Artificial Intelligence (AI) is changing the way we work–not just for chatbots in customer service or conjuring fake images. AI excels at repetitive and redundant tasks; it can increase workflow efficiency and speed up data analysis. These types of tasks are not limited to office work; even the work of protecting natural resources from invasive plants can benefit from AI.
The invasive miconia plant infests thousands of acres across East Maui. Its huge leaves shade out other plants, allowing it to dominate the landscape. The loss of understory plants and miconia’s shallow root system disrupt the forest’s ability to capture and store water, causing increased erosion and sedimentation that threaten the health of near-shore environments. The Maui Invasive Species Committee (MISC) has been working to keep miconia from spreading into higher-elevation watersheds. But to stop it, MISC first must find it.
For decades, MISC used helicopters to search low-elevation forests for the telltale green and purple leaves. A pilot and a team of three “spotters” methodically flew back and forth, low, and slow, heads hanging out of the ship, searching for miconia plants, and recording locations on a GPS. Exciting at first, the thrill wore off after a few hours, replaced by sore necks and sometimes nausea. It was tedious and expensive, but the best way to survey large swaths of land for miconia. Until now.
Similar to how software in your phone can recognize faces in photographs or act as a secure login, computer software can be used to identify plants from images. Ryan Perroy and Roberto Rodriguez with the Spatial Data and Visualization Lab at the University of Hawaii-Hilo (SDAV) have developed a process to collect footage of Hawaii’s forests and identify miconia.
The process still involves a helicopter flying back and forth in a lawnmower-like pattern, but instead of relying on human spotters, a digital camera mounted on the outside of the helicopter captures imagery destined for a computer in the lab. Small video cameras and equipment record the location, orientation, and speed of the helicopter. Rodriguez has trained the software to recognize miconia. The same visual cues that alert field crews to a miconia plant–leaf size and shape, the distinctive vein in the middle–cue the software to the presence of a plant. The computer then assigns GPS coordinates to the plant using data collected during the flight. A map is born.
Roberto Rodriguez from the Spatial Data and Visualization Lab at UH Hilo on a test flight collecting aerial imagery of East Maui forests. Rodriguez developed a computer program that identifies invasive miconia from photographs, saving field crews hours of work. — Adam Knox, MISC photo
“The software can recognize a single leaf poking through the forest,” Perroy explains. How do the two approaches hold up when compared against each other? The limits of the software are similar to those for a human spotter in some ways: leaves hidden by overhanging trees will go undetected using either method. But in trials comparing the overall speed and ability to detect miconia, the computer bested the humans–perhaps because it does not get tired, bored, or airsick.
The software developed by the SDAV Lab for East Maui is looking for miconia; by training it on different characteristics, it can be used to find other invasive plants, such as pampas grass, or trees that show symptoms of rapid ohia death, a fungal pathogen not known to be present on Maui. Perroy has already done this on Hawaii Island. “In the past we had a person review footage looking for symptoms of rapid ohia death. It was tiring, grueling work. A computer can do it in only a few hours.”
For Maui, initial efforts will focus on finding miconia along the borders of its known range–west of Hana to Huelo, upslope towards Hanawi, and around the southern edge to Kipahulu.
Woody Mallinson, Natural Resources Program Manager with Haleakala National Park, explains why the National Park Service is helping to fund this project. “Protection of forest bird habitat is the number-one natural resource priority for our park,” he says. “The threat of miconia getting into Kipahulu Biological Reserve is a concern.”
After some additional fine-tuning, the technique will be used to create a snapshot of miconia across East Maui–information that will help guide future work to ensure the long-term health of our forested watersheds.
AI and technology can’t replace our natural resources, but these tools can help us in our efforts to support the healthy environment on which we rely.
* Lissa Strohecker is the public relations and education specialist for the Maui Invasive Species Committee. She holds a biological sciences degree from Montana State University. Kia’i Moku, “Guarding the Island,” is prepared by the Maui Invasive Species Committee to provide information on protecting the island from invasive plants and animals that can threaten the island’s environment, economy, and quality of life.
Source: Maui News