The need for medium-range forecasting in early warning systems to improve risk outcomes
High-impact weather events – floods, storms, cyclones – threaten human life and property, as well as affecting the economy and inflicting significant societal hazards. Being able to predict these events accurately, and with sufficient lead times, enables people to prepare for them. It is also more cost-effective: a dollar invested in disaster preparedness – which reduces people’s vulnerability to the impacts – can prevent six dollars’ worth of disaster-related economic losses.
With the current trajectory of global climate change, countries face continual increases in the frequency and intensity of extreme weather events. Increasing attention is being paid to climate and disaster risk assessment, strengthening systems for early warning and early action, and identifying options for adaptation, mitigation and risk reduction.
Resilience to high-impact weather events requires an integrated risk management approach. This must include initial identification of a hazard, creation of a risk register, development of hazard warnings and risk communication, and preparation for and response to an event when it occurs.
How we measure the weather
Short- and medium-term weather forecasting models (up to 10 days), such as those that are available on smartphones, combine models of the atmosphere and oceans with current weather observations to make predictions.
By contrast, medium- to long-term seasonal outlooks try to predict how different the climate will be compared to normal over the next three months (i.e., whether it will be hotter or colder). These are based on slower changes in planetary patterns, which can drive weather over a number of months. Examples include intermittent oceanic warming patterns (e.g., El Niño) and the extent of sea ice coverage in the Arctic Ocean.
Medium-term forecasts are particularly challenging, as the initial information from short-term models is no longer as useful, while the long-term climate drivers associated with long-term forecasts are not yet apparent (WMO, 2012). However, advances in technology mean that the field is improving, as better computer models and new insights emerge about the atmospheric and oceanic patterns that drive weather over the long term (Witze, 2020).
Ensemble forecasting and early action
There is increasing demand for a longer lead times in predictions, as effective forecasting for hazards such as floods, droughts and cyclones has the potential to enable early actions that save lives and reduce the economic impact of an event. For example, the forecast-based financing approach anticipates disasters and reduces their impacts by enabling communities to access humanitarian funding for early actions, based on in-depth forecast information and risk analysis. This approach has been implemented by 30 Red Cross and Red Crescent National Societies (as of 2020), and has inspired similar anticipatory approaches by other humanitarian actors, including UN agencies and NGOs. The Anticipation Hub has a global map of anticipatory action initiatives.
Ensemble-based predictions are one way to enhance medium-range forecasts and increase lead times, both of which help to enable anticipatory action. Instead of using a single weather model, they draw on ensemble (group) weather models, giving a range of possible future weather outcomes. They also provide the opportunity to measure the weather for timeframes typically between a week and a month. Such probabilistic weather forecast systems are becoming increasingly common.
This type of modelling is commonly used to track tropical cyclone outcomes. As an example, 15% of ensemble members may track a tropical cyclone formed in the south-west Pacific towards northern New Zealand, while 85% of ensemble members show it will stay in the tropics. This split in outcomes helps decision-makers, as percentiles can be used to assess risk – something that would not be possible if a single weather model was used.
Humanitarian and development actors are already using ensemble forecasts to generate flood thresholds and trigger early actions (see Case Study 1). Indeed, many national forecasting services are moving towards multi-hazard impact-based forecasting and warning services that translate hazards into sector- and location-specific impacts and responses (WMO, 2018).
To be effective, impact-based forecasting requires collaboration with other decision-makers, including disaster and emergency managers, humanitarian agencies, stakeholders and communities (The Future of Forecasts, 2021). Every extra day of warning provides emergency managers and humanitarian agencies with more time to prepare communities for risks and reduce the impact of disasters. This results in people taking early actions, such as evacuating vulnerable communities, individuals and their livestock, the pre-deployment of flood barriers, and closing roads and bridges (WMO, 2020). For humanitarian agencies, these early actions can include the distribution of cash grants, evacuation of people and livestock, and distribution of hygiene kits. You can find a database of early actions on the Anticipation Hub here along with the lead-time required to implemnt them effectively.
Case study 1: Flood forecasting in Bangladesh
In Bangladesh, 1-10 day ensemble probabilistic flood-forecasting and early-warning systems have proven useful and helped communities to reduce disaster risks and increase their resilience. The 10-day forecasts provided farmers with a range of response options, such as changing cropping patterns or planting times.
For example, acommunity in Gibandha explored the shared economic benefits of an early warning system. The average economic benefit per natural hazard event per household in pilot areas was US$270 in household assets saved, US$485 for those with livestock, US$180 for those working in agriculture, and US$120 for those involved in fishing (Fakhruddin et al., 2015)
The Bangladesh Red Cresecent Society (BDRCS) activate their Flood Early Action Protocol if the 10-day probabilistic forecast indicates a greater than 50% probability of a 10-year flood lasting more than three days, whereas a full activation is triggered with a 5-day deterministic forecast. Ahead of the 2020 monsoon, the 10-day GLOFAS forecast triggered the UN OCHA CERF funding (US$5.2 million), allowing coordinated readiness activities by multiple humanitarian actors including OCHA, BDRCS, WFP, FAO and UNFPA.
Limits of forecasting
The usefulness of forecasts depends on both their accuracy and their relationship to users’ needs. As a result, this can be increased by systematic efforts to bring scientific outputs and users’ needs together. There is still an imbalance between users’ expectation and what forecasters can provide to a community. When these expectations are not met, users are less willing to use the information and act upon it.
Finding a balance between user expectations and the information that forecasters can provide is key. Managing these expectations, and fully explaining the technical limitations of forecasts, might help users to understand why they may not be able to get all the information they want, and why that information may be uncertain.
Medium-range forecasts are still inherently uncertain, due to the chaos in the atmospheric system. Forecasting skill also varies geographically, temporally and by climate parameter. Providing advance forecasts of high-impact weather events is particularly challenging, for several reasons:
- there can be a lack of clarity to simulate hazard scenarios dynamically in a forecast model
- model biases in representing storms can become increasingly pronounced in extreme scenarios
- there can be difficulty in defining and verifying a high-impact event
- some weather patterns are not predictable, but may influence the weather on a sub-seasonal scale.
New technologies, such as machine learning and artificial intelligence, combined with international collaboration around research, may help improve medium-range forecasting. For example, the Madden-Julian Oscillation, an eastward-moving tropical disturbance of clouds, rainfall, winds and pressure crossing the planet every 30 to 60 days on average, has been researched to improve hazard predictions.
Another weather pattern which needs to be understood to improve sub-seasonal forecasting is a ‘sudden stratospheric warming’ above the Arctic or Antarctic. This happens every couple of years in the northern hemisphere, and somewhat less often in the southern hemisphere. When it occurs, it can affect the weather worldwide. A large southern stratosphere warming set up the weather events leading to the tinder-dry conditions in Australia that caused major bushfires in late 2019 and early 2020.
Case study 2: Climate forecasts in the Phlippines and Indonesia
The application of climate forecasts in the Philippines and Indonesia demonstrated the practical use of ENSO (El Niño/Southern Oscillation) forecasts in designing drought risk-management strategies for climate-sensitive activities, particularly agriculture and water resources. The programme was applied to enable societies to deal with climate variability, offered an opportunity to educate public and policy-makers about long-term climate change, drought risk and mitigation options. A customized drought early-warning system offered a platform for advocating that the best way sto deal with drought risk given current climate variability.
One key achievement in both countries has been the establishment of institutional mechanisms that connect hydro-meteorological communities, risk-management institutions and societies. For example, a pool of meteorologists has been formed and trained to provide tailored climate information for drought risk management.
Another achievement is the development of institutional and community-level dissemination channels in demonstration sites, which have been built primarily through climate field schools, climate forums and community-level workshops. Drought forecast applications for disaster mitigation were internalized and owned by local governments involved in the programme, which can save lives and produce tangible economic benefits (Fakhruddin, 2017).
We are directly connected to the BMKG to access medium-range weather and flood forecasts and send them to the CBAT (Community-Based Disaster Preparedness) Team, the Indonesian Red Cross network at the Provincial and District City levels, communities and volunteer networks to carry out emergency alerts for initial planning and early response efforts, livelihood protection, and other planning. Based on the forecast, we have also reactivated contingency plans into disaster operations preparedness plans and mobilized community efforts so that the impact of disasters can be minimized.
The way forward
Weather never stops: it is ever evolving, and never quite the same twice. It affects almost every aspect of society in some capacity, from daily life to the economy. However, it is not ordinary weather conditions that produce the greatest societal impacts, but intense or rare weather events. It is crucial for communities, regions, countries, businesses and other sectors to have accurate forecasts of the risks, and with sufficient lead times to prepare.
Medium-range forecasts can provide humanitarian and development actors with additional lead times to implement a more diverse set of early actions and allow them to reach more people. As climate change causes global temperatures to rise – and more extreme weather events to occur – researchers will need to continue improving forecasting and modelling, not only for accuracy, but also to provide as great a lead time as possible.
This blog was written by Bapon Fakhruddin, Technical Director, Tonkin + Taylor, New Zealand, Chair, Risk Interpretation and Action (RIA)-IRDR, International Science Council. It is a summary of the article ‘Creating resilient communities with medium-range hazard warning systems’, which will appear in Progress in Disaster Science Volume 12, December 2021.
Key references
Fakhruddin, B.S.H.M. and Eslamian, S. (2017) ‘Analysis of drought factors affecting the economy’, in Handbook of drought and water scarcity, S. Eslamian and F. Eslamian (eds) Boca Raton: CRC Press, Routledge Handbooks Online. Accessed 27 October 2021.
Fakhruddin B.S.H.M., Kawasaki, A. and Babel, M.S. (2015) ‘Community responses to flood early warning system: Case study in Kaijuri Union, Bangladesh’, International Journal of Disaster Risk Reduction, 14(4): 323-331, ISSN 2212-4209, here.
Fakhruddin, B.S.H.M. and Schick, L. (2019) ‘Benefits of economic assessment of cyclone early warning systems - A case study on Cyclone Evan in Samoa’, Progress in Disaster Science 2(100034). Retrieved from here.
IFRC (2021) Forecast-based Financing. Retrieved from here.
WMO (2018) Multi-hazard Early Warning Systems: A Checklist. World Meteorological Organisation Retrieved from here.