Title: Oxygen migration through a cover with capillary barrier effects colonized by roots

Complete List of Authors: Proteau, Alex; Université du Québec en Abitibi-Témiscamingue Campus de Val-d'Or, Institut de recherche en mines et en environnement IRME Guittonny, Marie; Université du Québec en Abitibi-Témiscamingue, Institut de recherche en mines et en environnement IRME Bussière, Bruno; Université du Québec en Abitibi-Témiscamingue, Institut de recherche en mines et en environnement IRME Maqsoud, Abdelkabir; Université du Québec en Abitibi-Témiscamingue Campus de Val-d'Or, Institut de recherche en mines et en environnement IRME

One of the most significant environmental risks in mine tailings storage facilities (TSF) is the 23 formation of acid mine drainage (AMD). AMD occurs due to the oxidation of sulfide minerals 24 when they are exposed to water and atmospheric oxygen. The AMD is generally characterized by 25 low pH values and high metal concentrations (Kleinmann et al. 1990;Blowes et al. 2003;Demers 26 et al. 2009). In humid climates such as that of Québec, where the average annual precipitation is 27 around one meter per year (Government of Canada 2016), preventing water infiltration into tailings 28 can be challenging. Therefore, limiting oxygen migration to the reactive tailings is usually the 29 preferred reclamation approach in such regions (Aubertin et al. 2000;Johnson and Hallberg 2005). 30 Several techniques currently exist for controlling oxygen migration, including: water covers (e.g. 31 Aubertin et al. 1999;Yanful et al. 2004;Awoh et al. 2012) and engineered soil covers (Yanful et 32 al. 1999;Aubertin et al. 1995) such as monolayer covers with elevated water table (e.g. MEND 33 1996;Ouangrawa et al. 2006;Demers et al. 2008), geosynthetic clay liners (Adu-Wusu and Yanful 34 2006;Renken et al. 2005) and covers with capillary barrier effect (CCBE) (Nicholson et al. 1989;35 Yanful et al. 1993a,b;Wanful et al. 1999;Aubertin et al. 1995;Aubertin et al. 1999;Lundgren 36 2001;MEND 2001;MEND 2004;Bussière et al. 2003;Bussière et al. 2006;Molson et al. 2008). 37 A CCBE is an oxygen barrier cover that is comprised of a minimum of three layers (Aubertin et 38 al. 1995). Its effectiveness as an oxygen barrier is based on its capacity to maintain a (fine-grained) 39 moisture-retaining layer (MRL) at a high degree of saturation (Sr). Because oxygen migrates 10 3 40 to 10 4 times more slowly in water than in air (Hillel 1998;Chapuis and Aubertin 2003), the 41 saturation of the MRL limits oxygen fluxes. High saturations are maintained in the MRL by placing 42 layers of coarse-grained materials above [protection layer (PL)] and below [capillary break layer 43 (CBL)] the MRL (Aubertin et al. 1995;Bussière et al. 2003;Morel-Seytoux 1992). CCBEs 44 D r a f t 4 reported in the literature typically aim to maintain Sr values in the MRL greater than 85% and have 45 O2 fluxes below 20 to 40 g.m -2 .yr -1 (Nastev and Aubertin 2000;Ricard et al. 1997;Dagenais et al. 46 2001). CCBE design is performed using the hydrogeological properties of the cover materials (e.g., 47 porosity, saturated hydraulic conductivity, and water retention curves) and the boundary conditions 48 applied to the cover system (e.g., precipitation, evaporation, transpiration and water table level). 49 Studies on the design and in situ testing of CCBEs have been documented in numerous 50 publications (e.g., Nicholson et al. 1989;Yanful et al. 1993a,b;Yanful et al. 1999;Aubertin et al. 51 1995;Aubertin et al. 1999;Lundgren 2001;MEND 2001;MEND 2004;Bussière et al. 2003;52 Bussière et al. 2006;Molson et al. 2008). 53 Following construction, CCBEs can support vegetation, which can be established either through 54 active revegetation that is part of a reclamation design (Tordoff et al. 2000), or through natural 55 recolonization Smirnova et al. 2011). It has been 56 demonstrated that CCBEs meet or exceed performance expectations in the short and medium term 57 (<10 years) (Dagenais et al. 2001;Bussière et al. 2006;Bussière et al. 2009;Mbonimpa et al. 58 2011), despite possible influences from vegetation. In the longer term, roots could have significant 59 and contrasting impacts on important parameters controlling the effectiveness of CCBEs. For 60 example, root colonization could modify the porosity of the cover's material (Bodner et al. 2014;61 Sasal et al. 2006;Angers and Caron 1998), create preferential pathways (Scanlon and Goldsmith 62 1999), change the materials' water retention curve (Jotisankasa and Siritannachat 2017), decrease 63 the degree of saturation or increase suction through water pumping (Yan and Zhang 2015), and 64 finally consume oxygen via respiration (Lambers et al. 2008). However, the overall influence of 65 vegetation on the performance of CCBEs has not yet been documented, thus there is a strong need 66 to better understand potential long-term impacts. 67 The effect of plants on CCBE effectiveness likely could change with plant type and species. Since 68 most herbaceous plants have shallow root systems located in the first 50 cm of soil (Jackson et al. 69 1996), a thick protection layer at the surface of a CCBE should be able to limit the impacts of such 70 species. In contrast, trees and shrubs grow deeper root systems ) that can reach 71 depths of 75 to 200 cm in dry soils of the boreal forest ( Strong and La Roi 1983). Thus, woody 72 roots could potentially penetrate the MRL of a CCBE in forested environments. However, some 73 root growth can be limited by highly saturated soils (Boggie 2016). In such cases, species with 74 tolerances to high saturations, such as willows and poplars (Jackson and Attwood 1996;Gong et 75 al. 2007), could still be considered as threats to the MRL. Fine roots (< 2 mm diameter) will absorb 76 most of the water (Lambers et al. 2008), and their total length in a given soil volume is a good 77 indicator of overall water consumption (Zhang et al. 2009). Oxygen consumption rates generally 78 increase with root length density (RLD) (Cook et al. 2007) but respiration rate of roots decreases 79 while in highly saturated soil (Cook and Knight 2003). 80 The main goal of this study was to characterize the geotechnical and root colonization properties 81 of a 17-year-old CCBE at a mine site in Québec, Canada. The performance of the CCBE was 82 evaluated with respect to its ability to limit oxygen fluxes passing through the cover to values 83 lower than the established design criteria of 20 to 40 g of O2/m 2 /year (Nastev and Aubertin 2000). 84 Oxygen consumption tests (OCTs) were used to estimate oxygen fluxes. Following its 85 construction, the CCBE was gradually recolonized by the surrounding vegetation of the mixed 86 forest. The impacts of different vegetation types and species (i.e., herbaceous, broadleaf, 87 coniferous) are considered here, with a focus on woody species. The relationships between the 88 morphological parameters of roots colonizing the MRL (i.e., RLD, biomass, and diameter) and the 89 parameters influencing O2 diffusion into the materials [i.e., degree of saturation (Sr), reactivity 90 D r a f t 6 coefficient (Kr), effective diffusion coefficient (De), and porosity (n)] are also studied. The present 91 work links geoenvironmental engineering with plant biology and will enable better integration of 92 the effects of vegetation into the design of CCBEs. 93 The main research hypotheses were: i) roots exist primarily in the upper portion of the MRL due 94 to the high Sr in deeper parts of the MRL, and under typically deep-rooted species like Salix sp.; 95 ii) a negative correlation exists between RLD, and the Sr of the MRL; iii) roots colonizing the 96 MRL consume oxygen, resulting in increases in observed Kr values; iv) a positive correlation exists 97 between RLD and Kr. 98

Experimental site 100
The Lorraine mine site is located near Latulipe-et-Gaboury in the Témiscamingue region of 101 Québec (47° 24' 00' N, 79° 00' 00' W). Gold, silver, copper, and nickel were extracted for a total 102 of four years (from 1964 to 1968), after which the mine was abandoned leaving behind 103 approximately 600,000 tons of acid-generating tailings over an area of 15.5 ha (Dagenais et al. 104 2001;Genty et al. 2016). In 1999, the Québec Ministry of Energy and Natural Resources decided 105 to reshape the tailings storage facility, build a CCBE over the tailings, install limestone drains to 106 passively treat acidic effluents (Nastev and Aubertin 2000;Potvin 2009) (Smirnova et al. 2011) andlater in 2015 (Guittonny-Larchevêque et al. 112 D r a f t 7 2016a) to identify the dominant plant species as well as their cover and density levels. The average 113 monthly precipitation at the site varies between 36 mm (February) and 96 mm (August) for a yearly 114 total of 837 mm . The water table is close to the surface in the northern portion of the CCBE, but  115 can reach up to 2 m deep in the southern portion (Bussière et al. 2009). The average daily 116 temperature ranges between -15 °C in January and +18 °C in July (Government of Canada 2016). 117 The main technical objectives for the CCBE were to maintain a minimum Sr of 85% in the MRL 118 at all times, and to have maximum oxygen fluxes reaching the tailings of 20 to 40 g/m 2 year 119 (Dagenais et al. 2001). The materials used in the construction of the CCBE were characterized for 120 their grain-size distribution (GSD), Effective diffusion coefficient (De), saturated hydraulic 121 conductivity (ksat), water retention properties, and dry density. Using these properties, a three-122 layered design was selected that was comprised, from bottom to top, of: (1) a 30-cm thick uniform 123  (Table 1). More information on the design and the configuration of the 127 CCBE can be found in Nastev and Aubertin (2000), Dagenais et al. (2001)

Experimental design 142
Five squares monitoring zones (50×50m) were randomly selected in the southern portion of the 143 CCBE ( Figure 1). The southern part was selected because the water table is below the CCBE, 144 while in the northern part the water table is inside the MRL. Thus, in the northern part the 145 performance of the CCBE may be independent of the presence of vegetation. In each zone, five 146 plots corresponding to five different vegetation types were randomly selected. The five vegetation 147 types corresponded to one of the four dominant woody species (Populus balsamifera, Alnus 148 rugosa, Salix Sp. and Picea mariana) or to dominant herbaceous vegetation. Young woody 149 individuals were occasionally found on the herbaceous plots, but this was assumed to have no 150 impact on the MRL. All plots dominated by a woody species needed to have at least one individual 151 of the selected species with a minimum age of five years old. Other species could also be found 152 on each plot, but the targeted species were clearly dominant (> 50%) in terms of aerial biomass. 153

Oxygen consumption tests 154
Oxygen consumption tests were first developed by Elberling et al. (1994). They consist in a closed 155 and monitored gas chamber over tailings and are used as a rapid and precise method for 156 D r a f t 9 determining instantaneous sulfide oxidation rates. This method assumes that the oxygen migration 157 into tailings is mainly due to diffusion (Nicholson et al. 1989). Fluxes entering tailings are 158 therefore modeled using Fick's second law with first order reaction kinetics (Eq. 1). 159 where De is the effective diffusion coefficient of oxygen, C is the oxygen concentration, z is depth, 161 and Kr is the reactivity coefficient of the material. Assuming the boundary conditions C(z) = C0 at 162 z = 0 and C(z) = 0 as z → ∞, the steady-state oxygen flux at the top of the tailings can be estimated 163 by Eq. 2 (Elberling et al. 1994). 164 where C0 is the initial concentration of oxygen in the chamber's headspace. When the area (A) and 166 volume (V) of the headspace are known, KrDe can be obtained from the slope of the plot of ln � C C 0 � 167 vs. time (t). For further details on this OCT method, the interested reader is refered to Elberling et 168 al. (1994) and Elberling et al. (1996). 169 In the present study, OCTs were performed to assess the impact of roots on the oxygen fluxes 170 migrating through the CCBE at Lorraine. Whereas the original OCT method proposed by Elberling 171 et al. (1994) was intended to determine oxygen fluxes passing through reactive tailings at the 172 surface, Mbonimpa et al. (2003) and Dagenais et al. (2012) proposed a modified method to 173 evaluate oxygen fluxes reaching reactive tailings underneath oxygen barrier covers, such as 174 CCBEs. Their method uses longer cylinders to penetrate through the depth of a cover, as well as 175 longer measurement periods (typically from 3 to 5 days). Since longer measurement periods 176 usually violate the assumption of a steady-state flux, oxygen fluxes must then be calculated using 177 D r a f t a numerical model instead of the simplified Fick's second law (Eq. 2). At the Lorraine site, OCTs 178 were performed using a 10-cm diameter aluminum cylinder that was inserted into the CCBE such 179 that the whole depth of the MRL was penetrated ( Figure 2). The cylinder was then sealed with a 180 cap to create a headspace with a height between 1 and 10 cm. Decreases in the concentration of 181 oxygen in the headspace were monitored over a period of 3 to 5 days just after cylinder insertion 182 (Proteau et al. 2019). Eventual roots enclosed in the cylinder were assumed to respire O2 during 183 the test at the same rate as before cylinder placement. However, after the aerial part of the plant 184 was removed and roots were cut, in this case by the cylinder placement, the rate of oxygen 185 consumption could vary. Nevertheless, it has been shown to remain the same (Makita et al. 2013) 186 or to decrease slowly during the first couple of days after the section of roots (Lipp and Andersen 187 2003;Marshall and Perry 1987). The assumption of constant respiration by roots despite cylinder 188 placement thus produces realistic yet slightly conservative results for oxygen consumption by roots 189 in the MRL. 190 Oxygen and CO2 concentrations in the headspace were measured periodically using a syringe and 191 a portable gas chromatograph (GC; HDTS 3000 Micro GC Gas Analyzer, ± 10 ppm). Volumetric 192 water contents (θw; VWC) were measured using an ECH2O EC-5 probe (± 0.03 m 3 /m 3 ) that was 193 inserted vertically at the surface of the MRL (0 -5cm). A matrix-specific calibration was 194 performed on the EC-5 probe in the lab using a soil sample from the CCBE. The volumetric water 195 content was measured every 5 minutes, and CO2 and O2 concentrations were measured every 2 196 hours. For each zone, five OCTs were performed and each test was conducted in plots with 197 different vegetation types. This process was repeated for the five zones (25 OCT in total). The where made). Oxygen flux through the CCBE was determined by using the POLLUTE software 202 (Rowe et al. 1998). This software solves the second Fick's law, with oxygen concentration data 203 from the headspace and the MRL's material properties as the input data. The following parameters 204 were also required: equivalent porosity (θeq ), Darcy's velocity (v = 0), the half-life degradation 205 (t*1/2), and the coefficient of hydrodynamic dispersion (D* = Deθeq -1 ). The equivalent porosity (Eq. 206 3) is a parameter that was introduced to account for the oxygen transport that occurs in both air-207 and water-filled pores (Aubertin et al. 1999;Aubertin et al. 2000). 208 where θa is the volumetric air content, θw is the volumetric water content and H is the dimensionless 210 Henry's equilibrium constant ( ̴ 0.03 at 20 °C). The De parameter was estimated using a semi-211 empirical expression (developed by Aachib et al. 2004) that is based on the porosity and the 212 volumetric air and water contents of the material (Eq. 4). 213 where Da 0 and DW 0 are the O2 diffusion coefficients in air and water, respectively; and pa and pw 215 are related to the tortuosity of the interstitial gas and liquid phases, respectively. As suggested by 216 Aachib et al. (2004) values, pa = pw = 3.4 were used. 217 In the model, a parameter is included to define the reactivity of a soil called the reactivity 218 coefficient (Kr). Because the MRL is made of an inert silty material, in the initial model Kr=0 was 219 used. Roots, on the other hand, could consume oxygen during their respiration (Lambers et al. 220 2008), thus creating a biologic-based Kr. Since respiration also produces CO2, oxygen depletion 221 from respiration should be coupled with an increase in CO2 concentrations. After the initial 222 D r a f t simulation of the OCTs using Kr = 0, an iterative method was used to find a Kr to explain the 223 differences between the in situ measurements and modeled values. In POLLUTE, Kr cannot be 224 directly inserted but is related to a half-life constant as shown in Eq. 5. 225 Using these parameters and Eq. 6, it is possible to calculate the oxygen flux reaching the tailings 227 at the bottom of the MRL (Mbonimpa et al. 2003). 228 where Fs,L is defined by Eq. 7, Kr * is defined in Eq. 8, De * is defined in Eq. 9, and L is the depth of 230 the MRL. Diffraction, ± 0.02 µm, ASTM designation: D422-63) and gravimetric water content (oven drying 256 for 48 hours at 60 °C; ASTM D2216-10). Degrees of saturation (Sr) were calculated at various 257 depths using mass-volume relationships. Organic matter (OM) contents were also measured using 258 the calcination method (burning at 375 °C for 16 hours and weighing before and after; MA. 1010 259 -PAF 1.0). This method had a limit of detection of ~ 1%. All samples used for OM analyses were 260 taken at 50 cm from the center of each plot. 261 D r a f t

Statistical analyses 262
Linear relationships were analysed with Pearson correlation tests with a p-value < 0.05 263 significance level between root parameters (RLD, root biomass density, and root volume density) 264 and soil parameters (Kr, n, and Sr). The effects of sampling depth on the soil parameters used in 265 the CCBE performance assessment with POLLUTE (n and Sr) and root length density were tested 266 with one-way ANOVA analyses using depth as the fixed factor and zone as the random factor. 267 Analyses were performed with XLSTAT with an α of 0.05. 268 269

Soil properties of the MRL 271
Of the 25 OCTs that were performed, 18 produced data that were interpreted. Table 2  D r a f t

Oxygen consumption tests 284
After comparing the results of field measurements with the results obtained by numerical 285 modelling (using the measured soil properties), the plots were separated into two distinct groups 286 (A and B). Plots in Group A, which included A2, A5, E1, E2, E5, H1, H2, H3, H5, P1, P3, P5, S2, 287 and S5, generally showed slower or similar modeled rates of oxygen consumption with respect to 288 the field measurements. In contrast, Group B, which included plots A3, E3, S1, and S3, had 289 modelled oxygen consumption rates that were faster than those calculated from field 290 measurements. Further details on each group are provided below. 291

Group A 292
For Group A (N = 14), it appeared that oxygen was migrating on par with or slightly faster than 293 what was predicted by POLLUTE (see Figure 4). All group A plots had Sr values above 0.85 (see 294 Table 2). One assumption to explain these discrepancies could be the consumption of oxygen by 295 roots. Including a Kr value > 0 in the numerical model to take into account oxygen consumption 296 by roots improved the fit of the model with respect to measured O2 concentrations ( Figure 4). The 297 Kr values were inferred and varied between 1.2E-8 sec -1 and 1.0E-5 sec -1 , with a mean value of 298 1.4E-6 sec -1 . The De values were calculated and varied between 1.94E-11 m 2 sec -1 and 4.09E-9 299 m 2 sec -1 with a mean value of 7.14E-10 m 2 sec -1 (Table 3). These values for De are close to what 300 was expected and previously measured (Dagenais et al. 2001) whereas Kr values imply an oxygen 301 reactivity. Those results are discussed further in the discussion. 302 CO2 production in the OCT cylinders was used to validate that the observed reactivity was due to 303 biological activity, whether from the respiration of roots or decomposition of organic matter. 304 Results from four stations typical of Group A are presented in Figure 5. The CO2 concentrations 305 D r a f t increased significantly from atmospheric values (≈ 400 ppm) to values between 3000 and 8000 306 ppm after 40 hours, thus confirming the presence of biological activity. The production of CO2 did 307 not appear to be due to the presence of organic matter in the soil since the OM concentrations in 308 the MRL were lower than the detection limit of the method, regardless of the sampling depth. Only 309 two plots had OM concentrations that were slightly over the detection limit: 1.11% for E3 and 310 1.02% for P1. Therefore, the variability in OM concentrations could be attributed to In both groups, RLD was clearly greater at the 0-10 cm interval than in deeper parts of the MRL 332 at a 0.95 confidence interval (p-value < 0.0001) for both groups, as shown in Figure 9. At the 333 Lorraine site, these measurements showed that root impacts were mainly limited to the first 10 cm. 334 The linear correlation analysis revealed that RLD and Sr were negatively related ( Figure 10 Table 4). 347 As was the case with Group A, CO2 was produced during the OCTs for Group B. Results from the 348 four plots of Group B are presented in Figure 11. CO2 concentrations increased significantly from 349

Discussion 365
Here will be discussed the colonization of the MRL by roots, its impact on Sr, Kr and the overall 366 impact it has on the oxygen fluxes that go through the MRL. 367

Root colonization in the MRL 368
As suggested by our first hypothesis, root colonization that occurred in the MRL was mostly  root parameters, such as root biomass (R 2 = 0.53) and root volume (R 2 = 0.49), were also tested 397 but were less significant. Root length density is a parameter that is commonly used to represent 398 the extent of root colonization in soils (Angers and Caron 1998; Lipiec and Hatano 2003;Joslin et 399 al. 2000). All roots that were analysed in the MRL were fine roots (diameter < 2 mm). Since water 400 uptake by roots primarily occurs in fine roots (Lambers et al. 2008), the significant correlation 401 between root colonization density and degree of water saturation could be expected. The mean 402 RLD was 130 m/m 3 in the middle (20 to 30 cm) of the MRL and 97 m/m 3 in the lowest part of the 403 MRL (35 to 45 cm); RLD values did not exceed 610 m/m 3 at these depths. Since these values were 404 relatively low, it is unlikely that roots exerted a significant influence over the degree of saturation 405 at the base of the MRL. Using direct measurements (i.e., gravimetric water contents), it was 406 observed that Sr values were indeed higher than the 0.85 threshold in the deeper parts of the MRL. 407

Root colonization and reactivity coefficients (Kr) 408
In previous studies of the Lorraine CCBE, OCTs were performed while not taking into account 409 the potential for oxygen consumption since it was expected that only diffusion would be occurring 410 in the first few years (Dagenais et al. 2012). Generally, calibration of the model used to interpret 411 OCT results consists of fitting the modeled results to the observed results by manually adjusting 412 the diffusion coefficients (De and D * ). In the present study, some of the measured soil parameters 413 were either too high (Sr) or too low (De) to explain the rate at which oxygen concentrations 414 decreased during the OCTs. In fact, in some cases, an unrealistic adjustment of the D * of two orders 415 of magnitude (from 10 -11 to 10 -9 ) would have been required to better fit the empirical data with the 416 D r a f t 21 model. Introducing a Kr value to the model seemed to be a more realistic approach given that CO2 417 effluxes were observed during the OCTs (indicating root respiration and O2 consumption). 418 As mentioned previously, all roots in the MRL were fine, and such roots contribute the most to 419 oxygen consumption by respiration (Lambers et al. 2008;Makita et al. 2009). Accordingly, a good 420 level of correlation was observed between the RLD and Kr (R 2 = 0.65), which was the fourth 421 hypothesis. Soil biological oxygen consumption is primary driven by two processes: autotrophic 422 respiration, which is performed by roots, and heterotrophic respiration, which is performed by 423 microorganisms that break down organic matter (Bond-Lamberty et al. 2004;Chen et al. 2017;424 Olsson et al. 2005). In the present study, organic matter concentrations were too low to measure 425 in the MRL silt (< 1%). This suggests that autotrophic respiration is likely dominant and would 426 help to explain the high linear correlation between Kr and RLD. 427 In the four plots where Sr was under 0.85, it was not possible to measure a Kr with the observed Sr 428 value. Since the degree of saturation measured at the top of the MRL was low, the expected 429 diffusion coefficient was high, meaning that the expected measured oxygen depletion rate should 430 have been high. However, the bulk consumption/diffusion rates of O2 that were recorded on site 431 at those four plots were similar to those modeled with a Sr > 0.85. This is mainly due to the high 432 degree of saturation of the lower portion of the MRL. This portion of the MRL controls oxygen 433 migration through the cover, with the less saturated upper portion having a negligible impact. 434 Therefore, using a plausible Sr value representing the whole MRL, it was possible to calculate a 435 De and then extract a Kr value; values obtained using this method were in the same range as those 436 modeled for the other plots. 437 D r a f t

Impact of roots on oxygen flux and the overall performance of the CCBE 438
Using the calculated Kr values, it was possible to evaluate the oxygen fluxes through the CCBE. 439 Root colonization appeared to have a noticeable impact on oxygen fluxes reaching the base of the 440 MRL. Using eq.4 and considering a non-reactive silt (without the integrated Kr), calculated O2 441 fluxes were lower than the performance criteria of 20-40 g O2/m 2 /year (except for two plots where 442 they were close). These results show that the CCBE works relatively well even if the vegetation is 443 taken out of consideration. However, when considering reactivity due to root presence (with the 444 calculated Kr), calculated oxygen fluxes are reduced further. Plant roots actually consume part of 445 the oxygen that is slowly flowing through the MRL. For stations with an estimated oxygen flux 446 greater than 10 g O2/m²/year, the average reduction in flux ( ℎ − ℎ ) 447 was approximately 34.2 g O2/m 2 /year due to root respiration. 448 It is important to take into account that increases in RLD were generally matched by decreases in 449 Sr. Therefore, more significant root colonization could affect the degree of saturation of the MRL 450 and be harmful to the CCBE's performance. Nevertheless, 83% of root biomass is usually found 451 in the top 30 cm of soil in boreal region (Strong and La Roi 1983). This shows the importance of 452 having a thick protective layer that can help to lower evapotranspiration and confine plants roots 453 to the upper layer of a CCBE. For the Lorraine CCBE, where root colonization occurred mainly 454 in the top of the MRL (10 cm), vegetation did not impair the CCBE's effectiveness, even 17 years 455 after construction. Even if water uptake by roots could increase with greater RLDs (Zhang et al. 456 2009), it is not likely that roots would colonize the whole depth of the MRL in a boreal context. 457 As mentioned earlier, plants have difficulties to elongate their roots in the absence of air in the soil 458 where they grow (Grable and Siemer 1967)  D r a f t

Conclusion 482
Evaluating the effectiveness of engineered covers used to limit oxygen flow and control acid mine 483 drainage production is essential to validating cover designs. In this study, some biological effects 484 of plant roots were included for the first time in the evaluation of a CCBE's effectiveness. The 485 authors present a modified approach to the oxygen consumption test interpretation that considers 486 the presence of roots in the cover. The main modification to the OCT was the relation of root 487 colonization with the oxygen reactivity coefficient. Using POLLUTE to interpret the OCT made 488 it possible to create a root-based oxygen reactivity coefficient. It was shown that it is possible to 489 assess the impact of the quantity of roots present in a CCBE on the oxygen fluxes passing through 490 it. Root length density showed strong positive correlations with Kr and negative correlations with 491 Sr, thus suggesting overall effects of root colonization on a CCBE's effectiveness. Since the depth 492 of root colonization remained limited 17 years after construction, it was concluded that the 493 integrity and effectiveness of the CCBE was maintained. The results showed that the fluxes of 494 oxygen through the CCBE stayed under the design target and that roots helped in lowering this 495 flux. The presence of roots could be considered in future oxygen barrier cover designs and could 496 help to limit oxygen fluxes. However, since roots could also have impacts on other soil parameters, 497 such as hydrogeological properties, further research is necessary. 498

Acknowledgements: 499
The authors thank the Research Institute on Mines and the Environment (RIME UQAT-500 Polytechnique, www.RIME-IRME.ca\EN) for providing funding for this project.                    D r a f t Figure 6. Reactivity coefficient (Kr) as a function of root length density (RLD) for group A plots, with black triangles to highlight plots E5, P3, P5 and H3 which were presented in previous graphs D r a f t Figure 7. Linear models fitted with POLLUTE to the O2 concentration data measured in situ from the OCTs for plots A3, E3, S1, and S3 (circles). Black lines represent models using only the physical data as input parameters (Kr = 0). Dotted lines represent models with adjusted Sr values (Sr = 0.93) and inferred Kr values (Kr ≠ 0).