Creating forest inventories with drones and artificial intelligence
Aerial photograph showcasing a stretch of mangrove forest located in Utría National Park along Colombia's Pacific shoreline. Photo credit goes to Daniel Schürholz from the Leibniz Centre for Tropical Marine Research.
Scientists at the ZMT in Bremen have come up with a technique using drone pictures and artificial intelligence (AI) to pinpoint individual trees in a forest. Not only can this method determine the height and width of each tree, but it can also assist in compiling an inventory of the various species in forests like mangroves. Additionally, the technique provides information about the amount of carbon stored in these forests. This ground-breaking study was recently published in the renowned journal Remote Sensing.
Mangrove forests possess the remarkable ability to retain significant quantities of CO2 and other greenhouse gases within their plant matter and soil. Consequently, they are recognized as crucial ecosystems in the battle against climate change. Specialists approximate that these tidal forests house an astonishing four to twenty billion tons of organic carbon globally. Nevertheless, the magnitude of carbon stored fluctuates significantly across different regions and individual mangrove habitats.
Accurate assessments of the carbon reserves in the diverse mangrove regions of the tropics have been lacking so far. Typically, on-ground surveys are conducted in a limited number of small areas, where measurements of tree height and diameter are obtained. Then, utilizing the wood density and composition of the tree species, an estimation of the above-ground biomass is made.
The challenge posed by mangrove forests is quite demanding. Typically, they are located in secluded areas, making it hard to reach them. One may find themselves sinking deep in mucky sediments up to their waist while having to walk or swim through narrow channels, all while being pestered by hoards of mosquitoes. Despite all these hardships, estimating their true extent remains uncertain due to the inherent fluctuations in nature.
How can scientists improve the methods for measuring carbon stocks in expansive remote mangrove forests? This was the challenge that the Data Science and Technology group at the Leibniz Center for Tropical Marine Research (ZMT) set out to tackle. "We needed to find innovative approaches that would allow us to efficiently survey the entire forest and track changes over time," explains Daniel Schürholz, a doctoral candidate and the study's primary author. "The accuracy of our carbon storage calculations relies heavily on the level of detail we have about the trees in the forest."
The group explored the potential of precisely charting vast mangrove forests that have thick layers of vegetation. An assessment of each tree's specific details, such as its height, position, and extent of canopy coverage, enables a more precise estimation of the carbon stored within them while also facilitating the monitoring of the forest's well-being. "Considering the recent advancements in artificial intelligence, we opted to examine state-of-the-art methods that could automatically identify individual trees within the forest," Schürholz commented.
Located on the Pacific shoreline of Colombia, Utría National Park is primarily made up of dense mangrove forests. To capture the essence of this remarkable ecosystem, the park rangers, along with researchers from Universidad del Valle in Colombia, employed the use of aerial drones to obtain captivating photographs of the elevated forest canopy.
Once they returned to Bremen, the ZMT researchers utilized photogrammetric techniques to produce highly detailed forest mosaics, surpassing the level of detail captured by satellite images. They subsequently devised an artificial intelligence workflow capable of categorizing the extensive mosaics according to various habitat types. Moreover, this workflow enabled the identification and estimation of the height and diameter of individual trees belonging to the native mangrove species. By possessing information about the species, height, and diameter of a specific tree, one can then approximate the amount of above-ground biomass it contains.
Arjun Chennu, a specialist in habitat mapping at ZMT and co-author of the research, explains that with the help of our AI process, we were able to determine that there are approximately 19,717 trees of the unique mangrove species Pelliciera rhizophorae in the particular area that was examined. This estimation would have been incredibly challenging to attain using traditional methods.
"He mentions that the combination of inexpensive drone pictures and artificial intelligence technology has the potential to be utilized in various other aspects apart from identifying blue carbon reserves. These include areas like identifying illegal logging, detecting invasive species, or monitoring changes in animal and plant communities."
The scientists aspire to safeguard precious mangrove forests by offering decision-makers with more dependable information to advocate for the preservation or revival of these forests. Schürholz emphasizes the need to extend the current excitement surrounding advanced AI algorithms to address environmental concerns and enhance our comprehension of the environment. With the rapid advancements in AI, we will gain deeper insights into the intricate workings of nature and improve our ability to safeguard and responsibly oversee it.
Other types of algorithms could be employed to chart various ecosystems like coral reefs or moderate forests, and to recognize creatures and monitor their mobility. The technique could also have advantages for German forests. To achieve this, Schürholz mentioned that the essential step is to customize the AI algorithm to suite the species that inhabit these forests. "We are offering a solid model for a system that can be implemented worldwide," he concludes.
Additional details: The research conducted by Daniel Schürholz and his team titled "Perceiving the Whole Picture: Charting Coverage and Numerating Trees through Aerial Pictures of a Mangrove Forest Utilizing Artificial Intelligence" can be found in the publication Remote Sensing (2023). The research article can be accessed through the following DOI: 10.3390/rs15133334.
Offered by the Leibniz Center for Tropical Marine Research (ZMT)
Title: Generating Forest Inventories using Drones and AI Source: "Constructing forest surveys utilizing unmanned aerial vehicles and AI" (2023, August 16) obtained on August 18, 2023, from https://phys.org/news/2023-08-forest-drones-artificial-intelligence.html
The contents of this document are protected by copyright. You are not allowed to reproduce any part of it without obtaining written permission, except for fair use for private study or research purposes. The information provided in this document is for informational purposes only.
