Constructing and validating a transferable epidemic risk index in data scarce environments using open data: A case study for dengue in the Philippines.

Epidemics are among the most costly and destructive natural hazards globally. To reduce the impacts of infectious disease outbreaks, the development of a risk index for infectious diseases can be effective, by shifting infectious disease control from emergency response to early detection and prevention. In this study, we introduce a methodology to construct and validate an epidemic risk index using only open data, with a specific focus on scalability. The external validation of our risk index makes use of distance sampling to correct for underreporting of infections, which is often a major source of biases, based on geographical accessibility to health facilities. We apply this methodology to assess the risk of dengue in the Philippines. The results show that the computed dengue risk correlates well with standard epidemiological metrics, i.e. dengue incidence (p = 0.002). Here, dengue risk constitutes of the two dimensions susceptibility and exposure. Susceptibility was particularly associated with dengue incidence (p = 0.048) and dengue case fatality rate (CFR) (p = 0.029). Exposure had lower correlations to dengue incidence (p = 0.193) and CFR (p = 0.162). Highest risk indices were seen in the south of the country, mainly among regions with relatively high susceptibility to dengue outbreaks. Our findings reflect that the modelled epidemic risk index is a strong indication of sub-national dengue disease patterns and has therefore proven suitability for disease risk assessments in the absence of timely epidemiological data. The presented methodology enables the construction of a practical, evidence-based tool to support public health and humanitarian decision-making processes with simple, understandable metrics. The index overcomes the main limitations of existing indices in terms of construction and actionability.

Forecasting transitions in the state of food security with machine learning using transferable features.

Food insecurity is a growing concern due to man-made conflicts, climate change, and economic downturns. Forecasting the state of food insecurity is essential to be able to trigger early actions, for example, by humanitarian actors. To measure the actual state of food insecurity, expert and consensus-based approaches and surveys are currently used. Both require substantial manpower, time, and budget. This paper introduces an extreme gradient-boosting machine learning model to forecast monthly transitions in the state of food security in Ethiopia, at a spatial granularity of livelihood zones, and for lead times of one to 12 months, using open-source data. The transition in the state of food security, hereafter referred to as predictand, is represented by the Integrated Food Security Phase Classification Data. From 19 categories of datasets, 130 variables were derived and used as predictors of the transition in the state of food security. The predictors represent changes in climate and land, market, conflict, infrastructure, demographics and livelihood zone characteristics. The most relevant predictors are found to be food security history and surface soil moisture. Overall, the model performs best for forecasting Deteriorations and Improvements in the state of food security compared to the baselines. The proposed method performs (F1 macro score) at least twice as well as the best baseline (a dummy classifier) for a Deterioration. The model performs better when forecasting long-term (7 months; F1 macro average = 0.61) compared to short-term (3 months; F1 macro average = 0.51). Combining machine learning, Integrated Phase Classification (IPC) ratings from monitoring systems, and open data can add value to existing consensus-based forecasting approaches as this combination provides longer lead times and more regular updates. Our approach can also be transferred to other countries as most of the data on the predictors are openly available from global data repositories.

Improved Transferability of Data-Driven Damage Models Through Sample Selection Bias Correction.

Damage models for natural hazards are used for decision making on reducing and transferring risk. The damage estimates from these models depend on many variables and their complex sometimes nonlinear relationships with the damage. In recent years, data-driven modeling techniques have been used to capture those relationships. The available data to build such models are often limited. Therefore, in practice it is usually necessary to transfer models to a different context. In this article, we show that this implies the samples used to build the model are often not fully representative for the situation where they need to be applied on, which leads to a "sample selection bias." In this article, we enhance data-driven damage models by applying methods, not previously applied to damage modeling, to correct for this bias before the machine learning (ML) models are trained. We demonstrate this with case studies on flooding in Europe, and typhoon wind damage in the Philippines. Two sample selection bias correction methods from the ML literature are applied and one of these methods is also adjusted to our problem. These three methods are combined with stochastic generation of synthetic damage data. We demonstrate that for both case studies, the sample selection bias correction techniques reduce model errors, especially for the mean bias error this reduction can be larger than 30%. The novel combination with stochastic data generation seems to enhance these techniques. This shows that sample selection bias correction methods are beneficial for damage model transfer.

A framework for comparing permanent and forecast-based flood risk-reduction strategies.

Flood risk can be reduced at various stages of the disaster management cycle. Traditionally, permanent infrastructure is used for flood prevention, while residual risk is managed with emergency measures that are triggered by forecasts. Advances in flood forecasting hold promise for a more prominent role to forecast-based measures. In this study, we present a methodology that compares permanent with forecast-based flood-prevention measures. On the basis of this methodology, we demonstrate how operational decision-makers can select between acting against frequent low-impact, and rare high-impact events. Through a hypothetical example, we describe a number of decision scenarios using flood risk indicators for Chikwawa, Malawi, and modelled and forecasted discharge data from 1997 to 2018. The results indicate that the choice between permanent and temporary measures is affected by the cost of measures, climatological flood risk, and forecast ability to produce accurate flood warnings. Temporary measures are likely to be more cost-effective than permanent measures when the probability of flooding is low. Furthermore, a combination of the two types of measures can be the most cost-effective solution, particularly when the forecast is more skillful in capturing low-frequency events. Finally, we show that action against frequent low-impact events could more cost-effective than action against rare high-impact ones. We conclude that forecast-based measures could be used as an alternative to some of the permanent measures rather than being used only to cover the residual risk, and thus, should be taken into consideration when identifying the optimal flood risk strategy.