Mapping Jellyfish

| |

Large blooms of jellyfish may seem like a nuisance to many of us, perhaps closing beaches we enjoy, but perhaps many of us would not consider them a major concern comparable say to climate change or other global crises.

On the one hand, jellyfish, particularly large blooms could be an indicator of climate change, as large blooms are likely to form in warmer water. Jellyfish can also be damaging to everything from power plants that depend on water to fisheries that help feed us. There could be a significant economic cost associated with jellyfish blooms.

Jellyfish also can deplete fish populations due to their feeding behavior that often targets young fish or larva. Some jellyfish species could also be deadly to us.

As increasingly warmer waters will likely mean larger and more frequent jellyfish blooms can be expected in coming year, new tools are needed to spot these blooms before they cause significant health or economic damage.

Not all jellyfish produce the same harmful effects; therefore, we may also need to better identify given species more precisely.[1]

A photo of small translucent white jellyfish against a flowing deep blue background.
Moon jelly (Aurelia labiata). Photo: Caitlin Dempsey.

Monitoring jellyfish with satellite imagery and UAV

One tool created to monitor and predict jellyfish blooms is JellyX, which uses satellite imagery and forecasts blooms and movements using ocean current data. JellyX provides web mapping to monitor large-scale jellyfish swarms and forecast where they are likely to end up.

The tool is created by ColomboSky, a company supported by the European Space Agency (ESA), with data derived from the Copernicus Marine Service. The Copernicus program uses the Sentinel satellites created by the ESA to monitor Earth systems.[2],[3] 

Some scientists and researchers have used unmanned aerial vehicles (UAVs) to monitor jellyfish, given the near real-time nature of the data that can aid models forecasting and observing jellyfish blooms. While UAV data could be more timely for monitoring, and even more useful to correcting models showing where blooms are likely to spread, they can be limited in range or not always operable in given conditions. Identification of given jellyfish species may also be better underwater.[4]

Underwater jellyfish monitoring systems

Instead of UAVs, alternative approaches have included using underwater monitoring systems, even underwater drones (ROVs), that provide near coast monitoring. By using underwater monitoring systems, the intent is also to improve jellyfish identification that helps determine the particular species in blooms.

For instance, the box jellyfish family is perhaps the most venomous type of jellyfish or even among the most venomous animals. The identification of these species is critical for public safety, while other forms of jellyfish are far less harmful or even not harmful to humans.

A group of tan colored jellyfish against a deep blue background.
Black Sea nettle (Chrysaora achlyos). Photo: Caitlin Dempsey.

Underwater imagery can often be foggy or unclear. Using deep learning algorithms, which have preprocessing that enables image enhancement, could improve species identification, which enables researchers to know how harmful given jellyfish blooms are.

Algorithms that deploy defogging, adaptive histogram equalization, and multi-scale enhancements enable imagery to be improved and more useful for species identification. This helps to remove noise and other background factors that limit image quality underwater.

Deep learning classification, using artificial neural networks, can then be deployed for species identification that is automated and can estimate the scale of blooms.

One algorithm, the YOLOv3, has been used along with image enhancement techniques; the approach is useful for species identification. Such algorithms used along with ROVs could potentially be useful for differentiating the scale and threat jellyfish blooms present to the public and/or economy, giving researchers and monitors more information to calibrate their responses.[5] 

In general, and analysing different approaches researchers have taken, it has been shown that the GoogLeNet, a 22 layer neural network model, and Faster R-CNN algorithms, with R for region and convolutional neural network (CNN), demonstrate some of the best results in species identification. The accuracy of these networks range around 96-75%.[6]

Remote sensing has improved the mapping and tracking of jellyfish swarms

Our ability to spot large swarms of jellyfish has improved greatly through a combination of remote sensing data, both UAV and satellite data, and algorithms that forecast jellyfish movement based on current data. JellyX is perhaps the state of the art when it comes to large jellyfish bloom monitoring. On the other hand, such a tool may not be easily used for species identification, where some jellyfish have the potential to be more harmful to our health. The use of ROVs and deep artificial neural network tools has helped to automate detection of given jellyfish species and better understand the level of threat given blooms may pose to us. 


[1]    For more on the damage that jellyfish cause, see:

[2]    For more on JellyX, see:

[3]    For more on the Copernicus Marine Service, see:

[4]    For more on monitoring jellyfish through UAV imagery and tools, see:  Yang, Z., Yu, X., Dedman, S., Rosso, M., Zhu, J., Yang, J., Xia, Y., Tian, Y., Zhang, G., Wang, J., 2022. UAV remote sensing applications in marine monitoring: Knowledge visualization and review. Science of The Total Environment 838, 155939.

[5]    For more on YOLOV3, see:  Gao, M., Bai, Y., Li, Z., Li, S., Zhang, B., Chang, Q., 2021. Real-Time Jellyfish Classification and Detection Based on Improved YOLOv3 Algorithm. Sensors 21, 8160.

[6]    For more on deep artificial neural networks used for jellyfish detection and classification, see:  Han, Y., Chang, Q., Ding, S., Gao, M., Zhang, B., Li, S., 2022. Research on multiple jellyfish classification and detection based on deep learning. Multimed Tools Appl 81, 19429–19444.

Share this article

Enter your email to receive the weekly GIS Lounge newsletter: