The Red Cross Red Cresent is among the organizations with the longest and most extensive experience with forecast-based action. We present the findings of recently-published research based on interviews with 139 stakeholders involved in Red Cross Red Crescent (RCRC) AA programs in 18 countries. We find that the organizaitonal benefits of forecast-based ation include capacity building, more proactive operations, and expedited humanitarian response. Forecast-based action can also help to overcome common challenges in climate services by providing a framework and decision-making and resources for early action. Despite these benefits, AA practitioners struggle with challenges common to climate services, development, and humanitarian aid, including local project ownership, capacity and infrastructure, integration with existing systems, data availability, forecast uncertainty, and monitoring and evaluation. We conclude that forecast-based action systems can only be sustainble if they address these perennial challenges and focus on building capacity and ownership. Furthermore, donors can play a major role in facilitating these shifts by providing funding designed to support long-term multi-stakeholder processes.

How to cite: Tozier, A., Castro Jr., E., Rahaman, H., Heinrich, D., Clatworthy, Y., and Mundorega, L.: Lessons from Red Cross Red Crescent Anticipatory Action, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-9345, https://doi.org/10.5194/egusphere-egu23-9345, 2023.

In March 2022, the United Nations set as a five year target that every place on Earth should be served by Early Warning Systems (EWS) for natural hazards. Such an EWS provides emergency alerts when a natural disaster is imminent and can support local or international (aid) organizations to take effective action early on. Places most vulnerable to natural disasters are often those where little local data and capacity is available to locally develop and operate such a system. As local EWS are not yet available everywhere, robust and reliable global approaches and collaboration initiatives are needed as initial and possible fallback solution.

We propose an innovative flood hazard mapping method based on globally available data that can spatially indicate oncoming floods and thereby inform on preparatory actions to take, such as required emergency stocks, needed shelter capacity, clearing of evacuation routes, and strategic protection of vulnerable people and assets. It instantaneously calculates forecasted flood extents based on global precipitation forecasts and the terrain’s natural drainage network. Its functioning is demonstrated for a selection of historical flood events and shows to good agreement with satellite-observed inundated areas, even where flood extents have gone beyond catchment boundaries. The method can easily be scaled-up to other areas around the world and can be expanded to issue automated warnings and provide impact estimates.

How to cite: huthoff, F., van den berg, K., and wegman, C.: Rapid global hazard forecasting to support early action in data poor regions, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-3876, https://doi.org/10.5194/egusphere-egu23-3876, 2023.

This study investigates the importance of assessing ensemble flood forecasts with action-relevant scores to support flood preparedness actions by analyzing the discrepancies between traditional general scores that focus on the overall accuracy with other more specific flood event-based scores. Popular general scores such as the Kling-Gupta Efficiency (KGE) or the Continuous Ranked Probability Score (CRPS) are widely used in hydrological modeling and forecasting, but they aggregate different aspects of model quality into a single overall score. On the other hand, flood event-based scores, such as Flood Timing Error (FTE), False Alarm Ratios (FAR) and Probability of Detection (POD), provide more specific verification measures of forecast quality that should be more informative to decision-makers. Both classes of overall accuracy and event-based scores include either deterministic or probabilistic scores, focusing on either the ensemble mean (or quantiles) or on probabilities. 

Results are presented for ten catchments in Uganda with different morphological and hydrological characteristics. An evaluation of extended-range re-forecasts from the Copernicus-Emergency Management Service Global Flood Awareness System (GloFAS) has been carried out against observed streamflow data, contrasting overall performance scores, including the KGE and the CRPS, and event-based scores, including the FTE, FAR and POD for forecasts at different lead times (< 45 days). The relative performance of two different versions of GloFAS (2.1 and 3.1) is assessed by this multi-criteria verification setting. Results show that the relative ranking of forecast performance across model versions and catchments may vary based on the scores considered, suggesting that a multi-criteria and event-based evaluation is needed to inform flood preparedness actions.

How to cite: Mulangwa, D., Ficchi, A., Nyenje, P., Sempewo, J., Speight, L., Cloke, H., Harrigan, S., Zaake, B., and Stephens, L.: Assessing ensemble flood forecasts with action-relevant scores to support flood preparedness actions: An application to the Global Flood Awareness System in Uganda., EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-9862, https://doi.org/10.5194/egusphere-egu23-9862, 2023.

Floods pose an increasing challenge for societies in West Africa; causing loss of lives, damaged infrastructure, and food insecurity. Improving flood management is hence paramount for the region, which several initiatives aim to contribute to. Hydrological forecasting systems can help, but only if they lead to appropriate action.

This presentation focusses on how a flood forecasting system has been used to save lives and property in West Africa within the FANFAR project (www.fanfar.eu). The system was co-designed and co-developed together with hydrological services, emergency management agencies, river basin organisations, and regional expert centres in 17 countries. The pilot system was launched early in the project, producing new forecasts every day. This enabled operational staff at national and regional agencies to utilize the system during the current rainy season, for every season since 2019.

During 2020, Nigeria experienced severe flooding. The Nigeria Hydrological Services Agency (NIHSA) hence decided to utilize FANFAR to warn the population of forthcoming flood risks, which resulted in 2 500 lives saved on one occasion, and minimisation of property damage on another. In the presentation we describe these events, and how NIHSA acted together with other institutions to entice action.

FANFAR was also used in Ivory Coast during the 2022 rainy season. Operational staff at SODEXAM – the meteorological services of Ivory Coast – utilized the system to inform two flood-prone communities of forthcoming flood risks. This resulted in on-the-fly construction of a drainage ditch, which reduced impacts on the nearby community. In the presentation we describe the event and also the approach SODEXAM took to build trust and communicate with the communities.

We also briefly describe the FANFAR system that employs a daily forecasting chain including meteorological reanalysis and forecasting based on HydroGFD, data assimilation of gauge observations, hydrological initialisation and forecasting with the HYPE model, flood severity assessment, and distribution through e.g. web visualisation. 

How to cite: Andersson, J. C. M., Ibrahim, A. T., Soumahoro, A. L., Fofana, V., Ali, A., and Arheimer, B.: How a flood forecasting system saved lives and property in West Africa, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-15732, https://doi.org/10.5194/egusphere-egu23-15732, 2023.

Due to its geographical location, the Philippines is prone to tropical cyclones (TC) which produce strong winds, accompanied by heavy rains and flooding of large areas, resulting in heavy casualties to human life and destruction to livelihoods and properties. To reduce the humanitarian impact of TC, the Philippine Red Cross with the German Red Cross and 510, an initiative of The Netherlands Red Cross, designed and implemented a machine learning impact-based forecasting model based on XGBoost, which is used operationally to release funding and to trigger early action. The model predicts the percentage of houses that will be completely damaged due to a TC using predictive features for the hazard (wind speed, rainfall, storm surge and landslides), exposure (such as ruggedness and population density) and vulnerability (such as housing material and poverty) . However, this model is not easily transferable to other countries, due to its use of country specific data from the Philippines.

Here, we develop upon this line of research around the XGBoost model, in three ways. First, we evaluate multiple ML algorithms for classification and regression of impact data of tropical storms. Secondly, we perform a sensitivity analysis on the predictive features, replacing where possible those features for which only Philippines-specific data sources can be used with features for which data from global open data sources are available. Thirdly, the XGBoost model provides predictions at the aggregated geographical level of a municipality. Our research centres on transforming it to a grid based model with a resolution of 0.1 x 0.1 latitude-longitude degrees. For all experiments, due to the scarcity and skewness of the training data (algorithms are trained on only 40 historical typhoon events), specific attention is paid to data stratification, sampling and validation techniques. 

We find that XGBoost slightly outperforms random forest and that regression is more suitable to detect outliers than classification. Furthermore, we show that we can limit the predictive features from the original model to a subset of 20 features. The transformation to a grid-based model was possible by de-aggregating the impact data using OpenStreetMap housing data obtained from Humanitarian Data Exchange. Preliminary results show that the ML model performance worsens when going from municipality to grid-based level. This is likely caused by a larger error variance between the individual grid cells of a municipality which get averaged when aggregated. To conclude, relying on globally available data sources and working at grid level holds potential to render a machine learning based impact model generalisable and transferable to locations outside of the Philippines impacted by TCs. Future research will focus on validation with data for other countries. Ultimately, a transferable model will facilitate the scaling up of anticipatory action for tropical cyclones. 

How to cite: Kooshki, M., van den Homberg, M., Kalimeri, K., Kaltenbrunner, A., Mejova, Y., Milano, L., Ndirangu, P., Paolotti, D., Teklesadik, A., and Turner, M.: Towards a global machine learning based impact model for tropical cyclones, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-14435, https://doi.org/10.5194/egusphere-egu23-14435, 2023.

The Philippines is one of the countries most at risk to natural disasters. Amongst these disasters, typhoons and its associated landslides, storm surges and floods have caused the largest impact. Due to increased typhoon intensity, the country’s high population density in coastal areas and rising mean sea levels, the coastal flood risk in the Philippines is only expected to increase. The 510 initiative of the Netherlands Red Cross uses an Impact Based Forecasting (IBF) model based on machine learning to anticipate the impact of an incoming typhoon to set early action into motion. The IBF model underperformed in regions that are susceptible to storm surges. Most notably, it showed a poor performance for Super-Typhoon Haiyan (2013), which caused storm surges to reach up to over five meters high. The goal of this research is to evaluate how the IBF model can be improved by applying a fast hydrodynamic modelling approach that can forecast storm surges and coastal flooding associated with typhoons. First, the accuracy of the Global Tide and Surge Model (GTSM) in simulating Haiyan’s coastal water levels was examined. GTSM was forced with two different meteorological datasets: a gridded climate reanalysis dataset, ERA5, and observed track data combined with Holland’s parametric windfield model. Second, GTSM’s water levels were used as input for a hydrodynamic inundation model to simulate the flood depth and extent in San Pedro Bay, which was subjected to a widespread coastal flood during Haiyan. This was explored both with and without the inclusion of wave setup. Our results show that Haiyan’s flood cannot adequately be indicated using the ERA5 reanalysis dataset as meteorological forcing, as it underestimated Haiyan’s extreme wind speeds with ~60 m/s. By applying the Holland parametric wind field model, more accurate flood predictions and storm surge simulations can be made. Additionally, GTSM’s temporal resolution influences the models performance substantially. By increasing the 1 hour resolution to a 30 minute resolution the prediction of the overall flood extent improved by 16%. In future research we recommend examining the applicability of the Global Tide and Surge Model when using a higher spatial resolution to help better represent local processes. Additionally, exploring the accuracy for other typhoons that struck the Philippines and the applicability in operational setting using forecasted track data can contribute to further improving forecast-based early action systems in anticipating coastal flood occurrences.

How to cite: Barendregt, A., Benito Lazaro, I., Muis, S., van den Homberg, M., and Teklesadik, A.: Using hydrodynamic flood modelling to support impact-based forecasting: a case study for Super-Typhoon Haiyan in the Philippines, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-1894, https://doi.org/10.5194/egusphere-egu23-1894, 2023.

The Horn of Africa faces an ongoing multi-year drought due to five consecutive failed rainy seasons, a novel climatic event with unpreceded impacts. Over 50 million individuals in the region are expected to be highly food insecure by the end of 2022 and early 2023. The severity of these drought impacts call for the urgent upscaling and optimisation of early warning systems that trigger early actions. However, drought research focuses mainly on meteorological and hydrological forecasting, while early action is seldom addressed specifically. This leads to a gap between early warning and early action, which heavily reduces the effectiveness of these systems.

To address this gap, this study investigates the effectiveness of early action for droughts by using seasonal ensemble forecasts from the European Centre for Medium-Range Weather Forecasts (ECMWF) SEAS5 system, predicting rainfall for the March-April-May (MAM) and October-November-December (OND) rainy seasons. We show that these seasonal rainfall forecasts reflect major on-the-ground impacts, which we identify from 9 years of monthly drought surveillance data from 21 counties in Kenya. Subsequently, we show that the SEAS5 drought forecasts with short lead times have substantial potential economic value (PEV) when used to trigger action before the OND season across the region (PEV _max = 0.43). Increasing lead time to one or two months ahead of the season decreases PEV, but the benefits of early action still persist (PEV _max = 0.2). Highest value for early action is found for the OND season in Kenya and Somalia, with excellent PEV _max  of around 0.8 in Somalia. This indicates exceptional potential for early action to reduce impacts in this drought-prone country. The potential for early action is relatively low for the MAM season across the region, due to the season’s lower predictability. To illustrate the practical value of this research, we showcase how our methodology can be used by a pastoralist in the Kenyan drylands to effectively trigger livestock destocking ahead of a drought using SEAS5 forecasts.

These results are making headway to the development of concrete early action triggers for drought-prone regions, which are urgently needed to translate early warning to early action for droughts. It also emphasizes the need to expand historical datasets of drought impacts and early actions to support future research and policy development. Therefore, this work supports early decision-making and the development of early action protocols across the different countries in the Horn of Africa.

How to cite: Busker, T., de Moel, H., van den Hurk, B., and C.J.H. Aerts, J.: Impact-based seasonal rainfall forecasting to trigger early action for droughts, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-803, https://doi.org/10.5194/egusphere-egu23-803, 2023.