Comparison of 1D-1D and 1D-2D Urban Flood Models PDF

15th International Conference on Environmental Science and Technology Rhodes, Greece, 31 August to 2 September 2017 Comparison of 1D-1D and 1D-2D urban flood models Kourtis I. M.1, *, Bellos V.1, Tsihrintzis V. A.1 1 Schoo...

15th International Conference on Environmental Science and Technology Rhodes, Greece, 31 August to 2 September 2017 Comparison of 1D-1D and 1D-2D urban flood models Kourtis I. M.1, *, Bellos V.1, Tsihrintzis V. A.1 1 School of Rural and Surveying Engineering, National Technical University of Athens, 9 Iroon Polytechniou St., Zografou 15780 Athens, Greece *corresponding author: e-mail:[email protected]; [email protected] Abstract: The present study aims to compare two 2001; Elliot and Trowsdale 2007; Neelz and Pender different modelling approaches in the assessment of 2009). Two of the most well-known and most used urban flooding. Α real case study is used, which is a models are SWMM (Rossman 2010) and MOUSE small urban catchment located in the center of (DHI 2016a). Athens, Greece (Ano Patisia, Kypseli). In the first Conventional modelling approaches (1D and 1D-1D) modelling approach (1D-1D), the combined sewer are able to simulate quite accurately the drainage system and the surface system are coupled using the network. However, in cases of major rainfall events, Storm Water Management Model (SWMM), which these types of models are not able to simulate simulates flow both in the storm sewer system and on inundation depth in built-up areas and to visualize the surface (streets). SWMM solves the 1D Shallow flood extent (Bisht et al. 2016). For the representation Water Equations (1D-SWE) in both sewer and of the surface flooding depth and extent, more surface systems as a set of links and nodes. In the accurate models than 1D-1D are needed, such as the second modelling approach (1D-2D), the surface and 1D-2D models, which are based on the 2D-SWE and sewer system are coupled using MIKE URBAN and are solid tools for modelling and simulating flooding MIKE FLOOD. The coupled model solves the 2D- in urban areas (Leandro et al. 2009). SWE in the surface system and the 1D-SWE in the sewer system. The results show the importance of This paper presents and assesses two different considering the interaction of sewer and surface modeling approaches for the assessment of urban system when modelling urban drainage networks. flooding in a small urban catchment located in the The 1D-2D coupled models can be a very useful tool center of Athens, Greece (Ano Patisia, Kypseli). In in simulating flood extent and flood inundation in the first modelling approach (1D-1D), the combined urban areas. The comparison provides an insight into sewer system and the surface system are coupled the limitations of 1D-1D models in simulating flood using the Storm Water Management Model extent and flood inundation, problems that can be (SWMM), while in the second modelling approach overcome by using 1D-2D coupled models. (1D-2D), the surface and sewer system are coupled using MIKE URBAN and MIKE FLOOD. Due to the Keywords: Urban flooding, 1D-1D model, 1D-2D fact that the site is ungauged and there are no flow model, SWMM, MIKE URBAN-MIKE FLOOD measurements in the pipes of the system, the simulations took place using the following Intensity 1. Introduction Duration Frequency (IDF) curve of the area (Mimikou et al. 2000): In highly populated urban areas, floods are among the most common and catastrophic natural hazards as i  15.39T (0.276) d ( 0.725) (1) they affect most parts (economic, social etc.) of human life and infrastructure (Tsakiris 2013; Pistrika et al. 2014; Bellos and Tsakiris 2015). 2. Materials and Methods Nowadays, several pieces of software, both commercial and non-commercial, are available for 2.1 Case Study Area the hydrologic and hydraulic simulation of the stormwater runoff quantity and quality. Many The oldest part of the drainage network of Athens researchers have reviewed the applicability of various comprises a set of conduits and facilities that collect rainfall-runoff models for urban areas (e.g., Zoppou and drain the combined flow of stormwater and wastewater (combined sewers), which is divided into computations the momentum, mass and energy the following subcatchments: B, C, D, E, F, Z1, Z2, conservation laws (Rossman 2010). SWMM was I-H, H1, H2 and Th. It covers a total area of 1310 ha primarily developed for urban areas and can be used and the wastewater drains in the Central Sewerage for the design, analysis and planning of drainage Pipeline (CSP) while the stormwater drains in Kifisos systems, and for the simulation of runoff quality River and in the stream of Prophet Daniel. The (e.g., Zhu et al. 2016; El-Sharif and Hansen 2001; subcatchment modeled in the present study was D Hsu et al. 2000; Tsihrintzis and Hamid 1998). (89 ha) (Fig. 1). The study area is a highly urbanized In our study, the Dynamic Wave (DW) model was and impervious area, located in the region of Ano used for the hydraulic calculations and infiltration Patisia, Kypseli (Athens, Greece). The combined was calculated using the Curve Number (CN) sewer network consists of 112 nodes and 79 method. The main parameters for a subcatchment in combined sewer pipes, with a total length of about 5 SWMM software are: (i) area (ha); (ii) width (m); km. The drainage system comprises either circular (iii) slope (%); (iv) percent impervious; (v) pipes (newest part of the network) with diameters Manning’s n for pervious and impervious areas; (vi) ranging from 0.3 to 0.6 m, or egg-shaped pipes depression storage (mm) in pervious and impervious (oldest part of the network) with depths ranging from areas. Parameters (i), (iii) and (iv) were determined 0.9 to 2.4 m. The slopes of the pipes range from 0.6 using the appropriate tools in ArcGIS 10.3.1. The to 10.8 % (with an average of 3 %). Fig. 1 presents width of the subcatchments (parameter ii) was the aerial photo of Zone D of the combined drainage determined as the area divided by the average system of Athens (left) and the combined sewer maximum length of the subcatchment (Rossman system, the boundary of the study area and the land 2010). Finally, for parameters (v) and (vi), typical uses (right) (Corine 2006). values from the literature were used (ASCE 1992; Rossman 2010). The model parameters and their 2.2 SWMM Model variation ranges are shown in Table 1. EPASWMM5 is a fully dynamic rainfall-runoff simulation model employing in hydraulic Figure 1: Aerial view of Zone D of the combined drainage network of Athens (left); Representation of the combined drainage system of subcatchment D and land uses (right) (Corine 2006) Table 1: Key model parameters involved in this study No. Parameter Description Range No. Parameter Description Range Manning’s N for Con- Manning’s N for the 1 N-Imperv 0.013 6 0.013~0.014 impervious area Manning conduits-roads Manning’s N for Average percentage 2 N-Perv 0.10 7 Slope-sub 0.082~9.037 pervious area surface slope (%) Depth of depression Dstore- Subcatchments width 3 storage on impervious 2 8 Width-sub 7.93~162.77 Imperv (m) area (mm) Depth of depression Dstore- %Imperv- Subcatchments percent 4 storage on pervious 5.51 9 45~90 Perv sub of impervious area (%) area (mm) 5 CN Curve Number 77~94 2.3 MIKE URBAN-MIKE FLOOD Models 3. Results and Discussion MIKE URBAN is a hydraulic pipe flow model based 3.1 1D-1D vs 1D-2D Model on the MOUSE/MIKE11 engine which solves the full form of the 1D-SWE (DHI 2016a). Moreover, MIKE Due to the fact that the site is ungauged, simulations URBAN incorporates the SWMM engine. The main in both models (1D-1D and 1D-2D) took place using advantages of MIKE URBAN over SWMM5 are that MIKE URBAN offers GIS integration, and the IDF curve presented in Eq. (1). The simulations, moreover, it offers the capability for 2D simulations took place for two synthetic design storms of 1-hour for the overland flow paths through the coupling with duration and for return periods of 10 (29.06 mm) and the MIKE FLOOD software (DHI 2016a; DHI 25 years (37.42 mm). The rainfall distributions for 2016b.) the synthetic design storms were developed using the Alternating Block Method (USBR, 1977). MIKE FLOOD is a hydrodynamic surface flow model based on the MIKE21 engine which solves the The most representative results for both models (1D- 2D-SWE in a structured grid (DHI 2016c). MIKE 1D and 1D-2D) are reported here, which include URBAN and MIKE FLOOD are coupled in order to runoff produced from the subcatchments (hydrologic cope with the interaction between the underground model), flow in combined sewers, flow in the combined drainage system flow and the flow on the overland surface system (only for the 1D-1D model) surface system spilled by the manholes of the system. and the extent of flood inundation (only for the 1D- The main components for an integrated 1D-2D 2D model). simulation are: (i) data for the combined drainage Fig. 2 shows the runoff produced from one network (e.g., subcatchments area, pipes shape, subcatchment of the system (hydrologic model), for length, Manning coefficient, manhole location and return periods of 10 and 25 years and for duration of ground and surface elevation etc.); (ii) Digital 1 h, for the 1D-1D and the 1D-2D models. As it can Elevation Model (DEM); (iii) definition of 2D model be observed, runoff predicted from the two models area and resolution; (iv) specification of flooding and (Fig. 2) is in good agreement. The time of peak and drying depth; (v) Manning number for overland the peak runoff are comparable, while the runoff surface paths; (vi) equation for the flow exchange at volumes produced from the subcatchment differ only the inlet between the 1D and 2D models (orifice slightly. The runoff calculated with the 1D-1D was equation, weir equation or exponential equation). 2.3 m3 and 4.0 m3, while the volume calculated with In our study, for the hydrologic calculations in each the 1D-2D model was 3.2 m3 and 4.1 m3, for return subcatchment, the Kinematic Wave method was periods of 10 and 25 years, respectively, and for selected. The main parameters were the area of each storm duration of 1 h. Fig. 3 presents the flow in a subcatchment, the percent of imperviousness of the combined sewer of the drainage network (1D-1D and subcatchment and the subcatchment slope (Table 1). 1D-2D model) for return periods of 10 and 25 years For the loss model, Horton’s equation was the only and for storm duration of 1 h. The flow in the sewers choice and the default parameters were used for simulated by the two models show differences. This simplicity. For the exchange of flow between the 1D is probably because the 1D-1D model uses for the and the 2D models, the orifice equation was used. hydraulic computations the SWMM engine, whereas Finally, it should be mentioned that the 2D model is the 1D-2D model uses the MIKE11 engine. based on a 5x5m grid square-shaped cell, obtained Moreover, overflow from the manholes of the system from the same DEM. is simulated with the orifice equation in the 1D-2D model, whereas in the 1D-1D model any flow in excess of the sewer pipe capacity is automatically Leandro et al. (2009) the above procedure is diverted to the open surface system (Gironás et al. considered inaccurate. The simulations of the 1D-1D 2009). In Fig. 4, one can see the flow in the overland model showed a maximum depth of water in the surface system (roads), simulated with the 1D-1D surface network about 0.10 m and 0.13 m, maximum model, for return period of 10 and 25 years and storm velocity about 3.1 m/s and 4.1 m/s and a maximum duration of 1 h. The sewer system is surcharged and flux of about 1.3 m3/s and 2.7 m3/s, for return periods the water level is high enough to cause water to flow of 10 and 25 years and for duration of 1 h, out from the drainage system to the overland surface respectively. On the other hand, with the 1D-2D system (roads). Water flow at the downstream end of model the mean maximum depth of water, in the the surface system is zero. This is not the case, but an surface network, was about 0.17 m and 0.19 m, the artifact of the way that SWMM5 model draws the mean maximum velocity was about 0.43 m/s and water surface profile within an open channel (Gironás 0.60 m/s and the mean maximum flux was about 0.07 et al. 2009). Regarding the flood-inundation maps, m3/s/m and 0.11 m3/s/m, for return periods of 10 and only those obtained from the 1D-2D model are 25 years, respectively, and for duration of 1 h. The presented in Fig. 5. It is possible to obtain flood- maps are obtained from the 1D-2D model and present inundation maps from the 1D-1D model (Zhu et al. only water depths in excess of 0.1 m. 2016), but according to Mark et al. (2004) and 0.6 0.8 1D-1D 1D-1D 0.7 0.5 1D-2D 0.6 Runoff (m3 /s) 1D-2D Runoff (m3 /s) 0.4 0.5 0.4 0.3 0.3 0.2 0.2 0.1 0.1 0 0 0:00:00 0:15:00 0:30:00 0:45:00 1:00:00 1:15:00 0:00:00 0:15:00 0:30:00 0:45:00 1:00:00 1:15:00 Time (h:min:s) Time (h:min:s) Figure 2: Predicted runoff hydrographs produced from for one subcatchment of the combined drainage network for return periods of 10 years (left) and 25 years (right) and storm duration of 1h 25 25 1D-1D 1D-1D Flow (m3/s) 20 20 1D-2D 1D-2D Flow (m3/s) 15 15 10 10 5 5 0 0 0:00:00 0:15:00 0:30:00 0:45:00 1:00:00 1:15:00 0:00:00 0:15:00 0:30:00 0:45:00 1:00:00 1:15:00 Time (h:min:s) Time (h:min:s) Figure 3: Predicted flow in one sewer of the combined drainage network for return periods of 10 years (left) and 25 years (right) and storm duration of 1h  15th International Conference on Environmental Science and Technology Rhodes, Greece, 31 August to 2 September 2017 Figure 4: Predicted flow in the overland surface system (roads), simulated with the 1D-1D model, for return period of 10 (top) and 25 (bottom) years and storm duration of 1 hour Differences between the two models are reasonable. models are not in good agreement. It can be assumed Model structure differences include: (i) regarding that, in case where there were flow measurements in computer time, on a CPU Intel Core i5-3210M the pipes of the system and calibration/validation of 2.50Ghz and 6 GB RAM, the 1D-2D model takes both models was possible, the differences would not about 4 h to run, whilst the 1D-1D model only takes be of that extent. Moreover, it was found that only about 10 s; (ii) the 1D-1D model and 1D-2D model the 1D-2D model is able to simulate flood extent and are using different methods for the calculation of flood inundation. In the 1D-2D model, buildings are infiltration in each subcatchment; (iii) the overland represented by increasing DEM elevation by 20 m flow paths in the 1D-1D model are defined by the and so the model cannot give results for water depth modeler while in the 1D-2D model the overland flow and flux velocities at those locations. According to paths and velocities are simulated based on the Bellos and Tsakiris (2015), this approach can cause inserted DEM. numerical errors near the buildings. Moreover, the problem with 1D-1D and 1D-2D model, mainly in Based on the results, it is shown that the integrated urban areas, is the lack of real data, which can be 1D-2D model is able to give a more accurate overcome by calibrating a 1D-1D model with the prediction of overland flow paths and flood extent results of a 1D-2D model (Leandro et al. 2009). But than the 1D-1D approach. The flows in the sewers of this is the case only when a model is needed in short the combined drainage network predicted by the two time.  15th International Conference on Environmental Science and Technology Rhodes, Greece, 31 August to 2 September 2017 Figure 5: Flood-inundation maps for 10(top) and 25 (bottom) years return period and duration of rainfall of 1 h. 4. Conclusions and Recommendations Focusing on the 1D-2D model, some ideas can be proposed for further research. The first one is the This paper presents the comparison of 1D-1D urban implementation of rainfall-runoff monitoring in order flood model (SWMM) with 1D-2D (MIKE URBAN- to be able to calibrate and validate the model. The MIKE FLOOD) in order to demonstrate the model second one is to carry out a sensitivity analysis for structure importance. The two models were used and the 1D-2D model. Moreover, it is proposed the compared for the simulation of a small urban experimentation, in the 1D-2D model, with different catchment (Zone D) in Athens. The simulations took approaches for the representation of buildings. place for return periods of 10 and 25 years and 1 hour rainfall duration. The results showed that the 1D-1D Acknowledgements model is faster than the 1D-2D model, but it cannot The authors would like to thank the Athens Water simulate as accurately the flood extent and flood Supply and Sewerage Company (EYDAP S.A.) for inundation. supplying background information on the combined drainage network of Athens. The authors would also Leandro J., Djordjević S., Chen A.S. and Savić D.A., like to thank DHI for providing the software and free (2009), Comparison of 1D/1D and 1D/2D coupled licences for MIKE URBAN and MIKE FLOOD. (sewer/surface) hydraulic models for urban flood simulation, Journal of Hydraulic Engineering, 135(6), 495-504. References Mark O., Weesakul S., Apirumanekul C., Aroonnet S. B. Aguado J., Arsuaga J.M., Arencibia A., Lindo M. and and Djordjevic S. (2004), Potential and limitations of ASCE (1992), Design and construction of urban 1D modelling of urban flooding. Journal of stormwater management systems, New York. Hydrology, 299(3-4), 284-299. Bellos V. and Tsakiris G., (2015), Comparing various Mimikou M., Baltas E. and Varanou E., (2000), A study of methods of building representation for 2D flood extreme storm events in the Athens greater area. The modelling in built-up areas, Water Resources Extreme of Extremes: Extraordinary Floods, Management, 29, 379-397. Proceedings of a symposium held at Reykjavik, Bisht D.S., Chatterjee C., Kalakoti S., Upadhyay P., Sahoo Iceland, July, IAHS Publ. no. 271. M. and Panda A., (2016), Modeling urban floods and Neelz S. and Pender G., (2009), Desktop review of 2D drainage using SWMM and MIKE URBAN: a case hydraulic modelling packages, Environmental study, Natural Hazards, 84(2), 749-776. Agency, Bristol. DHI (2016a), MIKE URBAN CS – MOUSE, User guide, Pistrika A., Tsakiris G. and Nalbantis I., (2014), Flood Danish Hydraulic Institute, 430p. depth-damage functions for built environment, DHI (2016b), MIKE URBAN Model Manager, User Guide, Environmental Processes, 1(4), 553-572. Danish Hydraulic Institute, 414p. Rossman L.A., (2010). Storm Water Management Model DHI (2016c), MIKE FLOOD 1D-2D Modelling, User User's Manual, Version 5.0. Cincinnati, OH: National Manual, Danish Hydraulic Institute, 150p. Tsakiris G., (2013), Flood risk assessment: concepts, Elliott A. H. and Trowsdale S. A., (2007), A review of modelling, applications, Natural Hazards and Earth models for low impact urban stormwater drainage, System Sciences, 2(1), 261–286. Environmental Modelling and Software, 22, 394–405. Tsihrintzis V.A. and Hamid R., (1998), Runoff quality El-Sharif A. and Hansen D., (2001), Application of prediction from small urban catchments using SWMM to the flooding problem in Truro, Nova SWMM, Hydrological Processes, 12(2), 311-329. Scotia, Canadian Water Resources Journal, 26(4), USBR (U.S. Bureau of Reclamation), (1977), Design of 439-459. Arch Dams, U.S. Government Printing Office, Gironás J., Roesner L.A., Davis J. and Rossman L.A. Denver. (2009), Storm Water Management Model Zhu Z., Chen Z., Chen X. and He P., (2016), Approach for Applications Manual. Cincinnati, OH: National Risk evaluating inundation risks in urban drainage Management Research Laboratory, Office of systems, Science of the Total Environment, 553, 1-12. Research and Development, US Environmental Zoppou C., (2001), Review of urban storm water models. Protection Agency, 180p. Environmental Modelling and Software, 16, 195–231. Hsu M.H., Chen S.H. and Chang T.J., (2000), Inundation simulation for urban drainage basin with storm sewer system, Journal of Hydrology, 234(1), 21-37.