OPERATIONAL FLOOD FORECASTING AND FLOOD RISK MANAGEMENT IN GRONINGEN

Transcription

1 OPERATIONAL FLOOD FORECASTING AND FLOOD RISK MANAGEMENT IN GRONINGEN JAN GOOIJER 1 AND KLAAS-JAN VAN HEERINGEN 2 Waterschap Noorderzijlvest, Groningen, The Netherlands Deltares, Delft, The Netherlands Abstract Waterboard Noorderzijlvest is responsible for all aspects of water management in the northern parts of the Dutch provinces of Drenthe and Groningen. One task is to protect the area against inundation from the regional water system. Variations in elevation and soil types, only 3% surface water and problems like soil subsidence and climate change make this an increasingly challenging task. A 1700 hectare storage basin combined with nature reserve will be realized in 2012 to be able to handle extreme rainfall events. Although this measure will improve the robustness of the water system, additional measures are needed. One measure that is currently being worked out is what we call anticipatory water management based on flood forecasting. The article describes why the waterboard requires a forecasting system and how they have implemented this system at waterboard Noorderzijlvest, making use of observed and predicted meteorological, fluvial and tidal data. The flood forecasting system has been operational since the summer of 2010 and its functionality is currently being assessed. Based on the forecasting system a new strategy is being developed in which the focus is shifted from traditional structural measures that improve the water safety to a more just in time approach with a focus on emergency response and temporary measures during a peak discharge. Critical factors in this process are uncertainties in predictions, response times required for effective actions and uncertainty based communication. Introduction to the waterboard The responsibilities of a Dutch waterboard are to ensure surface water quality, sound ecological water conditions, water supply in times of drought and water safety in periods with excessive precipitation. Another responsibility is the collection of sewage to waste water treatment plants and sanitation of waste water. In 1995 waterboard Noorderzijlvest was formed out of several smaller waterboards. The catchment area managed by waterboard Noorderzijlvest is about hectares and serving about 375,000 peoples. For the first time in Dutch history one waterboard was responsible for water management in the entire Electra catchment that runs from the swamp area Fochteloërveen near Assen up to the Waddensea in the north of the Netherlands. Figure 1 shows this catchment area, the adjacent catchment Fivelingo and the polder systems Noordpolder and Spijksterpompen, which directly discharge water to the Waddensea. Both Electra and Fivelingo are polder-boezem systems with a brook system and different polders discharging their water 1 ir. Jan Gooijer, hydrologist, waterboard Noorderzijlvest, PO Box 18, 9700 AA Groningen, The Netherlands, J.Gooijer@noorderzijlvest.nl 2 ir. Klaas-Jan van Heeringen, senior consultant operational water management, Deltares, PO Box 177, 2600 MH Delft, The Netherlands, Klaas-Jan.vanHeeringen@deltares.nl

2 into the boezem. The boezem is a network of water courses that receive water from polders and brook systems and transport that water to the sea. The Electra boezem is in fact a three level cascade system separated by pumping stations. As a consequence of spatially heterogeneous soil subsidence due to mining of natural gas, different water levels in the Electra system have to be maintained. In the deepest boezem part, the so-called shell 1 a water level of m AD is maintained; in shell 2 this is m AD and in shell 3 this is m AD. As shown in the South-North cross-section of waterboard Noorderzijlvest in Figure 2, the southern part of waterboard Noorderzijlvest is slightly higher. Brooks draining this area discharge freely into the boezem system. The total open water surface is 3% of the total area. The boezem water courses are usually about 20 m wide. As Figure 3 shows, with obstacles like bridge heads and old sluices that narrow the waterway, water level slopes of up to 50 cm over the full length of the boezem system (which is approximately 20 kilometres) during a peak discharge are not unusual. A natural structural measure that mitigates flood risk in the Electra catchment is the large parts of the catchment that are directly connected to the boezem water courses without levees. As a result, the lower parts of the catchment function as a natural water storage area which inundate temporarily during a peak discharge. Inundation of these areas is accepted by inhabitants because this catchment has functioned in this way for centuries. Figure 1: waterboard Noorderzijlvest with the different hydrologically separated catchments, the main water courses and pumping stations.

3 Figure 2: a long side cross section of the waterboard Noorderzijlvest surface levels, from south (left) to north (right), relative to mean sealevel (datum). Figure 3, water levels at pumping stations De Waterwolf (red) and Leutingewolde (blue) during the 1998 storm. The storms of 1998 The summer and autumn of 1998 were relatively wet. With mm measured in De Bilt in the period June-August, 1998 rates sixth in the summer list of largest precipitation depth in the period With mm measured in De Bilt in the period September-November, 1998 rates first in the autumn list of largest precipitation depth in the period (KNMI, 2011). As a result, soils were saturated. In the period October 24 November , a deep depression accompanied with heavy rain storms moved over the north-east of the Netherlands (Figure 4). Statistical analysis by the Dutch research institute for waterboards (STOWA) later

4 water level (m AD) rainfall [mm/day] cumulative rainfall [mm] revealed that the return period of the precipitation depths in many parts of the area was about one hundred years (Smits et al., 2004). Although the Electra catchment was not within the area that received most precipitation, water levels surged over the highest alarm level (set at m AD) at for example Leutingewolde (see Figure 1) where a peak was recorded of m AD (Figure 3). Thanks to the natural peak storage area in the boezem and emergency measures like the deployment of sand bags, polders did not flood, although it was a close call at some locations. The meteorological depression also caused a storm surge in the Wadden Sea. Under normal conditions with high tides, water is pumped from the Electra catchment into the Lauwers Lake. From the Lauwers Lake water flows through large spilling gates into the Waddensea during low tide. As a result of the storm surge at that time, it was not possible to spill enough water from the Lauwers Lake into the Waddensea for a period of 5 days. At the same time the influx of water from the Electra catchment and its neighbouring catchment into the lake was large as a result of the rain storm. This combination caused extremely high water levels in the lake area which even surged above the severe flood warning level (Figure 4) daily average cumulative Tidal level Level at Lake Lauw ers Severe Flood Warning Flood Warning Flood Watch Figure 4: Storm of 1998 October and November Structural measures This 1998 event shows that waterboard Noorderzijlvest was not prepared for large storm events at that time. At a national scale an agreement was made to prepare all Dutch water systems for water levels with a return period of one hundred years in 2015 (Waterakkoord, 2003). Around 2000, in line with the national approach to flood management, flood risk management at waterboard Noorderzijlvest focused on structural measures such as river restoration in upstream parts of the catchment and the development of a large water storage area. The largest structural measure taken is the development of a peak storage area in the Eelder- and

5 Peizermaden (Figure 5). Since 1989 nature conservation organisations have tried to buy and exchange land in the Eelder- and Peizermaden area, south-east of Leutingewolde (Figure 1). The Eelder- and Peizermaden are a low lying peat dominated area on the southern fringe of the Drentse Hondsrug, a moraine formed during the last ice age. The area is very wet and seepage dominated. The process to change the Eelder- and Peizermaden from low-quality farming soils into a seepage dominated nature reserve was hampered by a lack of money and government support. In 2000 waterboard Noorderzijlvest decided to focus its effort to improve the robustness of the water system on the development of a large peak storage area combined with nature development in this area, in cooperation with the provincial government, nature conservation organisations, local stakeholders and farmers. Extensive calculations with an elaborate hydrologic and hydraulic SOBEK-model of the entire Electra catchment indicated that a water storage area of 1700 hectares directly connected to the water system would suffice to lower peak water levels enough to limit the necessity of improving and heightening the existing levees to a minimum. Hence, this project would suffice to ensure the minimum required safety in the entire catchment in The area needed for this water storage area was bought by the province, waterboard and nature conservation organisations and is now being turned into a seepage dominated wetland. This project is the largest in the Netherlands where water storage is combined with nature (except for projects focusing on the main rivers in the Netherlands). Total costs are 38 million including recreational facilities. The project will be finished in Figure 5, the project area south-east of Leutingewolde (see Figure 1). Computational flood risk management The governing board and specialists from waterboard Noorderzijlvest expected to comply with national safety standards with the realisation of the peak storage area. Although it was anticipated that climate change would make more measures necessary after 2015, research in 2010 and 2011 revealed that more measures were needed to even ensure water safety standards in 2015 because higher water levels were calculated. (van Heeringen & Heynert, 2010; Bosch, 2011). One explanation for this result is the development in quantitative flood risk management research in the last decade. In the Netherlands, stochastic analysis of return periods of extreme

6 water levels based on independent parameters has become an important technique to analyse flood risk in a model environment (Bossenbroek, 2004; Bosch et al., 2006). To accommodate this, statistical analysis of climate data has improved as well (Smits et al., 2004). Consequently, uncertainty in the outcome can be made visible. Applying these techniques to the Electra and Fivelingo catchment revealed that events with a return period of one hundred years are more extreme than calculated earlier. Another explanation is that the hydrological and hydraulic models of the water systems of waterboard Noorderzijlvest have been improved recently (Bosch, 2011). The models were expanded with more detailed information and calibrated with recent water level and discharge measurements. Preparing for new measures Accepting the recent change in what we understand about flood risk in the Electra and Fivelingo catchments, an immediate consequence is that new measures are needed to improve water safety. It is anticipated that it will be more difficult this time to work out structural measures that improve water safety: 1. In the Eelder- and Peizermaden, waterboard Noorderzijlvest was able to elaborate on an existing plan with a good deal of support for land-use change already present. No such area is available anywhere else within the waterboard area. 2. Earlier plans to use certain polders as an emergency peak storage area encountered fierce opposition. As the need to use these polders was diminished when land-use change in the Eelder- and Peizermaden area was considered a sufficient measure, these plans were quickly abandoned. Hence, revitalising these plans will lead to an old discussion. It is unlikely to deliver significant progress. 3. In the Electra catchment, which is characterized by its long narrow canals and the accompanying water slopes during peak discharges (Figure 3), a water storage basin potentially increases flood risks, because water accumulates in low lying areas far away from the Lauwers Lake and only slowly reaches De Waterwolf (Figure 1). This leads to prolonged periods with high water levels, which is a risk when it continues to rain. 4. The economic crisis of last years has grossly limited the available budget for expensive land-use changes that are necessary for the implementation of structural measures. The crisis has also diminished public support for spending tax money on nature reserve development. 5. The natural water storage area which has existed for centuries in the boezem is a potential problem. Farming grounds in this area flood more often than is acceptable according to the standards set by the national agreement (Waterakkoord, 2003). However, there is no compensation available, as there has been no formal acknowledgement for this multiple land-use which has already existed for a long time. 6. In 2025 the standards for water safety will increase from safeguarding against water levels with a return period of one hundred year to a return period of three hundred years. Waterboard Noorderzijlvest is committed to meet standards set for water safety and to find and realise measures that are efficient, cost-effective and supported by the public. However, it is clear that innovative solutions are needed. A strategy focusing on land-use change or other structural measures may not be sufficient. Focusing on flood mitigation and accepting current flood risks is not an option as there is national agreement on the standards set for water safety. Operational Water Management and Flood Forecasting The strategy waterboard Noorderzijlvest has chosen to reach the necessary solutions is to

7 initially focus on increasing its water system knowledge based on monitoring and analysis of daily operational water management. Because it is still a recent development that the entire Electra catchment is managed by one organisation, water system research on the entire catchment is relatively new as well. It is expected that more tailor made measures can be implemented when water system knowledge is constantly developed. In addition, it is expected that support for necessary measures can be increased when the need for measures is substantiated with knowledge gained from operational water management. Finally, as there may not be sufficient structural measures to ensure the required water safety, more emphasis is needed on operational flood risk management during rain storms that cause peak discharges. One necessary element in operational flood risk management is a thorough insight into water system behaviour in order to optimize the allocation of people and emergency measures. Loosely based on a just in time approach, operational water system knowledge improves the preparedness of the emergency response teams and enables them to act fast and focused. As an initial step in this strategy, waterboard Noorderzijlvest has decided to implement an operational flood forecasting system. In 2004 a rather basic instrument was configured (Lobbrecht, 2004) that automatically made simulations with an hydrological and hydrodynamic SOBEK model. Since 2010 a new and more advanced forecasting system has been operational, based on the Delft-FEWS software of Deltares (Werner et al.,2004) 3. The FEWS application at waterboard Noorderzijlvest collects real-time observations and forecasts, like: Observed tidal and fluvial water levels and structure operation Rainfall data from calibrated radar images Forecasted rainfall from HiRLAM (HiRLAM, 2011) and ECMWF (ECMWF, 2011) Forecasted tidal water levels All these data are collected, validated and processed automatically and in real-time (Figure 6). Via a standardized public interface the numerical SOBEK model is run with all the available data. The data is processed and provided to the model with all kinds of data hierarchies and fallback options for missing data: when to interpolate gaps in time series, when to fall back to other data sources etc. The model runs are presented to the end-user as five-day forecasts of water levels in the boezem at about 30 locations. To assess the quality of the model simulations, historical simulations are shown together with the observed water levels. All source data that is fed into the model is presented in an easy way. It is also possible to make long-sections of the the main water courses to assess the water slope at different stages during a flood event. 3 The implementation of the Delft-FEWS software at waterboard Noorderzijlvest was co-funded by the European Regional Development Fund under the Interreg IVB North Sea Region Programme

8 clients servers import validation processing alarms SOBEK model - database server - computational servers telemetry meteo tide control FEWS Noorderzijlvest web reports clients intranet operators organisation, public Figure 6: processes carried out in FEWS FEWS automatically runs tasks like data collection and processing and making forecasts with standard operation rules. However, it is also possible to run scenarios which give the operators insight into the effects of possible measures or insight into variations of the imposed boundary conditions like the meteorological conditions. It is possible for example to calculate the effect of a scenario with 20% more precipitation than forecasted. The effects of measures can be assessed such as turning all polder pumping stations off at some point during a peak discharge. All FEWS results are accessible for the entire organisation. Colleagues who manage the water system on a daily basis, like the chief engineer of De Waterwolf, are encouraged to frequently check the results, also outside periods with peak discharges, in order to validate the model results with their personal experiences. This works in two ways: model results that consistently deviate from measurements can indicate that the model is not performing well in a certain location, but it is also possible that the measurement is not correct. Hence, operating FEWS on a daily basis ensures that the water system is monitored critically stimulating water system knowledge and a quick response to problems. Building on a FEWS foundation Based on an operational FEWS system new steps are currently worked out to develop the chosen strategy of structurally improving the water system based on operational knowledge: 1. The control of structures like De Waterwolf should be equal in the SOBEK-model and in practice. However, the operational control system is outdated as it is only fed with information about the past and current status of the water system. In daily practice the engineers also take the weather forecast into account when they manage their pumping stations. As forecasts are available on a regular basis and through scenarios it is possible to calculate the consequences of different operational measures. This enables to optimizing the control of structures. It is expected that an early anticipation of an expected peak discharge may lower peak water levels. This project is currently in progress. 2. Modelperformance is continuously calculated and provided to the FEWS operators so they are able to correct the numerical forecasts with observed and known bias, both in time and place. In a project which is planned for this year, uncertainty in model predictions is analysed and related to uncertainty in supplied forecasted data and model performance. The goal of this project is to operate the emergency response teams based

9 on predicted water levels instead of fixed alarm levels. Based on a minimum lead time required to take necessary steps balanced with a minimum level of certainty that a certain water level will indeed be reached, emergency response allocation can be optimized both in time and place. This can reduce the need for structural measures. For example, structural measures may not be needed when known weak spots can be secured with temporary measures which are in place only during the most extreme peak discharges. This may reduce the investment costs significantly and lead to more public acceptance. In order to be successful in this strategy it is however essential that these temporary measures are in place in time and only when really needed. This requires sound communication within the emergency teams and clear choices as to how known uncertainty is effectively translated into concrete actions. Hence the just in time qualification which may be appropriate for this risk management based approach. 3. Detailed knowledge on the functioning of the water system gained with constant monitoring with Delft-FEWS, combined with knowledge from calculated scenarios, will allow that strategic studies on water safety can be carried out with more focus and yield a wider array of possible measures. In 2011 such a strategic study will start in which possible measures are investigated that will make sure that water safety norms are met in Conclusions As this paper has shown, a new strategy in flood risk management is applied at the regional scale of waterboard Noorderzijlvest. The concept of stimulating land-use change has contributed significantly to improving the robustness of the Electra catchment. However, set safety standards are still not met. The continuation of the strategy of land-use change seems unlikely due to a lack of both money and support. Accepting current flood risks and focusing on mitigation measures is not an option as safety standards are nationally agreed on. Therefore, the applied strategy is to invest in water system knowledge with a real-time flood forecasting system in which uncertainties are quantified and translated into concrete decisions, whether on temporal measures during an emergency or on structural measures in a strategic study. Hence, the flood forecasting system is not only used to predict water levels, but also as a tool to analyze and evaluate water system behaviour combining abstract model-based knowledge with practical knowledge of field operators. In this way the full potential of the forecasting instrument is used: a model that has proven its value in flood forecasting will be trusted more in scenario studies. In addition a wider array of possible measures to lower the peak water levels can be assessed as the spatial effects of location-specific measures are better understood. References 1. Bosch, S., Hakvoort, H., Diermanse, F., Verhoeve, C., 2006, Verantwoord omgaan met de nieuwe neerslagstatistiek, Stromingen 12 no. 1, Utrecht, The Netherlands. 2. Bosch, S., 2011, Upgrade simulatiemodel Waterschap Noorderzijlvest, Siebe Bosch Hydroconsult, The Hague, The Netherlands. 3. Bossenbroek, J.C., 2004, Statistiek vóóraf of statistiek achteraf, thesis TU Delft, Delft, The Netherlands. 4. ECMWF.int, model results provided by the Royal Dutch Meteorological Institute (KNMI), page visited 25 th of February Heeringen van, K.J., Heynert, K., 2010, Analyse T=100 boezemmodel Noorderzijlvest, Deltares, Delft, The Netherlands. 6. Het Nationaal Bestuursakkoord Water, 2003, The Hague, The Netherlands. 7. HiRLAM.org, model results provided by the Royal Dutch Meteorological Institute (KNMI); page visited 25 th of February KNMI klimatology desk, page visited 25 th

10 of February Lobbrecht, A.H., 2004, realisatie BOS Hoogwater Waterschap Noorderzijlvest, Hydrologic, Amersfoort, The Netherlands. 10. Smits, I., Wijngaarden, J., Versteeg, R., Kok, M., 2004, Statistiek van extreme neerslag in Nederland, STOWA, Utrecht, The Netherlands. 11. Werner, M.G.F., van Dijk, M. and Schellekens, J., 2004, DELFT FEWS: An open shell flood forecasting system, Proceedings of the 6th International Conference on Hydroinformatics, Liong, Phoon and Babović (Eds.), World Scientific Publishing Company, Singapore,