Author(s): Rebecca Hinsberger; Alpaslan Yoruk
Linked Author(s): Rebecca Hinsberger
Keywords: Soil erosion rill erosion Govers approach flash flood modeling rain-on-grid modeling HydroAS model
Abstract: 1. Introduction Frequent heavy precipitation events with increasing rainfall lead to flash floods and can result in adverse consequences for humans (IPCC, 2021; Nunes & Nearing, 2011). In addition, soil erosion can damage infrastructure and lead to soil loss. In particular, single heavy precipitation events cause a high proportion of erosion (Parkin et al., 2008). Soil erosion remains a challenge to farmers and other ecosystems. Therefore, predictive models have been developed to forecast and implement measures for reducing soil erosion. Soil erosion modeling has been conducted for nearly 50 years. Empirical (e. g., USLE) and process-based (e. g., WEPP and EUROSEM) models have been developed. Process-based models use a hydraulic approach to estimate the water forces acting on the soil. As stated by Morgan et al. (1998b), soil erosion predictions are only as accurate as the hydraulic results. Regardless, the aforementioned models use simplified hydraulic approaches to calculate the forces acting on the soil. Thus, errors are part of the simulation due to the imprecise hydraulic results. In addition to soil erosion models, hydraulic models have been used to simulate water flow in rivers and overland flows. There is a wide range of models with low levels of accuracy and quick results up to those with higher levels of accuracy that are associated with longer computing times. A balanced relationship between the accuracy and speed can be achieved using two-dimensional (2D) hydrodynamic numerical models that are state-of-the-art models used for flood and flash flood simulations in Germany. These models solve complete shallow water equations and consider the turbulence and acceleration terms. Accurate consideration of hydraulics is important and is associated with an accurate roughness approach. Studies have demonstrated that roughness coefficients influenced by water depth are significant for vegetated surfaces, and constant values are suitable for surfaces with low microreliefs (Hinsberger et al., 2022). The objective of this study is to combine the simulation of erosion on arable land and flash floods that occur as overland flows. Existing soil erosion models consider simplified hydraulics. However, a detailed hydraulic approach using a 2D model is essential for accurate calculations. This study aimed to accurately model heavy precipitation-induced linear and sheet erosion. An erosion approach suitable for overland flow on arable land was selected, implemented in a 2D model, and evaluated using measured natural erosion data. These data were collected during the study and represent rill erosion caused by heavy precipitation at the field scale. The combined novel approach and two existing erosion models were applied to erosion fields. A comparison of the simulation results demonstrated the suitability of the model combination. 2. Materials and Methods The combination of a precise hydraulic approach and soil erosion model can lead to a better estimate of soil erosion quantities and correct mapping of rill distributions. To evaluate this hypothesis, a transport capacity approach suitable for overland flows was selected and implemented in the 2D HydroAS model (Section 2.1). To evaluate this advanced and combined model, erosion due to natural, single heavy precipitation events was recorded and analyzed with respect to the spatial distribution and erosion quantity of the rills (Section 2.2). 2.1. Combined soil erosion and hydraulic modeling The existing transport capacity approaches are derived from experimental data and are valid only for the data ranges in which they were developed and calibrated. The Govers (1990) approach was derived for the overland flow. The slope reached 21 %, and the grain diameter of the soil ranged from 0.058 to 1.098 mm. A comparison of the validity ranges of different transport capacity approaches for the grain diameter and slope parameters clearly indicates that the Meyer–Peter–Muller (MPM) (Meyer-Peter & Muller, 1948), Ackers–White (AW) (Ackers & White, 1973), and Engelund–Hansen (EH) (Engelund & Hansen, 1967) approaches possess validity ranges that differ from those of the Govers approach for both parameters. The MPM, AW, and EH approaches are suitable for low slopes (<4%), and this typically occur under river conditions, whereas the Govers approach reaches 21% and is also suitable for overland flows. Regarding the grain diameter, the Govers material lies in the range of fine sand, whereas the other approaches exhibit a wider range and cover a coarser material. This perception is consistent with that of Wang et al. (2019), who stated that the Gover approach is the best method for simulating cropland soil. Therefore, the Govers approach was selected as the transport capacity approach suitable for the simulation of flash floods. HydroAS is a two-dimensional (2D) hydrodynamic numerical model used for flood and flash flood simulations. It can simulate discharge and rain-on-grid models. An add-on module for river erosion (HydroAS GS/FT) was already implemented in the model. Using this module, detachment, transport, and sedimentation were simulated, and the topography of the model was changed at each time step. In this study, the GS module was extended using the Govers approach to consider soil erosion on arable land. The combined model is referred to as HydroAS GS–Govers. 2.2. Data for plausibility checks and calibration Heavy precipitation often leads to rill and gully erosion. In several studies, erosion and sedimentation quantities and their spread have been investigated using rainfall simulators in limited-size laboratory and in-situ test plots (Aksoy et al. 2013; Polyakov et al. 2018; Romkens et al. 2001; Zhang et al. 2021). However, linear erosion developed from rainfall-induced overland flow and occurred only after a sufficient flow length. Therefore, erosion data that occurred after a single heavy precipitation event were collected and analyzed at the field scale. To collect data of erosion due to heavy precipitation, locally occurring extreme events in arable land areas were detected and inspected on-site. Over 3 years, a total of 456 fields were investigated. A study of the influence of vegetation cover on erosion indicates that erosion occurred only on fields with a cover ≤ 25% (Hinsberger, 2024). When erosion tracks were visible, the croplands were recorded using an unmanned aerial vehicle (UAV) DJI Phantom 4 RTK (P4 RTK). During post-processing, photographs taken by the UAV were used to generate digital elevation models (DEM) and orthophotos of the erosion field. Rill erosion was apparent in the DEMs and orthophotos. Rill areas were identified, and the erosion volume was derived using the DEMs of difference method (DoD) with pre-erosion and recorded DEM. Bulk density can be used to determine the quantity of erosion. In total, aerial surveys and analyses of erosion quantity were conducted for seven cornfields, and 32 rills were analyzed (Hinsberger, 2024). Preliminary tests were conducted using a UAV on arable land to estimate accuracy. The measured heights of the tilled and untilled soils were compared using aerial and terrestrial surveys. Focusing on linear erosion structures, aerial and manual measurements of the rill width and depth were conducted and compared. 3. Results and Discussion Erosion due to single events was recorded using a UAV and analyzed at the field scale to provide data suitable for plausibility checks and calibration. First, the accuracy of the UAV was investigated, and rill erosion quantities and distributions were analyzed based on aerial survey data (Section 3.1). The novel combined model was applied to recorded erosion fields and evaluated by comparing the measurements and simulation results of existing soil erosion models. 3.1. Accuracy of UAV measurements and erosion data The UAV used in this study was a DJI P4 RTK. Comparisons of the accuracy of aerial and terrestrial surveys revealed average absolute deviations of 1.5 cm and 3.9 cm for tilled and untilled soil, respectively. The erosion fields recorded in this study always exhibited erosion due to heavy precipitation on corn or potato fields, and thus, on tilled soil. Therefore, the aerial surveys showed good agreement with the terrestrial measurements. Comparisons of the rill width and depth derived from the aerial surveys with manual measurements demonstrated that the rill width is easily recognizable in the UAV generated DEM, and the measurements fit well with a root mean square error (RMSE) of 10.7 cm for rills that are up to 350 cm wide. However, the rill depth is often underestimated. Here, the RMSE was 2.11 cm for rills that are up to 20 cm deep. Thus, erosion depth can be underestimated by up to 20 % when UAV are used. Therefore, the erosion quantities derived from the UAV indicated the minimum amount of erosion. Based on this fault tolerance, the measured erosion may be higher than the suggested comparative values. This limitation must be considered when evaluating the quantity of erosion in recorded fields. Thirty-two rills in the recorded fields were investigated. Erosion volume, quantity, and rate were calculated using DEMs and orthophotos derived from the aerial survey, and the bulk density. These data were generated to investigate the plausibility of soil erosion models for the application of single heavy precipitation events. 3.2. Evaluation of the HydroAS GS–Govers approach To evaluate the HydroAS GS–Govers model, it was applied to laboratory and plot experiments from the literature and to measurements of natural erosion at the field scale of this study. The results revealed that sheet and rill erosions were simulated in the model. In principle, low erosion rates were well-reproduced in the experimental flumes/plots and at the field scale. For framework conditions in which the grain diameter was higher than the valid range of the Govers approach, low erosion rates (almost zero) were simulated and did not correspond to the measured values. A possible explanation is that the threshold value that is constant in the Govers approach is unsuitable for grain diameters larger than those within the validity range. The erosion rills were analyzed based on their spatial distributions and quantities. The spatial distribution showed good agreement with the orthophotos of natural events. The rills that appeared after natural events were largely reproduced. Comparisons of the simulated and measured rill quantities showed plausible results for small rills and an underestimation for large rills. A sensitivity analysis of different grid resolutions (0.25 m and 1 m) indicated a significant influence of the resolution. Higher grid resolutions tended to overestimate small rills, as the actual rills were smaller than the available grid width. Additionally, the RUSLE2 and EROSION-3D models were applied to the erosion fields to compare the simulation results to those of the HydroAS GS–Govers model. The RUSLE2 model applied to the rill catchment area exhibited similar results to those of the HydroAS GS–Govers model for erosion quantities. However, this empirical model did not provide information regarding the spatial distribution of rills. In contrast, EROSION-3D did not generate as many rills as did the HydroAS GS–Govers model, and it delivered lower erosion rates for the generated rills compared to the measurements and other models. 4. Conclusions The HydroAS GS–Govers model accurately calculates the hydraulics and represents an improvement in the simulation of erosion due to single events. The spatial distribution of the simulated rills was in good agreement with the recorded rills after a natural event, and more rills were generated compared to the existing models. However, the erosion quantity of the large rills was underestimated in the simulation. With the background information on the UAV accuracy that the aerial survey measurements underestimated the real erosion quantities, the simulation results do even more underestimated the naturally occurring erosion. An adaptation of Govers’ approach may overcome this limitation. However, further studies are needed to confirm this hypothesis. Furthermore, the erosion on bare soil was investigated. As mentioned in the Materials and Methods Section, erosion can occur in croplands with a land cover of up to 25 %. For vegetated fields with little coverage, the soil resistance is higher, and the erosion approach must be extended by a vegetation term or the threshold value of the approach must be adapted accordingly.
Year: 2025