Comparative Google Earth Engine Workflows for Flood and Landslide Diagnostics in Lower Silesia, Poland, and Batang Kali, Malaysia
This study demonstrates that Google Earth Engine offers a transferable computational framework for post-event geospatial diagnostics of contrasting natural hazards, successfully applying a unified workflow to map the 2024 Lower Silesia flood and the 2022 Batang Kali landslide while revealing that diagnostic accuracy is highly dependent on the specific hazard type and remote sensing indicators used.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
When a flood swells a river or a mountain slope gives way, the ground beneath our feet changes in ways that are often impossible to see from the road. In the hours and days after such a disaster, the terrain is frequently too dangerous to enter, and the weather may be too poor for a pilot to fly overhead. Yet, people need to know where the water has risen or where the earth has shifted to save lives and plan recovery. This is where a different kind of eye comes into play: satellites circling high above the Earth, capturing images of the surface day after day. By comparing a picture of a landscape taken before a disaster with one taken after, scientists can spot the differences that tell the story of what happened. The challenge has always been how to do this quickly and reliably across vast areas, especially when the two disasters look nothing alike. One involves water spreading over land, while the other involves the ground itself sliding down a hill.
A researcher from Poland and Malaysia set out to test a single, unified digital system to handle both of these very different problems. They used a powerful cloud-based platform called Google Earth Engine, which acts like a massive library of satellite images and a supercomputer that can process them all at once. Instead of downloading thousands of photos to a local computer, the researcher wrote a set of instructions that could run directly on the platform. They applied this same set of instructions to two distinct events: a major flood that struck Lower Silesia in Poland in September 2024, and a devastating landslide in Batang Kali, Malaysia, that occurred in December 2022. The goal was not to prove that floods and landslides are the same, but to see if the same digital workflow could be adapted to find the specific signs of each.
For the flood in Poland, the researcher looked at optical images, which are similar to the photographs a camera takes. They focused on a specific signal that reveals where water is present. Water absorbs light in a way that makes it stand out clearly against dry land and vegetation. By calculating the difference between the water signal before the storm and the signal after, they created a map showing exactly where the water had spread. The system worked with remarkable clarity for this event. It identified an area of 70.142 square kilometers that had been flooded. When the researcher checked their work against 240 specific points they had marked on the map, every single point matched their expectations. However, the author is careful to note that this perfect score came from checking the map against other satellite images, not from going out into the field to verify the water levels with their own eyes. While the result is encouraging, it suggests the method is highly effective for spotting floodwaters in this type of environment, even if the numbers might be slightly optimistic without ground truth.
The situation in Malaysia was far more complicated. The Batang Kali landslide happened in a dense, tropical rainforest where clouds often block the view of optical cameras. To see through the clouds and the trees, the researcher switched to radar, a type of sensor that bounces radio waves off the ground. When a landslide occurs, it strips away vegetation and smooths out the rough surface of the hill, which changes how the radar waves bounce back. The researcher looked for areas where this bounce-back signal dropped sharply, indicating a sudden disturbance. The system was incredibly sensitive; it found every single one of the 11 known landslide locations they tested. But this sensitivity came with a heavy cost. The system also flagged 109 other spots as landslides that were actually stable ground. In total, the map was correct only about 9 percent of the time when it claimed a landslide had occurred.
This stark difference in results tells a clear story about the limits of a one-size-fits-all approach. The same digital workflow successfully moved from one disaster to the other, proving that the computer architecture is flexible enough to handle different types of data. However, the physical nature of the disaster dictates how well the method works. For the flood, the water created a clear, distinct signal that the system could isolate easily. For the landslide, the signal was muddied by the complex environment of the tropical forest, where rain, soil moisture, and tree cover can all look similar to a landslide to a radar sensor. The researcher concluded that while the radar method is excellent for a rapid first look—telling emergency teams, "Look here, something has definitely changed"—it cannot yet be trusted to draw a precise line around a landslide without further checks.
The study demonstrates that a single, repeatable process can serve as a backbone for disaster response, but the final interpretation must always respect the specific conditions of the event. In the Polish flood, the method provided a clean, usable map of the affected area. In the Malaysian landslide, it provided a high-alert screening tool that points to areas of disturbance but requires human experts to filter out the false alarms. The value of this work lies in showing that open, free satellite data and shared computing power can be used to build these diagnostic tools quickly. It offers a way for scientists and emergency responders to get a rapid, consistent view of a disaster zone, whether it is underwater or buried under mud, while acknowledging that the final step of understanding the danger still requires careful, specific analysis.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.