Introduction

Predicting the long-term hydraulic performance of engineered clay barriers, such as landfill caps, is important in order to ensure their durability and the ultimate safety of the environment. In general, effectiveness of the caps can be evaluated on the basis of the water balance under different climatic conditions of a proposed landfill site. Simulation models such as the Hydrologic Evaluation of Landfill Performance (HELP) can be used as efficient tools during the designing and planning stage (Berger 2000, 2015, 2022). The HELP model was originally developed by the United States Army Corp of Engineers, but is now being used globally (Yalcin and Demirer 2002; Chabuk et al. 2018). It is a quasi-two-dimensional hydrological simulation model, especially designed to evaluate the water balance or hydrological performance of landfill components, such as caps and liners (Schroeder et al. 1994). Simulation runs using the HELP model allow the designers and permit writers to compare different design of landfill caps and to decide which of the design alternatives are hydraulically effective in the long-term. It also helps the contractors to select the best design alternative in terms of performance and cost savings.

The HELP model produces better long-term predictions when compared to other similar modelling tools (Khire et al. 1997; Dho et al. 2002), provided that the limitations of the model are understood. In the past, field measurements have been compared to HELP simulation results during validation of the model and to study the reliability of the HELP model in predicting the long-term water balance of landfills (Peyton and Schroeder 1988; Field and Nangunoori 1993; Berger 2000; Yuen et al. 2001). As with several other water balance models, the findings of these studies have shown significant differences in HELP model simulations compared with field data, ranging from over predicted flow in humid climates to under predicted flow in arid regions in the short-term. In some short-term simulations, HELP model simulations were found to be unrealistic; however, it appeared to give better simulation results for the long-term predictions (Dho et al. 2002).

Berger (2000) highlights that the HELP (v 3.04) model does not account for the aging of the cap, development of desiccation induced cracks, and root penetration into the capping layers; therefore, HELP cannot directly be used to predict the long-term performances effectively. It has been suggested that the combined effect of the above factors create macro-pores and leave the cap with wide opened cracks that act as preferential flow paths (Sinnathamby et al. 2014; Berger 2000). Therefore, in order to incorporate this process in HELP simulations, the present study used an approach in which the HELP model is calibrated for macro-pore development by using the crack-propagation and associated hydraulic conductivity data obtained from two physical landfill cap models that were monitored over a year (Sinnathamby et al. 2014). Additionally, Sinnathamby et al. (2014) found that roots tend to penetrate downward through physical landfill cap models through these cracks. To the author’s knowledge, there are no studies in the literature that explicitly assess the long-term performance of landfill caps accounting for changes in precipitation due to climate change. The requirement for these types of studies has been pointed out by Berger (2000).

This objective of this study is to investigate the long-term hydrological performance of landfill cap designs under different climate conditions. The climate conditions analysed included: (1) present conditions characterised by normal precipitation events (NP), and (2) possible future conditions characterised by climate change precipitation events (CCP). The location selected for climate data was the North West of England, Cumbria England, UK with relevance to nuclear facilities in the UK, including a Low-Level Waste Repository. Simulations were performed using the HELP model (v 3.04) by incorporating the results obtained from a year-long experimental study carried-out on large-scale physical landfill cap models by Sinnathamby et al. (2014).

Model development

Conceptual model and basic assumptions

Two cap designs were considered in this study. Both caps have a 30 cm thick topsoil layer (TSL) rich in organic matter content that supports vegetation (Italian ryegrass [Lolium multiflorum], an annual/biennial grass with an extensive root system). Both caps have a 30 cm thick compacted subsoil layer (SSL) composed of sandy clay layer (1:1 clay:sand mix) directly below the TSL. Two types of liners are being used: (1) a compacted clay layer (referred to as low permeable layer [LPL]) varying in thickness from 100 cm, 150 cm to 200 cm (cap design 1); (2) a composite liner which comprises of a 25 mm thick GCL underlain by a compacted clay layer that varied in thickness as mentioned above from 100 to 200 cm (cap design 2) (Fig. 1a, b). Detailed information of the material properties of the individual capping layers can be found in Phillips et al. (2011) and Sinnathamby et al. (2014). Mathematical models of these two landfill caps were created in HELP (v3.04). Simulations were undertaken for 4 scenarios for each of the cap designs (Table 1). Scenario variants included the climate and cap hydraulic conductivity (initial and maximum values, kin and kmax, respectively). Hydraulic conductivity data was generated by Sinnathamby et al. (2014). The kin represents the initial state of the capping layers soon after the construction (i.e. layers are intact and no aging), while the kmax represents the state of the capping layers after 1-year of operation after construction (i.e. partially/fully cracked layers that represents the aging). The kin and kmax data were used in simulations of cumulative infiltration and to assess the performance of the caps under a worst-case scenario (i.e. cracked or aging cap). Total landfill cover area of 1 ha and an average slope of 5% were used in the models. Depending on the modelling scenario, different vegetative conditions, evaporative zone depths (EZD) and soil properties were used in the models.

Fig. 1
figure 1

Cross section of the alternative landfill cap designs: a Cap design 1 (TSL-30 cm; SSL-30 cm; LPL-100/150/200 cm), b Cap design 2 (TSL-30 cm; SSL-30 cm; GCL-25 mm + LPL-100/150/200 cm)

Table 1 Different landfill cap modelling scenarios (simulation runs)

Scenarios and calculations cases

Four scenarios were simulated for the each of the two cap designs considered (i.e. 2 × 4 = 8 simulation cases) with different combinations of precipitation events (NP & CCP) and hydraulic conductivity of layers (kin & kmax) (Table 1). These scenarios were used to determine the influence of the climate change precipitation events and the hydraulic conductivity of the LPL on the amount of long-term cumulative infiltration through the capping layers.

Future weather predictions indicate that most parts of the UK could experience wetter winters (approximately 15% increase in precipitation) and dryer summers (approximately 15% deficit in precipitation) (Fowler and Kilsby 2004). The scenarios considered in this study incorporate these predictions; however, changes in temperature, solar radiation, relative humidity and other associated climatic factors are not considered because this current study was part of a larger investigation that examined physical landfill cap models under simulated climate change precipitation to determine the impacts of crack and root growth in a greenhouse where these factors could not be integrated into the physical models (Sinnathamby et al. 2014).

Assumption was made for the precipitation data recorded at the Sellafield weather station (Sellafield No1: Open 1950–1995, 5.6 km from Drigg, heading 317 degrees) from 1971 to 2010 that it would represent the typical rainfall from 2011 to 2060 (Fig. 2). It was also assumed that the surface conditions of the landfill caps, such as vegetation, surface runoff and evapotranspiration, remained constant throughout the simulation runs (Tables 2, 3, and 4).

Fig. 2
figure 2

Average annual Normal (NP) and Climate Change Precipitation (CCP) for Sellafield weather station (Sellafield No1: 5.6 km from Drigg, heading 317 degrees; From 1950 to 1995—Met office, UK)

Table 2 Input precipitation data with mean monthly temperature 20 °C for both landfill cap calibration models (same precipitation data were also used in HELP simulation runs in order to synthetically generate daily precipitation data)
Table 3 Modelling scenarios and some of the input parameters
Table 4 Evapotranspiration input specifications for scenario 1, 2, 3 & 4 (simulation runs)

Model calibration

Several studies have shown that the HELP model has a tendency to over and under predict the infiltration through landfill caps depending on the length of the simulation period. Dho et al. (2002) reported that the reason for the difference between the HELP model predictions and the measured leachate production is because most of the studies have not considered the aging properties of capping materials, such as change in hydraulic conductivity, during the operational period. As the hydraulic conductivities of the capping materials have direct influence on leachate productions, simulation runs should accommodate the periodic changes of hydraulic conductivities of these capping layers or, in other words, should accommodate aging of caps for realistic predictions.

Therefore, as the first step of the modelling in this study, the base case HELP model was calibrated according to Dho et al. (2002), using data obtained from a year-long experimental study conducted by Sinnathamby et al. (2014). Two calibration models were developed to incorporate the NP and CCP models in Sinnathamby et al. (2014) and to continue the simulations over the long-term on the basis of these calibration models. Annual infiltration through the LPL was considered as the calibration target. Initial conditions from the physical models in Sinnathamby et al. (2014) were used as input into the HELP model and the annual infiltration through the LPL was obtained. Using the annual infiltration as calibration target, the average hydraulic conductivity of the LPL at the end of the testing period was calculated and compared with the final hydraulic conductivity values obtained from Sinnathamby et al. (2014). This provided values for hydraulic conductivity of the LPL in both cap designs. For cap design 2, it is therefore assumed that the permeability of the LPL layer is the same as cap design 1, as the experimental models in Sinnathamby et al. (2014) did not utilise GCL due to the complex nature of sampling.

Several iterations were performed by changing the evaporative zone depth (EZD) and the maximum leaf area index (LAI) until the model gave best fit with the experimental value. Parameters considered in these model calibrations are shown in Tables 2, 5 and 6. Data such as precipitation, temperature, wind speed, leaf area index (LAI) and soil data from the controlled environment where the experimental models were tested are used in the calibration models. Detailed input specifications are explained below.

Table 5 General landfill cap design data
Table 6 Evapotranspiration specifications for both alternative cap calibration models at the end of calibration

Input specifications for the calibration models

HELP v 3.04 requires several parameters to perform the water balance simulations, such as soil data, climatologic data and design data. Soil layer data in the form of default soil textures available in HELP modelling library were selected and changes were made for measured hydraulic conductivity and porosity values. A typical multi-layered cap was considered as shown in Figs. 1a. Initial porosity and saturated hydraulic conductivity values of all representative layers were obtained from Sinnathamby et al. (2014) (Table 7). The major differences between the two calibration models are the daily precipitation, initial hydraulic conductivities, and the initial porosities of soil layers. Table 2 shows the mean monthly precipitation to which both experimental models (NP & CCP) were subjected and used as the input into the calibration models as well as in simulation runs.

Table 7 Default soil layer data and modification for both top cap designs under NC and CCP

Evaporative zone depth (EZD) was taken as the variable during calibration. For the NP case, as a reasonable assumption, the iterative process started at an EZD of 30 cm. After several iterations, an EZD of 20 cm yielded the best fit for the annual infiltration (basic calibration criteria) through the capping barrier. On the other hand, for the CCP case, EZD was initially assumed as 25 cm and the best fit was found at 50 cm. This demonstrates that the HELP model was able to represent the effects of the cracking of subsoil layer and associated increase in the EZD below the topsoil layer (> 30 cm) which occurred due to the combined effect of desiccation and root penetration into the capping layers as reported by Sinnathamby et al (2014). It should be noted that, in the latter case, the EZD extended into the SSL and HELP was able to represent this since the SSL was modelled as “layer type 1”.

Dates of starting and ending of the growth season were calculated based on the daily mean temperature. Under normal conditions, + 10 °C temperature is considered as the beginning of the growing season. As the models were tested in a controlled environment, where the temperature was maintained at + 20 °C (± 1 °C), the start and end of the growing seasons have been considered as 0 and 365, respectively, for both calibration models.

Normal average annual wind speed was assumed as zero for both models since the experimental models were tested in closed premises. Normal average relative humidity was measured on site and found to be constant at 90% throughout the testing period for both models. Daily precipitation was generated synthetically by the HELP model from the average monthly rainfall for Sellafield. For the NP case, precipitation history from the Sellafield No 1 weather station was collected from the Met office, UK, for the period from 1971 to 2010. For the CCP case, predictions were made on the basis of a 15% increase in the winter precipitation, a 15% deficit in the spring precipitation, an extremely dry summer with 2–5% precipitation, and a 15% increase in precipitation in the autumn (Table 2). Mean daily temperature data were entered manually as ASCII files for both of the calibration models. Default solar radiation for Olympia, Washington was selected as the best fit as it was considered to reflect similar conditions to that in Sellafield for both of the calibration models.

Input specifications of simulation scenarios

The input data used in the simulation runs are listed in Tables 2, 3, 4, 5 and 7. Simulation runs were carried-out for the period from 2011 to 2061; as a 50- year period was considered long enough to establish trends in the infiltration through the cap over the lifetime of a typical landfill. January 1, 2011 is considered as the initial date of the simulation runs for all scenarios and the results from the 2010 calibration model (e.g. EZDs) are used as the input in the simulation runs, respectively. Changes in the EZD during the operational period due to desiccation cracking and root penetration are not considered in these simulation models as long-term monitoring of physical models are not available.

Due to the unavailability of site-specific daily weather data, they were synthetically generated using the available mean monthly data by HELP. Normal average quarterly relative humidity values were obtained from a similar study which was carried-out at Risley landfill, Birchwood, Manchester in which data from Ringway weather station was used (Paksy 2002). Solar radiation data for all scenarios were synthetically generated by HELP as site specific data are unavailable. These are considered to have a minor impact on simulated results for UK climate. The city of Olympia, Washington, USA was selected from the HELP database as the best match to generate the solar radiation data.

Results and discussion

Simulation results for cap design 1 under NP and CCP events

The long-term performance of landfill cap designs in this study are described on the basis of the cumulative infiltration through the barrier liner for the projected operational period of 50 years from 2011 to 2061. Results from HELP simulations using kmax in cap design 1 showed a greater difference in cumulative annual infiltration between the NP and the CCP events than the results from simulations with kin values (Fig. 3A, B). However, simulation results under the NP and CCP events with kin values did not show such a noticeable difference in the cumulative annual infiltration during the same length of time (Fig. 3A, B), which means the cumulative infiltration estimated by HELP model is clearly sensitive to the input “k” values. It should be noted that, in the latter case, the cap under the NP yielded slightly higher cumulative annual infiltration than the cap under CCP. This is purely because the kin values of all layers of the cap under NP obtained from Sinnathamby et al. (2014) were slightly higher than the cap under CCP which resulted in a higher cumulative infiltration.

Fig. 3
figure 3

Cumulative infiltration of Cap Design 1 under A NP and kin, kmax, B CCP and kin, kmax

Under the kmax and CCP scenario, at the end of the first year, cap design 1 yielded a 211%, 246%, and 267%, higher cumulative infiltration than the kmax and NP scenario with LPL layer thickness of 100 cm, 150 cm and 200 cm, respectively. This remained above 200% after first year and was found to be 242%, 276%, and 293% at the end of the 50-year period which is an increase of 31%, 30%, and 26%, respectively, in the difference of cumulative infiltration. Under the kin and NP, CCP scenarios, the cap models showed an 11%, 7%, and 6%, higher annual cumulative infiltration at the end of the first year and a slight increase of 12%, 7%, and 6% cumulative infiltration at the end of the 50-year simulation period for simulation runs with a LPL layer thickness of 100 cm, 150 cm and 200 cm, respectively. It is understood that the majority of structural changes in compacted clay barriers occur during the first year of operation (Kodikara et al. 1999). Therefore, results from the study at the end of one year can be considered as a reasonable input to the HELP model despite continued structural changes that can happen inside compacted clay barriers. Simulation results from this study also show that the annual cumulative infiltration under NP and CCP events with kin values used in simulation models with 100 cm, 150 cm and 200 cm thick LPLs were about 1–2% of the annual precipitation of the relevant year. This was maintained at this range during the entire 50-year simulation period. In reality, this may not be true in the long-term, even though it can be acceptable in the short-term depending on how well the capping layers are preserved. However, annual infiltration from simulations under NP and CCP events with kmax values were much higher than the caps with kin values. A maximum cumulative annual infiltration of 12%, 11%, and 10% under the NP and 70%, 69% and 68% under the CCP of the annual precipitation was observed in the models with LPL thickness of 100 cm, 150 cm and 200 cm, respectively. This is because the kmax represents the condition of the capping layers after 1-year of operation after construction, where partially/fully cracked layers would be present. Due to the cracking the annual infiltration rate would be notably higher under the CCP despite lower precipitation compared to NP (Fig. 2).

HELP simulation results for NP and CCP events with kin values indicate that the projected changes in the precipitation conditions did not influence the long-term cumulative infiltration between both precipitation events (Fig. 3A, B). However, HELP simulations showed relatively high cumulative infiltration when the kmax values were used as an input under similar precipitation events. For NP, Fig. 3A shows that the difference in the cumulative infiltration is due to the differences in the thickness of the LPL, while the hydraulic conductivity (kin, kmax) did not appear to be a noticeable influence. However, for CCP (Fig. 3B) the thickness of the LPL did not appear to play a role in the cumulative infiltration, perhaps due to the formation of partially/fully cracked layers. The effect of the increased hydraulic conductivity (i.e. the influence of the desiccated capping layers), as an input to the HELP model, is clearly seen when the different scenarios are compared. In other words, when aging and other phenomena that affect the integrity of capping layers (i.e. kmax) were incorporated in HELP simulations, there was a clear difference between NP and CCP events. In comparison, when the original properties of capping layers (i.e. kin) were used in HELP simulations there were no significant difference between NP and CCP events.

Simulation results for cap design 2 under NP and CCP events

When comparing infiltration differences obtained over the entire operational period for both cap designs, it is obvious that inclusion of the GCL made a significant difference in the cumulative infiltration in all simulations. At the end of the operational period, as a result of inclusion of the GCL, the reduction in the cumulative infiltration under NP simulations varied between 7 and 13%, whereas under CCP simulations it varied between 8 and 20% (Table 8). Nevertheless, when the GCL was added to the simulations with kin values, the decrease in infiltration was not significant regardless of the varied LPL thickness under both the NP and CCP events (Fig. 4A, B). This is because the physical landfill caps in the study were prepared identically; therefore, they had very similar kin values in every layer (Sinnathamby et al. 2014). The simulations with kmax values can be the best representation of caps in a real world scenario. This is because, over the operational period of the landfill caps, the LPL tends to deteriorate as a result of a various processes (e.g. desiccation, intrusion of plant roots and animals into the barrier, etc.) which cause increased hydraulic conductivity. HELP simulations from this study show that by adding a GCL to NP and CCP caps a minimum of 8% and 10% of the infiltration can be reduced, respectively. Infiltration was reduced by 5–10% when LPL thickness was varied and a GCL was added to NP simulations with kin; however, under the same conditions with kmax, infiltration was reduced by 9–10%.

Table 8 Cumulative percolation through caps and the percentage decrease in percolation due to the inclusion of a geosynthetic layer
Fig. 4
figure 4

Cumulative infiltration of Cap Design 2 under A NP and kin, kmax, B CCP and kin, kmax

Furthermore, there was a great reduction of infiltration when LPL thickness was varied and a GCL was added to CCP simulations (i.e. 10–20% at kmax and 8–11% at kin). This illustrates that neither the desiccation of the capping layers (regardless of kin or kmax) nor the climate change precipitation (regardless of NP or CCP) made any significant difference in cumulative infiltration when a GCL was present in the cap. It is considered that the HELP model under-predicts the cumulative infiltration when a GCL is used in composite barrier liners. However, in reality, as a result of possible desiccation, shrinking and swelling of the barrier liner and other natural soil layers, the effectiveness of the GCL will be lost. Therefore, the effectiveness of the composite liner will degrade significantly in a long-term (Berger 2000).

Effectiveness of the simulated cap designs

HELP simulation results for the cumulative annual infiltration through the barrier liner from Berger (2000) are in agreement with the simulation results for the cumulative annual infiltration in this study. In Berger (2000), HELP simulated infiltrations through cap designs were compared to actual measured infiltration collected over a period of 8 years (1988–1995) through a cap liner at Georgswerder landfill site in Hamburg, Germany. When input parameters were fixed (i.e. no account for aging of capping layers) and the simulation was run for several years, the HELP model predicted almost a linear relationship between cumulative infiltration and time (Fig. 5A, B). From the simulations in this study, it is evident that inclusion of a GCL under both NP and CCP events was very effective in minimizing the amount of infiltration through the caps, especially for layer parameter kmax and NP CD1 kin (Fig. 5A, B). Simulation results under the NP with kmax showed that the composite barrier liner in cap design 2 with 100 cm thick LPL and a 2.5 cm GCL, could minimize infiltration almost as good as a single compacted clay liner with twice the thickness of the LPL in cap design 1. Under the CCP and kmax, at the end of the simulation period, cap design 2 with a 100 cm thick LPL yielded an infiltration of 193,706 cubic metres, whereas cap design 1 with 200 cm thick LPL yielded 212,376 cubic metres. This is almost a 10% reduction in cumulative infiltration between the proposed cap designs. On the other hand, apart from NP CD1 kin, even though simulations under NP and CCP with kin values show reduction in infiltration, it is not advisable to consider those as they do not represent real world scenarios. Also, as aforementioned, the HELP model tends to over-predict the performance of composite liners; however, in reality the GCL would lose its effectiveness as soil layers deteriorate.

Fig. 5
figure 5

Cumulative infiltration comparison of Cap Designs (CD) 1 & 2 under A Normal Precipitation (NP) with kin & kmax (LPL thickness 100 cm), B Climate Change Precipitation (CCP) with kin & kmax (LPL thickness 100 cm)

Suitability of the HELP model in assessing the long-term performance of landfill caps

The long-term HELP simulations in this study were performed by having most of the layer parameters fixed, assuming that the majority of the soil macro-structural changes occur during the first few cycles of shrinking and swelling. Therefore, the difference in the cumulative infiltration predictions is caused only by the difference in the input weather data such as daily precipitation, temperature and solar radiation. However, in real world applications, as the caps undergo aging, capping materials lose their integrity. Soil thin sections and hydraulic conductivity test results from Sinnathamby et al. (2014) in conjunction with this study showed significant variations throughout the testing period of twelve months. As observed at the end of the testing period, the CCP cap model experienced severe desiccation cracking during the extreme drought in the summer which did not reseal after subsequent wetting (Sinnathamby et al. 2014). Even though macro-pore development in the capping layers during the testing period was accommodated in HELP through calibrations prior to simulation runs, further changes in the structure (possible cracking due to desiccation) of capping layers during the simulation period of 50 years is not accounted. It is also important to note that the HELP model assumes that the barrier liner is always saturated which is not always the case; therefore, the predictions may not be close to the actual leakage that is measured (Berger 2000; Luellen and Brydges 2005).

Berger (2000) states that exact matches between the HELP simulated infiltrations and in situ measurements should not be expected as the HELP model does not account for real world processes that happen in capping material during its lifetime; however, better long-term predictions can be achieved by calibrating the HELP model through more extensive and longer-term in situ measurement/monitoring of landfill cap properties such as porosity and hydraulic conductivity of capping layers, and effect of vegetative cover. HELP simulations have to be updated with measured in situ parameters, to allow them to give better and reliable leakage predictions.

Therefore, the HELP model can be used as an effective tool to assess the long-term performances of cap designs under different climatic conditions, provided that the aging of the capping layers is carefully monitored in the field and the HELP model calibrated accordingly prior to simulation runs. These types of monitoring and predictions would be beneficial during the planning stage of landfill caps where climate change events, such as precipitation, can be accounted for in the design.

Conclusions

Simulation models such as HELP could assist landfill engineers in assessing cap designs under different climatic conditions, especially under climate change conditions, and for different lengths of operational periods by incorporating progressive changes in individual capping layers. HELP simulations from this study show that the hydraulic performance of similar (close to identical) landfill cap models comprised of natural barrier layers alone (cap design 1) is highly sensitive to climate conditions; therefore, it could differ greatly in cumulative infiltration in a long-term.

The effects on the cumulative infiltration by the inclusion of a GCL into the cap models was also studied by adding a 2.5 cm thick GCL between the SSL and LPL of the cap design 2. The HELP model simulations of cap design 2 with varied LPL thickness resulted in a 7–13% reduction in cumulative infiltration under NP and an 8–12% reduction in cumulative infiltration under CCP, which is considered as a significant reduction in the context of a field-scale landfill cap. More interestingly, when the observed maximum hydraulic conductivity values were used in the simulations to represent a more realistic cap performance, cap design 2 with a 1 m thick LPL under CCP produced almost 9% less leakage than cap design 1 with 2 m thick LPL under the same conditions. The simulation results suggests that a GCL can effectively be used in landfill caps, especially in caps that are susceptible to desiccation induced cracking, in order to minimize infiltration into the disposal facility, thereby reducing the amount of leachate produced. Importantly, inclusion of a GCL reduced the amount of infiltration by as much as 10% between cap designs 1 and 2 under CCP events with observed kmax criteria. This means that a GCL could be an effective element in cap designs under climate change precipitations. However, it should be noted that the HELP model tends to over-predict the performance of composite liners by neglecting the loss of effectiveness of the GCL as a result of the ageing of other natural barriers.