AI predicts sea surface temperature cooling during tropical cyclones

Warm tropical oceans form extreme weather events known as tropical cyclones. These cyclones are characterized by low atmospheric pressure, strong winds, and heavy rainfall. These cyclones can cause the ocean surface to cool, potentially reducing the fuel for cyclones and causing them to slow down or stop altogether. A new study published in Geophysical Research Letters has used machine learning to model the impacts of these cyclones on sea surface temperatures.

Hongxing Cui, a doctoral researcher at Guangdong Laboratory of Southern Marine Science and Engineering in China, and colleagues used a machine learning-based random forest method to predict sea surface cooling in the Pacific Ocean basin. The team used data from a 20-year period starting in 1998 to train the system. They focused on the northwest Pacific Ocean, one of the most active areas for tropical cyclones.

The researchers used 12 characteristics of tropical cyclones and pre-storm conditions to make their predictions. These included the intensity of the cyclone, the speed and direction at which the cyclone was moving, the smallest radius of the cyclone reaching a speed of 30 knots, and various oceanic conditions.

The researchers trained the system using data from a 20-year period beginning in 1998 and employed the machine learning-based random forest method to help predict the evolution of sea surface temperature over time and space within the northwest Pacific Ocean (equator to 30°N, 100–160°E), which is one of the most active zones for tropical cyclones.

Key Findings

The study found that tropical cyclone intensity, speed and size, pre-storm mixed layer depth and sea surface temperature had the most significant influence on subsequent surface temperature patterns observed in the ocean. The cooling effect began two days before the event, intensified during the passage of the tropical cyclone, but actually peaked a day after the event.

More intense, larger, and slower-moving tropical cyclones in areas with a shallow mixed marine layer had a greater cooling effect on surface waters. The intensity and speed of the storm had more of a local impact, while the overall size of the cyclone, the depth of the pre-storm ocean mixed layer, and sea surface temperature influenced the cooling effect over a larger area.

The two days prior to the event saw the research team observe the beginning of cooling, which intensified during the passage of the tropical cyclone. However, the day after the event witnessed the peak, with a decline in sea surface temperature of >1.3°C (reaching 2°C for category 3–5 hurricanes).

The cooling expands to affect a larger proportion of the ocean surface over the same period. Additionally, researchers found that the maximum cooling occurred offset to the direct track of the tropical cyclone, 50km to the right. The oceans began returning to normal conditions by days two to four with a relatively rapid warming, followed by slower recovery thereafter to day 14 when the cooling effect reduced to just 0.4°C above the local average.

Implications for Future Research

By comparing actual data with predictions from their model, the research team found a good correlation between results. This gives them confidence in their model’s ability to predict future tropical cyclone events.

The model can use to explore impacts in other ocean basins on a global scale and can help to explore how tropical cyclones affect primary producers in oceans such as photosynthetic algae.

Machine learning can be used to understand complex environmental phenomena, as underscored by the study. It also highlights how human activities like climate change could potentially alter these phenomena in ways that have far-reaching implications for ecosystems worldwide.

Keep Reading

Part 1: Cloudburst  in Ganderbal's Padabal village & unfulfilled promises