A deep learning-based super-resolution DEM model for pluvial flood simulation

Abstract. The rapid progress of lidar technology has made the acquirement and application of high-resolution digital elevation model (DEM) data increasingly popular, especially in regards to the study of floodplain flow. However, high-resolution DEM data pose several disadvantages for floodplain modeling studies; e.g., the data sets contain many redundant interpolation points, large numbers of calculations are required to work with data, and the data do not match the size of the computational mesh. Two-dimensional (2-D) hydraulic modeling, which is a popular method for analyzing floodplain flow, offers highly precise elevation parameterization for computational mesh while ignoring much of the micro-topographic information of the DEM data itself. We offer a flood simulation method that integrates 2-D hydraulic model results and high-resolution DEM data, thus enabling the calculation of flood water levels in DEM grid cells through local inverse distance-weighted interpolation. To get rid of the false inundation areas during interpolation, it employs the run-length encoding method to mark the inundated DEM grid cells and determine the real inundation areas through the run-length boundary tracing technique, which solves the complicated problem of connectivity between DEM grid cells. We constructed a 2-D hydraulic model for the Gongshuangcha detention basin, which is a flood storage area of Dongting Lake in China, by using our integrated method to simulate the floodplain flow. The results demonstrate that this method can solve DEM associated problems efficiently and simulate flooding processes with greater accuracy than simulations only with DEM.

Soil erosion in the agricultural area of a hill slope is a fundamental issue for crop productivity and environmental sustainability. Building terrace is a very popular way to control soil erosion, and accurate assessment of the soil erosion rate is important for sustainable agriculture and environmental management. Currently, many soil erosion estimations are mainly based on the freely available medium or coarse resolution digital elevation model (DEM) data that neglect micro topographic modification of the agriculture terraces. The development of unmanned aerial vehicle (UAV) technology enables the development of high-resolution (centimeter level) DEM to present accurate topographic features. To demonstrate the sensitivity of soil erosion estimates to DEM resolution at this high-resolution level, this study tries to evaluate soil erosion estimation in the Middle Hill agriculture terraces in Nepal based on UAV derived high-resolution (5 × 5 cm) DEM data and make a comparative study for the estimates by using the DEM data aggregated into different spatial resolutions (5 × 5 cm to 10 × 10 m). Firstly, slope gradient, slope length, and topographic factors were calculated at different resolutions. Then, the revised universal soil loss estimation (RUSLE) model was applied to estimate soil erosion rates with the derived LS factor at different resolutions. The results indicated that there was higher change rate in slope gradient, slope length, LS factor, and soil erosion rate when using DEM data with resolution from 5 × 5 cm to 2 × 2 m than using coarser DEM data. A power trend line was effectively used to present the relationship between soil erosion rate and DEM resolution. The findings indicated that soil erosion estimates are highly sensitive to DEM resolution (from 5 × 5 cm to 2 × 2 m), and the changes become relatively stable from 2 × 2 m. The use of DEM data with pixel size larger than 2 × 2 m cannot detect the micro topography. With the insights about the influencing mechanism of DEM resolution on soil erosion estimates, this study provides important suggestions for appropriate DEM data selection that should be investigated first for accurate soil erosion estimation.

Regular-grid Digital Elevation Model (DEM) data are fundamental for viewshed analysis, with direct effects on both computational accuracy and efficiency. High-resolution DEMs yield more detailed results, but computational costs can be prohibitively high, creating a dilemma of balancing computational accuracy and efficiency. To address this, we propose a multi-resolution viewshed (MRVd) method that couples varying resolutions of terrain data based on distance from the viewpoint: high-resolution DEM data are used near the viewpoint, while low-resolution DEM data are used further away. Experiments conducted across mountainous, plain, and urban terrains show that MRVd can achieve better performance than fixed-resolution viewshed approaches with only marginal additional computation time. Accuracy improvements were 2–3 times on average and up to 6–7 times in some cases. Furthermore, MRVd effectively mitigates chunk distortion by leveraging the localized high-resolution DEM data, producing results similar to those obtained using uniformly high-resolution DEM data. These findings suggest that coupling multi-resolution terrain data provides a scalable, resource-efficient computational strategy to balance accuracy and efficiency in viewshed computation. The MRVd approach enhances the robustness and reliability of viewshed calculation for practical applications in landscape assessment, spatial planning, archaeology, and related fields.

The Digital Elevation Model (DEM) is an important basic datas for analyzing and exploring the natural environment, hydrology, and geology. Since the cost of improving the DEM accuracy directly from the hardware device is high, it is envisaged to improve the DEM accuracy from the algorithm. Deep learning provides new research ideas for DEM super-resolution. Inspired by the use of implicit neural representation in image super-resolution, this paper combines implicit neural representation with DEM super-resolution, and proposes a DEM super-resolution algorithm based on implicit neural representation, which aims to convert low-resolution DEM (LRDEM) into high-resolution DEM (HRDEM). This paper presents the process of super-resolution of DEM data based on implicit neural representation. The experiment first maps the cropped normalized data to a high-dimensional space through Fourier transform, and then uses the encoder to extract features from the DEM data to generate a feature sequence. The decoder takes the feature sequence and DEM coordinates as input, and then predicts the coordinates of the DEM value. This paper trains and adjusts a super-resolution implicit neural expression model that generates DEM data through a large number of experiments, where the loss function is the norm loss of the predicted value and the true value. The experimental test results are compared with traditional interpolation methods, such as bicubic interpolation, nearest neighbor interpolation and bilinear interpolation. According to the evaluation index data of MSE and PSNR, arbitrary multiple super-resolution based on implicit neural expression can effectively solve the DEM super-resolution problem.

Digital elevation models (DEMs) are important in flood modelling as they present surface information and obstacles giving the direction of water during a flood event. All hydraulic modelling software needs both channel and floodplain elevation data to simulate the water flow from upstream to downstream. Based on the accuracy and representative capacity of the surface, modelling results can be evaluated in terms of hazard and risk analysis. Developments in remote sensing and GIS technology provide many sources to freely obtain DEM data of different resolution and accuracy. However, access to and the processing of high-resolution data associated with large areas is difficult and time-consuming. This study, therefore, focuses on determining the availability of different resolutions and DEM sources in flood modelling by testing unmanned aerial vehicle (UAV) data. For this, freely available DEMs such as ALOS DEMs (12.5 and 30 m), EU-DEM (25 m), ASTER GDEM (30 m), SRTM (30 and 90 m), MERIT DEM (90 m) and TANDEM-X (90 m) were chosen to investigate part of Biga River (Turkey) which features a natural channel. Then, 1D flood modelling was examined using different DEM resolutions, Manning’s values and flood data of various frequencies (10, 100 and 1000 years). Results were tested using high-resolution DEM data (5 cm) gathered from a UAV. It was concluded that EU-DEM is the most suitable dataset in terms of flood extent, depth and velocity. However, depending on the boundary characteristics of the data, ALOS 12.5 DEM and MERIT DEM (90 m) data are the most suitable in their extent, depth and velocity for high-resolution and low-resolution flood modelling worldwide extent.

Separability of soils in a tallgrasss prairie using SPOT and DEM data

An initial assessment of interstate runoff impacts to a receiving water body yielded approximately 2 years of groundwater, surface water, and streamflow measurements upstream and downstream of a section of an interstate within a rural/forested watershed. It became clear that stormwater runoff is larger downstream of the interstate, but analysis of area-averaged runoff datasets have resulted in a variety of inconclusive findings. In order to better understand the hydrology of the interstate and its associated flow features (medians, ditches, realignment of channels, compacted soils, etc.) a detailed hydrological model (Gridded Surface Subsurface Hydrologic Analysis (GSSHA)) is being set up with high-resolution digital elevation model (DEM) data, soil characteristics, and coupled with measured field data. Research questions include: Do the interstate’s features (ditches, median, soil types) encourage or limit infiltration, flow velocity, and groundwater fluctuations? Of interest is the variation in surface-water groundwater interactions at the interstate stream crossing, as well as interactions within the interstate median, interception of subsurface runoff by ditches, permeability of the interstate subsurface (as compared with the surrounding area), and the runoff-rainfall ratio of the areas. To answer these questions, a resolute, physics-based model (GSSHA) is used to resolve the varying hydrologic features of the interstate. By resolving the interstate runoff features, work in the future may be done including comparing the models effectiveness of predicting land use change impacts to other widely used and available hydrological models (such as Stormwater Management Model (SWMM5)).

The spatial variability of input parameters plays a crucial role in the interpretation of geomorphic indices, with digital elevation models (DEMs) being the primary data source. However, the influence of DEM resolution on these indices has rarely been investigated. This study investigated the influence of DEM resolution on the assessment of tectonic activity using the normalized stream length–gradient (SLk) index, which reflects variations along river profiles. The SLk index is sensitive to changes in river gradients that may result from active faulting or differential uplift, making it a valuable tool for identifying zones of active tectonic deformation. Therefore, understanding the impact of DEM resolution on SLk analysis is critical for accurately detecting and interpreting subtle tectonic signals, particularly in intraplate regions where deformation is slow and geomorphic expressions are faint and discontinuous. By comparing high-resolution LiDAR-derived DEMs (L-DEMs) and low-resolution topographic map-derived DEMs (T-DEMs), we analyzed the SLk index distributions along the Yangsan Fault, Korean Peninsula, an intraplate setting with Quaternary activity. According to the results, SLk anomalies derived from L-DEMs had a continuous distribution along the fault, closely aligning with known surface ruptures and indicating active tectonic deformation. In contrast, SLk anomalies derived from T-DEMs were sporadic and less continuous, especially in low-relief landscapes such as alluvial fans and floodplains, highlighting the limitations of T-DEMs in detecting fault-related features. High-resolution DEMs were better able to capture finer-scale geomorphic features, such as fault scarps, deflected streams, and lineaments associated with active tectonics, providing a more comprehensive view of fault-related deformation. This discrepancy highlights the importance of resolution choice in tectonic assessments, as low-resolution DEMs may underestimate the tectonic activities of intraplate faults by missing subtle topographic variations. While the choice of DEM resolution may depend on study area, scope, and data availability, high-resolution DEMs are critical for identifying tectonic activity in intraplate regions where geomorphic features of faulting due to slow deformation are subtle and dispersed.

Effects of DEM sources on hydrologic applications

In mountainous areas, surface solar radiation (SSR) exhibits high spatiotemporal variation at different slopes and aspects due to its great topographic relief. To get mountain SSR spatial distribution, remote sensing-based methods have been popularly used, which separate SSR into direct solar radiation, diffuse sky radiation, and adjacent terrain radiation. However, the methods are highly depended on the atmospheric and angular information derived from different data sources. To clearly address the uncertainties associated with the estimation, this letter conducted a preliminary comparison study by using different atmospheric transmittance models and digital elevation model (DEM) data to retrieve SSR in the Mt. Gongga region. The comparison results indicated that the uncertainty of the atmospheric constituent data greatly limited the performance of the physical atmospheric transmittance models. The resolution of DEM data also played an important role in SSR determination because of the determination of surface angular information. High-resolution (30-m) DEM data showed better performance than low one (90 m). In addition, the systematic underestimation of SSR estimation with the empirical model was significantly improved by using the averaging method with nearby pixel values. It indicated that the geometric errors of satellite image and DEM data should be considered in the estimation.

Using low resolution digital elevation model (DEM) data, reconstruction of cartographic contours is focused on. With a small number of control points, we aim to draw a smooth curve as well as to compute the area and perimeter of a closed contour accurately. The input data forms an oriented contour, each control point of which consists of three values: x, y, and tangent angle. An interpolation method based on arc spline is developed. The method can reproduce G1 continuous curves,in which the exact area and perimeter can be calculated. In a numerical experiment for analytically defined curves, accurate computation of the area and perimeter was achieved with a smaller number of control points. In another experiment using a DEM data, the reconstructed contours were smoother than those by a conventional method.

Deep learning-based automated terrain classification using high-resolution DEM data

Advancement in landslide detection can be largely attributed to the introduction of deep learning, particularly semantic segmentation. Susceptible regions can be identified using satellite imagery and Digital Elevation Model (DEM) data. This study explores multi-modal data to improve the identification and detection of landslides. The U-Net model serves as a baseline that is further enhanced by the introduction of an attention mechanism that refines pixel-level predictions. Evaluation of various loss functions resulted in increased performance optimization. The Bijie landslide dataset, featuring high-resolution satellite images, DEM data, and ground truth masks, was used for training and evaluation. Precision, recall, F1 score, accuracy, mean intersection over union (mIoU), and Area Under the Curve (AUC) were metrics used to evaluate the performance. The model incorporating the attention mechanism achieved the highest mIoU of 0.76, F1 score of 0.74, and accuracy of 0.94, surpassing the base model. Attention mechanisms concentrate on critical regions and thus improve feature extraction by enhancing segmentation precision. The integration of multi-modal data and diverse loss functions contributes to better landslide detection.

Terrain parameters like topographic horizon and sky view factor (SVF) are used in numerous fields and applications. In atmospheric and climate modelling, such parameters are utilized to parameterise the effect of terrain geometry on radiation exchanges between the surface and the atmosphere. Ideally, these parameters are derived from a high-resolution digital elevation model (DEM), because inferring them from coarser elevation data induces a smoothing effect. Computing topographic horizon with conventional algorithms is however slow, because large amounts of non-local terrain data have to be processes. We propose a new and more efficient method, which is based on a high-performance ray tracing library. By applying terrain simplification to remote topography, this allows the application of the new algorithms also with very high-resolution (< 5 m) DEM data, which otherwise would induce an excessive memory footprint. The topographic horizon algorithm is accompanied by a SVF algorithm, which was verified to work accurately for all terrain – even very steep and complex one. We compare the computational performance and accuracy of the new horizon algorithm with two reference methods from literature and illustrate its benefits. Finally, we illustrate how sub-grid SVF values can be efficiently computed with the newly derived horizon algorithm for a wide range of target grid resolutions (1–25 km).

<strong class="journal-contentHeaderColor">Abstract.</strong> Terrain parameters like topographic horizon and sky view factor (SVF) are used in numerous fields and applications. In atmospheric and climate modelling, such parameters are utilised to parameterise the effect of terrain geometry on radiation exchanges between the surface and the atmosphere. Ideally, these parameters are derived from a high-resolution digital elevation model (DEM) because inferring them from coarser elevation data induces a smoothing effect. Computing topographic horizon with conventional algorithms, however, is slow because large amounts of non-local terrain data have to be processed. We propose a new and more efficient method, which is based on a high-performance ray-tracing library. The new algorithm can speed up horizon calculation by 2 orders of magnitude relative to a conventional approach. By applying terrain simplification to remote topography, the ray-tracing-based algorithm can also be applied with very high-resolution (<span class="inline-formula">&lt;5</span> m) DEM data, which would otherwise induce an excessive memory footprint. The topographic horizon algorithm is accompanied by an SVF algorithm, which was verified to work accurately for all terrain – even very steep and complex terrain. We compare the computational performance and accuracy of the new horizon algorithm with two reference methods from the literature and illustrate its benefits. Finally, we illustrate how sub-grid SVF values can be efficiently computed with the newly derived horizon algorithm for a wide range of target grid resolutions (<span class="inline-formula">1</span>–<span class="inline-formula">25</span> km).