Science of the Total Environment 475 (2014) 97–103

Contents lists available at ScienceDirect

Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv

On the utilization of hydrological modelling for road drainage design under climate and land use change Zahra Kalantari a,⁎, Annemarie Briel a, Steve W. Lyon b, Bo Olofsson a, Lennart Folkeson a a b

Department of Land and Water Resources, Royal Institute of Technology/KTH, SE-10044 Stockholm, Sweden Department of Physical Geography and Quaternary Geology, Stockholm University, SE-106 91 Stockholm, Sweden

H I G H L I G H T S • • • • •

The magnitude of water level changes varied with the storm size and seasonality. The increase in runoff responses were more related to season rather than storm size. The dimensions of the studied structures were not sufficient. Upgrading is needed to handle increase in runoff generated by climate and land use changes. The approach has potential to assess the appropriateness of current road structures.

a r t i c l e

i n f o

Article history: Received 29 October 2013 Received in revised form 9 December 2013 Accepted 26 December 2013 Available online 27 January 2014 Keywords: Clear-cutting Extreme storm events Runoff Road infrastructure dimensioning MIKE SHE

a b s t r a c t Road drainage structures are often designed using methods that do not consider process-based representations of a landscape's hydrological response. This may create inadequately sized structures as coupled land cover and climate changes can lead to an amplified hydrological response. This study aims to quantify potential increases of runoff in response to future extreme rain events in a 61 km2 catchment (40% forested) in southwest Sweden using a physically-based hydrological modelling approach. We simulate peak discharge and water level (stage) at two types of pipe bridges and one culvert, both of which are commonly used at Swedish road/stream intersections, under combined forest clear-cutting and future climate scenarios for 2050 and 2100. The frequency of changes in peak flow and water level varies with time (seasonality) and storm size. These changes indicate that the magnitude of peak flow and the runoff response are highly correlated to season rather than storm size. In all scenarios considered, the dimensions of the current culvert are insufficient to handle the increase in water level estimated using a physically-based modelling approach. It also appears that the water level at the pipe bridges changes differently depending on the size and timing of the storm events. The findings of the present study and the approach put forward should be considered when planning investigations on and maintenance for areas at risk of high water flows. In addition, the research highlights the utility of physically-based hydrological models to identify the appropriateness of road drainage structure dimensioning. © 2014 Elsevier B.V. All rights reserved.

1. Introduction Significant changes in climate are predicted worldwide. Such changes are expected to bring about a higher frequency of more intense storm events in many regions. Scandinavia has been identified as one of the most vulnerable regions in Europe with regard to intense rainfall events (Green Paper EU, 2007) and climatic shifts. The Rossby Centre at the Swedish Meteorological and Hydrological Institute (SMHI) has, for example, predicted that the mean temperature in Sweden will increase by approximately 2–3 °C by 2050, which exceeds the predicted global ⁎ Corresponding author. Tel.: +46 8790 7377; fax: +46 8790 6857. E-mail addresses: [email protected] (Z. Kalantari), [email protected] (A. Briel), [email protected] (S.W. Lyon), [email protected] (B. Olofsson), [email protected] (L. Folkeson). 0048-9697/$ – see front matter © 2014 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.scitotenv.2013.12.114

average increase. This is likely to be coupled with an approximately 40–50 mm increase in mean annual precipitation by 2050 with the largest increase during late autumn and winter (according to a coupled atmosphere–ocean regional climate model, RCAO (Doscher et al., 2002; Persson et al., 2007)). In addition, more precipitation will likely fall as rain instead of snow across Scandinavia and northern Europe (Green Paper EU, 2007) increasing the likelihood of rain-on-snow events. A major societal concern of these anticipated changes is the subsequent increase in severe flooding (i.e., Christensen and Christensen, 2003). In addition to threatening safety and endangering human lives, such increased flooding poses a considerable threat to infrastructure. This is particularly true for road infrastructure where flooding can bring about severe obstruction of traffic and costly repair bills (Hansson et al., 2010). Under-dimensioned culverts and bridges can cause damages to roads and generate extensive costs for reconstruction,

98

Z. Kalantari et al. / Science of the Total Environment 475 (2014) 97–103

operation and maintenance (Vägverket, 2002). For example, during the period 1995–2007, the total cost of road infrastructure damage due to high flows and landslides in Sweden were an estimated 1200 million SEK (Holgersson et al., 2007). Most of this damage occurred in the regions already prone to flood damage (Västra Götaland, Värmland and Mellersta Norrland) (Holgersson et al., 2007). As such, there is the potential that future climate change will have large consequences in the form of washed-out roads and embankments and damaged bridges. Road systems are regarded as being especially vulnerable to changes in climate (Koetse and Rietveld, 2009; Kalantari and Folkeson, 2013). This is because, in addition to future shifts in extreme floods, there is a lack of knowledge on the adequate adaptation of infrastructure to climate change (e.g. Eisenack et al., 2012; Kalantari and Folkeson, 2013) both over the short and the long term (e.g. Adger et al., 2009). Longterm planning is particularly important for current and future infrastructure since roads and bridges have long lifespans. To address this, many countries in Europe have modified their guidelines for the design of new road-related constructions in response to predicted changes in climate (Hansson et al., 2010). In addition to shifts in climate, however, changes to land use and land management can also change a landscape's hydrologic response and flooding frequency and amount (Jarsjö et al., 2012; Kalantari et al., 2014). It is, thus, the coupled impact of future climate and land use changes that needs be taken into consideration when developing the guidelines for planning around current and future roads and bridges. This is already done to some extent in countries such as Sweden where current legislation demands that future planned forest clearcutting and future potential climate both be taken into account when designing new and developing maintenance plans for existing road structures (Vägverket, 2005). While on the one hand this may be applauded as a progressive strategy, there is unfortunately no clear path forward on how to implement this strategy. According to the Swedish Transport Administration this process is still very much on a project level. Currently, road drainage structures such as culverts and bridges in the rural areas of Sweden are dimensioned for flows with a return period of 50 years adjusted to a changing climate (Vägverket, 2008). These 50-year flows are calculated using the non-processbased Rational Method (Benzvi, 1989; Maidment, 1993). This largely empirical method does not (and arguably cannot) take into account important factors such as topography, soil conditions and land use that can influence hydrological processes. This can lead to under-dimensioning (Arvidsson et al., 2012). Further, the climate change adjustment implemented is nothing more than a regionally-defined factor-of-safety that cannot account for coupled hydroclimatic interactions or changes in land management. So, even though the legislation to consider coupled land and climate change impacts on road structure design is in place, there is no tool available capable of meeting the needs of authorities to satisfy the legislative requirements. The present study puts forward a process-based modelling experiment to examine the potential impacts of future rain events on roads with the focus on culverts and pipe bridges to explore the adequacy of current designs. This is done through scenario analysis involving extreme weather conditions and clear-cutting of forest analysed for a particular set of road drainage structures within the catchment of the river Hakerud in southwest Sweden. The main novel aspect here is the application of a physically-based hydrologic modelling environment (MIKE-SHE coupled to MIKE 11) to test if current road structures (designed using non-process based approaches to estimate hydrology) are adequate to handle potential coupled land cover and climate changes. Our hypothesis is that the current structures are inadequately sized as coupled land cover and climate changes will lead to amplified hydrological responses not fully anticipated using non-process based approaches. A specific aim of the analysis was to identify the design requirements/dimensions necessitated by intense rainfalls across different seasons to see if the anticipated design “inadequacy” is constant and/or systematic across various conditions (i.e., under differing

dominant hydrologic processes). Further, we utilize the physicallybased hydrological and hydraulic modelling environment to demonstrate a potential “way forward” for road authorities to better consider coupled land and climate change impacts. Based on the results obtained for the region considered, the modelling approach was then appraised regarding its suitability for use in other regions with similar conditions where under-dimensioning of road culverts and bridges can be expected and where adaptation to climate change is needed. 2. Materials and methods 2.1. Study area The study area, the river Hakerud catchment, is located in the municipalities of Vänersborg and Färgelanda in the county of Västra Götaland around 100 km north-east of Gothenburg and close to Lake Vänern, Sweden (Fig. 1). The catchment comprises 61 km2 of about 60% agricultural land and 40% forest land. The river Hakerud discharges into Lake Östra Hästefjorden within the main catchment of the river Göta. The main soil in the area is silty clay mixed with some river sediment. Long-term (1961–2012) mean annual temperature in the area is 6.5 °C with a minimum of −4.4 °C in January/February and a maximum of 16.7 °C in July. Mean annual precipitation (1961–2012) is 862 mm with a minimum of 22 mm in March/April and a maximum of 148 mm in October. This area was chosen because Västra Götaland has been identified as being especially vulnerable to flooding and related road damage (Holgersson et al., 2007). The focus of the study is on a major road culvert and two pipe bridges occurring at stream crossings of major roads (Fig. 1; Table 1). These types of structures are commonly found at stream/road intersections in Sweden. 2.2. Simulation systems There are a large number of models suitable for representing hydrological processes and estimating discharge in catchments of different sizes currently available (Beven, 2012). Here, we chose the physicallybased hydrological model MIKE SHE coupled with the hydraulic model MIKE 11 (DHI Software, 2008) because it is able to simulate the whole hydrological cycle including geographically distributed land use changes. This model combination has also been used in previous studies regionally (Refsgaard et al., 2010; Kalantari et al., 2014) and shown to be appropriate for the task at hand. In the following, we provide a brief overview of the models. 2.2.1. MIKE SHE MIKE SHE is a distributed, dynamic, deterministic and physicallybased model which describes the main hydrological processes in the land phase of the hydrological cycle (DHI Software, 2008). In the present application, the following components were applied: (i) evapotranspiration, including canopy interception, which is calculated according to Kristensen and Jensen (1975); (ii) overland flow, which is calculated with a 2D finite difference diffusive wave approximation of the Saint-Venant equations, using the same 2D mesh as the groundwater component. Overland flow interacts with water courses, the unsaturated zone and the saturated (groundwater) zone; (iii) channel flow, which is described through the river modelling component, MIKE 11, using the Saint-Venant equations (1D); (iv) unsaturated water flow, which is described as a vertical soil profile model that interacts with both overland flow (through ponding) and the groundwater model (the groundwater level is the lower boundary of the unsaturated zone). The 1D Richards equation model approach was used to calculate vertical flow in the unsaturated zone; and (v) saturated (groundwater) flow, which is described mathematically by the 3D Darcy equation and solved numerically by an iterative implicit finite difference technique. For a detailed description of the

Z. Kalantari et al. / Science of the Total Environment 475 (2014) 97–103

99

Fig. 1. The Hakerud catchment NE of Gothenburg, Sweden with the studied structures. A Culvert 1; B Pipe bridge 1; C1 Pipe bridge 2 downstream; and C2 Pipe bridge 2 upstream.

processes included in MIKE SHE, see e.g. Refsgaard and Storm, (1995), Graham and Butts (2005) and DHI Software (2008). 2.2.2. Model parameterisation The calculation grid of MIKE SHE was set to a horizontal resolution of 40 m × 40 m. This resolution applies to all spatial components of the model including overland flow and flow in the unsaturated and saturated zones. The resolution was chosen for reasons of model compatibility in the relatively large catchment considered and to balance spatial resolution with time limitations (e.g., to keep computer run times reasonable on a desktop machine). Parameters representing the vegetation associated with different land uses, such as Leaf Area Index (LAI), root depth (RD) and crop coefficient (Kc), were set up to follow seasonal vegetation patterns similar to Kalantari et al. (2014). Actual evapotranspiration was calculated within the model using a reference potential evapotranspiration from a reference surface (i.e., a hypothetical grass crop with no limitation on water availability (Allen et al., 1998)). The reference potential evapotranspiration was multiplied by Kc to get the potential evapotranspiration for a specific crop. Roughness coefficients for overland flow were taken from Chow (1959) and Arcement and Schneider (1989). Soil data were obtained

Table 1 Characteristics of the pipe bridges and culvert in the Hakerud catchment study area. Structure

Shape

Dimensions (diameter) [m]

Material

Culvert 1 Pipe bridge 1

Circular Circular

0.90 2.47

Pipe bridge 2

Circular

3.83

Concrete Galvanised sheet metal Galvanised sheet metal

from the Geological Survey of Sweden (SGU) and defined to a depth of 10 m in the model. Initial values for the unsaturated conductivity of the soil were taken from Knutsson and Morfeldt (2002). Since no survey data on stream cross-sections were available for the river Hakerud, the elevation reconditioning system for the DEM (horizontal resolution of 2 m × 2 m) was used to obtain an approximation of the channel network dimensions. This was necessary to set up the MIKE 11 model. Data on the location of the culvert, including dimensions and materials, were provided by Metria and drawings of the two pipe bridges by BaTMan (Bridge and Tunnel Management). These were implemented explicitly into the MIKE 11 environment.

2.3. Historical events The intensity-duration-frequency (IDF) curve and recorded precipitation data in the SMHI database 1961–2012 were used to define storm size and rainfall intensity for precipitation events corresponding to the 2-, 5-, 10- and 50-year return periods (Table 2). From these, one observed storm event from the 1961 to 2012 period was chosen for each return period as a representative example considered within the hydrological modelling. The model was set up to perform continuous simulations of the hydrology (see following section) and the runoff hydrographs resulting from each of the four rainfall events were extracted for comparison. These were spread across the various seasons thus allowing us to explore the potential influence of season on hydrological response (similar to Kalantari et al., 2014). Specifically, the modelling included a rainfall–runoff event from autumn 2011 chosen to represent a 2-year storm, an event from winter 2001 to represent a 5-year storm, an event from autumn 2000 to represent a 10-year storm and an event from summer 2004 to represent a 50-year storm.

100

Z. Kalantari et al. / Science of the Total Environment 475 (2014) 97–103

Table 2 Characteristics of the historical events considered, including number of storm events, peak rainfall intensity (mm day−1) recorded in the Swedish Meteorological and Hydrological Institute (SMHI) database 1961–2012 and Q-peak discharge (m3 s−1) at the outlet of the Hakerud catchment.

Number of events during 1961–2012 Event used in simulation Peak rainfall intensity (mm day−1) Q-peak intensity (m3 s−1)

2-year storm

5-year storm

10-year storm

50-year storm

42

11

7

1

6 Sept 2011 25

8 Jan 2001 34

30 Oct 2000 43

9 Jul 2004 54

Q2 = 9.45

Q50 = 12.47

Q50 = 12.74

Q10 = 11.8

Each rainfall event was used to generate peak discharges at the outlet of the Hakerud catchment (Table 2). 2.4. Calibration and sensitivity analysis Model calibration was carried out to find suitable modelling parameters for describing the Hakerud catchment. Daily discharge series 1999–2012 for the catchment were obtained from SMHI and the model was run for a 13-year simulation period from 1 January 1999 to 1 January 2012. This period was selected as it corresponds to a period of high data quality for both rainfall and runoff. The time step and numerical interaction criteria in the model were set to obtain a reasonable compromise between actual simulation times and numerical stability (DHI Software, 2008). In this case, the modelling time step was set at 1 h. Furthermore, parameter values were defined based on the available data and previous studies (e.g. Kalantari et al., 2014) and model stream flows compared against the observed SMHI flow data to establish a base model setup. A manual calibration was then performed on the 50-year storm event with the focus on the most sensitive parameters in the model. This event was selected as it was assumed to provide the widest range of hydrologic response across the catchment. A sensitivity analysis similar to Kalantari et al. (2014) was carried out to investigate the effect of altering certain model parameters on model performance. In this study, the model was found to be more sensitive to the drainage time constant (Tc) and soil saturated hydraulic conductivity (Ks) than other

parameters determining flow, e.g. in the unsaturated zone, drainage and evapotranspiration from vegetation. This is consistent with the findings regionally from Kalantari et al. (2014). Once the main sensitive parameters were identified, the model was calibrated to the 50-year storm event in 2004 (Table 2) by manually adjusting the Tc and Ks parameters in order to obtain the best possible match between the calibrated model and observed discharged. Here, “best” refers to the set of model parameters giving the least difference peak discharge (defined as the difference between simulated and observed peak stream flows) while maximizing the Nash–Sutcliffe efficiency (Nash and Sutcliffe, 1970) between modelled and observed stream flow. 2.5. Validation and model evaluation In order to evaluate the “goodness of fit” of the model to observed flows, the final calibrated model was validated using the other three rainfall–runoff events considered in this study. These were those occurring on 6 Sept 2011, 8 Jan 2001 and 30 Oct 2000 corresponding to the 2-, 5- and 10-year storms, respectively (Fig. 2). In general, there was good agreement between simulated and observed discharge over the calibration and validation events. The peak error (defined as the difference between simulated and observed peak stream flows), the Nash–Sutcliffe Efficiency (Nash and Sutcliffe, 1970) and the coefficient of determination between hourly observed and modelled stream flow (Table 3) were assumed to comprise a good combination of likelihood measures for evaluating the accuracy of both the magnitude and timing of predicted discharge (e.g. Andersen et al., 2001; Beven, 2012; Kalantari et al., 2014). Calibration improved the model performance, although a perfect fit could not be achieved (as should be expected in any modelling study). However, the validation results indicate that the model can be considered suitable for simulating the impacts of climate change and forest clear-cutting on peak discharge and water level at different road structures in the Hakerud catchment. 2.6. Scenario description The calibrated model was run for different scenarios representing changes in climate and land use over the 13-year simulation period from 1 January 1999 to 1 January 2012. All scenarios were compared against the calibrated model result, which was used as the “base case

Fig. 2. Simulated (MIKE SHE) and recorded discharge (SMHI) during a calibration period and three validation periods between 1 January 1999 and 1 January 2012.

Z. Kalantari et al. / Science of the Total Environment 475 (2014) 97–103

101

Table 3 Likelihood measures used for evaluating model accuracy: Peak error (m3 s−1) = Q Peak (sim) − Q Peak (obs), R2 = coefficient of determination, NSE = Nash–Sutcliffe simulation efficiency before and after calibration of discharge during two periods. Period

Likelihood measure

Before calibration

After calibration

Calibration period 9 July 2004 Peak error R2 NSE

−0.11 0.49 0.34

0.51 0.69 0.67

Validation period 6 Sept 2011 Peak error R2 NSE 8 Jan 2001 Peak error R2 NSE 30 Oct 2000 Peak error R2 NSE

−0.93 0.41 0.16 −0.70 0.44 0.20 −0.96 0.68 0.36

0.63 0.53 0.25 0.88 0.64 0.36 0.73 0.86 0.60

scenario”, to evaluate the changes due to the scenario implementation. To assess climatic change impacts, changes in precipitation were implemented based on the results given in Olsson and Foster (2013). These were designed assuming an increase in precipitation of 5% by 2050 and of 20% by 2100 in the region. For temperature changes, the increase predicted by Lind and Kjellström (2008) of 2 °C by 2050 and 4 °C by 2100 was considered. The projections of future climate change in Sweden were provided by RCA3 (Kjellström et al., 2005; Samuelsson et al., 2011), a regional climate model run under, and driven by, a number of scenarios and the Atmosphere–Ocean Global Circulation Models (AOGCMs) (Persson et al., 2007). These were used in the IPCCs fourth assessment report (IPCC, 2007). To analyse the impact of land use changes, total clear-cutting of the area currently forested (40% of the whole catchment) was simulated by altering the vegetation parameters in the model (LAI, RD and Kc) similar to the approach outline in Kalantari et al. (2014). The roughness coefficient (M), which determines overland flow, was also altered for the area covered with forest by adopting the value for grassland. In addition to separate climate and land use changes, the 2100 climate change and forest clear-cutting scenarios were combined and the coupled impact on stream flow modelled. 3. Results and discussion 3.1. The impact of climate and land use changes Clear-cutting only led to a minor increase in peak discharge. There was less than a 5% increase over the flooding under present land use conditions resulting from a 50-year summer storm event (Fig. 3). The lower intensity storms (shorter return periods) resulted in even less increase in peak flows. The frequency of changes in peak flow varied with time (seasonality) and storm size, but the magnitude of the peak flow and runoff responses in the Hakerud river catchment appeared to be relatively small in response to clear-cutting. This is (of course) dependent to some extent on catchment size, as the effect of forest clearcutting is relatively larger for smaller systems (catchments) (Kalantari et al., 2014). Still, this influence of catchment characteristics on hydrological response is typical of Scandinavian systems (Lyon et al., 2010). The 2100 climate change scenario increased peak discharge by 36, 29, 4 and 32% for the 50-, 10-, 5- and 2-year storms, respectively. The 2050 climate change scenario resulted in a smaller increase in peak discharge (11, 9, 3 and 11%, respectively). It is noteworthy that the increase was three-fold higher with the 2100 scenario over the 2050 scenario for all storms studied except for the 5-year return period storm. The reason for this difference in flooding for the 5-year storm event is because this was a winter storm event. Further, when considering the coupled climate and land use changes, the resulting increase in peak discharge

Fig. 3. Percentage change in simulated peak flow during historical storm events in the Hakerud catchment in the different climate change, forest clear-cutting and combination of both scenarios used in MIKE SHE simulation.

was 41, 32, 5 and 38% for the 50-, 10-, 5- and 2-year storms, respectively, with the 5-year storm event (again) standing out. The changes in peak discharge, brought about by both the climate change scenarios and the coupled climate change and forest clear-cutting scenario, indicate that the magnitude of peak flow and the runoff response are more highly related to season rather than storm size. In the present case, the runoff response to increased summer and autumn rainfall intensity was much greater than that to increased winter rainfall intensity. This was due to fluctuations in infiltration capacity as a function of soil moisture. 3.2. Implications for road structure design Comparison of water depth at the three road structures highlighted in Fig. 1 for each storm showed that with the current land use scheme the water depth does not exceed the road structures' dimensions (Table 4). However, all simulated scenarios resulted in an increase in water depth at all structures. For all storms modelled, the water depth in the four potential scenarios exceeded the dimension of Culvert 1 (Table 4). A picture of the culvert taken on 21 May 2013 provides some anecdotal evidence that the structure can easily reach its capacity during a rainfall event (Fig. 1-A). Data obtained from the closest weather station, at Kroppefjäll–Granan, showed that this exceedance of design capacity corresponded to an earlier intense precipitation event with a 2-year return period. The climate change scenarios for the water depth at Pipe bridge 1 during the largest storm events (10- and 50-year events) demonstrated that there is an under-dimensioning problem for this region under future potential climatic change. For Pipe bridge 2, none of the climate scenarios led to major changes in water depth (at least there was no exceedance of the design capacity). The clear-cut scenario resulted in a minor increase in water depth during all storms. When clear-cutting and climate change occurred simultaneously, the water level exceeded both Pipe bridge dimensions in almost all cases, most significantly during the largest storm events. Upgrading of Culvert 1 and Pipe bridge 1 by increasing the dimensions would be necessary to accommodate future climatic changes anticipated using the physically-based modelling environment. In contrast, the simulated water depth at Pipe bridge 2 indicated that its dimensions are sufficient except when looking at the combined climate and land use change scenario with the largest storm events (10- and 50-year). It is important to note here that the results in Table 4 should be regarded as somewhat conservative due to the model limitations. The model did not consider, for example, debris accumulation and associated blockage of drainage structure inlets (e.g. due to forest clearcutting) brought about through lacking maintenance. The maintenance of any structures' inlets is crucial to their functionality. This can again be

102

Z. Kalantari et al. / Science of the Total Environment 475 (2014) 97–103

Table 4 Highest simulated water depth during historical storm events in the Hakerud catchment in the different climate change and forest clear-cutting scenarios used in MIKE SHE simulation; Water depths exceeding the structures' dimensions in bold. Structure

July 2004 (50-year storm) Oct 2000 (10-year storm) Jan 2001 (5-year storm) Sept 2011 (2-year storm)

Culvert 1 Pipe bridge 1 Pipe bridge 2 Culvert 1 Pipe bridge 1 Pipe bridge 2 Culvert 1 Pipe bridge 1 Pipe bridge 2 Culvert 1 Pipe bridge 1 Pipe bridge 2

Dimensions (m)

0.90 2.47 3.83 0.90 2.47 3.83 0.90 2.47 3.83 0.90 2.47 3.83

seen anecdotally from photos collected during a field visit on 22 May 2013 where the inlet of Pipe Bridge 2 was partly blocked (Fig. 1-C2). During this visit, it was evident that a relatively large area must have been flooded, since a rim of debris material could be found half-way up the road embankment. Consequently, the water must have applied pressure causing considerable strain on the road embankment, although it is not known when the blockage happened and for how long the area was flooded. This underlines the conclusions drawn by e.g. Arvidsson et al. (2012) and Kalantari and Folkeson (2013) that regular maintenance is of utmost importance, especially during and after extreme discharge events. Further, the two pipe bridges studied were built in 1995 and presumably dimensioned for 40 years. This means that in 2035 they would need to be replaced and potentially resized. Since 2008 the Swedish Transport Administration (STA) has started to imply a correction factor (generalised over large areas) to adjust the design flow (50-year) in order to consider the effects of a changing climate. This factor is not based on the catchment specific topography, soil conditions and land use. Thus, the current STA method might not be appropriate to calculate accurate dimensions of culverts and pipe bridges. This is particularly true when considering the effects of potentially coupled land use and climate changes as these can be seen to have amplifying effects based on the simulations provided by the physically-based modelling environment (Table 4). 3.3. Potential for extension to other regions and possible uncertainties According to Magnusson et al. (2009), the risk of roads being washed away could be significantly reduced at a reasonable cost by identifying and rectifying weak points in the infrastructure system. The findings of the present study provide an approach to be considered when planning investigations in areas at risk of high water flows in order to identify whether the dimensions of current (and future) road drainage structures are adequate. This leverages the process representation available in the physically-based modelling approach considered. Of course, when prescribing a physically-based representation of catchment-scale hydrological response, much data are required. Prior to more general use, adjustments should be made to the model set-up. This would include collection of survey data on channel geometry to approximate the cross-sectional dimensions in order to increase the accuracy (a large potential weakness of this current study). Furthermore, grid size has a large impact on the simulation time (Xevi et al., 1997). There is a clear trade-off between spatial resolution and computational runtime. Moreover, the river network in MIKE 11 is interpolated to the edges of the MIKE SHE model grid representing the exchange between surface runoff and saturated water flow (Graham and Butts, 2005). Since the grid size considered in this current study is set relatively large, this could lead to a model which does not

Highest simulated water depth (m) Base

2050

2100

Clear-cut

2100+ Clear-cut

0.87 2.26 3.34 0.90 2.41 3.43 0.88 2.35 3.39 0.83 2.13 3.16

1.03 2.49 3.40 1.05 2.55 3.48 0.97 2.40 3.42 0.99 2.40 3.26

1.11 2.58 3.56 1.13 2.63 3.61 1.04 2.57 3.50 1.07 2.51 3.38

0.92 2.34 3.36 0.93 2.47 3.43 0.91 2.41 3.43 0.90 2.22 3.21

1.23 2.61 3.84 1.21 2.61 3.97 1.09 2.59 3.65 1.17 2.57 3.66

represent an accurate solution to the original equations (Beven, 2012). Besides the technical uncertainties, the model does not necessarily take account of changes in the channel dimensions (e.g. caused by erosion), although these could be considerable when making predictions up to the year 2100. In general, when using models and their results for design purposes, one has to bear in mind that they are only “sketches” and not perfect representations of reality. Finally, there are general uncertainties related to future climate change, especially regarding climate conditions on regional and local scale. The results presented here simply show different possible scenarios of which the reality may be quite different. Since these uncertainties exist in all climate modelling, the relevant authorities must develop plans that remain as flexible as possible when formulating guidelines. There is, of course, the need for monitoring and continued development to allow for adjusting design and maintenance guidelines accordingly in order to secure transport infrastructure in the future. Still, taken altogether, the results presented here show that the current road structures, designed using a method lacking any real process representation, cannot adequately handle the future flooding changes brought about by coupled climate and land use changes. In light of the potential shortcomings and uncertainties of the modelling considered, the case can clearly be made that more process-based representations of hydrological responses are necessary for road structure design. This becomes even more relevant as countries implement legislation, such as Sweden, seeking to design and maintain road structures in the face of coupled land and climate changes. 4. Conclusions Most of the risks to the stability of Swedish road constructions relate to precipitation and runoff in different ways. Future climate change can lead to more frequent and more intense rainfall events that result in higher peak discharges possibly causing damage to road embankments. Land use change (e.g. forest clear-cutting) can also alter the hydrological response of a catchment. Hydrological models can be applied to simulate and quantify these changes for future scenarios. Application of the MIKE SHE model to the Hakerud catchment revealed increased peak discharge and water levels in different climate and land use change scenarios. The magnitude of changes varied with the storm size and seasonality while the increase in peak flow and water level were more related to season rather than storm size. In this study, the insufficient capacity of road drainage structures to handle large water volumes anticipated under climate and land use change scenarios indicates the need of local increases to drainage structure dimensions. Clearly, relevant authorities must acquire more knowledge about future changes in discharge behaviour associated with climate and land use changes before making decisions about adaptation of road drainage systems. As part of this, the model approach described here has potential as a

Z. Kalantari et al. / Science of the Total Environment 475 (2014) 97–103

science-based path forward to assess the appropriateness of current road structures. Acknowledgements We gratefully acknowledge funding and data provided by the Swedish Transport Administration (STA). We also thank DHI Support and in particular Mona Sassner and Sten Blomgren for their outstanding support with reviewing of the model set-up. Furthermore, we thank Jonas Olsson and Lennart Wern from SMHI for providing us with meteorological data. The results of this study will be submitted to the STA and used in the research project ‘Adaptation of road drainage structures to climate change’ and other on-going projects coordinated by STA. References Adger WN, Dessai S, Goulden M, Hulme M, Lorenzoni I, Nelson DR, et al. Are there social limits to adaptation to climate change? Clim Change 2009;93:335–54. Allen RG, Pereira LS, Raes D, Smith M. Crop evapotranspiration. Guidelines for computing crop water requirements, FAO Irrigation and Drainage Paper; 1998:56. Andersen J, Refsgaard JC, Jensen KH. Distributed hydrological modelling of the Senegal River Basin — model construction and validation. J Hydrol 2001;247:200–14. Arcement Jr GJ, Schneider VR. Guide for selecting Manning's roughness coefficients for natural channels and flood plains. U.S. Geol. Surv. Water Supply Pap; 1989. p. 2339. Arvidsson AK, Blomqvist G, Erlingsson S, Hellman F, Jägerbrand A, Öberg G. Klimatanpassning av vägkonstruktion, drift och underhåll. (in Swedish). Vti publication, 771. Linköping: Swedish National Road and Transport Research Institute (Vti); 2012. Benzvi A. Toward a new rational method. J Hydraul Eng Asce 1989;115:1241–55. Beven K. Rainfall–runoff modelling, the primer. 2nd ed. West Sussex, England: John Wiley & Sons, Ldt; 2012. Chow VT. Open channel hydraulics. New York: McGraw-Hill; 1959. Christensen JH, Christensen OB. Severe summertime flooding in Europe. Nature 2003;421:805–6. DHI Software. MIKE SHE — user manual. Hørsholm, Denmark: DHI Water & Environment; 2008. Doscher R, Willen U, Jones C, Rutgersson A, Meier HEM, Hansson U, et al. The development of the regional coupled ocean-atmosphere model RCAO. Boreal Environment Research 2002;7:183–92. Eisenack K, Stecker R, Reckien D, Hoffmann E. Adaptation to climate change in the transport sector a review of actions and actors. Mitig Adapt Strat Glob Chang 2012;17:451–69. Graham DN, Butts MB. Flexible integrated watershed modeling with MIKE SHE. In: Singh VP, Frevert DK, editors. Watershed Models. CRC Press; 2005. p. 245–71. Green Paper EU. Adapting to climate change in Europe — options for EU action; 2007. Hansson K, Hellman F, Grauert M, Larsen M. Methods to predict and handle flooding on highways. The blue spot concept. Report 181. Danish Road Institute, Road Directorate; 2010. Holgersson B, Hedlund T, Ahlroth S, Frost C, Rosenqvist P, Thörn P. Klimat- och sårbarhetsutredningen. SOU; 200760 (in Swedish). IPCC. Climate Change 2007: The Physical Science Basis. In: Solomon S, Qin D, Manning M, Chen Z, Marquis M, Averyt KB, Tignor M, Miller HL, editors. Contribution of working

103

group I to the fourth assessment report of the intergovernmental panel on climate change. Cambridge, United Kingdom and New York, NY, USA: Cambridge University Press; 2007. (996 pp.). Jarsjö J, Asokan SM, Prieto C, Bring A, Destouni G. Hydrological responses to climate change conditioned by historic alterations of land-use and water-use. Hydrol Earth Syst Sci 2012;16:1335–47. Kalantari Z, Folkeson L. Road drainage in Sweden: current practice and suggestions for adaptation to climate change. J Infrastruct Syst 2013;19:147–56. Kalantari Z, Lyon SW, Folkeson L, French HK, Stolte J, Jansson P-E, et al. Quantifying the hydrological impact of simulated changes in land use on peak discharge in a small catchment. Sci Total Environ 2014;466–467:741–54. Kjellström E, Bärring L, Gollvik S, Hansson U, Jones C, Samuelsson P, et al. A 140-year simulation of European climate with the new version of the Rossby Centre regional atmospheric climate model (RCA3). Reports meteorology and climatology, 108. Norrköping, Sweden: SMHI; 2005. (SE-60176, 54 pp.). Knutsson G, Morfeldt CO. Grundvatten–Teori & tillämpning (in Swedish). 3rd ed. Stockholm: Svensk Byggtjänst; 2002. Koetse MJ, Rietveld P. The impact of climate change and weather on transport: an overview of empirical findings. Transp Res D Transp Environ 2009;14(3):205–21. Kristensen KJ, Jensen SE. A model for estimating actual evapotranspiration from potential transpiration. Nord Hydrol 1975;6:70–88. Lind P, Kjellström E. Temperature and precipitation changes in Sweden, a wide range of model-based projections for the 21st century. Reports Meteorology and Climatology, 113. SMHI; 2008. Lyon SW, Laudon H, Seibert J, Mörth M, Tetzlaff D, Bishop K. Controls on snowmelt water mean transit times in northern boreal catchments. Hydrol Process 2010;24:1672–84. http://dx.doi.org/10.1002/hyp.7577. Magnusson K, Berglind-Eriksson M, Knutz A, Löfling P. Metod för att hitta och åtgärda vägavsnitt med höga risknivåer till följd av stora nederbördsmängder (in Swedish). Technical Report 490 60 002. Vägverket; 2009. Maidment DR. Handbook of hydrology. New York: McGraw-Hill; 1993. Nash JE, Sutcliffe JV. River flow forecasting through conceptual models part I — a discussion of principles. J Hydrol 1970;103:282–90. Olsson J, Foster K. Extrem korttidsnederbörd i klimatprojektioner för Sverige. (in Swedish). Klimatologi, 6. SMHI; 2013. Persson G, Bärring L, Kjellström E, Strandberg G, Rummukainen M. Climate indices for vulnerability assessments. Report in Meteorology and Climatology, 111. Sweden: SMHI, SE-60176 Norrköping; 2007. 64 pp. Refsgaard JC, Storm B. MIKE SHE. In: Singh VP, editor. Computer models of watershed hydrology. Colorado: Water Resources Publications; 1995. Refsgaard JC, Storm B, Clausen T. Systeme Hydrologique Europeen (SHE): review and perspectives after 30 years development in distributed physically-based hydrological modelling. Hydrol Res 2010;41:355–77. Samuelsson P, Jones CG, WillÉN U, Ullerstig A, Gollvik S, Hansson ULF, et al. The Rossby Centre Regional Climate model RCA3: model description and performance. Tellus A 2011;63:4–23. Vägverket. Ökade vattenflöden - Behov av åtgärder inom väghållningen. (in Swedish). Vägverket publikation, 156. 2002. p. 8–14. Vägverket. ATB Väg 2005, Kapitel D Avvattning och dränering (in Swedish). Technical report. Borlänge: Vägverket; 2005. p. 112. Vägverket. VVMB 310 Hydraulisk Dimensionering (in Swedish). Method description. Borlänge: Vägverket; 2008:61. Xevi E, Christiaens K, Espino A, Sewnandan W, Mallants D, Sørensen H, et al. Calibration, validation and sensitivity analysis of the MIKE-SHE Model using the Neuenkirchen catchment as case study. Water Resour Manag 1997;11:219–42.