Scientists are currently developing adaptation and management strategies to anticipate the future effects of climate change in Europe. Irstea researchers participated in the European ClimAware project, focusing on the impact of climate change on water resources on a continental and regional scale. This article will focus on their study of the 4 reservoirs in the Seine river basin, where there is an increasing risk of drought levels by 2050.
In order to avoid a repeat of the 1910 Seine flood and to provide support during periods of drought, large reservoirs were built along the Seine basin between 1949 and 1991. The 4 main structures were the Pannecière (on the Yonne river), the Seine, the Marne and the Aube reservoirs. These 4 artificial reservoirs can store around 800 million m3 and meet 2 aims :
- Preventing/limiting floods
- Providing support during periods of drought .
The reservoirs are managed by the Seine Grands Lacs Local Public Basin Establishment, a partner of the ClimAware project. Although the reservoirs currently work towards reducing the effects of flooding and ensuring a continuous supply (water for households, industry and agriculture etc.) during periods of drought and in response to climate change, the question of whether they do so efficiently remains.
The ClimAware project  looked at reservoir management and basin hydrology up until 2050. Teams from Irstea's Montpellier and Antony centres  modelled potential outcomes using past (1965-1991) as well as future (2045-2065) climate scenarios, and meteorological and flow data. 25 flow measuring stations were distributed along the rivers in the basin to divide it into sub-basins and incorporate spatial differences.
Breaching the vigilance threshold
The project falls under the remit of the European Water Framework Directive, which aims to create good ecological status for waterways by 2015. How periods of drought are dealt with is crucial to this aim : in order to avoid any pollution or eutrophication of ecosystems, following output from waste water treatment plants, there needs to be enough water within these systems to dilute any pollutants. However, as highlighted by the conclusions of the ClimAware project, due to climate change, the natural hydrology of the basin will evolve : "We anticipate drought flows that are 30 to 40 % lower as well as longer drought periods (summer and autumn)", explains David Dorchies, hydraulics engineer at the Irstea Montpellier centre. Under current reservoir management regulations , the impact will be felt quickly. "It will be difficult to maintain a high enough flow towards Paris to guarantee the quality of the water." Therefore, increasingly, the vigilance threshold will be breached .
The research project revealed there will be more frequent and more severe droughts. Surprisingly, modelling flood risks proved to be harder : extreme rainfall scenarios vary and there do not seem to be any overarching flood trends. "Significant flooding is rare, 1 in every 100 years. Floods are therefore rarely tested despite being particularly interesting. However, our scenarios and data are based on time scales that are too short."
Representing Seine Grands Lacs
In addition to modelling and creating scenarios, the challenge was to propose adaptation strategies that would improve management of the reservoirs. "We had to become Seine Grands Lacs and see what we could improve", notes Dorchies. Scientists worked towards 2 levels of adaptation.
- Tactical. The aim was to modify target fill curves for the reservoirs. These curves determine the status of a reservoir at any given moment (during dry seasons the reservoir should be full ; during wet seasons it should be empty enough to absorb flood waters). However, climate change will change the seasons. "Our method was to work backwards," explained Dorchies. "We know the future so we started from the end of our period and calculated the amount of water that should have been saved to not breach the threshold." This minimises the risk of not achieving set targets.
- Operational. Aim : real time management of the reservoirs, using predictive controls . "A model was run over 9 days and this 9-day prediction was used to optimise reservoir activity." Researchers used real meteorological predictions for rain (from 2005 to 2008) to reveal prediction errors in relation to actual data, in order to apply these to the models and make them as accurate as possible.
2 approaches were used for 2 very different objectives. The drought process is particularly slow (with little variation in flow rate). For this reason, predictive controls, which focus on the short-term, are of little benefit when compared to fill curves. For floods, however, the opposite is true : predictive controls help define the precise peak flood moment and, therefore, when water needs to be held back.
Work was finalised by the development of a tool that will soon be transferable to managers. The tool will be used to build optimised fill curves and to evaluate the risks taken by managers in relation to targets defined downstream for specific volumes from any reservoir on any given day of the year. "This is this most promising adaptation", highlights Dorchies. "However, the implementation of predictive controls has been more complicated as these need a centralised system that can operate in real time." Currently, the reservoirs are managed according to local water regulations. They are sized so that their impact is felt locally and downstream from the Ile de France region. Floods are therefore blocked upstream, well before arriving in Paris. Until now...…
This type of project has its limitations : climate scenarios include many uncertainties, most of which are regional in nature. "With unbiasing (correcting scenarios by comparing past scenarios to historical data) we have values that match our present reality even more closely (in terms of past vs. present)", concludes Guillaume Thirel, hydrologist and research associate at the Antony centre. "We can apply this error margin to simulated data for the future. We are, however, concentrating on future developments rather than absolute values as we know that things are not always so precise."
For more information
- Consult the web pages of the Irstea Montpellier and Antony centres.
- The Water resource management, actors and uses (G-EAU) research unit in Montpellier and Hydrosystems and Bioprocesses (HBAN) research unit in Antony.
 A drought period is defined as any period in the year when the water level in a river is at its lowest point. This should not be confused with regular seasonal fluctuations in water level, even if the former is an exacerbation of the latter.
 ClimAware Project (September 2010-December 2013), in the context of the 2nd Joint IWRM-NET Call for Research on Integrated Water Resource Management. Partners: University of Kassel (Germany), Irstea (France), EPTB Seine Grands Lacs (Seine Grands Lacs Local Public Basin Establishment, France) and the International Centre for Advanced Mediterranean Agronomic Studies (CIHEAM-IAMB, Bari, Italy).
 The G-EAU joint research unit in Montpellier with mainly David Dorchies and the HBAN research unit in Antony with Charles Perrin and Guillaume Thirel.
 Current management regulations for reservoirs : between 1st November and 1st July, the reservoirs are to be filled according to a predefined curve; between 1st July and 1st November, they are emptied. Maximum flow thresholds for rivers leading away from the reservoirs are defined to limit flooding.
 There are 4 official drought thresholds: vigilance threshold (no restrictions, but the point at which problems with industrial and waste water output dilution may occur), alert threshold, sustained alert threshold (watering ban, restrictions to drinking water etc.), and crisis threshold (no one but nuclear power plants is allowed to use the water etc.)
Tree Based Model Predictive Control (TB-MPC), developed in collaboration with the Delft University of Technology (The Neterlands) and Politecnico di Milano (Polytechnic University of Milan - Italy)