Providing Valuable Impact-Based Information to Disaster Response Teams

As the world deals with growing problems caused by climate change, the number and intensity of natural hazards have increased a lot. The planet is getting warmer, weather patterns are changing, and we're seeing more and more very serious events.

In recent years, the rise of various innovative forecasting approaches, including the utilization of big data, has significantly bolstered disaster management for a wide range of natural hazards. These approaches leverage cutting-edge technologies and interdisciplinary collaborations to enhance the accuracy and timeliness of hazard predictions, thereby minimizing the potential impact on vulnerable communities. 

However, obviously, it cannot prevent such an event. Over the past few decades, natural disasters have shown a worrying trend of causing more damage, bringing devastating consequences for both human lives and the environment (World Meteorological Organization, 2014). Several key factors contribute to this troubling increase. First and foremost, climate change plays a pivotal role in intensifying the frequency and severity of extreme weather events. Rising global temperatures lead to more frequent and intense heatwaves, droughts, hurricanes, and storms (IPCC, 2022). As a result, there has been about a threefold increase in the occurrence of relevant natural hazards worldwide in the last 35 years (Hoppe, 2016).

In the case emergency response teams need to take action the timely availability of accurate, accessible, and comprehensible information increases effectivity of the action and in the end it can save lives. The last decade we have witnessed a transition from flood early warning (flooded area) to more impact based early warning (roads blocked/ buildings flooded). Now new opportunities arise with new technologies such as AI, Big Data, and 3D visualization. 

This article delineates certain ongoing advancements in the arena of disaster response, with a specific focus on flood forecasting, even though this is a subject that spans across various hazards.

Improve calculation time of expert models 

Traditional expert models (physics based) are high demanding models. They need lots of (meta)data, expert knowhow (calibration validation) and calculation time.  The last 20 years major improvements are introduced for example ‘subgrid’ a method to compute on a coarse grid considering high-resolution bathymetry and roughness (Volp, 2013). But nowadays the world is pixelating in a rapid speed. The amount of global available digital data is exploding. Stretching the limitations of current physics-based models. Luckily science is again pushing the boundaries.

One of the most promising developments to minimize calculation time is leveraging the power of big data and AI. But a common fact is that it is difficult to train AI on Extreme natural events. Especially because every year records are broken, and events occur even more extreme than ever experienced. How can you train AI with data on events that we never experienced before. Here is where physics-based models come in. These models when build and calibrated can product lots of output. Feeding them with a whole variety of possible (over dimensioned) weather conditions. They can create big data training sets to train AI models and prepare them even for the worst-case scenario. And with that introducing a new generation of very fast AI expert models. 

Another development is optimizing physics-based models. For example, running them on GPU and or apply parallelisation on either GPU or CPU. A more subtle development is an optimisation within sub grid introducing grid cloning (Casullo, 2019). In short, an approach to optimize the computational efficiency of models based on sub grid. This new approach is applied within TEMA on the sub grid based hydrodynamic model 3Di (http://3diwatermanagement.com). It will be validated in use cases in Germany, Greece, and Italy.

Harvest real-time ground truth and automate ‘evaluation and optimization’.

Traditionally gathering ground truth is time consuming. You have multiple sources (satellite, social media, monitoring data), coming in on different timeframes. And most of the time there is a specialist and/or expert that via data management combines all sources into a structured dataset. Which than is compared with the results of the expert model (either AI or physics based) to determine the quality of the impact-based information created.

Within TEMA, extensive research is being conducted on deploying automotive drones to collect ground truth, thereby enhancing situational awareness. These drones, equipped with advanced on-board AI, have the capability to furnish accurate ground truth for evaluation purposes. The process involves employing data fusion to extract valuable insights concerning the quality of the impact-based information and the expert model. When this information is acquired in near real-time, it further aids in automated evaluation and optimization. As a result, there is a marked enhancement in the subsequent forecast run and newly produced the impact-based information.

Common Operational Picture, supporting a more effective dissemination.

In disaster response management a common operational picture is essential. Multi-disciplinary teams need to collaborate. Instant decision making with potential high impact requires a common understanding of what is happening and equally important what is going to happen. To achieve this common understanding 3D platforms are a promising environment to represent in time the natural hazard. It is common knowhow that if you can digitally represent a flooding close to how it is experienced in real life one is able to understand the impact of such an event much better. This makes the dissemination more effective.

References

​Casullo V. Computational grid, sub grid, and pixels. Int J Number Meth Fluids. 2019; 90:140–155. https://doi.org/10.1002/fld.4715

Hoppe, P. (2016). Trends in weather related disasters – Consequences for insurers and society. Weather and Climate Extremes, 11, 70-79. Doi: HTTPs://doi.org/10.1016/j.wace.2015.10.002

IPCC. (2022). Summary for Policymakers. In H. O. Partner, D. C. Roberts, M. Tignor, E. S. Polychasia, K. Moneyback, A. Alegría, . . . B. Rama (Red.), Climate Change 2022: Impacts, Adaptation, and Vulnerability. Contribution of Working Group II to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change (p. In Press). Cambridge, UK: Cambridge University Press.

Volp, N. D., B. C. van Proin, and G. S. Stelling (2013), A finite volume approach for shallow water flow accounting for high-resolution bathymetry and roughness data, Water Retour. Res., 49, 4126–4135, doi:10.1002/wrcr.20324