|Tuesday, June 01|
Short-term forecast of spring freshet peak flow using GAM model
* Véronique Dubos, INRS, Canada
Taha Ouarda, Canada
André St-Hilaire, Canada
"For rivers with natural hydrological regimes, the spring flooding risk is determined by peak flow intensity. Although many methods have been derived to predict flood volume, especially for dam management, short-term forecasting of peak flow intensity is subject to a lot of uncertainty and depends largely on the ongoing specific snowmelt condition. An operational method for short-term forecasting of spring freshet peak flow is presented. The objective of the method is to be applied by any stakeholder during flood-watch in order to help decision-making according to emergency plan. The method uses the General Additive Model (GAM) to forecast peak flow from daily observed data, publically available on a daily basis. The model was tested on five rivers in Québec, with drainage basins varying between 350 km2 and 1700 km2. Peak flow is linked to both snow accumulation on the watershed and specific melting conditions. To predict the peak flow, two types of variables were used, those related to hydrograph characteristics and those related to climate and meteorological data. The best predictive variables were specific to each of the 5 rivers tested. Nonetheless, they were constituted by a combination of snow accumulation, previous air temperature, rain accumulation and rate of flow increase. A comparison of the forecasted peak flow was done between GAM and the determinist model Hydrotel, used by the Centre d’Expertise Hydrique du Québec (CEHQ), for flood-watch in the Province of Quebec. To assess further the operational forecast capacity of GAM, simulations of real time peak flow forecasts are presented. The results showed that 24-hour peak flow forecasts compared advantageously to the determinist model forecasts currently used. Contrary to the determinist model, which requires time consuming calibration on each watershed and numerous input variables, the GAM model necessitated fewer, easily accessible data in real time during flood-watch and an automatic fitting. The model can be applied by non-specialists. Although results were good for the short-term prediction, inclusion of longer term meteorological forecasts may allow for longer term (e.g. 7-day) peak flow forecasting. "
Accumulating Flood Risk
* Seth Bryant, University of Alberta, Canada
Evan G. R. Davies, Canada
"Responding to rising disaster damages and shrinking budgets, Canadian cities are looking for ways to maximize protection from flood damages while minimizing costs. To inform such mitigation investments, quantitative decision-making tools have been used for decades to evaluate structural protection measures (e.g. levees), where estimating the current risk using a ‘static view’ (where vulnerability does not change through time) is thought to be adequate. However, no such tool exists to evaluate time-sensitive mitigation measures such as Flood Hazard Regulations (FHRs). Unlike structural protection measures, FHRs mitigate flood damage over time as flood-specific building rules guide new development towards less vulnerable buildings. Despite widespread application of FHRs in Canada since 1975, decision makers have no method to quantitatively evaluate FHRs or answer questions like: how much flood risk do FHRs mitigate? --- To address such questions, this presentation summarizes recent work to implement a dynamic view of flood risk through the development of the novel Stochastic Object-based Flood damage Dynamic Assessment model framework (SOFDA) that conceptualizes the intersection of increasing flood hazard and heightened urban vulnerability as driving an accumulation of flood risk. SOFDA builds on the 2014 Rapid Flood Damage Assessment Model (developed by the Government of Alberta) to create a framework that includes: stochastic uncertainty; property level mitigation measures; and urban redevelopment. After presenting the SOFDA framework, a case study on single-family homes in the Sunnyside/Hillhurst neighborhood of Calgary, Canada will be discussed that uses a SOFDA simulation experiment to: 1) evaluate the shortcomings of the traditional static view of risk; and 2) explore the potential for optimizing Calgary’s current FHRs. --- Model results suggest that, for structural protections in redeveloping neighborhoods (like Sunnyside/Hillhurst), traditional static methods underestimate the present value of flood risk mitigations. Further, a novel FHR that avoids onerous building restrictions was evaluated and shown to improve the risk mitigation of the current FHRs by 9.7%. --- This case study demonstrates the significance of vulnerability dynamics for flood risk assessments while the SOFDA framework unlocks the quantification of FHRs and other time-sensitive mitigations for decision makers. This should foster plans that optimize from a wider range of flood defenses, including more tailored and effective FHRs. With this new tool, decision makers can move past a static view of risk focused on today’s communities — and towards more robust and resilient mitigations for the uncertain risks of tomorrow. "
On estimating long period precipitation return levels from annual maxima
* Mohamed Ali Ben Alaya, Pacific Climate Impacts Consortium, Canada
Francis W. Zwiers, Canada
Xuebin Zhang, Canada
Statistical extreme value theory (EVT) is a fundamental, heavily used tool for estimating the frequencies and intensities of hydro-meteorological extremes by analyzing time series of annual maxima. A key implicit assumption in the application of EVT is “max-stability”, i.e., that the statistical behavior of annual maxima is predictive of that of maxima calculated over multi-decadal or longer intervals. This assumption cannot be tested using available observational records of length a few decades to perhaps a century at most. Here we use a recent large ensemble simulation of the North American climate with the Canadian regional climate model CanRCM4 to assess whether the max-stability holds for annual maxima of extreme precipitation. We find that in the CanRCM4 simulated climate, annual maxima tend not to be max-stable. We discuss reasons why the annual maxima of precipitation extremes may not be max-stable and explore the implications of our findings on the estimation of very long period return levels.