Votre recherche

A systems-oriented review of machine learning (ML) over the entire flood management spectrum, encompassing fluvial flood control, pluvial flood management, and resiliency-risk characterization was undertaken. Deep learners like long short-term memory (LSTM) networks perform well in predicting reservoir inflows and outflows. Convolution neural networks (CNNs) and other object identification algorithms are being explored in assessing levee and flood wall failures. The use of ML methods in pump station operations is limited due to lack of public-domain datasets. Reinforcement learning (RL) has shown promise in controlling low-impact development (LID) systems for pluvial flood management. Resiliency is defined in terms of the vulnerability of a community to floods. Multi-criteria decision making (MCDM) and unsupervised ML methods are used to capture vulnerability. Supervised learning is used to model flooding hazards. Conventional approaches perform better than deep learners and ensemble methods for modeling flood hazards due to paucity of data and large inter-model predictive variability. Advances in satellite-based, drone-facilitated data collection and Internet of Things (IoT)-based low-cost sensors offer new research avenues to explore. Transfer learning at ungauged basins holds promise but is largely unexplored. Explainable artificial intelligence (XAI) is seeing increased use and helps the transition of ML models from black-box forecasters to knowledge-enhancing predictors. © 2025 by the authors.

ABSTRACT Urbanization is leading to more frequent flooding as cities have more impervious surfaces and runoff exceeds the capacity of combined sewer systems. In heavy rainfall, contaminated excess water is discharged into the natural environment, damaging ecosystems and threatening drinking water sources. To address these challenges aggravated by climate change, urban blue-green water management systems, such as bioretention cells, are increasingly being adopted. Bioretention cells use substrate and plants adapted to the climate to manage rainwater. They form shallow depressions, allowing infiltration, storage, and gradual evacuation of runoff. In 2018, the City of Trois-Rivières (Québec, Canada) installed 54 bioretention cells along a residential street, several of which were equipped with access points to monitor performance. Groundwater quality was monitored through the installation of piezometers to detect potential contamination. This large-scale project aimed to improve stormwater quality and reduce sewer flows. The studied bioretention cells reduced the flow and generally improved water quality entering the sewer system, as well as the quality of stormwater, with some exceptions. Higher outflow concentrations were observed for contaminants such as manganese and nitrate. The results of this initiative provide useful recommendations for similar projects for urban climate change adaptation.

Combined sewer surcharges in densely urbanized areas have become more frequent due to the expansion of impervious surfaces and intensified precipitation caused by climate change. These surcharges can generate system overflows, causing urban flooding and pollution of urban areas. This paper presents a novel methodology to mitigate sewer system surcharges and control surface water. In this methodology, flow control devices and urban landscape retrofitting are proposed as strategies to reduce water inflow into the sewer network and manage excess water on the surface during extreme rainfall events. For this purpose, a 1D/2D dual drainage model was developed for two case studies located in Montreal, Canada. Applying the proposed methodology to these two sites led to a reduction of the volume of wastewater overflows by 100% and 86%, and a decrease in the number of surface overflows by 100% and 71%, respectively, at the two sites for a 100-year return period 3-h Chicago design rainfall. It also controlled the extent of flooding, reduced the volume of uncontrolled surface floods by 78% and 80% and decreased flooded areas by 68% and 42%, respectively, at the two sites for the same design rainfall.

There is mounting pressure on (re)insurers to quantify the impacts of climate change, notably on the frequency and severity of claims due to weather events such as flooding. This is however a very challenging task for (re)insurers as it requires modeling at the scale of a portfolio and at a high enough spatial resolution to incorporate local climate change effects. In this paper, we introduce a data science approach to climate change risk assessment of pluvial flooding for insurance portfolios over Canada and the United States (US). The underlying flood occurrence model quantifies the financial impacts of short-term (12–48 h) precipitation dynamics over the present (2010–2030) and future climate (2040–2060) by leveraging statistical/machine learning and regional climate models. The flood occurrence model is designed for applications that do not require street-level precision as is often the case for scenario and trend analyses. It is applied at the full scale of Canada and the US over 10–25 km grids. Our analyses show that climate change and urbanization will typically increase losses over Canada and the US, while impacts are strongly heterogeneous from one state or province to another, or even within a territory. Portfolio applications highlight the importance for a (re)insurer to differentiate between future changes in hazard and exposure, as the latter may magnify or attenuate the impacts of climate change on losses.

The Penman-Monteith reference evapotranspiration (ET0) formulation was forced with humidity, radiation, and wind speed (HRW) fields simulated by four reanalyses in order to simulate hydrologic processes over six mid-sized nivo-pluvial watersheds in southern Quebec, Canada. The resulting simulated hydrologic response is comparable to an empirical ET0 formulation based exclusively on air temperature. However, Penman-Montheith provides a sounder representation of the existing relations between evapotranspiration fluctuations and climate drivers. Correcting HRW fields significantly improves the hydrologic bias over the pluvial period (June to November). The latter did not translate into an increase of the hydrologic performance according to the Kling-Gupta Efficiency (KGE) metric. The suggested approach allows for the implementation of physically-based ET0 formulations where HRW observations are insufficient for the calibration and validation of hydrologic models and a potential reinforcement of the confidence affecting the projection of low flow regimes and water availability.

Abstract Recent flood events in Canada have led to speculation that changes in flood behaviour are occurring; these changes have often been attributed to climate change. This paper examines flood data for a collection of 132 gauging stations in Canada. All of these watersheds are part of the Canadian Reference Hydrometric Basin Network (RHBN), a group of gauging stations specifically assembled to assist in the identification of the impacts of climate change. The RHBN stations are considered to have good quality data and were screened to avoid the influences of regulation, diversions, or land use change. Daily flow data for each watershed are used to derive a peaks over threshold (POT) dataset. Several measures of flood behaviour are examined based on the POT data, which afford a more in‐depth analysis of flood behaviour than can be obtained using annual maxima data. Analysis is conducted for four time periods ranging from 50 to 80 years in duration; the latter period results in a much smaller number of watersheds that have data for the period. The changes in flood responses of the watersheds are summarized by grouping the watersheds by size (small, medium, and large) and also by hydrologic regime (nival, mixed, and pluvial). The results provide important insights into the nature of the changes that are occurring in flood regimes of Canadian rivers, which include more flood exceedances, reduced maximum flood exceedance magnitudes for snowmelt events, and earlier flood events. Copyright © 2016 John Wiley & Sons, Ltd.

ABSTRACT Large-scale disasters can disproportionately impact different population groups, causing prominent disparity and inequality, especially for the vulnerable and marginalized. Here, we investigate the resilience of human mobility under the disturbance of the unprecedented ‘720’ Zhengzhou flood in China in 2021 using records of 1.32 billion mobile phone signaling generated by 4.35 million people. We find that although pluvial floods can trigger mobility reductions, the overall structural dynamics of mobility networks remain relatively stable. We also find that the low levels of mobility resilience in female, adolescent and older adult groups are mainly due to their insufficient capabilities to maintain business-as-usual travel frequency during the flood. Most importantly, we reveal three types of counter-intuitive, yet widely existing, resilience patterns of human mobility (namely, ‘reverse bathtub’, ‘ever-increasing’ and ‘ever-decreasing’ patterns), and demonstrate a universal mechanism of disaster-avoidance response by further corroborating that those abnormal resilience patterns are not associated with people’s gender or age. In view of the common association between travel behaviors and travelers’ socio-demographic characteristics, our findings provide a caveat for scholars when disclosing disparities in human travel behaviors during flood-induced emergencies.

Abstract This study integrates novel data on 100-year flood hazard extents, exposure of residential properties, and place-based social vulnerability to comprehensively assess and compare flood risk between Indigenous communities living on 985 reserve lands and other Canadian communities across 3701 census subdivisions. National-scale exposure of residential properties to fluvial, pluvial, and coastal flooding was estimated at the 100-year return period. A social vulnerability index (SVI) was developed and included 49 variables from the national census that represent demographic, social, economic, cultural, and infrastructure/community indicators of vulnerability. Geographic information system-based bivariate choropleth mapping of the composite SVI scores and of flood exposure of residential properties and population was completed to assess the spatial variation of flood risk. We found that about 81% of the 985 Indigenous land reserves had some flood exposure that impacted either population or residential properties. Our analysis indicates that residential property-level flood exposure is similar between non-Indigenous and Indigenous communities, but socioeconomic vulnerability is higher on reserve lands, which confirms that the overall risk of Indigenous communities is higher. Findings suggest the need for more local verification of flood risk in Indigenous communities to address uncertainty in national scale analysis.