Assessing the perturbations of the hydrogeological regime in sloping fens through roads

Roads in sloping fens constitute a hydraulic barrier for surface and subsurface flow. This can lead to the drying out of downslope areas of the sloping fen as well as gully erosion. Different types of road construction have been proposed to limit the negative implications of the roads on flow dynamics. However, so far no systematic analysis of their effectiveness has been carried out. This study presents an assessment of the hydrogeological impact of three types of road structures in semi-alpine, sloping fens in Switzerland. Our analysis is based on a combination of field measurements and fully integrated, physically based modelling. In the field approach, the influence of the road was examined through tracer tests where the upslope of the road was sprinkled with a saline solution. The spatial distribution of electrical conductivity downslope provided a qualitative assessment of the flow paths and thus the implications of the road structures on subsurface flow. A quantitative albeit not site-specific assessment was carried out using numerical models simulating surface and subsurface flow in a fully coupled way. The different road types were implemented in the model and flow dynamics were simulated for a wide range of slopes and hydrogeological conditions such as different hydraulic conductivity of the soil. The results of the field and modelling analysis are coherent. Roads designed with an Ldrain collecting water upslope and releasing it in a concentrated manner downslope constitute the largest perturbations. The other investigated road structures were found to have less impact. The developed methodologies and results are useful for the planning of future road projects.

developed an empirical model for estimating road sediment erosion of roads 23 located in forested catchments in the Washington state (USA). They concluded that a heavily used road produced 24 130 times more sediment that an abandoned road. Wemple and Jones (2003)  The results showed that the road induced a slight increase of runoff and a decrease of surface-subsurface water 31 exchange around the road. Dutton et al. (2005) investigated the impact of roads on the near-surface subsurface 32 flow using a variability saturated subsurface model. They concluded that the permeability contrast caused by the 33 road construction leads to a disturbance of near-surface subsurface flow which may significantly modify the 34 physical and ecological environment. 35 Road construction can also impact the development of vegetation (Chimner, 2016). Von Sengbusch 36

57
While these studies clearly indicate that roads can have adverse effects on the surface and subsurface flow 58 dynamics and the associated ecosystems, a detailed study on how roads perturb the flow system and dynamics in 59 a sloping fen has not been carried out. In Switzerland, more than 20'000 ha are included in the national inventory 60 of fens of national importance (Broggi 1990), most of them are located in the mountainous regions of the 61 northern Prealps. Hence, the majority of Swiss fens is composed of sloping fens, which developed on nearly 62 impermeable geomorphological layers such as silty moraine material or a particular rock layer named "flysch". 63 Although organic, soils are not necessarily peaty and most of the time quite superficial, not exceeding a few 64 decimeters in thickness. Water flow is therefore mostly consisting of runoff and partly occurring in the shallow 65 part of the subsurface. The construction of a road in this kind of sloping fens removes completely the soil layer 66 in which subsurface flow occurs, thus constituting a major perturbation of the hydraulic regime. Construction 67 techniques to limit these adverse impacts have been proposed but their efficiency has so far not been investigated 68 Three road structures with various construction techniques and materials (hereinafter further detailed) were 69 developed in Switzerland to reduce the impacts of roads. These three road types are conceptually illustrated in 70

Study areas and fieldwork 102
Four sloping fen areas located in alpine or peri-alpine regions of Switzerland (Table 1) were selected. All 103 areas are situated in protected fen areas, and their selection was based on two main criteria: 104 1. The subsurface water flow must occur only in the topsoil layer and as runoff (as described in the 105 introduction). 106 2. The types of installed road structures (no-excavation, L-drain and wood-log). 107 To fulfil the first criteria, soil profiles were analysed to ensure that each area with different road types had the 108 comparable soil stratigraphy: It had to be composed of organic soil on top of a layer of impermeable clay and 109 similar hydraulic regimes (e.g., runoff and subsurface flow occurring only in the topsoil layer). In addition, to 110 ensure that subsurface water is forced to cross the road instead of flowing in parallel of the road (and thus not 111 being affected directly by the road), another important criterion for the selection of the study areas was that 112 subsurface flow is perpendicular to the road. 113 To evaluate the hydraulic connection provided by the roadbed structures, tracer tests were carried out. As 114 illustrated schematically in Figure 3, the upslope area was irrigated with a saline solution and the occurrence of 115 7 the tracer was monitored downslope the road. In the absence of surface runoff, the occurrence of a tracer 116 downslope demonstrates the hydrogeological connection through the road. Furthermore, the spatial distribution 117 of the tracer front reflects the heterogeneity of the flow paths. 118

120
Each area corresponds to an 8 x 20m rectangle that includes a 2.5 to 3.5m wide road segment. A network 121 of approximately 30 mini-piezometers on both sides of the road (Figure 3) was installed to monitor the hydraulic 122 heads and was used to obtain samples for the tracer test. 123 The mini-piezometers are high-density polyethene (HDPE) tubes no longer than 1.5m (ID: 24mm). Each 124 tube was screened with 0.4mm slots from the bottom end to 5cm below ground level. It was inserted into the soil 125 after extracting a core with a manual auger (diameter: 4-6cm). The gap between the tube and the soil was filled 126 with fine gravel and sealed on the top with a 4cm thick layer of bentonite or local clay. Hydraulic heads were 127 measured using a manual water-level meter (± 0.3cm). At each point, the terrain and the top of the piezometer 128 were levelled using a level (± 0.3cm), whereas the horizontal position was measured with a tape measure (± 129 5cm). 130 The tracer tests were conducted using two oscillating sprinklers designed to reproduce a 30mm rain event 131 during 2-3 hours. This is equivalent to an intense rain event. Prior to the experiment, the sprinklers were 132 activated for 15-60 minutes to wet the soil surface. Sodium chloride was added to the irrigated solution to obtain 133 an electrical conductivity of 5-10mS/cm which is approximately ten times higher than the natural electrical 134 conductivity of the groundwater. Then, the area (60m 2 ) upslope of the road (upslope injection area of Figure 3) 135 was irrigated with the salt solution using the two sprinklers. The electrical conductivity (EC) of soil water was 136 manually measured using a conductivity meter in all mini-piezometers prior to the experiment, immediately 137 after, and 24h later. An increase in EC in piezometers located in the downslope area indicates that the injected 138 salt water flowed from the upslope area to the downslope area below the road and clearly shows a hydraulic

183
To numerically solve the 3D flow equation, a 3D mesh was developed (Figure 5a). The mesh is 76m long 184 in the X direction, 20m in the Y direction and the mesh thickness is 1.2m. The top elevation was fixed at 2m on 185 the right side (x=76m) and varies from 9.6m to 24.8m on the left side (x=0) according to the slope of the model. 186 The mesh was made up of 24 layers, 127,200 nodes and 118,440 rectangular prism elements. To ensure an 187 appropriate level of detail, several mesh discretization refinements were made. Therefore, the element size varies 188 between 2m and 0.1m horizontally (in the X and Y directions) and 0.09m and 0.06m vertically. 189 The base case model and the three other models representing different road types have the same boundary 190 conditions and finite element meshes, however, modifications were made between coordinates 61<x<66 to 191 implement the different road types. Figure 5 depicts the differences between the base case model (Figure 5a and  192 b) and models with roads (Figure 5c

Model setup 202
The sensitivity analysis consists of the variation of model properties and parameters in order to 203 understand how they control the sloping fen dynamics. The sensitivities of the following parameters were 204 analyzed: fen slope, soil hydraulic conductivities and road drain hydraulic conductivities. These parameters were 205 selected because they govern the Darcy law (1) and consequently the groundwater dynamics. K is the hydraulic 206 conductivity of the soil and the drain and ∇H the gradient of the fens controlled by the slope. 207 q = K * ∇H (1) For each property, three different values were chosen (Table 2) their van Genuchten parameters are those of gravel. The slopes were fixed at 10%, 20% and 30% as observed 216 during the fieldwork. Note that the drain hydraulic conductivities of the wood-log (W-L) were assumed ten times 217 more conductive and more porous than gravel drain because of its particular structure (wood logs). The road 218 concrete is almost impermeable with a very low hydraulic conductivity and its van Genuchten parameters of fine 219 material. The road basement made with highly compacted fine material (sand and loam) feature a low hydraulic 220 conductivity and are assigned van Genuchten parameters corresponding to fine material. Finally, the 221 implemented soil and road surface flow properties correspond to a wetland and urban cover (Li et al., 2008). 222 In order to simulate each parameter combination, a total of 90 models were developed (27 models for 224 each road structures and 9 models for natural conditions). Models are run for 10'000 days (about 27 years) with 225 a constant flux equal to 380mm/y on the top representing the rainfall to reach a steady state. This precipitation 226 allows for the saturation of the downslope part of the model. Subsequently, subsurface flow rates in the soil layer 227 were extracted at each section with an area of 0.4m 2 (1m wide times the soil thickness) presented in Figure 6. 228 Flow-rates at any given location represent the cumulative vertical flow. Their spatial assessment allows to assess 229 to what extent the roads perturb the system, and further allows to assess the erosion risk associated with the 230 induced preferential flow. Therefore, a comparison of flow rates between each model was made to present the 231 effect of each road structure and sloping fen properties on the dynamics. 232  through the drain (e.g. through the dissolution of the construction material). Given that this initial distribution of 242 EC is not uniform, the comparison of EC after the sprinkling experiment has to be made in a relative manner 243 (Figure 7, 3rd column). 244 The heterogeneity of the hydraulic conductivity of the soil is apparent from the tracer tests (Figure 7, 3rd  245 column: EC 24 hours after injection). At all four sites, the front of the saline solution is not uniform but follows 246 the heterogeneity of the soil hydraulic conductivity. Nevertheless, road structures may create preferential flow 247 path that is particularly obvious at the SCH site where the front follows two preferential flow paths. One related 248 to the L-drain (right path) and the other on the left, unrelated to the L-drain, suggesting that the latter drains only 249 a part of the water and the other part follows a natural preferential flow path. At the HMD site, the saline 250 solution is far more concentrated on the left side of the plot, yet apparently not as a result of the road's structure. 251 Rather, the soil appears more permeable on the left side of the plot, both upslope and downslope of the road. 252 Finally, the decrease in EC observed 24 hours after injection at some locations might result from the following: 253 (1) the tracer injection induces, by "piston effect", the displacement of a small volume of local water with a 254 lower EC; (2) the tracer injection was preceded by a period of irrigation without tracer, which could have diluted 255 the pre-irrigation soil solution. 256 In each case, the irrigation experiments demonstrate the continuity of subsurface flow under the road for 257 all structures. For the no-excavation and wood-log type, the perturbation of the flow field seems to be controlled 258 by the natural heterogeneity of the soil and flow paths, and not by the road itself. Conversely, the field data 259 strongly suggest that the L-drain constitutes an important preferential pathway and consequently subsurface flow 260 is increasingly concentrated. In terms of wetland conservation, this flow convergence is a serious threat (gully 261 erosion, local drying up of the soil). Despite these strong indications, it is clear that with the field data alone no 262 conclusive analysis can be made as no data before the construction of the road are available. Fieldwork allows 263 for site-specific conclusions, but more general conclusions which are not specific to a site are impossible. 264 Therefore, numerical modelling was used to fill this gap.   Figure 8a shows the results of the models with a slope of 10%, Figure 8b with a slope of 20% and Figure  272 8b with a slope of 30%. In each dot chart, the groundwater flow rates (always in m 3 /d) are plotted with crosses 273 for the base case model, diamonds for the no-excavation type, squares for the L-drain type and circles for the 274 wood-log type. In addition, the maximum flow rate capacity of the soil calculated with the Darcy Law (1) and 275 the flow rate induced by the precipitation are also presented for the interpretation of the results. In following 276 paragraphs, the base case (natural conditions) results are presented and discussed, followed by the simulations of 277 the road structures. 278 In the base case model, groundwater flow rates vary from 0.003 (m 3 /d) to 0.069 (m 3 /d) for 10% slope, controlled by the hydraulic conductivities and therefore the slope plays a minor role. Differences between the 284 maximum and minimum hydraulic conductivity are two orders of magnitude, whereas changes between slopes 285 multiply by two (for a slope of 20%) or three (for a slope of 30%) the groundwater flow. Therefore the 286 groundwater flow is increased by a factor 3 between the model KS3 with a slope of 10% and model KS3 with a 287 slope of 30%. Finally, it can be seen that the maximum flow rate of the soil is reached and lower than 288 precipitation in all cases except if the hydraulic conductivity is high (KS1). This means that for KS2 and KS3 289 models, surface flow occurs and conversely the soil is able to infiltrate the precipitation in KS1 models. 290 In the no-excavation and wood-log type models, the effect of road structures is quite similar. The 291 groundwater flows vary from 0.01 (m 3 /d) to 0.069 (m 3 /d) for 10% slope, 0.01 (m 3 /d) to 0.069 (m 3 /d) for 20% 292 slope and to 0.010 (m 3 /d) to 0.069 (m 3 /d) for 30% slope. Compared to the base case model, results show that the 293 no-excavation and wood-log type structures have a minimal impact. The only marked difference is that 294 groundwater flow rates are slightly higher if the soil hydraulic conductivities are low (KS3) for each slope in the 295 wood-log type model. This can, to a certain extent, be explained by the fact that the hydraulic conductivity of the 296 base of the road (consisting of wood-logs) is higher than the hydraulic conductivity of the soil and therefore 297 facilitate the infiltration. Conversely, in the base case model, less water is infiltrated but more runoff occurs. 298 However, the process is limited, because at 3.5 m downstream the road (x=67m), the simulated flow rates of the 299 model KS3-KD1 and a slope of 20% for the wood-log and no-road model are equal (Figure 9). For the no-300 excavation model with a slope of 10%, results are not presented for technical reasons. For this specific geometry 301 and topography, a different structure of the mesh had to be generated which did not allow for a direct visual 302 comparison with the other models. In the 20% and 30% slope models, the results of the no-excavation model are 303 similar to the base case model. 304 In the L-drain type model, the effect is markedly different from the other road structures. The rates are slightly higher than the base case model. Finally, it can be seen that slope accentuates groundwater flow 314 rate differences along the transect. Therefore, an increase of groundwater flow differences in the same model is 315 observed for the 10% and 30% slope scenarios. The impact of L-drain may be further explored by extracting 316 groundwater flows lower than 3.5m to assess the extent of perturbations. Figure 10 shows additional simulated 317 groundwater flows for the most critical cases (i.e. KS1 with a slope of 10%, 20% and 30%) downstream the road 318 at 3.5m and 6.5m respectively and 2.5m upstream. It can be seen that at 3.5m the groundwater flows already 319 regain their upstream conditions. At 6.5m downstream the road, all observation sections are very close the 320 upstream flows except in section G where flows are still slightly higher. 321 The model results can be used to predict the risk of gully erosion and. Gully erosion may occur when 322 changes in surface flow dynamics induce runoff concentration (Nyssen et al., 2002;Valentin et al., 2005b). As 323 presented in Figure 8, the maximum flow rate capacity of the soil is small in caparison with precipitation. For all 324 model scenarios except for KS1, the soil capacity is lower than the precipitation amount which is already set 325 pretty low in the model. This means that runoff already occurs in sloping fens. However, the runoff may be 326 accentuated by subsurface perturbation caused by the L-drain structures. To illustrate this process, the simulated 327 surface flow velocities of each road structure downstream the road for the model KS2-KD2 and slope of 20% are 328 18 presented in Figure 11. In this case, maximum flow rate capacity of the soil is approximately equal to 329 precipitation, therefore runoff should not occur. However, it can be seen some runoff in the L-drain model which 330 is the consequence of the subsurface flow concentration. In this configuration, the soil infiltration capacity is too 331 small and consequently, the groundwater emerges and flows on the surface. Although the formation of gullies 332 depends a lot of other factors (Valentin et al., 2005b), such as soil type or the rain intensity, the model showed 333 that downstream L-drain structure may cause runoff concentration which is an important factor. A simple 334 recommendation can be made to avoid this runoff concentration. 335 1. If the maximum flow rate capacity of the soil is smaller than the flow rate induced by precipitation, 336 the installation of an L-drain structure should not be considered. 337 2. If the maximum flow rate capacity of the soil is larger than the flow rate induced by precipitation, an 338 L-drain may be considered only if the concentred flow calculated by multiplying the drainage area 339 by the precipitation is smaller than maximum flow rate capacity of the soil 340 Finally, the impact of road structure on the upstream road dynamics may be also assessed. Figure 12 shows the 341 same information as Figure 8 but at 2.5m upstream. It can be seen that for all models, upstream flows are similar 342 to the base case model. This means that all structures allow the groundwater to flow across the road. 343 The impact of the L-drain road structure which concentrates groundwater flow is clearly identified in the 344 numerical approach and is consistent with the field observations. For other road structures also, numerical 345 models are consistent with fieldwork results by showing relatively undisturbed groundwater flow downslope the 346 road. The development of models with various combinations of parameters also allowed for exploring a larger 347 parameter space than using field work only. For instance, the fact that the impact of an L-drain structure on the 348 water dynamics is less marked if the hydraulic conductivity of soil is low would have been impossible to identify 349 by using fieldwork only. However, a numerical model is always a simplified reproduction of reality. The main 350 model assumption is that hydraulic conductivity of the soil is homogeneous. Groundwater flow in fens can occur 351 along preferential pathways. Therefore, the models are not able to reproduce small-scale observations, i.e. the  Conclusions 373 This study assessed three road structures regarding their perturbations of the natural groundwater flow. 374 Two of these road structures were specifically developed to reduce the negative impacts of the road. The study is 375 based on two complementary approaches; field-based tracer tests and numerical models simulating groundwater 376 flow for the different road structures. 377 It is the first time that the performance of these road-structures has been investigated in the field. The 378 tracer tests showed that both sides of the road where hydraulically connected for all investigated road structures. 379 Groundwater flow was heterogeneous suggesting the occurrence of preferential flow paths in the soil. The 380 presence of a transversal drain (L-drain) beneath the road constitutes a preferential flow path, however, which is 381 of much greater importance than the naturally occurring preferential pathways. This was also confirmed by the 382 models. Groundwater flow rates 10 times larger than in the natural case were obtained in the numerical 383 simulations. This is not further astonishing as the drains were specifically designed for this purpose. The two 384 other road structures (wood-log and no-excavation) do not perturb the flow field to the extent of the L-drain. To 385 minimize the perturbation of flow fields, the wood-log and no-excavation structures are recommended. 386 The combination of fieldwork and the development of numerical models was fundamental to achieve the 387 goal of this study. The tracer test allowed for a better understanding of groundwater flow throughout road 388 structures and allowed for evaluating their effectiveness at a given location. However, the tracer tests are time-389 consuming and only a few field sites are available. The numerical approach, on the other hand, allows for 390 exploring any combination of slope, hydraulic properties or road structure, thus providing a more comprehensive 391 approach. In our study, the trends between the numerical and field approaches were consistent. 392 5