Modeling Consequences of Climate Changes on the Copenhagen Urban Drainage System Flood Vulnerability Mapping and Future Adaptive Measures

Transcription

1 Modeling Consequences of Climate Changes on the Copenhagen Urban Drainage System Flood Vulnerability Mapping and Future Adaptive Measures C. N. Nielsen, 1* T. S. Madsen and A. H. Mollerup 2 1 Ramboll, Hannemanns Allé 53, 2300, Copenhagen, Denmark 2 Copenhagen Energy, Oerestad Boulevard 35, 2300, Copenhagen, Denmark *Corresponding author cnn@ramboll.dk ABSTRACT Two models covering the catchment areas of the Damhusåen and Lynetten wastewater treatment plants in the city of Copenhagen in Denmark have been developed for the purpose of integrated optimization of sewers, wastewater treatment plants and recipients. Both models include detailed descriptions of recipients in the drainage area and links to the sewer systems. Manholes and recipients in the Municipality of Copenhagen and Frederiksberg were coupled to a DEM and used for simulating urban runoff, and simultaneous pipe and recipient flow coupled to numerical surface flooding simulations. Runoff inputs were 8 hour duration Chicago Design Storms (CDS) with return periods of 10 and 100 years. Current and future scenarios for runoff and sea levels for years 2010, 2060 and 2110 were modelled. Future municipal adaptive measures towards increased runoff inputs were modelled along with measures to preserve current discharge capacities with future sea levels. Flooded areas were mapped to locate areas vulnerable to flooding from sewers and recipients now and in future scenarios. The total area and volume of flooding was used as primary parameters to evaluate the overall consequences of predicted climate changes and the effect of adaptations/mitigations. KEYWORDS Climate change, urban flooding, adaptation, extreme rainfall, CBA. INTRODUCTION Two models covering the catchments of the Damhusaaen and Lynetten wastewater treatment plants in the city of Copenhagen in Denmark have been developed for the purpose of integrated optimization of sewers, wastewater treatment plants and recipients. The models cover in total 8 municipalities with a population of The total runoff area is 147 km 2 and the model network comprises nodes. Yet the model network has been simplified, thereby reducing the manholes density to about 20% while preserving 95% of the volume. Both models include detailed descriptions of recipients in the drainage area and links to the sewer systems. The aim of this work has been use these models to evaluate the consequences of climate change in Copenhagen by mapping areas vulnerable to flooding now and in future scenarios and to quantify the needs for adapting to the expected changes in precipitation patterns and sea level. Page 1 of 7

2 METHOD The global climate scenario A2 presented by the Intergovernmental Panel on Climate Change in 2007 has been used as input for a regional model of future rainfall intensities in Denmark. The modeling results were presented by the Danish Wastewater Committee in 2008 and included some general recommendations about future rainfall intensity increase s for use in the area of urban drainage planning. The recommendations are now used by Danish municipalities for planning and evaluation of drainage pipes and structures within a 100 year time horizon. In this work the climate s are used for scaling synthetic rain storms of type Chicago Design Storms first presented by Keifer & Chu (1957) and discretized for the purpose of numeric modeling. Modeling of flooding scenarios was done using a standard numerical hydraulic model for pipe flow and 2D flow based on the Saint Vernant equations. Manholes and recipients in the Municipality of Copenhagen and Frederiksberg were coupled to a digital elevation model with a spatial resolution of 2-4 meters. The two coupled models were used for simulating urban runoff, and simultaneous pipe and recipient flow coupled to surface flooding simulations. For model validation a historical rainfall from August 2010 was used with an estimated return period of more than 100 years. Runoff inputs to simulations were synthetic Chicago Design Storms (CDS) of 8 hours duration with return periods of 10 and 100 years. Current and future scenarios for years 2010, 2060 and 2110 were modeled. Predicted changes in sea level and predicted runoff patterns for scenarios 2060 and 2110 were implemented as boundary conditions. The expected mean sea levels in 2060 and 2110 were adopted based on consultations with The Danish Meteorological Institute. The present and anticipated future sea levels and climate dependent rainfall intensity increase s are in table 1. Future sea levels are adjusted for an isostatic land uplift rate of mm/y. Table 1: Now and future mean sea level and climate s for scaling rain storm intensities. Scenario Expected mean sea level [m] Climate T10 Climate T The model boundary recipient hydrographs were determined from four years of historic flow and water level data. Future municipal adaptive measures towards increased runoff inputs were modelled along with measures to preserve current discharge capacities with future sea levels. The anticipated adaptations aim to fully mitigate the future increase in storm intensity of a 10 year rain storm. Consequently the future climate adaptations to increased runoff inputs were implemented in the models by reducing CDS intensities by the climate of a 10 year CDS. E.g. a 10 year storm is expected to increase by a 1.3, which is successively adapted to by decoupling corresponding areas or equivalent means. The increase for a 100 year storm is anticipated to be 1.4 hence the increase for a 100 year return period after adaptation would be 1.4/1.3. Adaptations to higher future sea levels were modelled by simulating pumps at locations where discharge crest level was below sea level for a given scenario. Pump capacities corresponded to the maximum modelled local discharge during a 10 year CDS rain event since this is the existing and expected future level of service for the sewer company. Page 2 of 7

3 Flooded areas were mapped to locate areas vulnerable to flooding from sewers and recipients now and in future scenarios. The volume of water on terrain and the flooded area were used as primary parameters to evaluate the overall consequences of predicted climate changes and the effect of adaptations/mitigations. Sensitivity analyses were made to assess: To which degree the model was sensitive towards the simplification of the system. To which degree the model was sensitive towards outlet geometry in manholes etc. To which degree the model was sensitive towards the flow in the major recipient in the system. To which degree the flooding was sensitive towards an area reduction on runoff inputs. To which degree the model was sensitive towards lack of basement volumes. To which degree the spatial distribution of flooded areas was sensitive towards model network spatial density. The effect of omitting numerous manholes and smaller pipes in the simplified models was a decrease in available sewer volume by about 5-8%. A corresponding sewer volume was added to the simplified model for sensitivity analysis by increasing manhole diameters accordingly. The model sensitivity towards outlet geometry in manholes was checked by changing all round edged geometries to sharp edged geometries. This applied to 93% of the model manholes. For sensitivity analysis the major recipient inflow was varied between drought flow and the maximum recorded flow in a 4 year data series (7.3 m 3 /s), which was also used for flood calculations. It is known that the average intensity of a rainstorm with a given return period is diminishing with spatial scale (e.g. Morin et al. 2003). Since the area covered by the drainage system is 147 km 2 this effect had to be considered. Rainfall intensities from a number of rain gauges operating in the Copenhagen area were analyzed to find the spatial correlation between gauges for a 10 year return period rain storm. A spatial correlation of 75% was found to be a conservative best fit, when considering a variety of durations. Hence the intensities of a 10 year synthetic rain storm were scaled to 75% to evaluate the sensitivity to spatial scale. It is known that flooding of private basements can contribute substantially in preventing terrain-flooding, yet the model did not include basement volumes. In the sensitivity analysis basement volumes was added to the model by increasing manhole diameters. In this approximation basement depths were assumed 1.25 m. Basements in the Zealand municipalities of Copenhagen and Frederiksberg was assumed to be covering 80% of the floorplan, while it was assumed to be covering 50% of the floor plan in the rest of the Lynetten catchment. The network spatial density would presumably have effects on the predicted flooded volume and the predicted spatial distribution of flooding. The effect of model simplification on the spatial accuracy of prediction was assessed by detailing two sub-areas of the Lynetten catchment model: The area of Ryvang in northern Copenhagen and the municipality of Frederiksberg. The sub-areas were detailed to include all catchments, manholes and minor pipes within these areas, while keeping the coarser model resolution in surrounding areas. A Page 3 of 7

4 10 year return period CDS rain storm was used as runoff input to the Frederiksberg model, while a historic rain event from August was used as input to the Ryvang model. Sensitivity analysis towards model simplification, recipient flow, area reduction and manhole outlet geometry was conducted for the catchment of Damhusåen wastewater treatment plant and based on a 10 year return period CDS in scenario The remaining sensitivity analysis was based on the Lynetten model. RESULTS & DISCUSSION The sensitivity is evaluated by calculating the number of manholes spilling water on terrain using Mike Urban. Results are in Table 2. The model is very sensitive towards outlet geometry in manholes. Also the area reduction and flow in Harrestrup River plays a major role. The simplification of the model is of relatively low importance regarding the reduced volume. The sensitivity of model network spatial density on local flooded area distribution was found to be highly dependent on the type of capacity problems experienced. The position and extend of larger flood volumes is predicted relatively good in the simplified model, while smaller surface spills caused by local and/or short duration capacity problems are poorly predicted. In the Ryvang area the spatial overlap between the simplified and the detailed model in predicted flood location was 70% in both flooded volume and area. The flooded volume in the simplified model was 16% higher than in the detailed Ryvang model. In general flooding in the Ryvang area is caused by overall capacity problems in larger pipes and canals, and these volumes are well described in the simplified model. For the municipality of Frederiksberg the overlap in area and volume was only 9 and 6% respectively. The flooded volume predicted by the simplified model was 64% of the volume predicted by the detailed model. The disagreement can probably be ascribed to the fact that flooding in the municipality of Frederiksberg are scattered and caused by local short duration capacity problems. These parts of the drainage system are not described in the simplified model. Table 2: sensitivity analysis results Sensitivity Parameter Flooded manholes Baseline Flooded manholes Change [%] Simplification Manhole outlet geometry Area reduction Recipient inflow Basement volumes The model sensitivity is systematically overestimated by using a 10 year return period rainstorm for analysis, since the utilities in the Copenhagen area must ascertain that surface flooding from combined sewer occur with an annual probability of less than 10%. For that reason the water level in manholes will be very close to terrain in these events. The inclusion of basement storage volumes reduced the surface flooding to near null during a 10 year design storm. The results indicate that basement volumes in Copenhagen pose an important sewer storage reserve maintained at high private costs. More widespread private use of anti flooding devices such as non-return valves and pumps to prevent basement floodings will Page 4 of 7

5 reduce the total available volume in the future and most likely contribute to a higher frequency of surface level flooding. Regarding manhole geometry the model showed high sensitivity to the outlet geometry. However the loss used in the models is determined based on hydraulic calibration thereby reducing the actual uncertainty of the model considerably. Flood maps covering Copenhagen pinpointed several areas vulnerable to flooding and clarified visually areas where proactive mitigating investments can be considered. Also the consequences of adapting to climate changes compared to doing nothing, and the effect of sea level rise on the drainage system was mapped. The mapping made it possible to quantify the consequences of adapting to future climate compared to not doing so. For example during a 10 year return period CDS the Copenhagen drainage system is currently spilling about 10 5 m 3. Without adaptive measures this volume is slightly increased in 2060 and most likely doubled in The above calculation assumes an area reduction of 1, which is conservative by at least 20-30%, and does not account for existing private property storage volumes and flooded basements. Similar evaluations were done using lower area reduction s and for 100 year return period CDS rains. Table 3 summarizes the flooded areas and volumes. Blue areas are omitted while flooded green areas are included. Only flooding depths above 4 cm are included. The total area of the municipality of Copenhagen is ha. Table 3: Results Scenario [year] T [years] Area reduction Climate Adap- Tation Total Adaptation Flood area [ha] Flooded volume [10 4 m 3 ] / / / / / / Page 5 of 7

6 To mitigate effects of future increased runoff intensities and the keep the current level of service the utility must increase the retention volume by about 10 5 m 3 if using a conventional approach. Alternative approaches are to use green areas as retention volumes or to uncouple storm sewers. By the year 2110 it is anticipated that 31 weir crests will be below sea level. If the ability to discharge from these weirs must be preserved in year 2110 the utility will need to pump CSOs. Assuming that current discharges during a 10 year storm must be maintained the total necessary pumping capacity was found to be 100 m 3 /s. Other actions could be to increase critical weir crest levels and elevate low lying pipes and/or use moveable gates in the harbor and other recipients. The flood modeling results have been used as input to CBA and technical evaluations with the purpose of calculating present and future damage costs of urban flooding. The results are in Table 4. All results are in net present value (NPV) evaluated over a 100 year time span and must be seen in relation to the NPV of damages if nothing is done, which will total 2.9 billion US$. In all scenarios it is assumed that the present level of service must be maintained. Table 4: NPV in billion US$ of adaptation strategies. Net savings are compared to not adapting at all. Scenario Scenario 1 Scenario 2 Scenario 3 Scenario 4 Measures Sewer Sewer Sewer Source control Basement valves Basement valves Basement valves Flood control Flood Control Adaption costs Reduced damages Net savings If the adaptation is based exclusively on increasing sewer capacity and/or switching from combined to separate sewers the NPV of adaptation will be 1.9 billion US$, and yet the total damage in NPV will be 1 billion US$. Consequently the total adaptation and damage costs approximately equal the total damage costs of not adapting at all. If increased sewer capacity is combined with basement non return valves the NPV savings are 0.4 billion US$ compared to not adapting. Additional measures to control surface flooding to areas where the water does the least harm do not contribute positively to this adaption approach. The lowest costs and the highest benefits are achieved from scenario 4. In this scenario adapting is done by using basement valves and flood control in combination with source control, which involves local infiltration or use of roads and green areas for storm water retention. By this adaption approach the NPV savings are 1.4 billion US$ compared to not adapting. CONCLUSIONS Areas vulnerable to flooding are identified and the flooding maps will be an integral part of the long term urban drainage planning in Copenhagen. The area and volume flooded during a 10 or 100 year return period rain storm is likely to double in the next 100 years without future adaptive measures towards flooding. The adaptive measures can include a combination local infiltration and retention, increased sewer storage and conveying capacity, optimized usage of available system volume along with pumped CSOs and controlled flooding. Page 6 of 7

7 Increased future precipitation and higher mean sea level will over a 100 year period cause a net present loss of 2.9 billion US$ if no adaptations are done. Should the municipality of Copenhagen choose to adapt through increased capacity of the sewer system, the net loss for the society will be zero. However a combination of source control, basement valves and flood control can lead to a positive net present value of 1.4 billion US$ compared to not adapting to climate change. ACKNOWLEDGEMENT We would like to thank Agnethe Nedergaard Pedersen, Gitte Schnipper and Niels Bent Johansen for their contributions. REFERENCES The Danish Wastewater Comitee (2008) White Paper 29. Anticipated changes in extreme rain caused by climate change (in danish). IDA Keifer C.J. & Chu H.H. (1957) Synthetic storm patterns for drainage design. ASCE J. Hydraul. Div. 83 (HY 4) 1332/1-1332/25 Morin E., Krajewski W.F., Goodrich D.C., Gao X. and Sorooshian S. (2003) Estimating rainfall intensities from weather radar data: the scale-dependency problem. J. Hydrometeorology Vol. 4 pp Page 7 of 7