Author: Saut Sagala, Dekka Putra, Novia Anggraini (Intern)
This year, Indonesia commemorates the 20th year of the Indian Ocean / Aceh Tsunami since the devastating tsunami struck the western part of the country. In 2004, the catastrophic disaster, preceded by a Mw9.2 earthquake, swept through Aceh Province, claiming hundreds of thousands of lives (Lay et al., 2005). In Gampong Mon Ikeun, a village in Aceh Besar, the head of the village recounted that out of 5,000 residents, only around 16% survived the massive wave (Syamsidik et al., 2020). This tragic statistic underscores the importance of robust disaster mitigation systems to safeguard lives in the future.
Time and accuracy are critical factors in disaster mitigation efforts, particularly for tsunamis. The delay in issuing a warning for the 2004 tsunami and the underestimation of its height contributed to the significant loss of life (Satake, 2015). Faster and more accurate tsunami forecasting can save countless lives. Researchers have begun leveraging Artificial Intelligence (AI) in forecasting methods to enhance forecasting quality, which promises to reduce forecasting time and improve accuracy. Studies by Mulia et al. (2020, 2022) apply machine learning to forecast real-time tsunami inundation levels with higher resolution outputs, offering critical insights for disaster preparedness and response.
Advancements in Tsunami Forecasting
Research by Mulia et al. (2020) aims to produce high-quality maps detailing tsunami inundation distance and depth flow, which is critical for identifying potential impact zones. Using Deep Neural Networks (DNNs), the researchers transformed low-resolution tsunami simulations into high-resolution inundation maps, drastically reducing computational time by approximately 90% compared to traditional methods.
Figure 1. Tsunami Model in Otsuchi and Rikuzentakata
Source: Mulia et al., 2020
These maps, with an accuracy range of 85% to 99.7%, are essential for planning and developing evacuation routes and designing resilient infrastructure in high-risk areas.
Figure 2. Flowchart of the proposed method using DNNs to produce a high-resolution inundation map
Source: Mulia et al., 2020
Artificial Intelligence (AI) is enhancing tsunami modeling in Indonesia by enabling rapid predictions of tsunami characteristics, such as height and arrival time, through machine learning algorithms that analyze seismic and oceanographic data (Wibowo et al., 2023). AI-driven early warning systems improve the speed and accuracy of alerts, allowing for timely evacuations and disaster response. Additionally, AI models assist in predicting tidal patterns and identifying potential tsunamis in shallow waters, which is crucial for Indonesia's coastal areas (Dharmawan et al., 2024).
In 2022, Mulia et al. developed another approach that bypasses initial simulations, directly forecasting tsunami inundation levels from offshore data using Artificial Neural Networks (ANNs). This method processes data from observation stations, such as Japan's dense S-net network, and generates forecasts within 0.05 seconds—compared to 30 minutes required by physics-based models. Such advancements hold immense potential for integration into Indonesia’s Tsunami Early Warning System (InaTEWS), significantly enhancing speed and precision.
The Role of Infrastructure in Tsunami Mitigation
Infrastructure can play a number of roles in tsunami mitigation.
- First. Evacuation Infrastructure: High-resolution inundation maps can inform the design and placement of evacuation routes, shelters, and signage systems in tsunami-prone areas (UNDRR, 2015). AI-enhanced forecasts can help ensure that these facilities are located in safe zones and accessible within short evacuation times.
- Second. Coastal Defense Structures: AI and Machine Learning can help optimise the design of sea walls, breakwaters, and dikes, balancing cost and effectiveness. Combined with natural solutions such as mangrove restoration, these structures act as critical buffers against tsunami impacts (Bappenas & UNDP, 2005).
- Third. Critical Infrastructure Resilience: Hospitals, schools, and transportation systems in tsunami-prone regions must be designed or retrofitted to withstand tsunami forces. AI-powered simulations can forecast potential impacts, guiding the development of robust structures and recovery plans (Kelman & Spence, 2004).
Challenges and Opportunities
Despite its promise, AI-based tsunami forecasting systems face challenges in regions with sparse station distribution, such as Indonesia. Dense networks like Japan's S-net enable precise data collection, but many tsunami-prone areas lack sufficient observation points. Enhancing the spatial density of offshore sensors, improving bathymetry data, and integrating community knowledge are essential for reliable forecasting (Mulder et al., 2020).
Moreover, implementing these advanced systems requires collaboration across sectors, from government agencies to private companies and local communities. Investments in data infrastructure, capacity building, and technology transfer are critical to bridging existing gaps.
Building Resilience with AI and Deep Learning
The Resilience Development Initiative (RDI), through its Smart System Centre (SSC), is at the forefront of research, leveraging AI and machine learning to improve tsunami forecasting and disaster mitigation. At the AIWEST Conference, commemorating the 20th anniversary of the 2004 tsunami, the RDI team is showcasing innovative methods for accelerating tsunami forecasting and informing infrastructure planning.
By integrating AI-driven advancements into disaster risk reduction strategies, RDI aims to enhance the resilience of both natural and built environments. These efforts align with SSC's strategic roadmap, emphasising the need for adaptive infrastructure, early warning systems, and community engagement. Together, these initiatives will strengthen Indonesia's preparedness for future tsunamis, protecting lives and livelihoods in vulnerable coastal regions.
References
Bappenas & UNDP. (2005). Aceh and Nias: Two years after the tsunami – Building a better future.
Harig, S., et al. (2019). Tsunami Early Warning and Mitigation System in Indonesia: Current Status and Challenges. Natural Hazards and Earth System Sciences.
Kelman, I., & Spence, R. (2004). An overview of flood actions on buildings. Engineering Geology, 73(3-4), 297-309.
Lay, T., et al. (2005). The Great Sumatra-Andaman Earthquake of 26 December 2004. Science, 308(5725), 1127-1134.
Mulder, K., et al. (2020). Bathymetric data for improved tsunami modeling. Journal of Geophysical Research: Oceans.
Mulia, I. E., et al. (2020). AI-driven inundation mapping for tsunamis. Natural Hazards.
Mulia, I. E., et al. (2022). Offshore data to predict tsunami inundation levels using AI. Earth Science Informatics.
Satake, K. (2015). Advances in tsunami research since 2004. Nature Geoscience, 8, 829-836.
Syamsidik, et al. (2020). Post-tsunami recovery in Aceh. International Journal of Disaster Risk Reduction, 45, 101503.
UNDRR. (2015). Sendai Framework for Disaster Risk Reduction 2015–2030.
Wibowo., et al. (2023). Development of Real-Time Tsunami Early Warning System Dashboard Based on Tunami-F1 and Machine Learning in Sunda Arc, Indonesia. Ocean Engineering Technology and Innovation Conference: Ocean Observation, Technology and Innovation in Support of Ocean Decade of Science (OETIC), IEEE.
Dharmawan., et al. (2024). Tsunami tide prediction in shallow water using recurrent neural networks: model implementation in the Indonesia Tsunami Early Warning System. Journal of Reliable Intelligent Environments. Springer Nature.