Climate-Informed Flood Design for Dam Construction and Maintenance

Dams and dikes need to be able to withstand extremely large floods; how extreme depends on the damage that would be caused should the infrastructure fail. Depending on the hazard potential of a dam, regulations might require an infrastructure to sustain the maximum annual flood with a 1 in 1,000 or 1 in 10,000 chance of being exceeded.

Such design flood values are estimated by fitting a statistical extreme value distribution to observed river flows and computing percentiles. While apparently simple, this short description hides the myriad difficulties that professional engineers face in practice, including short observational records, missing data and measurement errors. On top of this, both recent historical observations and climate change simulations show an intensification of extreme precipitation events over Canada.

This project looks at how engineers can include information from climate change projections into the calculation of flood design values, and how this feeds into climate adaptation strategies.

Photo : K. Koenig, Manitoba Hydro

• Propose a pragmatic approach to incorporate climate change projections into flood design values for 1 in 1,000- and 10,000-year events;

• Propose framework to account for multiple sources of uncertainties in flood frequency analysis;

• Reflect on how to adapt existing infrastructures to an evolving risk environment.

• Review the current state-of the art in flood design estimation;

• Create working groups made of dam owners, regulators, engineering firms, climate scientists and professional associations;

• Co-design a relatively simple methodology providing an upper estimate for risk-averse decision makers; Co-design a more complex and data intensive methodology considering the various sources of uncertainties;

• Apply both methodologies to selected watersheds; Identify appropriate adaptation options and decision-making strategies to accommodate potential changes to flood design values under climate change.

The first Working Group (WG1) met monthly during the first year of the project to design a relatively simple approach to include hydroclimatic projections into the calculation of flood design values. WG1 chose to take advantage of Large Ensemble (LE) experiments, in which one climate model runs with multiple initial conditions in order to study the influence of natural climate variability. These simulations can then drive a hydrological model to yield daily streamflows, from which annual maxima can be extracted.

The benefit of using such large ensembles is that for every simulation year, there is a reasonably large sample of values to compute extreme flood frequencies and reduce sampling uncertainties. The methodology, shown in Figure 1, uses both CanESM2-LE (Canadian Earth System Model) and CESM1-LE (Community Earth System Model) to drive the GR4J hydrological model and compute annual maximum flows. Maxima over blocks of five years are fitted by a Generalized Extreme Value (GEV) distribution to estimate 1,000- and 10,000-year flood events.

The results from WG1 methodology for 10,000-year events are shown in Figure 2. For watersheds where data was available, the map shows the relative change in the flood design value estimated using projections from CanESM2-LE.

Figure 1: Methodology developed by WG1 to include climate change information in flood design values.

The second Working Group (WG2) also met monthly during the first year of the project to develop a framework to ingest multiple sources of data into the flood frequency analysis, accounting for the uncertainties affecting each source. Results from WG2 fed into WG1’s approach to ensure it met the objective of providing an upper estimate for risk-averse decision makers.

Finally the third Working Group (WG3) proposed a five-step approach to climate change adaptation, and performed a case study on the Winnipeg River. Results from WG3 examined the impacts of climate change, explored adaptation options and their performance/feasibility, and concluded by identifying the preferred adaptation strategy.

Figure 2: Relative change in the 90th percentile of 1:10,000 flood events estimated using WG1's approach using climate projections from CanESM2-LE. Watersheds where the sign of change is the same for CESM1-LE are hatched.

• Hydro-Québec

• Ministère de l’Environnement et Lutte contre les changements climatiques (MELCC)

• Ontario Power Generation

• Rio Tinto