Catastrophe models are computer-assisted calculations that estimate financial losses resulting from natural hazard events. Created primarily for insurance purposes, catastrophe models quantify expected losses due to claims of policyholders affected by a particular hazard, such as a flood or earthquake. The information generated by catastrophe models is valuable to insurers for many reasons, including understanding their exposure to perils, informing risk-based premiums, and detecting areas that are uninsurable due to their high level of risk (Botzen and van den Bergh 2008; Lloyds 2014). Private sector catastrophe models are proprietary in nature, so access is restricted to insurers who are willing and able to invest in data and technology. As a result, they are often unavailable for public research studies and mass dissemination, and their use in the public domain is rare (Sampson et al. 2014).
However, the information generated by catastrophe models is a potentially valuable input for public policy. First, by using catastrophe models to identify areas that are particularly prone to flood damage, insurers generate damage and loss information that could be used to improve maps of at-risk communities (Surminski and Thieken 2017). Such maps could enable governments to prioritize investments in flood mitigation and encourage homeowners to purchase flood insurance. Second, catastrophe models generate loss estimates resulting from both frequent and rare floods, which could offer governments a basis to weigh the costs and benefits of flood mitigation investments (e.g., structural protections along rivers), regulate land use to reduce property exposure, and determine ways to share flood risk among governments, private stakeholders and homeowners.
Similar to global trends, flooding is Canada’s largest contributor to disaster losses, estimated to account for 78% of federal disaster assistance costs (United Nations International Strategy for Disaster Risk Reduction (UNISDR), 2011; Parliamentary Budget Officer (PBO), 2016). Flood-related losses are influenced by several factors, including population growth and economic development in flood-prone areas, a reduction of permeable surfaces (e.g., wetlands) and the impacts of climate change (Kundzewicz et al. 2014). Climate change is a source of uncertainty for flood risk management, because physical components of the hydrological cycle are subject to change (e.g., extreme rainfall is expected to become more frequent). This changing flood regime demands accurate and up-to-date information on flood risk (Alexander et al. 2016), but existing flood maps in Canada are outdated, typically focus on a single type of flood hazard (e.g., riverine) and provide no information to estimate economic consequences (MMM Group 2014; Stevens and Hanschka 2014). However, catastrophe models that examine riverine and surface water flood risk have improved in recent years due in part to the introduction in 2015 of residential flood insurance as an optional coverage in select provinces (Calamai and Minano 2017; Meckbach 2016).
This study aims to contribute to a growing body of literature on the quantification of flood risk under the current and future climate (Bouwer et al. 2010; Feyen et al. 2009; Kundzewicz et al. 2014). For the purposes of this research, we adopted a relatively narrow definition of “flood risk” to mean potential direct economic losses associated with flooding. Re/insurance catastrophe models made available to the researchers were used to (1) quantify flood risk (i.e., direct economic losses) caused by rainfall-driven riverine flooding in residential areas, and (2) determine how the information generated can be used to inform Canadian public policy in the face of a changing climate. The models were tested in Halifax Regional Municipality, Nova Scotia—a mid-sized coastal city that has recently experienced riverine floods due to heavy rainfall (CBC 2014).
The paper begins by providing an overview of catastrophe models and Canada’s current disaster management policy landscape. It then describes the study methodology, including why the study area was selected, the procedure used to map present flood risk and how future flood-related losses were estimated. The third section presents the results, with a focus on how climate change could influence flood-related losses. The paper closes with a discussion of the relevance of the results for Canadian disaster management policy and for international scholarship on flood risk in a changing climate.
Climate change, catastrophe models and Canada’s disaster management policy
Climate change impacts regions across the globe in different ways, particularly in how it affects the hydrological cycle. The Intergovernmental Panel on Climate Change (IPCC) (2012), p. 13 states that “it is likely that the frequency of heavy precipitation…will increase in the 21st century over many areas of the globe”, particularly in high latitude regions like Canada. Greater precipitation will cause “increases in local flooding in some catchments or regions” (Intergovernmental Panel on Climate Change (IPCC), 2012, p. 13). The impact that climate change has had on historical flood-related losses (i.e., direct economic damages) is less clear. Socioeconomic development, rather than climate change, has been considered the primary contributor to increasing global natural hazard losses (Bouwer 2013), and increasing exposure of people and property in flood-prone areas is considered to be the primary driver of growing flood losses in recent decades (Kundzewicz et al. 2014).
Changes in the frequency and intensity of rainfall over the next century will likely create more opportunities for flood damages to occur in some regions (Rosenzweig et al. 2002; Bouwer 2013; Kundzewicz et al. 2014). Quantifying future flood risk is difficult, however, because there are many factors that influence flood losses, including uncertainties surrounding emissions pathways (Bouwer 2013). Flood risk is the product of exposure (i.e., assets likely to be affected by flooding), the frequency of occurrence (i.e., how often flooding impacts an area), and vulnerability (i.e., susceptibility to suffering damages) (de Moel et al. 2009). Climate change influences flood risk in that the probability of flooding events changes in response to greenhouse gas emissions. For example, one of the key impacts associated with different emissions scenarios is changes in the return period of extreme precipitation events (e.g., “a 1-in-20-year annual maximum daily precipitation event is likely to become a 1-in-5-year to 1-in-15 event by the end of the 21st century in many regions”, including Eastern Canada) (Intergovernmental Panel on Climate Change (IPCC), 2012).
Perhaps due to the complexity in quantifying flood risk, there is a relatively small body of literature that explores the impact of climate change on flood losses (Bouwer 2013). Bouwer (2013), for instance, states that “there are only few studies that have translated such changes in extreme weather to economic impacts, and very little quantification is usually given of how large the impact from climate change on extreme weather losses potentially is” (p. 916). Moreover, existing studies in developed nations focus primarily on European countries and the United States, with few studies from Canadian coastal regions (Bouwer 2013; Kundzewicz et al. 2014; Lemmen et al. 2016). Efforts that quantify flood risk are beneficial particularly for determining what is at risk, estimating increases in losses over time, and analysing how policymakers and decision-makers can address these growing risks (e.g., finding a balance between risk reduction costs and benefits).
Many of the factors that influence flood risk are found in catastrophe (CAT) models (Bouwer 2013), including exposure, hazard and flood probability reflected as a stochastic event set (Sampson et al. 2014). The stochastic event set is composed of thousands of flood event simulations that are informed by observational data but also capture events that exceed the magnitude of historical data to capture tail-end risks (i.e., rarely occurring floods that have potentially catastrophic consequences) (Sly and Ma 2013). The event set offers a “database of extreme precipitation events over the catchment(s) that drive fluvial or pluvial risk” (Sampson et al. 2014, p. 2306). CAT models also contain location-specific information, such as the siting and characteristics of property assets, and can therefore estimate damage due to certain flood depths (depth-damage functions) to determine which assets are most vulnerable to flood-related losses (Lloyds 2014; Sampson et al. 2014). However, research coupling climate change and CAT model data is sparse, meaning there is an opportunity to make better use of this technology to inform public policy and investments aimed at reducing flood risk.
Canada and 186 other countries have adopted the Sendai Framework on Disaster Risk Reduction (Henstra and Thistlethwaite 2017b). The framework identifies four “priorities for action”, which include using risk assessments to better understand disaster risk; strengthening governance to manage disaster risk; investing in disaster reduction and resilience; and enhancing disaster preparedness for effective response in order to “build back better” (United Nations International Strategy for Disaster Risk Reduction (UNISDR), 2015). Although these principles are not new—they build on a solid foundation of research that dates back more than 70 years (e.g., White 1945; Burton et al. 1978; White et al. 2001)—they have experienced a renewed emphasis as the costs of natural disasters have risen dramatically in recent years.
Despite Canada’s commitment under the Sendai Framework to adopt risk assessment as the basis for disaster risk reduction, many Canadian jurisdictions lack up-to-date flood risk information (MMM Group 2014). Although the Government of Canada has a renewed effort underway to improve flood mapping (Natural Resources Canada, 2017), a lack of information has impeded efforts to prevent or reduce flood consequences on people and property. The primary objective of this research was to use CAT models to estimate the influence of climate change on flood risk in a Canadian municipality, in order to inform flood risk management decisions.
Study area
The study area consisted of the urban core of Halifax Regional Municipality (HRM), Nova Scotia, Canada—a geographically large municipality with spatially distributed populated areas, particularly along the coastline (Fig. 1). In comparison to other parts of the municipality, the urban core has the highest population density, encompassing approximately 74% of the 400,000 people who reside in HRM (Statistics Canada 2016).
HRM has a relatively mild climate compared to other parts of Atlantic Canada, with a mean annual temperature of 7.5 °C and mean annual precipitation of 1468 mm (Environment and Climate Change Canada (ECCC), 2018). The city experiences few extreme hot and cold days, due to moderation by the Gulf Stream, but it faces occasional hydrological risks due to hurricanes, Nor’easter storms, and rainfall-driven riverine flooding. This research focused on rainfall-driven riverine flooding, since it had recently affected neighbourhoods in HRM and many other communities in the Canadian Maritimes (Harding 2017; Weeks 2017). Riverine flooding can occur when extreme rainfall or ice jams cause rivers to reach their capacity and overflow onto surrounding land. Since studies in Nova Scotia have focused primarily on coastal flood impacts (including sea level rise) (see Leys 2009; Lemmen et al. 2016), this study offered an opportunity to develop new knowledge about flood risks on Canada’s East coast. In addition, HRM staff had recently identified a need to update their riverine flood maps and floodplain regulations—HRM restricts residential development in the 20-year floodplain but allows flood-proofed development in the 100-year floodplain (Environment and Climate Change Canada (ECCC), 2013; Berman 2016; Irish 2016).
In the event of a disaster, governments provide compensation funds to those affected. The availability of reliable climate change scenarios for this part of Canada also supported its selection as a test site for projecting the influence of climate change on flood loss estimates in the near and long-term future. Finally, the study area covered a reasonably small catchment area, which allowed the researchers to maximize the likelihood that changes in rainfall conditions caused by climate change would produce relatively uniform responses in riverine flood conditions. The next section describes the study’s research design and methods.