Cloud Hosting

Computational science, including numerical simulation through high-performance computing, data analytics and visualization, can underpin the SDGs. This is especially true when deployed in collaboration with other scientific domains and as part of co-produced knowledge-generation processes with a range of urban stakeholders and end users. By acknowledging the systemic nature of the causes of disasters, such research must facilitate the inclusive engagement of scientists, engineers, policy-makers, economists, private sector groups and, critically, representatives of the citizens who will live in the cities experiencing rapid growth. The SDGs recognize that urban development plans made today will either brighten or blight the lives of citizens for centuries.

A three-part agenda for interdisciplinary science marks out how computational science can be used to underpin and catalyze this ambition. Each step in the agenda can stand alone or together form a structured process from better understanding to better action to reduce disaster risk in future cities.

Digitally capturing inclusive future visions

Many social science methodologies are available with which to solicit preferences for neighborhood or city-wide futures. A challenging task is enabling such methodologies to capture the subjective visions of the future of diverse urban stakeholders. Only by doing this can future cities disrupt established norms and consider what a safer city of the future looks like from different or multiple perspectives. The outputs of such methodologies are difficult to translate into policy options; they are often qualitative and can appear imprecise to policy-makers. However, such qualitative information has a huge potential to act as a basis for future urban scenario development if it can be assimilated into precise digital representations. Computational science can help here. Spatial components of the projections (such as desired land-use zones and their attributes) can be translated into land-use plans using geographic information systems and related computational tools. This information can be complemented with predicted patterns of urban growth, determined using machine learning algorithms that rely on remote sensing data. Spatial priorities emerging from stakeholder groups are thus rendered into high-resolution digital representations of possible urban futures8.

Such digital future cities also incorporate detailed attributes of people and assets. These include engineering characteristics for each building and infrastructure component and system, information on socio-demographics for each individual and household, and data on socio-physical interdependencies (for instance, where each person goes to work and where each child goes to school). The virtual representation is achieved using several computational models, including synthetic population-generation algorithms, human mobility methods, procedural modelling and optimization processes8,9.

Exposing digital futures to likely hazard events

The high-resolution virtual representations of possible future urban developments must be exposed to hazard events that are consistent with the hydrological and geophysical environment of the city. A series of hazard events can be selected to cover possible life-cycle experiences of the development. A key effort in the Tomorrows Cities project, for example, has been to code these events into high-resolution, physics-based simulations10 (Fig. 1), taking advantage of the latest developments in high-performance computing. A custom-developed web-based application merges site-specific intensity data from the hazard event with exposure and vulnerability information in the future development scenario to compute the likely impact of any particular event. These calculations use several underlying computational tools, including high-resolution, multi-hazard fragility models developed from detailed building-level numerical performance assessments11 and data-mining models that can distinguish the magnitude of disaster impacts on the basis of social vulnerability indicators12,13. Agent-based modelling is another powerful approach in the field of disaster simulation that allows researchers to simulate the dynamic behavior of individual entities (agents), with their socio-economic features, within a complex system.

Disaster reduction programs often depend on externally developed solutions imposed on specific local challenges. Computational science provides digital tools that can support innovative capacity strengthening, freeing possible futures thinking from the responsibility for real lives and encouraging experimentation with innovative planning solutions. a,b, Here, the virtual city of Tomorrowville is shaken by a virtual earthquake (a) and flooded by a virtual extreme rainfall event triggered by climate change (b), exposing spatial variability in exposure and driving reconsideration of spatially uniform building regulations.

Depending on the particular scenario, multiple impact metrics reflecting diverse aspects of the lived experience (for instance, number of deaths, number of displacements, number of injuries, hospital occupancy, lost days of production or school and total replacement costs) can be calculated and mapped, providing a detailed picture of the total impact of any disaster event resulting from the decisions and policies that generated the specific digital future being tested. Each of these metrics can be disaggregated in different ways, including by age, gender, income or any other attribute contained in the demographic dataset of the virtual future representation, providing an understanding of the consequences of the decisions made during planning and scenario building.

To complete the picture of the root causes of disaster impact for policy-makers, quantifying and mitigating social vulnerability (the susceptibility of an individual from a given group to the impacts of hazards) can help to build resilience to multiple types of hazard shock. So far, there is a dearth of disaggregated data recording disaster impacts and social vulnerability measures simultaneously, and the current priority is to collect longer data series. These might emerge, for example, from satellite remote sensing; computational methods in unsupervised learning and data clustering as well as deep learning (for instance, neural networks) could then be leveraged to refine quantitative modelling of social vulnerability. Exploring nonlinear and multi-scalar relations between exposure, vulnerability and disaster impacts is an important research ambition14.

Convening risk agreement and institutional learning

Impact is objective, but risk depends on personal or group priorities; the value of property replacement, for example, has a different priority depending on whether or not you own property. Computational science supports interactive representations of complex urban impact scenarios, facilitating the quantification of subjective risk priorities by generating impact-weighting matrices that include the voice of marginalized groups in the local definition of disaster risk. Equipped with weighted risk definitions, attention turns to exposing the root causes of such risk in the choices and decisions behind any development plan. Dynamic digital visualizations of the impact metrics produced by simulation-based tools could help to elucidate the distribution of risk inherent in development planning and to diagnose risk drivers, inverting complex causal chains and exposing the underlying flaws in decision-making. In the case of the Tomorrows Cities Hub, this is communicated to stakeholders through the web-based application. More formal inversions uncovering root causes from impact metrics are needed to clarify the diagnosis and reinforce evidence-based decision-making for risk reduction.

Focusing on the origins of risk in the decisions, policies and assumptions underpinning future development scenarios allows stakeholders to examine their choices and reflect on broader governance questions. Modifications to particular stakeholder priorities that are likely to lead to reduced risk are implemented in the digital development scenarios. These are then subjected to the same simulated hazard events to test the resulting risk reduction. The process is iterated, optimizing the future for lower risk, elucidating the effectiveness of governance processes and supporting evidence-based decision-making.

Original post:

Reducing disaster risk for the poor in tomorrow's cities with ... - Nature.com