Anthropocene flooding: Challenges for science and society

Global Institute for Water Security, School of Environment and Sustainability, Department of Civil, Geological, and Environmental Engineering, University of Saskatchewan, Saskatoon, Saskatchewan, Canada School of Geographical Sciences and Urban Planning, Arizona State University, Tempe, Arizona School of Civil, Environmental and Mining Engineering, University of Adelaide, Adelaide, Australia Water Institute and Department of Economics, University of Waterloo, Waterloo, Ontario, Canada Institute for Environmental Studies, Vrije Universiteit Amsterdam, Amsterdam, The Netherlands Department of Civil and Environmental Engineering, Imperial College London, London, UK

This event caused 76 fatalities and more than $2 billion in damage.
The combination of heavy rain and melting snowpack in the Ottawa River basin, which is regulated by a series of reservoirs, produced record-breaking flood peaks in the Canadian cities of Ottawa and Gatineau and surrounding communities in April. Record-breaking spring floods also occurred across the Mississippi River and its tributaries, a heavily managed water system in the United States, causing four casualties and more than $12.5 billion in damage.
Public discourse about these events centred on the efficacy of infrastructure design and management and the operational rules that govern day-to-day decision-making. Australians questioned whether the management of Ross River Dam contributed to flooding in Townsville when a large amount of water was released downstream (Smee, 2019). Iranians in many of the affected areas wondered why large existing reservoirs did not protect them from flooding (ISNA, 2019).
Reservoir operations also came under scrutiny in the Ottawa floods (CBC, 2019), in response to public concerns, and in the United States, where residents blamed the US Army Corps of Engineers for the mismanagement of hundreds of reservoirs in the Mississippi River Basin (Fernandez & Schwartz, 2019). In the latter case, local residents were critical of opening Keystone Dam and releasing a large amount of water into the Arkansas River, which had a particularly severe impact on the city of Sand Springs, Oklahoma.
These real-time operational concerns arose from deeper, more fundamental issues associated with infrastructure planning and design, human settlement in the floodplain and perceptions of risk.
Also, they raised serious concerns about the credibility of official floodplain risk maps. For example, in the Townsville and Ottawa flood events, many affected properties were uninsured because they were considered to be in flood-free zones, but now face the prospect of becoming uninsurable. In Iran, many of the flooded roads and properties were built in the last few decades in areas now part of the floodplain.
As we explain below, the above issues arise because traditional scientific solutions to flood risk management are unable to adapt to the Anthropocene. We highlight two grand challenges facing science and society and outline the tools required to meet these challenges.
The overarching goal is to inform decision-making related to land use, floodplain management, infrastructure design and operational protocols. The ability to capture the uncertainty and complexity of future flood risk and society's exposure and vulnerability to such hazards is especially relevant.

| EMBRACING UNCERTAINTY
Both scientific and engineering practice emphasizes deterministic solutions despite the fact that the magnitude and frequency of flood hazards are changing in a deeply uncertain manner due to compound effects from a range of natural and human drivers. Anthropogenic climate change may potentially alter the intensity, duration, frequency and spatial distribution of precipitation, as well as the resulting streamflow, coastal inundation, storm surge, tropical cyclones and hurricanes. Dams and reservoirs, while perceived to modify streamflow regimes in "known" engineered ways, often add significant uncertainty to flood risk. Their operating rules are typically confidential and can be ad hoc at the time of crisis to address emergencies such as dam security. Runoff generation and river routing mechanisms are changing in a variety of ways due to urbanization, land cover/land-use change, artificial drainage and river training, as well as climate warming effects such as changed regimes of snowmelt and rain-on-snow floods. The combined effects of changing human and physical drivers produce profound uncertainty because of our limited understanding of how they interact and evolve under conditions that differ from those experienced in the past. Milly et al. (2008) argued that changing climate and land-use patterns of the past are no longer an adequate basis upon which to predict the future, and that "stationarity is dead." They called for development of non-stationary probabilistic models of environmental variables to optimize water systems. The need for new models and tools is nowhere more apparent than for the stationarity-based notion of return period, which remains the basis for floodplain analyses and  1960 1962 1964 1966 1968 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998  F I G U R E 1 Annual frequency of disastrous floods across the world has been the largest among all types of natural hazards and increasing over time. The data shown were obtained from EM-DAT International Disaster Database (accessed on January 26, 2020). According to EM-DAT, an event is deemed disastrous if it conforms to at least one of the following criteria: 10 or more people dead; 100 or more people affected; the declaration of a state of emergency; or a call for international assistance lack of information (Stakhiv, 2011). The classic flood risk analysis paradigm, based on best-guess estimates of how the future might look, is obsolete in the face of radically uncertain climate, environment and society.

| DISCOVERING COMPLEXITY AND TRADE-OFFS
The design and operation of water systems and other public infrastructure often involve multi-dimensional trade-offs between upstream and downstream users, flood and drought protection, economic growth and endangered species protection, and rural and urban needs.
Engineering-based cost-benefit analysis methods focus on a small subset of these dimensions and often ignore the rest. Natural,

| NEXT STEPS
Recent flood events suggest that science and society are ill-prepared for the difficult decisions that lie ahead as the flood events and impacts of the Anthropocene unfold. These decisions require a new science-policy paradigm that grasps and systematically accounts for the uncertainty and complexity associated with future flood risk, based on the following three fundamental underpinnings.

| From predictions to scenarios
The concepts of flood probability and exposure need to be re-defined to accommodate alternative future scenarios, to address the limited predictive power arising from the deep uncertainties and profound complexity of coupled human-natural systems in the Anthropocene.
Scenarios are consistent stories about the future of systems that are too complex to predict (Wiek, Keeler, Schweizer, & Lang, 2013). They cover a range of plausible futures, including rare catastrophic conditions, and their development provides a process for stakeholders to share competing views of the future (Lempert, Popper, & Bankes, 2003). Scenario generation facilitates robust decision-making, the search for solutions that work across a range of future conditions (e.g., climate change, infrastructure development, operational decisions, land-use change and policy decisions). Sensitivity analysis tools can be used to guide scenario generation by identifying dominant controls of human-natural systems (Razavi & Gupta, 2015). Recent advances in the field of Decision Making Under Uncertainty (Lempert et al., 2019;Maier et al., 2016) can help us to move away from crisp solutions that induce a false sense of security, to focus on critical scenarios that need attention, characterize associated trade-offs and enable decision-makers to think beyond "return periods" and the likelihood of floods (Gober, 2018).

| Building a science-public interface
Broad participation of community stakeholders in modelling efforts, scenario development and cross-sectoral trade-off analysis is urgently needed to ensure that scientific tools represent the interests, beliefs and disagreements of local communities, economic sectors and public institutions, and to increase the likelihood that the best-available science will be used for flood management decision-making. Social scientists have called attention to the importance of close and iterative collaboration between stakeholders and scientists and more effective communication of science for decision-making (Dilling & Lemos, 2011). In this new social framework, society becomes an active force in developing and managing uncertain and complex systems exposed to flooding rather than a passive victim of flood damage. Efforts under the European