce these errors. This paper evaluates the accuracy of InSAR processed digital elevation models (DEM), against a high resolution DEM. InSAR DEMs were aligned and least-squares registered with USGS, 1/3 arc second NED. The accuracy evaluations after registration show InSAR DEMs alignment and registration can eliminate DEM errors caused by lack of accurate GCPs. In some cases, this process improves the accuracy of the InSAR DEM even more efficiently than including GCPs into InSAR processing.

GIScience Research at the Thirty‐second Annual Esri International User Conference

<p>The ability of identifying –based on numerical analysis– disconnected areas –in terms of overland flow pathways– depends on the digital elevation model (DEM) resolution, type of flow accumulation algorithm and DEM accuracy. On the other hand, tillage practices (in lowlands) and terrain preparation (at any slope gradient) may condition the occurrence of permanent/ temporal disconnected areas. In this study, the effect of DEM resolution and the presence of a drainage ditch and forest trails on the number, location and characteristics of disconnected areas is evaluated in a steep (mean slope gradient of 29%) farmland area of the Spanish Pyrenees. A new vineyard plantation (3785 m<sup>2</sup> and 5120 m<sup>2</sup> including the transit area; espalier system) and its upslope drainage area are evaluated. This site is located near Barbenuta village (Huesca province), at high elevation (1184-1260 m a.s.l.). Abandoned terraced fields and patches of natural vegetation (trees and shrubs) occupy the upslope area, where several forest trails cross from east to west. To protect soil against water soil erosion, farmers built a drainage ditch (total length of 137 m; ca. 0.30 m width; ca. 0.15 m depth) upslope the vineyard boundary, which minimizes runoff entrance into the field. A professional drone (senseFly© eBee X) was used to obtain –after point cloud processing– Structure-from-Motion (SfM)-derived DEMs at different spatial resolution, namely: 1, 0.5, 0.2, 0.1 and 0.05 m. We used combined information of the DEMs before and after filling the local sinks. As expected, the number (n=34, 341, 1079, 1272 and 1907) and size (mean=500, 60, 21, 18 and 12 m<sup>2</sup>; median=68, 15, 5, 4 and 2 m<sup>2</sup>; σ=920, 178, 69, 71 and 49 m<sup>2</sup>) of sub-basins increased and decreased, respectively, with decreasing the pixel size, due to fractal geometry and higher influence of micro-topography components (e.g. soil roughness, random local sinks) –higher ratios of 'residual topography (σ of slope) / pixel size': 0.2 (at coarser resolution), 1.8, 20.3, 113.6 and 636.8 (at finer resolution)–. The total area also varied with the different DEMs: 17010, 20514, 22398, 22852 and 22807 m<sup>2</sup>. The number (n=21, 292, 903, 928 and 1283) and area (41, 143, 118, 58 and 44 m<sup>2</sup>) of disconnected areas increased and decreased, respectively, with decreasing the pixel size, representing 0.24%, 0.70%, 0.53%, 0.25% and 0.19% of the total drainage area. Similar differences were observed in other topographic metrics like the drainage-boundary perimeter and maximum flow length. These results prove the impossibility of defining a unique overland flow pattern. Further research should be focused on the role of runoff depth and how the effect of man-made landscape elements (drainage ditch, forest trail) and practices (tillage) on disconnectivity may depend on rainfall depth and intensity, and indirectly on plant growth.</p>

The aim of this work is to develop an object-based methodology for semi-automatic geomorphological mapping. The study area is the municipality of São José dos Campos, located in the Paraíba do Sul Valley, São Paulo State, Brazil. A multilevel hierarchical semantic network driven by a set of multispectral, geomorphometric and texture variables is used. The geomorphometric and texture variables are extracted from a digital elevation model (DEM) generated from a stereo pair of advanced spaceborne thermal emission and reflection radiometer (ASTER)/Terra images. The DEM accuracy is evaluated and its root mean square error (RMSE) is found to be 9.38 m. The validation of the geomorphological map is accomplished by means of object-based statistical indices derived from an error matrix, obtained by a comparison between the classified scene and a reference map produced by visual interpretation. The object-based approach is found to be effective for geomorphological mapping.

Satellite sensors provide currently reliable elevation data over large areas, with spatial resolutions below 100 m and reported accuracies of approximately 10 m. However, the vertical accuracy of these quasi-global digital elevation models (DEM) varies depending on the terrain and land cover, decreasing in mountain areas, and may influence the quality of products derived from these data, like stream channel network or relief forms. In this paper, we investigated two global elevation data sets: SRTM-X DEM and ASTER GDEM, assessing their vertical accuracy and reliability of terrain features extracted automatically from the elevation data. The research was carried out in the Pieniny Mts., a small mountain range in southern Poland, with elevations ranging from about 400 m to 1000 m a.s.l. One particular attribute of this mountain range that may create several problems in deriving accurate terrain features is a deep, narrow and windy gorge of the Dunajec river. We found that the vertical accuracy of SRTM-X DEM is significantly better than of the ASTER GDEM (Root Mean Square errors of 12.05 m and 17.43 m, respectively). Both DEMs allow to extract terrain features with a similar accuracy in areas of moderately steep and gentle slopes, yet fail to properly represent the floor of the Dunajec gorge as relative vertical errors along the river may exceed 100 m. Therefore, we conclude that while both models provide statistically reliable description of the terrain surface, they require additional corrections for hydrological or terrain modeling.