Towards improving the hydrologic design of permeable pavements

Due to climate change, the amount and intensity of precipitation in Norway are expected to increase in the future (Hanssen-Bauer et al. 2017). Additionally, the expected increase in temperature will enhance snow melting. On the other hand, urbanization changes the permeable land use area into impervious surfaces which decreases the natural infiltration and hence increases surface runoff. The combined impact of climate change and rapid urbanization will increase the amount of runoff in the urbanized catchment and hence increase the risk of flooding in the future (Yazdanfar & Sharma 2015).

Permeable pavements (PPs) have emerged as sustainable stormwater measures that could mitigate the impact of climate change and restore the pre-development hydrology of urban catchments (Fassman & Blackbourn 2010). PP increases the permeability of urban catchments which enhances infiltration (Kuruppu et al. 2019). Moreover, PP improves the quality of stormwater runoff (Collins et al. 2010), reduces urban temperature (Liu et al. 2020), and enhances groundwater recharge (Yang et al. 2015), among other benefits (Kuruppu et al. 2019).

Regarding the hydrological functionality of PPs, they can be categorized into three main types. The first is the undrained-unlined PP that infiltrates the water into the native soil (type A). The second type is the drained-unlined PP which is equipped with a drainage pipe and at the same time allows for water to infiltrate (type B). The third type is the drained-lined PPs with a drainage pipe and a bottom liner that does not allow for water to infiltrate into the native soil (type C). In general, type A is suitable for locations with deep groundwater levels and high infiltration rates. Type C is preferred for sites with shallow groundwater while type B is suited for locations with low permeability.

The hydrological performance of PP can be quantified by determining retention and detention metrics. Retention is the permanent reduction of stormwater either due to infiltration to native soils beneath the pavement or due to the evaporation from the water stored in the pavement layers. The detention is the delay and attenuation of stormwater flows measured at the drainage network connected to the pavement. The detention is typically estimated by several indicators such as peak reduction, peak delay, and centroid delay (Stovin et al. 2017).

Several field experiments have been conducted in the literature to quantify the hydrological performance of the three permeable pavement types. For type A, high volume reduction rates (up to 90%) due to infiltration were reported, even under soils with low permeability (Dreelin et al. 2006). Ball & Rankin (2010) found surface runoff to only occur over type A pavement for rainfall events with intensity exceeding 20 mm/h. For type B, reported values range between 15 and 70% (Fassman & Blackbourn 2010; Braswell et al. 2018; Winston et al. 2018) depending on rainfall intensity and antecedent conditions. Furthermore, type B pavement was found to detain drainage flow effectively with an average peak reduction of around 80% (Braswell et al. 2018; Winston et al. 2018) and average peak delay values range between 20 min and 2 h (Brattebo & Booth 2003; Fassman & Blackbourn 2010; Braswell et al. 2018). For type C pavement, reported peak reduction values reach up to 98% and peak delay values up to 4 h (Pratt et al. 1995; Abbott & Comino-Mateos 2003; Støvring et al. 2018).

The hydrologic design of PP aims at optimizing parameters of PP which includes the sizes and material properties of PP's layers (i.e. pavement, subbase, and base layers) in order to enhance infiltration, storage, and attenuation of drain outflows, taking into consideration the local conditions (i.e. infiltration capacity of the native soil, connecting sub-catchments, drainage network properties, etc.) and climatic forcing (i.e. precipitation) (Weiss et al. 2019; Kia et al. 2021). The common approach for the hydrologic design of PP relies on synthetic design rainfall hyetographs, such as the Chicago hyetograph method (Keifer & Chu 1957). In this approach, the total storage capacity of the pavement layers is optimized to store a synthetic design rainfall event with a specific return period. Hydrological models could be applied to establish inflow–outflow relationships that can help to design the pavement (Swan & Smith 2010; Martin & Kaye 2014; Martin & Kaye 2015; Jabonero et al. 2018; Kia et al. 2021). However, this approach does not take into consideration the initial saturation of the pavement before rainfall events. Therefore, even if the pavement was designed to accommodate a 20-year return period of rainfall, much smaller rainfall events could cause surface flooding if they fall when the pavement storage is full. An alternative approach is to design PP based on long-term continuous simulation data. This approach, arguably, offers a more reliable approach that can incorporate the effect of climatic conditions on pavement performance. While some papers were found to evaluate the hydrological performance of PP using long-term simulation data (Liu & Chui 2017; Randall et al. 2020), no study was found to use long-term data for the hydrologic design (i.e. sizing) of PP.

The current study assessed the validity of the design rainfall approach for undrained permeable. Type A pavements were designed for five Norwegian cities, using design rainfall (25 years of return period). The reliability of these PP was tested using long-term simulation data (12–30 years) with high temporal resolution (1 min). Furthermore, the study attempted to provide methods for analysing long-term simulation data which can be used for the hydrologic design of permeable pavements.

The study applied the low-impact development module of the Storm Water Management Model (SWMM) (Rossman 2015). The model is widely used for modelling the hydrological performance of permeable pavements (Kim et al. 2015; Chui et al. 2016; Xie et al. 2017; Randall et al. 2020). The validity of the SWMM-LID module has been investigated in many studies by comparing the simulation with measured data. For instance, Liu & Chui (2017) confirmed the ability of the SWMM-LID model to simulate drainage flow from the permeable pavement with high accuracy (Nash–Sutcliffe efficiency higher than 0.9). Similarly, Palla & Gnecco (2015) reported Nash–Sutcliffe efficiency values higher than 0.75 of the simulated drainage flows of a calibrated SWMM-LID model for a laboratory permeable pavement.

In SWMM-LID, the pavement can be represented by five layers: surface, soil, pavement, storage, and drain (Rossman 2015). Each layer contains parameters that control the water movement and need to be adjusted from field measurements, calibration or from the SWMM manual. Precipitation water infiltrates through the pavement layer to the soil layer limited by the permeability of the pavement layer (Rossman 2015). The infiltrated water is stored in the soil layer until the soil moisture reaches the field capacity. After that, the water percolates into the storage layer following Darcy's law (Rossman 2015). The water accumulates in the storage layer and can either infiltrate into the native soil for type A and type B pavements or be discharged to the drainage pipe for type B and type C pavements.

The SWMM-LID assumes a constant value of the infiltration to the native soil (Rossman 2015), which can be set to the minimum infiltration capacity of a given location. Indeed, this assumption is conservative since the infiltration capacity is a function of the soil moisture (when the soil is dry the infiltration capacity is high). However, Martin & Kaye (2014) argued that such an assumption is appropriate since the soil beneath the storage layer takes a longer time to dry in comparison to the soil on the surface.

The soil layer of the LID module is optional and can be used to represent the bedding layer for permeable interlocking concrete pavements (Rossman 2015). The soil layer contains many parameters that are considered sensitive (Randall et al. 2020; Abdalla et al. 2021) which require careful measurement or calibration. On the other hand, for other types of permeable pavement, such as pervious concrete, the soil layer is not required (Jato-Espino et al. 2016; Wang et al. 2019). In this study, the pavement systems were modelled as pervious concrete without soil layers.

The parameters of the SWMM are presented in Table 2. The average values of the recommended ranges of the void ratio of the pavement and storage layers are selected from the SWMM manual (Rossman 2015), following the work of Zhang & Guo (2015). The berm height of the surface layer was selected as zero to allow for surface runoff to occur if the pavement is full while the pavement slope was selected as 1% (Zhang & Guo 2015). The impervious surface fraction parameter is selected arbitrarily as 0.9. This parameter, together with the void ratio of the pavement layer, is only used by the SWMM in clogging conditions (Rossman 2015). In this study, however, we assumed no clogging conditions during the pavement lifetime through frequent maintenance. Randall et al. (2020) found the void ratio of the pavement layer to be an insensitive parameter for the SWMM with no clogging condition. The permeability of the pavement was set as 3,600 mm/h, which is a typical value for permeable pavement that is designed with permeable materials (Zhang & Guo 2015; Vaillancourt et al. 2019).

The depth of the storage layer varied between 10 and 400 mm, which represents the design parameter, controlling the performance of type A pavement. The infiltration capacity of the native soil varied between 0.1 and 20 mm/h, following the study of Muthanna et al. (2018) in which they conducted serval field measurements for the infiltration capacity of urban areas in Trondheim city (Figure 1).

It should be noted that, during the winter months (November–March), most of the precipitation in Norway falls as snow. However, in the current study, snow modelling was not considered. Hence winter months were excluded from the analysis. In addition, all simulations were done for a permeable pavement with an area of 100 m² (8 × 12.5) that is not connected to any adjacent catchment

The hydrologic design of undrained permeable pavement (type A) aims at optimizing the depths of the different layers of the pavements to eliminate or minimize the occurrence of surface runoff (Jabonero et al. 2018; Kia et al. 2021). Indeed, surface runoff in urban catchments is highly undesirable due to the associated negative impact of floods. Additionally, the standing surface water in urban catchments creates odour and health issues (Kuruppu et al. 2019). In this study, the depth of the storage layer was optimized for the reduction of surface runoff.

Flow duration curves (FDC) of the surface runoff were plotted to analyse the performance of PP. This study applied a modified version of FDC that relates between the flow and the duration of time in which this flow value is exceeded, which is recommended by Abdalla et al. (2021). To improve the visualization, FDC were plotted in log-log scales.

In this study, we present two methods for analysing information from FDC, in order to obtain a proper hydrologic design of permeable pavements. In the first method, the probabilities of surface runoff durations above a defined threshold are extracted. In the second method, depth–duration plots are generated which show the relationship between the depth of the storage layer and the duration when the surface flow is higher than zero.

The rainfall events of the five cities were further analysed to compare the climatic conditions of the five cities. Rainfall events were separated from the long-term data sets with 6 h intra-event periods, similar to previous literature into sustainable stormwater measures (Stovin et al. 2012; Johannessen et al. 2018). For each rainfall event, four characteristics were determined: amount (mm), duration (min), average intensity (mm/h), and antecedent dry weather periods (ADWP) (h), which refers to the periods between each two consecutive rainfall events. Cumulative distribution functions of the four characteristics were plotted and compared between the five cities.

The design storm approach underestimated the storage layer depths, as it did not take into consideration the initial saturation of the pavement prior to the rainfall event. One could argue that the design storm approach can be applied if a proper initial saturation is selected. However, it is very difficult, if not impossible, to define a ‘proper’ initial saturation without continuous simulations. It can also be argued the design storm method can be used if the drawdown time is considered; if the drawdown time of a proposed pavement design is very long, the depth can be increased, and/or an underdrained pipe can be installed (Martin & Kaye 2014). However, what defines a ‘long’ drawdown time can be subjective without taking into consideration the long-term rainfall characteristics of the locations.

Another issue with the design storm method is that only one rainfall hyetograph is typically used in the design. Nevertheless, the shape of the design storm hyetograph can affect the flooding outputs of pavements. For instance, Hamouz et al. (2020) analysed the effect of changing rainfall hyetograph on the drainage flow of a green roof. Martin & Kaye (2014) acknowledged the effect of rainfall hyetograph and presented different design charts for each design rainfall hyetograph. However, natural rainfall hyetographs could differ significantly from the design rainfall ones and can be different among cities. Therefore, it can be considered more realistic to use long-term measured precipitation data for validating the effectiveness of a pavement design, which takes into consideration the influence of hyetograph shapes and initial conditions of the specific location.

Long-term simulation of surface runoff can provide valuable information for the PP design. However, it can be challenging to analyse the huge amount of data in order to obtain an appropriate pavement design. Hence, the study suggested two methods to summarize the information from FDC of the surface runoff. If the goal of stormwater management is to eliminate the occurrence of surface runoff, the depth–duration plots using the maximum values, as shown in Figure 9, can be used. If it is acceptable to allow for surface runoff up to a critical value, the probabilities of durations above that critical value can be extracted from the FDC of individual years and can be used for risk management.

In future studies, it is recommended to use long-term precipitation data generated from climate change models, to incorporate the effect of climate change in the design of permeable pavements. Climate change is expected to alter the rainfall characteristics (e.g. intensity, ADWP, durations, etc.). However, the temporal resolutions of the current climatic change models are daily time steps, which is too coarse for such analysis. Recently, Pons et al. (2022) evaluated different statistical approaches to downscaling climate change models from daily to 6-min resolution, which is suitable for stormwater analysis. Hence, the methodology applied in Pons et al. (2022) can be used to generate long-term precipitation data with high resolution, for designing permeable pavements, for future cities.

Undrained pavements (type A) were designed for five Norwegian cities using a design storm with 25 years of return period. The effectiveness of the designed pavements was tested using long-term simulation data (12–30 years) with high temporal resolution (1 min). The common design approach using design rainfall events resulted in underestimation of storage layer depths leading to the frequent occurrence of surface runoff in the pavements, which is considered a failure of the hydrologic design with respect to the functionality of undrained pavements.

Long-term simulation of surface runoff can provide valuable information for the PP design. Due to the difficulties of analysing long-term simulation data, the study proposed methods for summarizing the information from FDC of the surface runoff, which can assist in the hydrologic design of PP. In future studies, it is recommended to use long-term precipitation data generated from climate change models, to incorporate the effect of climate change in the design of permeable pavements.

The study is part of the Drensstein project (grant no. 269526) and the Klima2050 project (grant no. SFI Klima2050 (237859/030)) which are financed by the Research Council of Norway and the project partners.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.