The Role of AI in Disaster Management
A natural disaster is a serious disruption occurring over a short or long period of time that causes widespread human, material, economic or environmental losses. Disaster is categorized as the event(s) that exceeds the affected community's ability or society to cope with the use of its own resources. Therefore, disaster management must be implemented to prevent and reduce a disaster's impact on the community.
Generally, disaster management is involved in four phases: mitigation, preparedness, response, and recovery. Mitigation means the protection of future risk so that the future risk can be mitigated. Preparedness means to get ready and prepare necessary equipment for the response in case of disaster happen. The response phase refers to an immediate response or action to dangerous situations that saves lives. The last phase is recovery, which refers to the process of repair, reconstruction, and restoration. The following figure shows the disaster management cycle.
Now, the question is how can AI fit into each phase to support disaster management? With the advancement of computational intelligence, AI could fit into all four phases.
Let’s take a look at how AI can support each stage of disaster management.
Phase1: Mitigation is about prevention to reduce the severe impact on humans, lives, and properties. One of the significant functions of AI is the predictive analytic in which AI could support forecasting hazard and risk assessment analysis. The output of forecasting and risk assessment can support decision-makers or stakeholders to make a proper decision before a disaster happens. Additionally, evaluating possible impacts on the potential area of disaster would support authorities in better preparing and managing disaster strikes.
Plase2: Preparedness comes after mitigation in which authorities and people in the community are ready for the response during a disaster. In this phase, AI could play roles such as real-time monitoring and early warning systems, real-time prediction, and hazard detection. For example, at certain places where humans cannot access the observation, AI is capable of assisting.
Phase3: Response is when the disaster strike. Immediate response and action are vital to assist and rescue affected people or communities. AI application for the support of search and rescue activities such as robots that can access the building to search for humans. Additionally, understanding people’s emotions and reactions during disasters would help public authorities and humanitarian assistance agencies to assist accordingly.
Phase4: Recovery is the final stage in which damages are fixed. However, AI can be used to support impact assessment and the evaluation of losses.
Now, what are AI techniques that can be applied for disaster management?
In response to the above question, I would like to explain AI in a simple way. Artificial Intelligence (AI) is a fascinating area of research that involves using machines to think as humans do, to make a machine think like humans it needs to build or apply algorithms that involve automatically classifying, analyzing, and extracting information from data. Additionally, for AI to do the tasks mentioned above, AI needs data to learn from it and constantly improve the task given.
How can AI perform assigned tasks? In order to do the tasks, AI must have an engine to drive, that engine is call machine learning(ML). ML is used to identify the pattern of the dataset(s) and predict the output. Classical ML such as support vector machine(SVM), K-nearest neighbours (KNN), and random forest were used for disaster forecasting and risk assessment, impact estimation, and vulnerability assessment. These ML techniques are used to support the mitigation phase.
Other classical ML algorithms that can be used to support the preparedness phase are logistic regression, decision tree, KNN, and neural network. These algorithms are used in implementing early warning systems and real-time disaster monitoring and detection.
ML algorithms that applied in the response phase are the neural network, K-hierarchical clustering, K-mean clustering, and more. These algorithms are used in implementing event mapping and damages assessment.
The last phase is recovery, there are few ML algorithms that can be applied to support the tasks in this phase. They are linear regression, non-linear regression, SVM, random forest, and more. These algorithms are used for impact assessment and evaluation losses.
In line with the big data era, together with the advancement of the internet and computational capacity, deep learning(DL) has emerged. DL is basically a subset of ML, based on the architecture of artificial neural network(ANN) which contains input layer, hidden layer, and output layer. However, DL has multiple hidden layers, and it is suitable for big data analysis which combines different types of data format such as text, image, video, and sound. The application of DL in disaster management includes hazard forecasting and vulnerability assessment in the mitigation phase. Convolutional neural network(CNN) is one method under DL which a suitable solution for event mapping, damage assessment, disaster rescue, and relief. these problems are under the response phase. In the recovery phase, CNN is used in the impact assessment problem. Recurrent neural networks (RNN) is another technique under DL that can be used to solve the problem of track recovery.
In summary, the use of AI to solve the problem of disaster management has been carried out over the decades. However, some applications have been shown to be conceptual and still in research and development. End products of AI to support disaster management are not widely available because disasters are uncertain, difficult to predict, and data scarcity. These challenges continue to exist. Therefore, the key to successfully applying AI to support disaster management is data availability and its accessibility.
I hope this article would benefit those who are disaster management managers or decision-makers interested in applying AI for disaster management.
You have read to the end. Thank you.
References:
[1] Sun, W., Bocchini, P., & Davison, B. D. (2020). Applications of artificial intelligence for disaster management. Natural Hazards, 1–59.
[2] Simões-Marques, M., & Figueira, J. R. (2018, July). How Can AI Help Reduce the Burden of Disaster Management Decision-Making?. In International Conference on Applied Human Factors and Ergonomics (pp. 122–133). Springer, Cham.
