QUANTIFYING AND REDUCING THE PREDICTIVE UNCERTAINTY OF FLOODS – PROJECTS

Prediction and decision-making in aquatic environments requires a diversified approach to monitoring, modelling, and risk assessment. Theme 2 addresses the challenges associated with the reduction and quantification of predictive uncertainty in the management of water resources. To achieve this, the research will process uncertainty through a complete vertical chain of models and techniques exploiting operational meteorological ensemble forecasts, a sound hydrological ensemble prediction system, including a probabilistic streamflow assimilation scheme, and water resources management tools. This theme will contribute to the design of the Canadian flood forecasting system to be developed in Theme 3

THEME LEADER:

Professor François Anctil (Université Laval)
Department of Civil Engineering and Water Engineering
Email:  [email protected]

PROJECT 2-1

COMPARISON OF ENSEMBLE FORECAST METHODS FOR OPERATIONAL STREAMFLOW FORECASTING BASED ON A SINGLE MODEL

LEADER:

Dr. Bryan Tolson (University of Waterloo)
Email:  [email protected]

CO-INVESTIGATORS:

Dr. François Anctil (Université Laval)
Email:  [email protected]

Dr. Aaron Berg (University of Guelph)
Email:  [email protected]

Dr. Paulin Coulibaly (McMaster University)
Email:   [email protected]

Objective: Compare the performance and reliability of many probabilistic implementations of operational ensemble streamflow forecasting based on a single hydrological model.

Significance: Many Canadian agencies interested in flood forecasting have devoted substantial resources into developing one reasonably accurate, often distributed, hydrologic simulation model of their system that they want to utilize in a formal hydrological ensemble prediction system.  In these cases, agencies are often hesitant to consider multi-model ensemble prediction systems (see Project 2-2 below) and as such this project aims to identify the most robust single model based forecast method.  In particular, the work will focus on calibrating the model to multiple realizations of input forcing data (precipitation and temperature) to identify multiple parameter sets (rather than multiple models) to use in the ensemble forecast.    

Outcomes: Project 2-1 will compare multiple approaches for making ensemble forecasts with a single model.  In addition, we will develop a forecast system evaluation framework customized for how each partner organization will make decisions using an ensemble forecast. This project will also inform the development of CAFFEWS (Project 3-4) on the role of input uncertainty.

Publications

Liu, H., Tolson, B. A., Craig, J. R. and M. Shafii (2016) A priori discretization error metrics for distributed hydrologic modeling applications. Journal of Hydrology 543B, 873-891.

Liu, H. Thiboult, A., Tolson, B., Anctil, F., & J. Mai (2019) Efficient treatment of climate data uncertainty in ensemble Kalman filter (EnKF) based on an existing historical climate ensemble dataset. Journal of Hydrology. 568, 985-996.

Mai, J. & B. Tolson (2019) Model Variable Augmentation (MVA) for diagnostic assessment of sensitivity analysis results. Water Resources Research, 1-21.

PROJECT 2-2

COMPARISON OF ENSEMBLE FORECAST METHODS FOR OPERATIONAL STREAMFLOW FORECASTING BASED ON MULTIPLE MODELS

LEADER:

Dr. François Anctil (Université Laval)
Email:  [email protected]

CO-INVESTIGATORS:

Dr. Bryan Tolson (University of Waterloo)
Email:  [email protected]

Dr. Amaury Tilmant (Université Laval)
Email:  [email protected]

Objective: Compare the performance and reliability of many probabilistic implementations of operational ensemble streamflow forecasting based on multiple hydrological models.

Significance:  Because models are abstractions of real systems, it cannot be anticipated which specific model offers the greatest accuracy and predictive capability for specific catchments and hydrologic conditions. Multimodel prediction aims to extract as much information as possible from existing models. A multimodel approach emerged as a top priority for a large group of recently interviewed professional flood forecasters (Wetterhall et al., 2013).

Outcomes:  Project 2-2 will identify the advantages and disadvantages of ensemble prediction systems of various complexities.  This outcome will guide Canadian agencies responsible for flood warning in identifying an ensemble prediction system suitable for their needs. It will also guide FloodNet in developing the Canadian Adaptive Flood Forecasting and Early Warning System (Project 3-4).

Publications

Abaza, M., Anctil, F., Fortin, V. & R. Turcotte (2014) Sequential streamflow assimilation for short-term hydrological ensemble forecasting. Journal of Hydrology 519, 2692-2706.

Abaza, M., Anctil, F., Fortin, V. & R. Turcotte (2015) Exploration of sequential streamflow assimilation in snow dominated watersheds. Advances in Water Resources 80, 79-89.

Abaza, M., Garneau, C. & F. Anctil (2015) Comparison of sequential and variational streamflow assimilation techniques for short-term hydrological forecasting. Journal of Hydrologic Engineering 20(2), 04014042.

Abaza, M., Anctil, F., Fortin, V. & L. Perreault (2017b) On the incidence of meteorological and hydrological processors: Effect of resolution, sharpness and reliability of hydrological ensemble forecasts. Journal of Hydrology 555, 371-384.

Abaza, M., Anctil, F., Fortin, V. & L. Perreault (2017a) Hydrological evaluation of the Canadian meteorological ensemble reforecast product. Atmosphere - Ocean, 55(3), 195–211.

Anctil, F. & M.H. Ramos (2018) Verification metrics for hydrological ensemble forecasts. In Duan, Q., Pappenberger, F., Thirlen, J., Wood, A., Cloke, H. & Schaake, J. (eds) Handbook of Hydrometerorological Ensemble Forecasting. Springer, Berlin, Heidelberg.

Gaborit, E., Fortin, V., Xu, X., Seglenieks, F., Tolson, B.A., Fry, L., Hunter, T., Anctil, F. & A. Gronewold (2017) A hydrological prediction system based on the SVS land-surface scheme: Efficient calibration of GEM-Hydro for streamflow simulation over the Lake Ontario basin. Hydrology and Earth System Sciences 21(9), 4825-4839.

Thiboult, A. & F. Anctil (2015a) Assessment of a multimodel ensemble against an operational hydrological forecasting system. Canadian Water Resources Journal 40(3), 272-284.

Thiboult, A. & F. Anctil (2015b) On the difficulty to optimally implement the Ensemble Kalman filter: An experiment based on many hydrological models and catchments. Journal of Hydrology 529(3), 1147-1160.

Thiboult, A., Anctil, F. & M.-A. Boucher (2016) Accounting for three sources of uncertainty in ensemble hydrological forecasting. Hydrology and Earth Systems Sciences 20(5), 1809-1825.

Thiboult, A., Anctil, F. & M. Ramos (2017) How does the quantification of uncertainties affect the quality and value of flood early warning systems? Journal of Hydrology, 551, 365–373.

Xu, J. Anctil, F. & M.-A. Boucher (2019) Hydrological post-processing of streamflow forecasts issued from multimodel ensemble prediction systems. Journal of Hydrology, 578, 124002.

Xu, X., Tolson, B. A., Li, J. & B. Davison (2017) Assimilation of synthetic remotely-sensed soil moisture in Environment Canada's MESH model. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing 10(4), 1317-1327.

PROJECT 2-4

EVALUATION OF FLOOD WARNING BASED ON A HYDRAULIC MODEL WITH ASSIMILATION AND HYDROLOGICAL ENSEMBLE FORECASTS

LEADER:

Dr. François Anctil (Université Laval)
Email:  [email protected]

CO-INVESTIGATORS:

Dr. Bryan Tolson (University of Waterloo)
Email:  [email protected]  

Dr. Aaron Berg (University of Guelph)
Email:  [email protected]

Dr. Paulin Coulibaly (McMaster University)
Email:  [email protected]

Objective: Explore flood warning based on a hydraulic model with assimilation and hydrological ensemble forecasts, extending the hydrological ensemble prediction system tested in Project 2-2, with an additional vertical component.

Significance:  Flood warning relies on threshold-based decision rules that prescribe actions when streamflow exceeds a predefined value. In a deterministic world, the decision to act on forecast information is often guided by experience, especially when water levels are close to a threshold. It is then strictly up to decision makers to interpret the situation based on a qualitative appreciation of the uncertainty (experience). In a probabilistic world, access to a predictive distribution allows a better appreciation of the risks since the probability of exceeding a threshold may be estimated to be, for example, 20% or 70%.

Outcomes:  Project 2-4 will extend hydrological ensemble forecasts by issuing a distribution of water level forecasts at each time step.

Publications

Bessar, M.A., Matte, P. & F. Anctil (2020) Uncertainty analysis of a 1D river hydraulic model with adaptive calibration. Water, 12.

PROJECT 2-5

REAL-TIME RESERVOIR OPERATION BASED ON A COMBINATION OF LONG-TERM AND SHORT-TERM OPTIMIZATION AND HYDROLOGICAL ENSEMBLE FORECASTS

Leader: Dr. Amaury Tilmant
Email:  [email protected]

CO-INVESTIGATORS:

Dr. François Anctil (Université Laval)
Email:  [email protected]

Dr. Bryan Tolson (University of Waterloo)
Email:  [email protected]

Objective: Incorporation of hydrological ensemble forecasts into a large-scale, stochastic, optimization algorithm.

Significance:  Traditional stochastic optimization techniques are not computationally amenable methods for solving large-scale reservoir operation problems subject to hydrological ensemble forecasts (H-EPS). Most attempts found in the literature have adopted simplifications and/or case studies that are not representative of the real conditions faced by operators in Canada (e.g., Faber and Stedinger, 2001). Recent advances in the field of stochastic programming offer new opportunities in terms of modelling details, particularly with respect to the size of the system and to the treatment of hydrologic uncertainty (Tilmant et al., 2008; Marques and Tilmant, 2013). We propose a solution strategy based on the construction of a locally accurate approximation of the objective function instead of an exhaustive representation over the entire state-space. This approximate solution strategy is particularly attractive here because the use of frequently updated H-EPS implies that there is no need to explore the entire state space but only the most relevant subset considering the initial status of the system and the forecasts (Meier et al., 2012).

Outcomes:  Project 2-4 will propose an optimization framework that can solve the multi-reservoir operation problem using H-EPS. By integrating H-EPS in a stochastic optimization model, reservoir operators will be able to derive dynamic, risk-based, reservoir operation policies. This outcome will help Canadian agencies and hydropower companies responsible for operating large-scale water resource systems to improve their management tools.

Publications

Pina, J., Tilmant, A. & F. Anctil (2017) Horizontal approach to assess the impact of climate change on water resources systems. Journal of Water Resources Planning and Management 143(4), 04016081.

Pina, J., Tilmant, A. & P. Côté (2017) Optimizing multireservoir system operating policies using exogenous hydrologic variables. Water Resources Research, 53(11), 9845–9859.