When Seconds Matter: The Future of Predicting Natural Disasters

Climate change is leading to more frequent and intense extreme weather events that have a catastrophic effect on people, property, and ecosystems.

The World Meteorological Organization reports that tropical cyclones, floods, droughts, and other events resulted in a 16-year high in new internal displacements in 2024, with 45.8 million people forced to flee their homes in 163 countries and territories.

At the same time, global economic losses from natural disasters in 2024 were estimated to be between US$316 billion and US$368 billion

This underscores the critical need to provide quick and accurate forecasts that enable communities and emergency responders to act within the ‘golden warning window’ and can guide policy makers to develop long-term mitigation measures.

However, high-density, real-time monitoring of atmospheric and geological conditions is not an outcome that has been historically possible. General-purpose, ground-based water vapor observation stations are usually sparsely distributed, with currently only around 2,300 stations in China. Sites are far apart and coverage is low. Moreover, the satellite signals and data they receive suffer from delays and low accuracy, making it difficult to capture sudden, localized extreme weather events.

Forecasting the fury: Low-cost, precise, real-time, AI-powered

In partnership with China Mobile and the Meteorological Observation Centre/China Meteorological Administration (MOC/CMA),and Nanjing University of Information Science and Technology (NUIST) Huawei has developed and verified the world’s first 5G/AI-GNSS solution for high-precision water vapor retrieval to improve accuracy for fine-grained precipitation forecasts to support disaster prevention and mitigation.

Using thousands of China Mobile’s 5G base stations coupled with a Huawei-developed algorithm, the partners have created a super-dense monitoring infrastructure that can sense what is happening on the ground and in the sky.

Each 5G base station is natively equipped with Global Navigation Satellite System (GNSS) capability. The integrated dual-frequency, high-frequency BeiDou timing system can receive satellite signals in real time to achieve high-precision time synchronization.

For atmospheric monitoring to predict storms and floods, the 5G/AI-GNSS solution analyzes satellite signal delays on geo-positioning as they pass through the atmosphere, providing data on precipitable water vapor to an accuracy of 10 mm. The solution’s proprietary Edge-Native Real-Time PPP Engine uses Multi-Frequency Carrier-Phase Processing with Instantaneous Integer Ambiguity Resolution (PPP-AR) to precisely isolate delays from satellite clock and orbit errors with sub-millimeter fidelity. These micro-scale atmospheric signals are fed into a Physics-Informed Neural Network (PINN), which reconstructs a real-time, 4D tomographic model of the troposphere, turning raw signal noise into precise meteorological intelligence.

For detecting landslides and earthquakes through geological monitoring, minute crustal movements, structural deformation, and landslide creep can be monitored because the GNSS signal can pinpoint a base station’s real-time position to the sub-millimeter accuracy. Any changes to the position of the base station can provide critical data for earthquake precursor analysis, mudslide warnings, and post-disaster damage assessments.

Testing and verification

Testing on a single 5G base station in 2025 met requirements for integration into meteorological forecasting systems. The 5G-GNSS deduced precipitable water vapor from satellite signal delays to a sub-3 mm error margin.

Network testing of 1,400 5G base stations in Guangdong, also in 2025, verified that high-density 5G GNSS improves precipitation forecasting accuracy. Average scores for 4-day, 24-hour precipitation forecasts for precipitation above 50 mm improved the average Critical Success Index (CSI) by 25.04%, Probability of Detection (POD) by 29.33%, and Equitable Threat Score by 34.91%.

Testing in Guangdong, China, in 2025 covered 1,400 5G sites

CSI is a metric used to assess the forecast accuracy of a prediction model by measuring correct forecasts against the total number of forecasts made. POD divides the number of true positives by the sum of true positives and false negatives to evaluate the performance and reliability of a detection system. ETS removes random chance from the overall calculation.

Improving these metrics to a significant degree has huge potential for delivering accurate forecasts that can detect natural disasters.

5G/AI-GNSS: New capabilities, new possibilities

The 5G/AI-GNSS solution:

• Enables precise short-term precipitation forecasts from 3 hours to 48 hours by detecting gaseous water molecules at a 10-km altitude, instead of observing raindrops below 2 km like traditional ground stations.
• Leverages the high density of 5G base stations spaced at 0.5 km to 2 km apart to fill the coverage gaps of traditional weather radars, which are typically spaced tens or even hundreds of kilometers apart.
• Refines spatial resolution from the sub-kilometer level to the sub-100-meter level
• Boosts data acquisition frequency to 100 milliseconds, far surpassing the minute-level intervals of traditional stations. This reduces forecast errors, improving extreme weather prediction accuracy by up to 80% and pinpointing heavy rainfall locations with greater precision.
• Consumes less power than meteorological radar and requires no additional land or sites for deployment, achieving wide coverage at low cost.
• Creates a continuous field of atmospheric data. Precise GNSS data collected across a region serves as a high-fidelity baseline to infer the true state of atmospheric water vapor, enabling the effective prediction of natural disasters and timely emergency response.

Huawei combines its R&D and communications technology with 5G base stations and AI to integrate meteorological data, significantly boosting the accuracy of fine-grained weather forecasts and supporting disaster prevention and mitigation.Professor Zhi Xiefei
Nanjing University of Information Science and Technology (NUIST)

Leveraging the vast global coverage of 5G base stations, they are being converted into a low-cost, high-density water vapor detection network. This innovation has improved precipitation forecast accuracy scores by up to 45% in trial, significantly cutting missed severe rain warnings and enhancing disaster relief efforts.Liang Hong
Chief Scientist, CMA Meteorological Observation Centre

5G/AI-GNSS: In the field

Currently, the solution uses existing GNSS modules on 5G base stations in China. In 2025, the system was expanded to 1,400 sites across Guangdong Province (covering an area comparable to Belgium, the Netherlands, and Denmark combined).

In pilot testing on previously missed or incomplete forecasts, the solution successfully predicted debris flow caused by heavy rain in China’s Maoming in 2024, and correctly forecast when the Typhoon Butterfly, which hit Guangdong in 2025, would make landfall.

Following a dedicated discussion at a World Meteorological Organization (WMO) session in 2025, the solution is set to be integrated into the WMO's revised WIGOS Vision 2050 as a definitive solution for meteo-geological monitoring.

This low-cost, green solution is set to redefine how we predict and respond to natural disasters, and in the process save lives, protect property, and safeguard ecosystems and biodiversity.