**5. Results**

The arable land is influenced by the intensive erosion processes in the Svacenický Creek catchment. The most vulnerable areas are located at the bottom of the slopes and in the steepest parts of particular slopes (Figure 7a). The intensity of the erosion is often higher than 5.0 t/ha/yr. Together with inappropriate crop management (corn), these localities are exposed to extreme intensities of erosion processes (>10 t/ha/yr). Wide-row crops such as corn do not represent such a threat on short slopes as in the right part of the Svacenický Creek catchment. Nevertheless, narrow-row crops such as wheat or lucerne result in a lower rate of soil degradation (3–5 t/ha/yr). The intensities of the erosion look similar in both periods of time (Figure 7a). Nevertheless, there is a difference, which is more notable in the case of the volume of sediment (Figure 7b).

**Figure 7.** Results from the EROSION-3D model (calculated for a function A): (**a**) Erosion intensity; (**b**) Volume of sediment.

The intensive erosion processes are caused by intensive rainfall events. Table 5 presents the results for the rainfall-erosion events selected with a total rainfall amount higher than 12.5 mm [35–37]. In the first time period investigated (2015–2016), the rainfall-erosion events produced almost 23% of the eroded material compared to the total production of the time period in Table 6. On the other hand, only 2.4% of the sediments were produced by rainfall-erosion during the second time period investigated (2016–2017). The difference can be found in the total number of rainfall events, because the period 2016–2017 was richer in rainfall than the period 2015–2016, which made this period much drier and the soil more vulnerable to the erosion process.


**Table 5.** Results from the EROSION-3D model calculated for the erosion-rainfall events (total rainfall amount >12.5 mm; [35–37] and function A.


**Table 6.** Comparison between the bathymetric measured data and modelled data of the sediment budget in the Svacenický reservoir (2015–2017).

(1) Calculated for the initial soil moistures from the function A in Figure 4. (2) Calculated for the initial soil moistures from the function B in Figure 4.

The production of the eroded material is connected with areas of intensive erosion processes (Figure 7a). The total predicted sediment volume was 216.5 m<sup>3</sup> in the period 2015–2016 and 375.8 m3 in the period 2016–2017. The higher amount of sediment in the period 2016–2017 correlates to the higher total number of rainfall events compared to the period 2015–2016. Comparing the predicted (modelled) and observed (bathymetric measured) sediment volume in the Svacenický reservoir, the EROSION-3D predicted a lower amount of sediments (Table 6). More specifically, the predicted volume of sediment (calculated for a function A) in the period 2015–2016 is 76% lower and 26% lower in the period 2016–2017 than the observed volume of sediment in the reservoir. Based on the function B, the production of sediment increased, and the modelled volume of sediment is even higher in the period 2016–2017 compared to the bathymetry. It must be noted that the actual measured volume of sediment was reduced by 56.5% because of the assumed water content in the reservoir sediment [38].

The eroded material reached the catchment outlet and the amount was measured by the AUV EcoMapper (YSI Company, Yellow Springs, OH, USA). The measured sediment volume was reduced by considering a sediment water content of 56.5% according to [38].

## **6. Discussion**

The characteristics such as land use, and crop management (type of crop, crop rotation) belong to temporal dynamic features, while attributes such as slope morphology or type of soil represent relatively constant catchment characteristics [19]. The crop distribution with a higher amount of winter wheat pointed to a lower volume of sediments and generally less endangered soil [39]. The probable causes are the lower values for the parameters skin factor, surface roughness, and erosion resistance in comparison with corn. The higher amounts of the surface runoff (m3), the erosion and deposition rates (t/ha), and the volume of sediment (m3) in the Svacenický Creek catchment detected for the second period were the result of two main causes. The first one is the higher total number of rainfall events. The second cause lies in the crop management of the study area, where corn represents 54% of the arable land. The combinations of the effects of the management systems and rain conditions resulted in the higher amounts of erosion processes in the catchment. The type of crop has a great influence on the erosive process and soil losses [40]. There is a direct connection between the crop management system and soil losses; therefore, the crop management system is one of the main factors affecting soil erosion by water [40]. Appropriate land management can significantly help reduce erosion processes [41–43]. In the case of model parameters, the biggest influence on runoff and erosion processes has bulk density, soil moisture and skin factor. The skin factor is the parameter used to reduce the prediction error resulting from the simplified assumptions of the model because the EROSION-3D model considers simplifications of homogenous soil matrix, but the process of infiltration is influenced by many different factors, i.e., soil compaction, soil crusting, surface soil pores, biological activity due to rodent burrows or worm and it is necessary to include correcting parameter.

Although erosion resistance and hydraulic roughness are considered as important factors, it has been found that they do not affect model results to a significant extent. However, the general trend of decreasing resistance of erosion with increasing erosion intensity has been confirmed by experimental results [44,45].

The results also pointed to the importance of rainfall-erosion events that are able to result in high erosion of soil [46]. In this case, the role of such events could increase in the future because of the predicted increase in extreme rainfall events due to climate change [4,46–50]. According to the Slovak Hydrometeorological Institute, the number of rainfall events with durations of 5 to 240 min has been increasing during the last decade, while rainfall-erosion events occur 2.5 to 4.7 times per year; this amount is expected to increase in the future, especially during the spring and summer months [51,52].

When testing the application of the EROSION-3D model, some difficulties were found during the appropriate determination of the input parameters and when verifying the model results through the actual measured data. A model dependability heavily depends on its calibration [53] and the testing of a physically-based erosion model is considered as a necessary part of any scientific work in order to develop an understanding of what the model will demonstrate [54].

Due to the testing process, the strengths and weaknesses of a model can be discovered. According to [55], long-term results are generally simulated in the best possible way. Here, the results presented pointed at a notable difference between the modelled and observed sediment yields, which indicates the difficulty posed by the model simplification of the natural phenomenon of erosion [18].

As it is mentioned in [56], mathematical simulation models consist of three sub-models: (a) Rainfall-runoff submodel, (b) a soil erosion sub-model and (c) a stream sediment transport sub-model. All three sub-models should be considered, but in this case, it was not done, because relevant data for stream sediment transport sub-model was not possible to obtain. In spite of a fact that small streams exist in the polder basin, significant flow occurs only in case of intense precipitation. This fact largely limits the acquirement and measurement of the relevant data needed to calculate the flow of streambed erosion. In addition to that, it can be seen (Figure 2a), that majority of these small streams flow in low slope area, so flow velocity is also low and so, frame velocity is a weak one as well. So, for this reason, we assume this type of erosion has not to be taken into account. Besides all, the used simulation model cannot simulate it as well.

However, this part of the available sediments in the catchment should be taken into account [56], but because of the conditions mentioned above, there was not done. On the other hand, during the three-years monitoring time there was found depositions of sediment in one inlet part of the reservoir (the longest contributing creek), that confirm flow stream erosion, but their volume is not significant. From all of those, it can be supposed that omission of streambed erosion could be a part of differences between calculated and measured /observed sediment volume data.

As is mentioned, based on the field measurements of the bottom bathymetry which started in 2015, the current status of the clogging of the reservoir was evaluated. The results confirmed our theories about the on-going sedimentation processes in the Svacenicky Creek reservoir. According to the analysis, we determined that during the years 2012–2019, over 10.4285 m<sup>3</sup> of sediment on the area of the Svacenicky Creek catchment have accumulated and the polder volume capacity has decreased 6%. The average sediment transport during the years was estimated to be 1400 t/year. The sediment transport has lowered since 2016, which can be e.g., due to lower precipitations as also flood protection measures adopted in this area. According to Table 6, the lifetime of the reservoir varies between 303.7 years for observed sedimentation, 315.0 years for predicted sedimentation B and 728.7 years for sedimentation A.

Possible sources of errors in the predicted sediment volume by the EROSION-3D model can mostly be associated with the model parameters, values of the initial soil moisture before each simulated event, and the grid size of the catchment area. The parameters of the EROSION-3D model as the resistance to erosion, surface roughness, and skin factor were chosen from the Parameter Catalogue for EROSION-3D, which contains their tabularized values by the type of soil, land use and the specific crop and its growth phase in different months within the year. Because of the attempt to continual modelling based on the modelling of a sequence of individual rainfall-erosion events, it was not

possible to calibrate these parameters for each event. For the continual modelling, the crucial input parameters were the values of initial soil moisture before each simulated rainfall-erosion event. The values of initial soil moisture were estimated from the relationships between the soil moisture and the previous precipitation index, developed experimentally for the Plzen region in the Czech Republic (Figure 4, function A) and for the experimental plots in the Myjava region in Slovakia (Figure 4, function B). The data of initial soil moisture estimated from the function developed for the Myjava region in Slovakia, which is more representative for the case study area, improved the predictive sediment volumes in comparison with the measured sediment volumes—the relative error decreased to 40% for the period 2015–2016 and 30% for the period 2016–2017. On the other hand, the function B was developed only from seven pairs of experimental data, and we suppose that the extension of the data and improvement of this function would improve the performance of the model.

Also, the use of bathymetry to estimate trapped sediment in a reservoir brings out other uncertainties. According to [23], several problems with this method can be observed, e.g., the unknown effectiveness of the reservoir, the importance of a detailed analysis of the sediment cores; the precise location of sources of the sediment; the size of the reservoir and related discharges; and depreciation of the organic fraction of the sediment due to the measurements. Nevertheless, the bathymetric method is arguably the best way to determine the volume of sediment and can be carried out by several approaches [23,24,57]. The use of the autonomous underwater vehicle (AUV) to investigate the temporal evolution of changes in the bottom of the reservoir through the DEM was approved in this paper.
