Modelling of shallow groundwater levels and interaction with drainage system and streams under the impact of climate change

Urban areas in close proximity to coastal areas often have a high groundwater table, and with climate changes, the groundwater table can be expected to increase at these locations. The increase in the shallow groundwater level can be caused by the mean increase in the sea level, succeeding larger pressure heads at downstream boundaries, an increase in precipitation and groundwater recharge (depending on the infiltration capacity of the local soil), or an increase in water levels in connected water courses (Fletcher et al. 2013; Seidenfaden et al. 2022; Oude Essink et al. 2010; LaBianca et al. 2023; Schirmer et al. 2013; Rasmussen et al. 2023).

Furthermore, urban areas with drainage or sewer pipes that unintendedly allow groundwater infiltration are challenged if the existing water infrastructure is rehabilitated (Su et al. 2020). For example, in Denmark, it is common to convert aging combined sewer systems into water-tight separate sewer systems (Thorndahl et al. 2015). This rehabilitation obviously has positive environmental impacts, e.g., by minimizing combined sewer overflow volumes, improving treatment processes at wastewater treatment plants, as well as avoidance of oversizing of pipes, pumps, and other structures (Bertrand-Krajewski et al. 2006; Thorndahl et al. 2008; Staufer et al. 2012; Ahm et al. 2016), but can on the other hand cause local increase in groundwater levels due to the elimination of the sewer infiltration effect. In Denmark, current legislation does not allow water utility administrators to establish perforated drains to decrease groundwater levels in urban areas. In the autumn of 2023, a governmental proposal was published opting to change the legislation, which at the beginning of 2024 was not enacted.

The groundwater increase can result in local flooding from backwater effects and water exceeding the terrain level, local flooding due to limited infiltration during heavy rain (Nielsen et al. 2019), and an increase in infiltration to existing drainage systems causing capacity problems of existing water infrastructure (Boukhemacha et al. 2015; Thorndahl et al. 2016). Furthermore, more extended dry weather periods and potentially longer drought periods as a result of less precipitation during summer might lead to more considerable seasonal fluctuations in shallow groundwater levels in flat and coastal areas where the shallow groundwater level fluctuations are dominated by precipitation. In some areas, this might result in a larger potential for subsidence damage on buildings and structures (Sun et al. 1999).

With numerical modelling tools, it is possible to estimate groundwater dynamics in current and future hydrological conditions and to evaluate changes and potential local solutions to stabilize or decrease groundwater levels. These tools should be able to simulate groundwater levels and interaction among groundwater, surface water, coastal water, drainage system, and son on under different boundary conditions (Barron et al. 2013b). The difference between a non-critical and a very critical shallow groundwater level can vary within a few meters (and even less than a meter) in urban coastal areas. Compliance with this level of detail, therefore, demands high-resolution groundwater models and observational data, if groundwater levels should be simulated with a precision less than the difference between critical levels. Furthermore, modelling of shallow groundwater levels is challenging in urban areas due to (1) hindered infiltration through semi-pervious and impervious surfaces (Thorndahl et al. 2006; Tuyls et al. 2018; Nielsen et al. 2019; Rasmussen et al. 2023), (2) heterogeneous geology due to anthropogenic modification of the subsurface (Oude Essink et al. 2010; LaBianca et al. 2023), (3) groundwater interaction with constructed and natural surface watercourses (Barron et al. 2013a; Locatelli et al. 2017), (4) groundwater interaction with subsurface wastewater and stormwater systems (De Bénédittis & Bertrand-Krajewski 2005; Karpf & Krebs 2011; Barron et al. 2013b; Jeppesen & Christensen 2015; Su et al. 2020; Rasmussen et al. 2023), and (5) groundwater interaction with underdrains and perimeter drains (Wittenberg & Aksoy 2010; Hibbs & Sharp 2012).

In this article, we develop an integrated groundwater model in which the processes described earlier can be modelled with the accuracy required to determine critical shallow groundwater levels in urban areas. The purpose of the model setup is to simulate, evaluate, and compare current conditions, potential rehabilitation solutions, potential prevention solutions, and impacts of climate change in terms of shallow groundwater levels. The primary focus is to screen for potential decreases in the distance between ground level and groundwater level and thereby identify areas with critically high groundwater levels or groundwater flooding. Model simulations of multiple scenarios should help decision-making in terms of selecting the best technical solutions to avoid worsening of current conditions.

This article presents a case study from Aalborg, Denmark, where a small urban residential district, Kærby, is significantly challenged by a currently high groundwater level, pending sewer rehabilitation, and impacts of climate change. Besides focusing on the local challenges and potential solutions of the case study area, the article serves a higher purpose in describing the general requirements for modelling hydrological interaction processes and small variability of groundwater levels in urban areas.

To evaluate the effects of sewer rehabilitation, we develop a model setup based on different potential configurations of the future system. The principles of the solutions are to decrease groundwater levels by lowering water levels in natural waters in surrounding areas, by horizontal drainage through gravitation or pumping, or by vertical drainage through extracting groundwater in wells. As reference, and for comparison, we evaluate the current systems as well as worst-case scenarios where no technical solutions are implemented to reduce groundwater levels. We focus on the technical solutions of each configuration rather than the legislative or regulatory limits.

We evaluate the impacts of climate changes on each configuration by climate projecting to the near future (2041–2070) and the far future (2071–2100) with climate scenario RCP 8.5 (IPCC 2013), which represents a worst-case scenario. One could argue to include other climate scenarios with less climate forcing, e.g. RCP 4.5. However, we do span a complete range of potential outcomes between the current situation and RCP 8.5. This is further explained in Section 4.4.

The following configurations are defined:

• A.

  Current situation, reference (maintaining the current combined sewer systems with extraneous water infiltration).

• B.

  Separate sewer system (construction of a new water-tight drainage system with separate wastewater and stormwater pipes, which does not allow infiltration).

• C.

  Separate sewer system (as B) with elevated overflow discharge from two nearby lakes to decrease the water level and subsequently enhance groundwater exfiltration into the lakes.

• D.

  Separate sewer system (as B) with horizontal perforated drains in the same alignment as the sewer system (installation of a third parallel drainage pipe).

• E.

  Separate sewer system (as B) with a drilled vertical well lowering the groundwater by pumping.

• F.

  Separate sewer system (as B) with combined horizontal (D) and vertical (E) drains.

In this study, a groundwater model, MODFLOW 2005 (Harbaugh 2005), is used to simulate the current and future groundwater levels in the Kærby area. The groundwater model is a deterministic and fully distributed model that describes the most important flow processes in the hydrological cycle. It simulates three-dimensional groundwater flow in the saturated zone as well as one-dimensional drainage water inflow. The model is set up in the MODFLOW 2005 model code and is built via FloPy version 3.3.2 (Bakker et al. 2016; Hughes et al. 2023).

The model is run under steady-state conditions, i.e. stationary over time, which means that the model results show periodic constant conditions but cannot be used to simulate single events or extreme dynamic conditions. As described in Section 4.4, we select the period of the year with the highest observed groundwater levels to simulate the worst-case scenario within the annual cycle.

In the different scenarios, we consider the urban structure, river network, and so on to remain unchanged and limit the study to varying the external boundary conditions forced by climate change and the model scenarios.

In the following section, the input data and methodological background of the setup are described in detail.

The hydrogeological model describes the geophysical and hydraulic properties of the individual geological layers. The hydrogeological model is an interpretation of the spatial distribution of the geological layers depending on their hydraulic properties. The initial hydrogeological setup is based on the National Danish water resource model (DK model) (Stisen et al. 2020). Subsequently, the model is refined with a more detailed interpretation of the near-terrain geology according to the local geology in Kærby. The near-surface geology has been inspected through a series of samplings from 57 boreholes made in 2021 (Ziegler et al. 2023). The depth of the boreholes varies between approximately 2 and 3 m below ground level.

The model is three-dimensional with a horizontal resolution of and 11 geological layers with varying thicknesses corresponding to the top and bottom elevation of the geological layers. A small horizontal discretization is required to simulate the interaction with drainage system and streams (as also shown by Karpf & Krebs 2013; Thorndahl et al. 2015). The model topography is based on the DHM/Terrain raster at 2.5 m, aggregated to 5 m (Danmarks Højdemodel, https://dataforsyningen.dk/data/926).

The internal boundary conditions consist of an exchange between the groundwater and other parts of the water cycle. It covers seepage to/from streams and sewer systems. In the Østerå system, water flows northward through the valley of Østerå along with the ditches Vestre and Østre Landgrøft. These watercourses are included in a model to understand how groundwater and watercourses interact based on the water levels in the watercourses and the position of the groundwater table. The MODFLOW 2005 RIVER module is used for this purpose, allowing for the calculation of a pressure-dependent flux between the stream and groundwater (Harbaugh 2005). The size of the flux is determined by the pressure levels in the stream and groundwater, with the direction of the water flow between the two determined by the pressure gradient.

The level of the drainage system varies between 0.8 m and 3.1 m below ground level with a mean of 1.9 m.

To approximately determine the infiltration flow, the average production of wastewater is subtracted from an estimate of the total flow based on records of pump operation. The estimation is performed during dry weather to discard the contribution from the surface runoff. It is estimated that infiltration into the sewer system is in the order of four to eight times greater than the wastewater flow (Nielsen 2017). This range is based on yearly averages and indicates the difference between sub-catchments. By using the assumption of temporal stationarity, a mean value of six times the wastewater flow is used to estimate the conductance parameter for the four sub-catchments in the sewer system.

The boundary conditions for groundwater recharge and horizontal fluxes are obtained from the Hydrological Information and Forecasting System’s (HIP) database, covering both historical and future periods (Henriksen et al. 2020; Seidenfaden et al. 2022; Henriksen et al. 2023). HIP is a version of Denmark’s national water resource groundwater model, which is designed to produce a national map of groundwater levels under both historical conditions and future conditions under the influence of climate change. The setup applies a two-layer model, which calculates the groundwater formation (recharge) based on precipitation, current evaporation (depending on landuse), and the water content in the root zone. Thus, the unsaturated zone is not included in the groundwater model, which is assumed to be acceptable due to a small unsaturated zone in Kærby with a thickness of 1–1.5 m.

The available boundary condition product from HIP is constructed from 22 climate model runs. Seventeen of these models represent climate scenario RCP 8.5, and five represent climate scenario RCP 4.5 (Henriksen et al. 2020). The selection of climate scenarios is based on the Euro CORDEX initiative described by Jacob et al. (2014). From the 22 different climate model simulations, we extract boundary conditions from the specific simulation: the NorESM1-M/DMI-HIRHAM5 setup for Denmark. This is a RCP 8.5 scenario simulation of the global climate model NorESM1-M from the Norwegian Climate Centre (Bentsen et al. 2012) and the regional climate model HIRHAM5 version 2 from the Danish Meteorological Institute (Christensen et al. 2007). This specific climate model simulation according to Henriksen et al. (2020) is characterized as a median climate model, thus ranked in the middle between ‘wet’ climate models and ‘dry’ climate models. One could argue to select either a ‘dry’ or ‘wet’ climate model or, ideally, an ensemble of climate model simulations; however, in this study, we focus mainly on the relative differences between current, near future, and far future. Despite acknowledging the variability between climate models, for reasons of clarity, we do not study differences between individual climate models nor inter-annual variability and seasonal variability (since we use the periodically stationary model results).

Since the occurrence of snow cover for longer periods is marginal in Denmark due to climate change, we consider all precipitation to be liquid.

The boundary conditions (both fluxes and recharge) are available in horizontal resolution for the current climate and for the future climate scenarios (Schneider et al. 2022). Along the model area boundary outlined in Figure 2, we use the following gridded outputs: the horizontal flow, the head elevation in the saturated zone, and the groundwater recharge.

The groundwater model’s vertical extent is set to 50 m below the top of the limestone layer, where the deeper limestone is assumed to be practically impermeable (Freeze & Witherspoon 1967; Anderson et al. 2015). As a result, the boundary condition between the upper and the lower limestone is treated as a no-flux boundary.

At the northern boundary, the pressure level in the upper layer is set to the semi-stationary sea level at 0 m elevation in the current situation. The tide in Limfjorden is marginal (around 10 cm) and therefore not considered crucial to include. According to Thejll et al. (2023), the expected mean sea level rise in Limfjorden in RCP 8.5 is 0.24 m (uncertainty interval: 0.05–0.57 m) in the near future (2041–2070) and 0.56 m (uncertainty interval: 0.22–1.20 m) in the far future (2071–2100). In this article, the expected mean values are applied in the climate scenario simulations as a fixed head in the northern boundary; thus, values of 0 m for current conditions, 0.24 m for near future conditions, and 0.56 m for far future conditions are applied . It is assumed that no water exchange occurs across the Limfjorden estuary for the deeper layers because the aquifer is unconfined, and the salt intrusion is not considered due to low salinity in Limfjorden. The western, eastern, and southern boundaries maintain the pressure level in all layers, with the pressure level taken from HIP’s periodic stationary model results.

We assume that the calibration based on 1 year is valid for both the historical period and the future scenarios and that the simulated groundwater heads only depend on the specified boundary conditions and the model scenarios.

In simulations of the current state (Scenario A: 1990–2019) of the focus area (Figure 9 and Table 1), the DTW is greater than 1 m. However, 60.1 ha (42 %) of the area has a groundwater table 1–2 m below the terrain level, and a total of 110.0 ha (78 % of the area) has a groundwater table that is less than 3 m below the terrain level. Therefore, results show that there is currently a limited risk of surface flooding, but significant risk of seepage into basements. It is also evident that the most plausible reason for the absence of direct surface flooding is that the current drainage system maintains the groundwater level low due to the infiltration to the sewer system and that the area is not prone to pluvial flooding caused by insufficient surface permeability. As it is illustrated in the cross sectional profiles in Figures 10 and 11, the bottom level of the drainage system is located below the groundwater table in most parts of the focus area and therefore allows for infiltration into the drainage system.

The map in Figure 9 illustrates the spatial distribution of DTW in the current situation. In the vicinity of the Kærby focus area, both the eastern and western sides exhibit DTW exceeding 4 m due to the elevated topography in these areas. The elevated groundwater table is notably higher in the eastern section of the focus area, particularly near the Østerå stream, in contrast to the western part of the focus area. This discrepancy can be attributed to a general higher water level in Østerå due to damming and the distinction that Vestre Landgrøft is partly enclosed in a concrete canal, resulting in differing hydrological interactions between surface water and groundwater compared with those of the Østerå stream. Furthermore, the area east of the Østerå stream has generally a higher groundwater table compared to the focus area. This is due to the fact that this area has a rather new separate sewer system which do not allow infiltrating groundwater. In the simulations of the current situation, there is a decreasing pressure gradient from the Østerå stream towards Kærby and the focus area. This underlines the effect of the infiltration on the drainage system by being inverted compared with the expected gradient in a natural situation.

In Scenario A, the near and far future groundwater table is maintained at almost the same level despite both higher boundary heads and more extensive recharge to the groundwater compared with the current conditions. In the near future, there is an insignificant average increase in the groundwater table of 0.03 m and correspondingly 0.08 m for the far future (Table 1). This is explained by the fact that it is predominantly the hydraulic conductivity of the upper soil layers which is decisive for the fraction of the recharge, which can infiltrate into the sewer system. The conductance (Equation (1)) is thus not the limiting factor causing the excess recharge in future scenarios mainly to infiltrate the sewer system maintaining the groundwater level at a rather constant depth.

Scenario B, where the drainage system is replaced with a non-infiltrating system, is regarded as the worst-case scenario in terms of a potential rise in the groundwater level. Since no initiatives are implemented to decrease the groundwater level, and Scenario B can be considered the natural state of the groundwater level in the focus area. Results show that 26.5 ha (or 19% of the total focus area) in the far future situation has a groundwater table less than 1 m from the surface and 113.6 ha or (81% if the total focus area) has a groundwater table less than 2 m from the surface. Figures 10 and 11 illustrate that the groundwater table almost reaches the ground level (minimum DTW is 0.03 m). There is an average increase in the groundwater table of 0.41 m relative from the current situation to the near future and 0.73 m from the current to the far future. Figure 1 illustrates the distribution of basements across the residential part within the focus area. Contrasting this with Figure 12, which displays the spatial distribution of the depth to groundwater, it becomes evident that the majority of the residential area is characterized as having a groundwater table depth of less than 2 m from the surface, signifying an indeed critical condition in terms of seepage into basements.

In Scenario C, an internal boundary is constructed within the model to simulate the reduction of water levels in the two adjacent lakes positioned to the east and south-east of the focus area. This change in boundary conditions involves a fixed head in the lakes, although in reality, this fixed level would be upheld using pumps. In the simulations, the water levels are maintained at a level 0.05 m lower than the current situation (Scenario A, 1990–2019), essentially preventing the model from accounting for a groundwater level increase due to climate change in the proximity of the lakes.

The model simulations reveal that, within both near and far future scenarios, the water levels adjacent to the lakes remain consistent with their current levels. However, the impact is rather confined and does not significantly affect the groundwater levels in the focus area. This result can be attributed, in part, to the location of the lakes on the opposing side of Østerå. Hence, Scenario C is not is not considered any further.

Scenario D coincides with Scenario A, as a perforated horizontal drain is implemented into the model at the existing drainage system’s level. In practical terms, this drain could be implemented as a third additional drainage pipe within a separate drainage system (comprising one pipe for sewage, one for stormwater, and a third for infiltrating groundwater). Implementing this solution is also feasible from a practical standpoint since the drainage system is already integrated into the roads, allowing for straightforward implementation concurrently with the rehabilitation of the current combined system to a separate system. The difference from Scenario A is the application of a larger value infiltration coefficient that remains uniform across all parts of the drainage network, thus allowing for more infiltration than in the current situation. Despite a minor rise observed in the overall groundwater level in both the near future (0.05 m) and the far future (0.08 m), along with a slight expansion in the area characterized by a depth to the groundwater table (DTW) of less than 2 m compared to the current situation, it remains evident that the implementation of horizontal perforated drains effectively maintains the groundwater level at a lower state than simulated in the worst-case scenario (B).

In Scenario E, the model incorporates the use of three vertical drains, also referred to as extraction wells. The placement of these three vertical drains is determined through an assessment of viable public sites and their suitability to effectively lower the groundwater level in the most critical parts of the focus area. The vertical drains are implemented within the model as boreholes, each equipped with a perforated filter positioned 5 m beneath the surface. The anticipated discharge from each well is set at 150 ⁠. It is worth noting that the specific geographic locations and suggested yields are susceptible to alteration; they have been selected primarily for the purpose of screening the potential effectiveness of this type of solution.

A cone of depression forms around the three wells. The areas of influence around these wells have radii in the order of magnitude 100–150 m. As depicted in Figures 10 and 11, the local impact of pumping groundwater is evident in the creation of localized depressions in the groundwater table. Across the entire focus area, the critical region where DTW is less than 1 m experiences an 86% reduction, and the region with DTW less than 2 m is reduced by 5% when compared to the worst-case situation (Scenario B, far future). Despite representing an improvement over the worst-case scenario, the effects of Scenario E do not provide sufficient decrease in the overall groundwater table to balance the effects by rehabilitation of the drainage system and impacts of climate change.

Scenario F combines Scenarios D and E by incorporating both horizontal and vertical drains. By doing so, it becomes feasible to avoid any significant area with DTW less than 1 m in the far future, thus reducing the risk of surface flooding. In addition, Scenario F demonstrates the smallest rise in the average groundwater table compared with all the far future scenarios, as outlined in Table 1 and illustrated in Figure 8. Despite this, the groundwater table in Scenario F remains higher than the current situation due to the impact of climate change, i.e. increased recharge.

In this article, we have focused on screening different climate scenarios as well as scenarios to represent future technical solutions to remedy problems with an increasing shallow groundwater level. However, we have made some indeed very subjective choices of implementation of scenarios in the presented model setup, e.g. with regards to model calibration, boundary conditions, climate scenarios, annual variability, and technical solutions. These are discussed as follows.

Model calibration: A significant contribution to the model setup is the heterogeneity of the urban geology in the focus area. Alternating layers of sandy and clayey soils make the calibration difficult and lead to a high uncertainty when trying to model the vertical exchanges in the aquifers. The calibration primarily targets the accuracy of the upper layers, with a focus on shallow groundwater levels.

As stated in Section 4.5, the model setup is calibrated using a specific period in 2021 and with other boundary conditions than the ones presented in this article. We apply both some shallow boreholes in the focus area as well as some deeper boreholes in the model area in the PEST optimization. We do consider the model to be representative of the focus area and therefore emphasize the relative changes, using periodically averaged stationary boundaries, from the current situation (1990–2019) to the near future (2041–2070) and far future (2071–2100) situations. Focusing on the relative changes between simulated scenarios, a potential bias between observed and modelled groundwater levels becomes less important in comparing the different scenarios, since all the modelled scenarios are subject to the same bias. It is evident that a groundwater model is subject to significant uncertainties both conceptually as well as in parameterization. The uncertainty of a predicted groundwater level at a certain point is most certainly larger than the differences between scenarios due to local and random errors; however, the specific uncertainty becomes less important by focusing on relative comparison between scenarios relative to absolute values.

Boundary conditions: The boundary condition for the pressure head and recharge is available in a resolution for the current climate conditions, whereas for the future scenarios, the resolution is (Schneider et al. 2022). To investigate the focus areas’ independence of boundary effects and resolution of boundaries, we initially tested different locations of model area boundaries (for the southern boundary) as well as different spatial aggregations of the boundary condition. In the initial simulation testing, we did not see any significant effects of the location of the model boundary nor the data resolution on the simulated groundwater levels and therefore decided to use the boundary condition layers in the original resolution.

Climate scenarios: As argued in Section 4.4, we have selected a median climate model input which according to Henriksen et al. (2020) is ranked between ‘wet’ climate models and ‘dry’ climate models. This is indeed a subjective choice; nevertheless, screening for possible technical solutions to overcome problems with increasing groundwater levels in the focus area as presented here would not benefit further from simulations with the full ensemble. Once the more detailed design technical solutions are settled based on the initial screening, it would be relevant to include the full ensemble of climate simulations as boundary conditions, e.g. as presented by Seidenfaden et al. (2022) and Pastén-Zapata et al. (2022).

Annual variability: The choice of steady-state simulations of a median winter situation for current and future scenarios represents a possible worst-case scenario in regard to a high groundwater table. Both observations from Ziegler et al. (2023) and simulated groundwater heads from Henriksen et al. (2020) demonstrate that winter conditions lead to the highest groundwater levels, driven by increased precipitation and reduced evapotranspiration during this season. One thing which indeed also relevant to investigate in the more detailed technical design process is the annual variability of the groundwater level.

According to Thejll et al. (2023), precipitation is projected to increase in winter, autumn, and spring across all climate scenarios, with winter continuing to contribute the most in terms of high shallow groundwater levels. In contrast, summer precipitation is expected to decrease, leading to longer drought periods but also more intense heavy rainfall events.

In the focus area, significant variability in local groundwater levels between summer and winter has been observed in the past. For example, in extraordinarily dry summers experiencing drought periods, some residential houses in the focus area have had problems with subsidence and following damages to building foundations. It is, therefore, relevant to ensure that the annual fluctuations of the groundwater level are kept rather low in the implementation of technical solutions. Moreover, it would be interesting to conduct a comprehensive study on whether the climate scenarios, with potentially increased precipitation in winter as well as extended drought periods in summer, would have an impact on annual fluctuations of the groundwater levels.

Technical solutions: The five presented technical scenarios (B-F) represent solutions that are technically realistic to be implemented. It is, however, important to emphasize that each of the five solutions could be optimized in more detail to provide generally better performance in terms of lowering the overall groundwater level in the focus area. However, some solutions could indeed be optimized locally, either by allowing an adjustment of capacity in the suggested drainage solutions or by horizontal or vertical replacement of the drainage solutions. Again, we argue that this is subject to further investigation in the design phase.

In this article, a setup of a high-resolution groundwater model is demonstrated. The model can simulate shallow groundwater levels and the interaction with drainage systems and surface water in a small urban area. To detail the relative changes in groundwater level due to rehabilitation of the drainage system and to compare potential solutions under the impact of climate change, a high horizontal resolution is important. This is evident when comparing groundwater levels from the applied boundary conditions which has a resolution that is 100 times smaller than simulated groundwater levels in high resolution. Despite the, for this purpose, coarse resolution of the available boundary conditions, it is a large benefit to be able to implement a nested model in higher detail in an urban area as suggested here. Especially also given the fact that different climate model scenarios and climate model forcings are available from the national DK model Hydrological Information and Prediction (HIP) system.

In the simulations, we examine relative changes in shallow groundwater levels in the current situation, near future, and far future by applying the boundary conditions and inputs from one single model setup available in HIP database with forcing from one climate model. We acknowledge that there is large variability in the available ensemble of 22 groundwater/climate model setups, but since we focus primarily on screening the difference between the three simulated periods as well as the difference between interaction with groundwater and sources of drains, the span of the complete ensemble and the potential bias between climate scenarios becomes less important for the intercomparison of simulation results. However, for more detailed simulations of absolute groundwater levels, we recommend investigating the ensemble variability further. In this case of detail, it can also prove relevant to study annual and seasonal variability in more detail, rather than, as here, employing stationary averages over the three simulated periods.

For the specific case of Kærby, Aalborg, we have simulated the current situation (with climate forcing in an average period 1990–2019). The groundwater level in the focus area is currently maintained low due to the seepage of groundwater into the existing combined drainage system. Simulation clearly shows that the area is highly at risk of experiencing higher groundwater tables in both the near and far future, if the current drainage system is rehabilitated and exchanged with a separate drainage system which do not allow infiltration of groundwater. We have simulated a total of 12 future scenarios ranging from the worst-case scenario where the system is rehabilitated as described earlier to scenarios where different drain solutions are implemented. Results show that it is not possible to avoid the impacts of climate change and that the focus area will experience a generally higher groundwater level. It is, however, possible by implementing horizontal perforated drains in the same alignment as the current drainage system, to reduce impacts of increasing groundwater levels. Implementation of horizontal drains proves the best solution over vertical drainage and reduction of water level locally in lakes. The details of implementing the horizontal perforated drains and their capacity can be still optimized in the detailed planning and design of solutions. Currently, the establishment of drains is not possible within the Danish legislation to be implemented publicly by the water utility service, which is responsible for the reconstruction and operation of the sewer system network.

The authors would like to acknowledge the funding provider of the project GRAVA: VUDP (https://www.danva.dk/viden/vudp/projektuddelinger/grava/) and partners in the GRAVA project: Aalborg Forsyning (Aalborg Public utilities): Anja Sloth Ziegler, Aalborg Municipality, and Niras: Morten Westergaard. Figures 6, 8, 9, 12, and 13 are based on the color scheme DEVON (Crameri et al. 2020) https://doi.org/10.5281/zenodo.8409685.

Data cannot be made publicly available; readers should contact the corresponding author for details.

The authors declare there is no conflict.