TEMPO.CO, Jakarta - The Transient Artifact and Continuous Learning System (TACLS) project by the United States' National Aeronautics and Space Administration (NASA) serves as the latest example of Artificial Intelligence (AI) use for flash flood prediction systems. Wahyu Eka Styawan, Urban Campaign and Spatial Policy Officer at WALHI, said the early warning system—which traditionally relied on sensors—has now transitioned into machine learning models, marking an essential step in disaster mitigation.
"This transition is a strategic necessity to transform the flood management paradigm from reactive to proactive, predictive, and evidence-based," he told Tempo on Tuesday, July 7, 2026.
What Are the Advantages of AI-Backed Warning Systems?
According to Wahyu, the accuracy rate of conventional flood analysis ranges from 70 to 80 percent. AI-based models can increase this accuracy to 90 percent.
Additionally, conventional systems take longer to process data updates, leading to delayed responses. Physical and empirical models are also deemed incapable of capturing the complex relationships between land cover changes, river dynamics, rainfall patterns, and the impact of climate change on flood potential.
Artificial intelligence, he went on, provides earlier warnings through modeling of rising water levels that integrates various data sources such as satellite imagery, field sensors, and meteorological information. With longer warning times, there is a greater chance for government and communities to prepare mitigation and evacuation measures.
Another advantage of AI in early warning systems is its ability to process large amounts of data simultaneously. Artificial intelligence can process Internet of Things (IoT) sensors, satellite images such as Sentinel-1 SAR, and historical weather data collected over years.
In addition to enhancing predictive capabilities, Explainable Artificial Intelligence (XAI) can also clarify the factors influencing flood potential, such as land use changes, river sedimentation, and extreme rainfall intensity. The Digital Twin technology and hydrodynamic models are also capable of processing various flood scenarios, supporting mitigation planning and evacuation routes.
The Extent of AI Use in Flood Warning System
WALHI believes that AI use for disaster management must not be limited to time- and place-predictions but expanded into identification of factors that increase flood risks. With smart systems, lawmakers should be able to anticipate the impact of changes in forest cover, river basin damage, changes in infiltration areas, expansion of mining and plantations, and development that ignores environmental carrying capacity.
"If it is only used to speed up evacuations, the technology will only make the country more adaptive to disasters without solving the root of the problem. AI should not only generate early warnings, but also serve as the basis for policy evaluation, enforcement against environmental destruction, and prevention of potentially harmful permits that increase disaster risks," Wahyu stated.
Also covered in Tempo's premium report: Bagaimana AI Memprediksi Banjir Lebih Presisi (How AI Predicts Floods More Accurately), NASA's TACLS machine learning was trained with 30 years of historical data from the global satellite network (GNSS). The results of its analysis were translated into a visual model for interpretation by meteorological researchers.
"The system enhances existing methods to reduce the amount of time it takes for a human analyst to determine whether to issue a flash flood warning," as stated by NASA.
The entire TACLS analysis process is expected to be rapid and operated in near real-time. After being tested with initial data from past extreme weather conditions, precisely from the period 2017-2023, TACLS proved capable of predicting 93 percent of flash flood incidents.
Will Indonesia Implement AI-Based Early Warning Systems?
In the US, the National Weather Service has started integrating TACLS into the flash flood forecasting system in Southern California. The TACLS software, including its training data, will be provided as open source for use by other researchers.
The Chief Secretary of the Meteorology, Climatology, and Geophysics Agency (BMKG), Guswanto, said the agency is also pushing for the digitization of early flood warning systems. However, implementing AI-based systems in Indonesia is not without its challenges. "One of the challenges is the quality and availability of data," he said.
He also mentioned that system integration is a major task. AI systems cannot operate independently; they need to be connected to systems used by BMKG, the National Disaster Management Agency (BNPB), and local governments. This synchronization is necessary for the information generated to be consistently utilized by all agencies involved in disaster management.
"Synchronization is necessary to prevent the AI system from operating separate from existing systems," he said.
AI and hydrology models also require strong computational support. These prediction models require servers with high capacity to process large amounts of data and run simulations quickly. Personnel capacity also determines the success of AI adoption into early flood warning systems.
Read: Drought and Floods: How El Nino Fuels Global Weather Extremes
Click here to get the latest news updates from Tempo on Google News
