Accelerating anticipatory action across the IFRC — a systems approach

A little history

The International Federation of Red Cross and Red Crescent Societies (IFRC) and its global network of 192 National Societies face numerous challenges when it comes to preparing for and responding to disasters brought on by natural — and particularly weather-related — hazards. As long ago as the 1920s, the Red Cross and Red Crescent Movement recognised the need to learn from past disasters in order to take effective early actions to mitigate risks and lessen the impacts of natural hazards that cannot be avoided.

In 2015, IFRC’s specific role was established in Principles and Rules for Red Cross and Red Crescent Humanitarian Assistance, a Movement-wide coordination and cooperation framework adopted by the National Societies. According to these Principles, the IFRC has specific responsibilities related to hazard monitoring and the use of risk information, evidence and tools n order to help National Societies prepare for future disasters:

3.4 The International Federation shall monitor hazard information and disseminate relevant warnings to National Societies and promote and share information on key disaster preparedness efforts, initiatives and good practices at national, regional and global levels.

3.5 The International Federation shall ensure that globally consistent policies, rules, standards, tools and procedures relevant to disaster response are developed, disseminated and updated, as well as promoting their use in local contexts.

In the past two years, IFRC’s efforts have focused on scaling up systems for early warning and anticipatory action, which are strongly recognised in the current IFRC ten-year strategy and its operational Plan and Budget 2021–2025.

IFRC’s role as curator of others’ risk and impact-forecast models

Rather than reinvent the wheel, IFRC and National Societies are learning from what’s already been achieved using predictive models in other sectors, such as public health. Furthermore, instead of developing its own risk models, IFRC is playing the role of facilitator-curator to ensure that the National Societies can access the forecasts with the best track record. Toward this end, IFRC has have begun collaborating with impact-forecasting teams from the University of Oxford, MeteoSwiss, the Netherlands Red Cross’s 510, the European Commission’s Joint Research Centre, the World Food Programme, UN OCHA, the Italian Red Cross and Government of Italy, and the Pacific Disaster Center — to name just a few.

In line with its curatorial role, IFRC first engaged with partners such as UNDRR and UN OCHA in order to develop a list of indicators for the modelling teams to predict for each hazard event (figure 1), and to establish common definitions for each impact metric:

• the number of people exposed to and likely to be affected by the hazard
• the number of people displaced
• the number of homes, schools and health facilities damaged and destroyed
• the amount of crop and livestock losses

Information on these indicators is useful for national reporting systems as well as informing humanitarian decision-making. Forecasted and real-time estimates of these impacts can indicate the likely magnitude of the crisis, and they indicate how often these impacts are likely to occur in a given location (figure 1).

Figure 1: Screenshot displaying the forecasted path of a hypothetical cyclone, its estimated impacts and the frequency these impacts are expected to occur

IFRC has been working behind the scenes with technical experts from the UN to standardise the way the data is shared — not just with IFRC but across the entire humanitarian sector. For example, for hydro-meteorological hazards, we are working with colleagues from the World Meteorological Organization’s Cataloging of Hazardous Events initiative and the WMO Coordination Mechanism (WCM) to support humanitarian activities to ensure alignment with internationally recognised standards and to ensure that we use data from authoritative sources. We have also begun developing guidance on how to share and visualise the data, including how to communicate the uncertainty of the predictions.

As a curator, IFRC is ensuring that users of this information have access to all the data they need in the most useful format. To ensure transparency, we will make all relevant information on how an impact forecast was produced and assessments of the historical accuracy of that particular model for a specific hazard or location.

Using risk information to scale up anticipatory action

The information and tools that IFRC will curate and develop are designed to support the implementation across the network of our operational framework for scaling up anticipatory action. One of the key requirements for scaling up — for both IFRC and the National Societies — is knowing in advance what questions the models need to answer. If the amount of action and resources are to increase, then the evidence base needs to be there to support it.

This initiative will support the localisation of anticipatory action by providing National Societies with more information that they need to identify the risks they face and take ownership of the way they will address them through anticipatory actions. The system and tools developed by IFRC will also help the National Societies leverage the best available evidence and risk information so that they can access the resources required to take action.

In order to know where to prioritise action, IFRC needs to access models to help answer questions about seasonal risks, such as “What kinds of crises are likely to occur over the course of the year, and where?” In order to establish triggers to respond to specific events, IFRC needs reliable, accurate information on both the risks and the expected impacts of forecasted events, such as “How many people are at risk of becoming displaced when the cyclone makes landfall?”

The use of this evidence requires knowledge of risk metrics as well as the ability to assess and weigh the uncertainty of each model. Thus, one of the key roles of IFRC will be to provide guidance and support, through training and workshops, on how to use risk information and impact-forecasts responsibly.

The evidence generated by impact-forecasts and other predictive models, and its responsible use, can help access more funding commensurate with the scale of the response required through the use of innovative forecast-based financing instruments.

In the last few years, we have made significant progress. For example, IFRC has contributed to the establishment of the Anticipation Hub, increased funding from its own Disaster Relief Emergency Fund, and launched a beta version of an alerting system using the common alerting protocol. At the same time, in order to achieve our ambition, we need to address two key challenges:

Challenge 1: Using models responsibly

At present, hazard-impact modelling is still in its infancy. There is a relatively small but rapidly growing number of institutions producing models, and these models are being used to predict various impacts such as the number of people displaced or hospitalised, or the amount of physical damage to homes, schools, hospitals and other critical infrastructure.

This work is still not yet well coordinated, which makes it difficult for potential users to compare the relative skill and accuracy of the different models, or to assess improvements in their accuracy over time. This can be dangerous. If users don’t know how to properly use the models at their disposal there can be significant unintended consequences — from failing to evacuate people in harm’s way to creating a panic over a relatively minor event. Misinterpretation or misuse of risk models can therefore cause harm for people affected by and responding to crises — with knock-on effects for the modelling community itself. The hype around predictive analytics and new approaches using artificial intelligence and “big data” has led to heightened expectations and exposed structural biases within the models. When these expectations are not met, the repercussions are felt by the loss of confidence not just in the model in question but in the entire approach.

Challenge 2: Coordinating and customising multiple models for multiple use cases

The second challenge relates to the mismatch between different model outputs and the need to support different use cases. If each model is using a different set of assumptions and metrics, we will struggle to compare accuracy, and ultimately be unable to compare and improve collectively.

Furthermore, the IFRC requires modelling outputs to align with agreed Federation-wide means of organising information for emergency response decision-making. It also requires providing space for National Societies to participate in the development of locally owned and recognised forecast-based financing instruments — but without duplicating what already exists.

In short, this challenge requires greater coordination of the modelling teams so that their estimates contribute a comprehensive understanding of the crisis at various points in time and from different perspectives. This coordination should aim to ensure their respective impact metrics align not only with one another but with IFRC’s approach to understanding and analysing the scale and scope of crises and how the severity and needs evolve over time. Using IFRC’s common analytical framework and standard risk and impact indicators makes it easier to compare crises and prioritise resources and attention.

IFRC’s approach

IFRC has adopted a holistic approach to address both challenges. First, we are building a system that will ingest and visualise predictions from multiple partners’ models for a standard set of impacts: the number of people likely affected and displaced, as well as the expected damages to homes, schools, hospitals and critical infrastructure. The crisis data bank will record the predictions for thousands and thousands of events and then ground-truthed observational data on those same indicators for those same events (figure 2).

Figure 2: Schematic diagram of the integrated IM system and the Crisis Data Bank’s place within it

Given the relative infancy of impact-forecasting modelling within the humanitarian sector, it is essential that we establish a baseline that transparently reflects the skill of these models at present. If all models are wrong, then new models are especially prone to err in their predictions. We need to accept that up front so that we can measure and then demonstrate improvements in the models over time, which should build confidence in the models. What we commit to is presenting the modelled estimates and the validated observations side by side, with full transparency, and to show rather than hide the uncertainty associated with the predictions.

To address the second challenge, IFRC is adopting a common crisis analysis framework, one which aligns the impact forecasts with the data collected by National Societies and partners in the field. This observational data will be fed into standard templates for requesting funds for disaster preparedness and response — and, crucially, it will be shared with the modelling teams and compared with their predictions. These last steps will enable users of the GO platform’s risk module to compare the accuracy of the different models and track if and how the models’ accuracy improves over time.

By adopting this approach, IFRC aims to improve all of the models, which will be a necessary step for scaling up anticipatory actions taken by National Societies and the communities of which they are part, as well as government agencies and their UN and civil society partners.

What’s next — and how to get involved

As one might imagine, developing this system is a large, long-term collaborative effort. We will kick things off this month with the testing and launch of the beta version of it on the GO platform. Over the coming months we will visualise more data — impact-forecasts from additional partners, more impact metrics and more data about uncertainty. We will also add new features and functionality to ensure that the system continues to meet the needs of its users — and over the coming months we will share updates here about how the system is being used and what’s coming.

Developing this system is also a highly collaborative initiative, and we are always on the lookout for new partners. Therefore, if you’d like to get involved there are a number of ways to do so, including but not limited to the following:

• testing the beta version of the system to help us improve what’s already there;
• providing impact estimates for forecast and detected natural hazards; and/or
• providing data on impacts of these hazardous events, such as the number of people displaced or injured; or the number of homes, schools and health facilities damaged and destroyed.

If any of this sounds interesting to you, we’d be delighted for you to get involved. You can do so by contacting Justin Ginnetti (justin.ginnetti@ifrc.org), Senior Officer for Information Management and Risk Analysis, and indicating how you’d like to participate.