Influence of urban expansion on the hydrology of

0 downloads 0 Views 238KB Size Report
Sep 16, 2011 - (Semadeni-Davies et al., 2008) and more particularly to simulate ... which describe the sewer system, the waste water treatment ... In order to describe not only the sewer system and the receiving water ... Peri-Urban Rivers) project (Braud et al., 2010). .... the lake inlet and by starting a new river at the outlet.

12th International Conference on Urban Drainage, Porto Alegre/Brazil, 11-16 September 2011

Influence of urban expansion on the hydrology of small catchments: development of the suburban PUMMA model by coupling of urban and rural hydrological models. S. Jankowfsky1*, F. Branger1 , I. Braud1, F. Rodriguez2, S. Debionne3 , P. Viallet3 1

Cemagref, UR HHLY, 3bis Quai Chauveau, CP 220, 69336 Lyon Cedex 9, France 2 LCPC, Division Eau et Environnement, BP 4129, 44341, Bouguenais, France 3 HYDROWIDE, 1025 Rue de la Piscine, 38600 Saint-Martin d’Hères, France *Corresponding author, e-mail [email protected]

ABSTRACT Urban expansion mainly affects suburban areas. These areas are subject to rapid modifications such as an increase of impervious areas or concentration of runoff in sewer systems. These changes have an impact on local hydrology and can induce floods, pollution or decrease of groundwater resource. Distributed hydrological models are useful tools for process understanding and water management in these areas. This paper presents the PUMMA model, simulating the rainwater flux both in urban and in rural areas. PUMMA follows an objectoriented approach. The landscape is discretized into cadastral parcels in urban areas and irregular hydrological response units, resulting from intersection of land use, geology, soil and sub-basin maps in rural areas. The considered hydrological processes are infiltration in natural soils and hedgerows, subsurface flow in the saturated layer and overland flow in rural and urban areas, storage in storm water retention basins and flow routing within networks consisting of ditches, natural rivers and sewer pipes. Several drainage networks can coexist and interact (e.g. via storm water overflow devices), which allows the modelling of complex suburban drainage systems with multiple outlets. The paper presents the model structure and the application to the Chaudanne catchment (2.7 km2) which is located in the suburbs of Lyon, France.

KEYWORDS Suburban, distributed hydrological modelling, object oriented, sewer system, ditches

INTRODUCTION Growing urbanisation causes an encroachment of built-up areas in the peripheries of cities, leading to a higher soil imperviousness. Furthermore, urbanization is often followed by installation of sewer systems and sewer overflow devices (SODs), artificial ditches, river regulations or construction of retention basins. These changes influence the hydrological processes in a catchment. Often, a faster catchment response is the result of these anthropogenic modifications, causing floods, a decrease of groundwater resource and a higher risk of water contamination. Distributed hydrological models, capable of simulating these processes, can thus contribute to a better process understanding and water management in suburban areas. They offer the possibility to take into account land use or climate change (Semadeni-Davies et al., 2008) and more particularly to simulate different scenarios of the drainage system (e.g. local storage or infiltration, different connections, etc.). Modelling of suburban areas arises from two different directions: small scale urban hydraulics stormwater

Jankowfsky et al.

1

12th International Conference on Urban Drainage, Porto Alegre/Brazil, 11-16 September 2011 models based on engineering principles, often having a time step of several minutes and large scale hydrologic models designed to describe natural processes, often with a daily time step. As suburban catchments consist of a patchwork of urbanised, natural and agricultural areas (Andrieu and Chocat, 2004) which implies different processes, both approaches are needed. Hence, the engineering tendency was to develop integrated models (Rauch et al. 2002; Muschalla et al., 2007; etc.), which describe the sewer system, the waste water treatment plant (WWTP) and the receiving water body. Concerning hydrological models, the urban area is often represented by means of imperviousness percentages (Jürgens, 2001; Franczyk and Chang, 2009; etc.). In order to describe not only the sewer system and the receiving water body, but also the rural part of the catchment Klawitter and Ostrowski (2005) developed a 2layer approach, coupling a grid-based hydrological model to a pollution load sewer model. This was further developed to the BlueSim model (Reußner et al., 2009). Here, we propose an object oriented approach based on the LIQUID® modelling framework (Viallet et al., 2006; Branger et al., 2010). The suburban catchment is divided in irregular model units. This allows the correct representation of suburban terrain, consisting of urban and rural objects. Each hydrological process is described by a different module and has its own variable time step. The computation time can thus be minimized, as computations are performed only when required. Feedback between process modules is easily possible. New process modules can be added straightforwardly due to existing templates, which allows the construction of customized models. This paper presents the Peri-Urban Model for landscape MAnagement (PUMMA) built within the LIQUID® framework and designed for the detailed simulation of water fluxes in small suburban catchments. The model is designed as a hypotheses testing tool to understand the role of the various landscape objects and to propose a hierarchy of their roles according to the climatic and hydrological context. First a description of the model discretization, retained for suburban areas is given, then the structure of the model and the model components are presented. The model is applied to the Chaudanne catchment, France, and the results are discussed.

DESCRIPTION OF THE PUMMA MODEL Model mesh The model mesh of the PUMMA model consists of a combination of hydrological response units (HRUs) (Flügel, 1995) in the rural area and urban hydrological elements (UHE) (Rodriguez et al., 2008) in the urban area. The HRUs are created by intersection of detailed land-use maps, geology, soil and sub-basin maps. The rural sub-basins are calculated using the stream-burning technique in order to obtain one sub-basin for each river and artificial ditch reach (Jankowfsky et al., submitted). The UHEs encompass one cadastral parcel (typically house plus surrounding yard) and half of the adjoining street. The object oriented model mesh consists of vector objects: polygons for HRUs/UHEs and lines for the drainage network. Model The PUMMA model was developed within the AVuPUR (Assessing the Vulnerability of Peri-Urban Rivers) project (Braud et al., 2010). It combines the BVFT model (Branger et al., 2008), which was developed for the assessment of hydrological impacts of landscape management practices in agricultural areas and the URBS model (Rodriguez et al., 2008), particularly designed for urban areas (see Fig. 1). For this reason, URBS was integrated into the LIQUID® framework. Furthermore, new developments were made describing processes found in suburban areas, such as SODs, retention basins and overland flow routing. As in the Jankowfsky et al.

2

12th International Conference on Urban Drainage, Porto Alegre/Brazil, 11-16 September 2011 BVFT model the choice of the modules in PUMMA depends mainly on the land-use properties. Five main modules are involved: URBS for urban hydrological elements, FRER1D for natural surfaces (forests, agricultural fields) and roads, HEDGE for vegetated field borders and riparian zones, SISTBA for retention basins and lakes and RIVER1D for the drainage network (natural river and artificial drainage network). Each model unit (HRU/UHE) represents one object, which has its own interactions and time step. The water flux inside the process modules is 1D and vertical. The lateral subsurface and surface flow exchange is modelled with interface modules which adds a second dimension to the model. The next two paragraphs explain the modules and their connections in more detail. Modules The URBS module (Rodriguez et al., 2008), calculates the components of the rainwater flux at the UHE scale. The UHE is divided into three land-use classes: street, impervious area and natural area (Rodriguez et al., 2008). Each land-use class is represented by four reservoirs (tree, surface, vadose zone and saturated zone). With this representation it is possible to describe the following processes: interception by trees, evapotranspiration, surface runoff, soil infiltration and infiltration into sewer pipes due to water tightness defects. The FRER1D module represents vertical infiltration over natural surfaces using a 1D resolution of the Richards equation. FRER1D can produce saturation overland flow when the soil column or the first horizon is fully saturated. It can also produce Horton overland flow, when the infiltration capacity is exceeded (typically on streets). As input of the FRER1D module, the evapotranspiration is calculated by an extra set of modules, integrating the effects of crop rotation and root extraction (Varado et al., 2006). The conceptual HEDGE module calculates the water table dynamics and evapotranspiration processes below hedgerows. Lateral flow in the saturated zone is represented in HEDGE, FRER1D and URBS modules by source/sink terms. The SISTBA (SImulation of STorage BAsins) module, developed especially for the PUMMA model, uses a simple linear reservoir equation to simulate retention basins or natural lakes. It has a lower outflow depending on the hydraulic head and an overflow. Evaporation, precipitation, irrigation and water exchange with the groundwater table are taken into account by source/sink terms. Natural river network, ditches and sewer system are modelled with the RIVER1D module (Branger et al., 2010), which calculates a solution of the one-dimensional kinematic wave approximation and the Manning-Strickler equation. The drainage network is divided into several reaches that can receive lateral surface and subsurface flows. Each reach can have a different cross section (rectangular, trapezoidal or triangular) and Manning coefficient value. Input modules allow the distribution of rainfall and potential evapotranspiration over the model units. Nearly all the current state variables such as soil moisture content, water levels of the saturated zone, river reach discharge and actual evapotranspiration can be extracted as distributed outputs. Spatial coupling The spatial coupling between different modules is realised by means of interfaces (Figure 1). They simulate the subsurface flow between adjacent model units. The Water Table Interface (WTI) uses the Darcy law to calculate the subsurface flow between agricultural fields, hedgerows and urban cadastral parcels. The water exchange between river, or lake and adjacent agricultural field, hedgerow or UHE is computed in the Water Table River Interface (WTRI). The calculated flow direction depends on the hydraulic head inside the model units and is thus bidirectional. Depending on the flow direction the computed discharge, which is taken into account in the receiving modules due to source or sink terms, is positive or negative. The interfaces are line objects defined by intersection between polygons. Ponding

Jankowfsky et al.

3

12th International Conference on Urban Drainage, Porto Alegre/Brazil, 11-16 September 2011 over FRER1D / HEDGE modules is extracted with the OLAF (OverLAnd Flow) interface. The overland flow is routed unidirectionally from model unit to model unit until it reaches a river branch, following topography based on the Manning-Strickler equation. The surface outflow and network infiltration of the URBS module are directly injected into the closest drainage network stream. Several rivers (instantiations of RIVER1D module) can coexist and exchange water via source and sink terms. This facilitates the representation of an artificial drainage network besides a natural river. SODs are simulated by the TDSO (Threshold Dependent Stormwater Overflow) interface between sewer network and river streams. The overflow is calculated with a weir or orifice equation based on a control water depth threshold value. The connection between river and lake is modelled by interrupting the river network at the lake inlet and by starting a new river at the outlet.

Figure 1. Structure of the PUMMA model and coupling between the modules.

MODEL APPLICATION AND RESULTS Catchment description and model parameterisation The model was applied to the upper part of the Chaudanne catchment (2.7 km2), which is located in the suburbs of Lyon, France. Its headwater consists mostly of rural areas, whereas the lower part of the catchment is covered by urban residential zones. The rural part is drained by roadside ditches and the natural river. Both combined and separated sewer systems, including ditches, can be found in the urban part. The combined sewer system is connected to the natural river via a SOD. The Chaudanne catchment has three outlets: the natural river, the pipe towards the WWTP and an overflow towards another sewer system controlled by a SOD (Jankowfsky et al., submitted). The precipitation is measured at one rain gauge located inside the catchment. Continuous discharge measurements are available at the gauging stations of Pont de la Barge upstream and downstream of SOD 1, SOD 1 and the pipe towards the WWTP (Figure 2) (however with large uncertainties for the station downstream of SOD 1). Furthermore, the water level is measured inside the first retention basin, located at the catchment outlet (Figure 2). The simulation outputs are compared to these measured values. For this first application of PUMMA the HEDGE module was taken for all HRUs and URBS for the UHEs. The soil with a silty and clayey sand texture (SIRA, 2010), was parameterized for the first 10 cm with data acquired during a field survey (Gonzalez-Sosa et al., 2010) and for the lower soil horizons with data from the DONESOL databank (SIRA, 2010). As no Jankowfsky et al.

4

12th International Conference on Urban Drainage, Porto Alegre/Brazil, 11-16 September 2011 information of the real soil depth was available, a homogenous soil depth of 1m was assumed in the rural zones. The depth of the urban soil was set to the depth of the sewer pipes. For each UHE the built, road and natural surfaces and the percentage covered by trees were calculated based on detailed land use data (Béal et al., 2009), see Figure 2. For the typical urban parameters, such as the permeability of roads and buildings and the size of their surface reservoirs the same parameters as Rodriguez et al. (2008) were used as default values. The section and Manning values needed for RIVER1D were obtained by field investigations. One value per river branch was collected. The slope and elevation were calculated with the digital elevation model, except the sewer slope which resulted from the sewer data. The SODs were parameterized with measured data as far as available. The rest of the parameters was estimated. The average transit time of the retention basin was adjusted using the retention basin data, as no measured value of the transit time was available. A typical summer storm event (27.8 mm) on the first of August 2008 was chosen for the simulation. The Chaudanne river, which is an intermittent stream, was completely dry prior to the storm event. The water level in the stream and the water table level in the soil was therefore initialised to zero. No waste water was simulated in the sewer system. For such an event, only the urban UHEs are expected to contribute to the discharge.

Figure 2. The model mesh of the Chaudanne catchment having as downstream limit the measurement stations at Pont de la Barge and the series of three retention basins. The distribution of the modules is shown, as well as the numbers of the SODs. Results and discussion Figures 3 and 4 show the rain event and the simulated discharge at the gauging stations of the Chaudanne catchment versus the measured discharge. As expected only the urban UHEs contribute to the discharge. Without any calibration, the model simulates relatively accurately the peak time and response time, and thus the urban dynamics. However, using the default values for the UHEs (link coeff = 1), the model overestimates the flow volumes and peaks. One explanation can be that we connected all water from the impermeable surfaces (including dispersed settlements of the rural part) directly to the closest drainage network. The low permeability and storage capacity of the built and road surfaces generates direct runoff. Jankowfsky et al.

5

12th International Conference on Urban Drainage, Porto Alegre/Brazil, 11-16 September 2011 However, the UHE connections are quite uncertain, especially in dispersed settlements (Jankowfsky et al., submitted ), and part of the runoff created on roads and roof tops is likely to re-infiltrate, whereas the default URBS values assume that it flows directly into the drainage system. As an illustration of the model capability to assess the relevance of various hypotheses, Figs. 3 and 4 also show the simulated discharges when the road link coefficient is set to 0.5, and the coefficient for the buildings to 0.6, instead of 1. This allows to consider the infiltration of runoff created on impermeable areas on the neighbouring permeable surfaces. The overflowing discharge could be relatively well represented. Depending on the link coefficients, two or three overflow events were detected. The maximum level of the retention basin was set to the real basin depth, whereas the transit time of the retention basin of one hour fifty was adjusted. The interpretation of the simulated water level in the retention basin should be done carefully, as the model does not take into account the real basin geometry (Fig. 4).

Figure 3. Simulated (for different link coefficients) versus measured flow data upstream and downstream of the SOD and the precipitation (variable time step) during the simulation run (1/8/2008). The simulated discharge corresponds to the RIVER1D output.

CONCLUSIONS AND PERSPECTIVES Hydrologic processes in suburban areas are quite complex. Natural flow paths are modified by the introduction of ditches, sewer systems, retention basins, etc. The distributed hydrological model PUMMA was developed in order to simulate these different processes. It represents the agricultural influence, such as hedges or ditches as well as the urban influence on small streams. Muschalla et al. (2007) showed that grid-based models are not well suited for urban zones. Hence, the object oriented, vector based approach of PUMMA facilitates the Jankowfsky et al.

6

12th International Conference on Urban Drainage, Porto Alegre/Brazil, 11-16 September 2011 representation of the suburban terrain, as each object or model unit is independent, but at the same time it can interact with all other objects. Therefore, the model can represent several connected urban areas, as well as dispersed settlements and parallel drainage networks with opposite flow directions. With PUMMA’s variable time step it is possible to simulate the slow natural processes and the fast urban catchment response at the same time. The model can be used for event-based as well as continuous long term simulations. Yet, because of the distributed character of the model, many input data are necessary and the model setup can be quite tedious. Subsequently, the model application remains restricted to small catchments of several square kilometres. However, an automatic pre-processing of the geographical data facilitates the preparation of the data.

Figure 4. Simulated (for different link coefficients) versus measured flow data in the sewer system downstream of SOD 1 and in SOD 1 and simulated versus measured water level in the first retention basin at Pont de la Barge during the simulation run (1/8/2008). A summer storm event was chosen for this first application. With the use of a maximum of a priori information, it was possible to correctly simulate the urban flow dynamics. Questions arise, such as how to identify the contributing zones to the stream flow (Jankowfsky et al., submitted) and which water contributes to the direct runoff. The model appears as a powerful tool for testing various functioning hypotheses as for example a different representation of the UHEs of dispersed settlements in the rural part and highlights the most sensitive parameters for which a better characterization should be necessary. The next steps consist of simulating a winter rainfall event and to undertake long term simulations. The use of the FRER1D and associated evapotranspiration modules should also better take into account the nature of vegetation cover and crop rotation in agricultural fields. Jankowfsky et al.

7

12th International Conference on Urban Drainage, Porto Alegre/Brazil, 11-16 September 2011

ACKNOWLEDGMENTS The development of the PUMMA model is part of the AVuPUR project funded by the French National Research Agency (ANR) under contract ANR-07-VULN-01. This work has also been supported by Cemagref and Hydrowide fundings. CCVL, IGN, Grand Lyon, SAGYRC, SIAVHY, Sol-Info Rhône-Alpes, UMR 5600 EVS/CNRS provided data used in the study. Special mention should be made of Pedro Sanzana, Florent Brossard and Yvan Paillé who created the model mesh for the Chaudanne catchment.

REFERENCES Andrieu H. and Chocat B. (2004). Introduction to the special issue on urban hydrology. Journal of Hydrology 299, 163-165. Béal, D., Gagnage, M., Jacqueminet, C., Kermadi, S., Michel, C., Jankowfsky, S., Branger, F., Braud, I. (2009). Cartographie de l’occupation du sol pour la modélisation hydrologique spatialisée du cycle de l’eau en zone péri-urbaine, Proc. 2ème atelier SIDE2009, Biramonte S., Miralles, A., Pinet, F. (Eds), Toulouse/France, 26 May, 23-32. http://eric.univ-lyon2.fr/~sbimonte/side2009.html, consulted on 29/3/2011. Branger F., Braud I., Viallet P. and Debionne S. (2008). Modelling the influence of landscape management practices on the hydrology of a small agricultural catchment. In Wang S.S.Y., M. Kawahara, K.P. Holtz, T. Tsujimoto and Y. Toda (Eds), Proceedings of the 8th International Conference on HydroScience and Engineering, Nagoya, Japan, 586-594. Branger F., Braud I., Debionne S., Viallet P., Dehotin J., Henine H., Nedelec Y. and Anquetin S. (2010). Towards multi-scale integrated hydrological models using the LIQUID® framework. Overview of the concepts and first application examples. Environmental Modelling & Software 25(12), 1672-1681. Braud, I. et al. (2009). The AVuPUR project (Assessing the Vulnerabiliy of Peri-Urban Rivers) : experimental set up, modelling strategy and first results, Proceedings of the 7th Novatech 2010 Conference, June 28-July 1 2010, Lyon, France, 10pp. Flügel W.A. (1995). Delineating hydrological response units by geographical information system analyses for regional hydrological modelling using PRMS/MMS in the drainage basin of the River Bröl, Germany. Hydrological Processes, 9(3-4), 423-436. Franczyk J. and Chang H. (2009). The effects of climate change and urbanization on the runoff of the Rock Creek basin in the Portland metropolitan area, Oregon, USA. Hydrological Processes 23, 805-815. Gonzalez-Sosa E., Braud I., Dehotin J., Lassabatère L., Angulo-Jaramillo R., Lagouy M., Branger F., Jacqueminet C., Kermadi S., Michell K. (2010). Impact of land use on the hydraullic properties of the topsoil in a small French catchment. Hydrological Processes, 24, 2382-2399. Jankowfsky S., Branger F. Braud I., Gironás J. and Rodriguez F. Integration of sewer system maps and field observations in topographically based subcatchment delineation in suburban areas. Submitted to Hydr. Proc. Jürgens C. (2001). Application of a hydrological model with integration of remote sensing and GIS techniques for the analysis of land-use change effects upon river discharge. Remote Sensing and Hydrology 2001. Proceedings of a symposium held at Santa Fe, New Mexico, USA, April 2000. IAHS Publ. No. 267, 2001. Klawitter A. and Ostrowski M.W. (2005). A 2-layer approach to simultaneously model rainfall runoff events in urban catchments and in their rural surroundings. Proc. 10th Int. Conf. On Urban Drainage, Copenhagen, Denmark, 21-26 August 2008.8pp. Muschalla D., Ostrowski M. and Klawitter A. (2007). Innovative simulation and optimisation tools for basinwide urban stormwater management. Int. Symposium on New Directions in Urban Water Management. UNESCO Paris, 12-14 September. 8 pp. Rauch W., Bertrand-Krajewski J.L., Krebs P., Mark O., Schilling W., Schütze M. Vanrolleghem P.A. (2002). Deterministic modelling of integrated urban drainage systems. Wat. Sci. Tech. 45(3). 81-94. Reußner F., Alex J., Bach M., Schütze M., Muschalla D (2009). Basin-wide integrated modelling via OpenMI considering multiple urban catchments. Wat. Sci. Tech., 60(5), 1241-1248. Rodriguez F, H. Andrieu, F. Morena (2008). A distributed hydrological model for urbanized areas – Model development and application to case studies. Journal of Hydrology 351, 268-287. Semadeni-Davies A., C. Hernebring, G. Svensson, L.G. Gustafsson. (2008). The impacts of climate change and urbanisation on drainage in Helsingborg, Sweden: Suburban stormwater, Journal of Hydrology 350, 114-125. SIRA. (2011). Sol Info Rhône-Alpes, http://www.rhone-alpes.chambagri.fr/sira/. Consulted on 2011/03/29. Varado N., I. Braud and P.J. Ross (2006). Development and assessment of an efficient numerical solution of the Richards’ equation including root extraction by plants, Journal of Hydrology 323(1-4), 258-275. Viallet P. S. Debionne, I. Braud, J. Dehotin, R. Haverkamp, Z. Saâdi, S. Anquetin, F. Branger and N. Varado (2006). Towards multi-scale integrated hydrological models using the LIQUID® framework. In Gourbesville, P., J. Cunge, V. Guinot and SY. Liong (Eds), Proc. 7th Int. Conf. on Hydroinformatics, Nice, France, 542-549.

Jankowfsky et al.

8

Suggest Documents