全球变化知识资源中心

The hazard data shown here are from the SSBN global flood model¹⁴, and show 1-in-100-year maximum flood depth for: a, all of Africa; b, River Niger at Onitsha, Nigeria; and c, Blue Nile at Omdurman, Sudan.

Of course, GFRMs are no panacea. A question regularly posed by users is what the exact risk is for a particular community, street, or asset. Such questions highlight the apparent mismatch between the resolution and accuracy of GFRMs and the detail of information required for local-level decisions. The accuracy of flood hazard maps is dependent on the quality of the elevation data (digital elevation models; DEMs) used. Although a near-global DEM of 30 m resolution is now available^{18, 19}, flood hazard map accuracy at such scales remains limited due to a number of factors, including: accuracy of boundary conditions used to force inundation models (for example, meteorological data)⁵; limitations in knowledge of river geomorphology (for example, river profiles and roughness); and the accuracy of DEMs themselves²⁰. To estimate risk at such detailed scales, local data on exposure (how many people, buildings, and other elements are located in harm's way) and vulnerability (how susceptible are those exposed elements to flooding) are also required. Obtaining detailed information on exposure and vulnerability at the global scale remains an open challenge.

While GFRMs are used to assess the effectiveness of DRM strategies at the large scale, the granularity of the input data means that they cannot be used to design individual DRM measures. For example, while we can assess damages avoided by protecting a country against a 100-year flood, we cannot use GFRMs to design the actual measures needed to achieve this. In the example of dykes, questions regarding location, height, material, and so forth, cannot be answered using GFRMs. These limitations must be clearly communicated to users at all stages of engagement, to avoid mismatches in expectations down the line. A key challenge is the lack of data on flood management strategies already in place. Flood management has already played a strong role in reducing vulnerability to flooding²¹; ignoring this leads to gross overestimates of risk⁵.

In addition, emergency responders require flood risk information in terms of the potential impacts of a flood event, which is currently not contained in global flood forecasting and monitoring systems. At present, most of these systems only provide physical flood thresholds, based on statistical analyses of predicted riverflow, since detailed global flood hazard modelling in near real-time remains a daunting task.

A priority for increasing GFRM accuracy is an improved representation of fundamental physical and socioeconomic processes leading to flood impacts. For hazard modelling, we need better elevation data²⁰. While the recently released 30 m DEM¹⁹ has improved horizontal resolution substantially, it did not improve the vertical accuracy required to improve global inundation models. An exciting commercial initiative towards such an improvement is WorldDEM, which is to feature a large increase in vertical accuracy. Efforts have been undertaken to yield river-width data at the global scale¹³, but more data are needed on other key geomorphological aspects, not to mention the presence of anthropogenic features in hydraulic systems, such as dykes and levees.

But without profoundly improved representations of exposure and vulnerability, gains from improved hazard modelling will not filter through to improved risk estimates. At present, GFRMs use simple aggregated land-use data to represent exposure, such as 'urban area' per cell, and vulnerability is commonly represented by a handful of depth-damage functions. These assumptions ignore huge spatial heterogeneities in exposure and vulnerability²². An exciting initiative that could revolutionize GFRMs is OpenStreetMap (OSM), a global crowdsourced database of buildings and infrastructure. OSM is being used to support local DRM worldwide²³. A challenge for the global risk modelling community is to find smart ways to harness this wealth of object-based exposure data for GFRMs.

In many of the world's most 'risky' places, such as rapidly urbanizing deltas, flood hazards are the result of the interaction of river discharge and coastal sea levels. We therefore need to move towards an integration of river and coastal flood risk modelling to accurately assess risk in these places. The improvements described above would help with this, but it would also require developing dynamical global tide and storm surge models, and linking these with river flood models. Further gains in granularity, accuracy, and model speed could be made by nesting local flood models within global models. Such nesting has a long tradition in global climate modelling²⁴, and could offer GFRMs the advantage of gaining more detail in areas where it is required.

Urgent efforts are required to develop databases showing current flood management worldwide. A database of flood protection standards per country is expected shortly, providing a valuable first step. Moving forward, the community requires information on which specific measures are taken and where, in terms of not only structural measures, but also non-structural measures such as flood zoning, building codes, and insurance. We could learn from crowdsourcing successes in other fields, such as OSM, by using such platforms to develop crowdsourced databases on the location and height of protection measures such as dykes and levees. Such a database could play an important role in monitoring DRM and climate change adaptation efforts.

Global-scale flood forecasting and monitoring systems need to go beyond the provision of solely information on exceedance above physical flood thresholds. Emergency responders require information on forecasted flood impacts, and should be able to link their action thresholds to global-scale flood forecasting and monitoring systems. Connecting the offline global flood risk information from GFRMs to global flood forecasting and monitoring systems is a key to further increasing the usability of both of these model families.

The effectiveness of risk information in supporting DRM is largely contingent on how the information is communicated and its timely availability. When a disaster occurs, information is required immediately — there is no time for lengthy periods to commission new studies, carry out new simulations, analyse and present results, and so forth. A range of tools has recently been developed to communicate risk in a clear and effective way. One example is InaSAFE, an open-source tool for impact assessment, now used extensively to support decision-making at national and sub-national levels in Indonesia and elsewhere. Successes like InaSAFE and the Aqueduct Global Flood Analyzer emphasize the demand for translating complex flood risk data into interactive and actionable information. While a first push in this direction is being made, further developments are required.

Not only should information be actionable, its (deep) uncertainties should be estimated and transparently communicated. At present, there is clear frustration regarding the lack of guidance from modellers on the reliability and uncertainty contained in model results. A better representation of uncertainty may lead to better decision-making²⁵. Techniques to communicate and visualize uncertainty in relation to decision-making already exist, such as adaptation pathways²⁵, but such novel techniques still need to be made fit for use with GFRM outputs.

Addressing this research agenda is a daunting task for any single organization. It requires interdisciplinary research, close collaboration with user communities, and grounding in a wider context of DRM and climate change adaptation. Launched in 2014, the Global Flood Partnership provides a platform for organizations involved in global flood risk, early warning, and observation to collectively identify and address challenges²⁶. As a bottom-up, voluntary partnership, it allows direct knowledge transfer from centres of excellence in global flood risk modelling to users, including hydrological and emergency management authorities, humanitarian response organizations, international development banks, and the private sector. Harnessing the knowledge and drive of this network should help us push towards a new generation of actionable GFRMs for contributing to real-world DRM and climate change adaptation.