From a drone flying above the waves to a camera just inches from the seafloor, University of Miami researchers are transforming how scientists observe coral reefs and measure changes that were once difficult, costly or time-consuming to document.
Two recently published studies from the University of Miami Rosenstiel School of Marine, Atmospheric, and Earth Science's Aircraft Center for Earth Studies (ACES) demonstrate how advanced imaging and artificial intelligence can offer both broad and highly detailed views of reef ecosystems. One maps an entire reef system from the air at the scale of individual coral colonies. The other extracts coral boundaries and measurements from routinely collected underwater photographs.
Together, the technologies could help scientists and resource managers detect reef decline sooner, assess damage after hurricanes and bleaching events, evaluate restoration projects and direct limited conservation resources where they are most needed.
"Coral reefs are changing at a pace that requires us to collect and analyze information much more efficiently," said Ved Chirayath, Vetlesen Endowed Chair of Earth Sciences and director of ACES, based in the Department of Ocean Sciences at the Rosenstiel School. "By combining advanced imaging with artificial intelligence, we can examine reefs across much larger areas without losing the fine-scale detail needed to understand what is happening to individual coral colonies."
Turning waves into a lens
In one study, researchers used NASA Airborne Fluid Lensing and artificial intelligence to create detailed maps of more than 5 square kilometers (about 2 square miles) of coral reef habitat in Tumon Bay, Guam.
Developed by Chirayath while at NASA and now advanced at ACES, Fluid Lensing addresses a longstanding challenge in aerial reef imaging: the constantly moving ocean surface. Waves bend and distort light traveling from the seafloor, causing reefs to appear blurred or displaced in conventional aerial images.
The technology captures thousands of images in rapid succession and uses the waves' natural magnifying effect to computationally reconstruct a clear view of the reef. The resulting imagery resolves the seafloor at the centimeter scale, down to features as fine as half a centimeter, in water up to about 20 meters deep.
Researchers surveyed Tumon Bay in 2022 and again in 2024, allowing them to compare the reef before and after Category 4 Typhoon Mawar struck Guam in May 2023 with winds of 140 mph.
The maps revealed a 59 percent decline in massive coral cover and a 35 percent decline in coral fore-reef habitat. Algae, which can compete with corals for space and impede reef recovery, increased by 105 percent.
Unlike conventional assessments based on a limited number of dive sites, the maps provided a location-by-location record of changes across the entire bay. This level of detail can help managers identify areas that sustained the greatest losses, locate sections of reef that survived and determine where restoration or additional protection may be most effective.
"We mapped an entire reef rather than sampling a small portion of it, and we did it at a resolution capable of distinguishing individual coral colonies," said Soufyane Bouchelaghem, a postdoctoral associate at ACES and the study's first and corresponding author. "Repeating the survey after Typhoon Mawar turned that map into a precise measurement of damage across Tumon Bay.
Artificial intelligence and citizen science helped make the analysis possible. Trained volunteers and marine biologists using NASA's NeMO-Net video game labeled about 1 percent of the 2022 imagery, identifying features such as branching coral, massive coral, algae, sand and rubble. Researchers used those labels to train a deep-learning model that classified the remaining imagery with 88 percent accuracy.
The model was then applied to the 2024 images without any retraining, even though it had not previously encountered them. Free-diver photoquadrat transects at ten locations in the bay provided independent field validation.
The study, "Cm-scale marine habitat mapping of entire Tumon Bay, Guam Coral Reef using NASA airborne fluid lensing and NeMO-Net pre (2022) and post (2024) Typhoon Mawar," was published April 23, 2026, in the journal Frontiers in Marine Science.
Measuring corals one photograph at a time
While Fluid Lensing provides a broad aerial view, a second study addresses another obstacle to effective reef monitoring: the time required to analyze the many photographs collected during underwater surveys.
Scientists traditionally examine these images manually, tracing coral colonies and calculating the percentage of the seafloor covered by living coral. Processing a large collection can take months and requires considerable expertise, delaying information that may be needed after a bleaching event, disease outbreak or hurricane.
ACES researchers, working with scientists at NOAA's Pacific Islands Fisheries Science Center and the University of Hawai'i at Manoa, developed an open-source artificial intelligence system called PICOGRAM (Prediction of Individual Coral Organismal Growth, Recruitment, And Mortality) to automate much of that work while retaining scientific oversight. The project was assessed at Application Readiness Level 8 by NASA, a rating indicating that a tool has been validated in an operational setting rather than remaining a research prototype, and NOAA has since tested PICOGRAM at operational scale.
PICOGRAM identifies coral colonies, traces their boundaries and estimates coral cover from standard underwater photographs. Rather than requiring scientists to manually outline thousands of corals to train the system, PICOGRAM learns from automatically generated training masks based on image characteristics such as color, texture, edges and spatial context.
The system also assigns a confidence score to each result. Scientists can review uncertain predictions and correct them through a simple interactive interface, allowing human expertise to remain part of the quality-control process.
Researchers trained PICOGRAM with underwater images collected through NOAA's National Coral Reef Monitoring Program and evaluated it using 137 photographs independently analyzed by marine ecologists. The test images included different reef environments, water depths and levels of visibility.
PICOGRAM achieved 87.5 percent overlap with expert-drawn coral outlines under familiar conditions and 82.5 percent at reef sites it had not previously encountered. Its coral-cover estimates closely matched expert measurements, with an average difference of 1.2 percentage points. Uncertain results required fewer than two corrective clicks, on average, to reach 90 percent agreement with expert annotations.
"Researchers are collecting more reef imagery than can reasonably be analyzed by hand," said Imad Eddine Tibermacine, a postdoctoral associate at ACES and the lead author of the study. "PICOGRAM helps turn those photographs into reliable ecological measurements while allowing scientists to quickly review and correct results when needed."
The technology can be incorporated into existing monitoring programs that already collect underwater photographs. Repeated images from the same locations could help scientists determine whether coral cover is increasing or declining and, with further development, track the growth, recruitment and mortality of individual colonies.
The study, "PICOGRAM - informing coral reef resilience-based management through prediction of individual coral organismal growth, recruitment, and mortality," was published July 14, 2026, in the journal Frontiers in Marine Science.
From observation to action
The two approaches operate at different scales but address the same fundamental challenge: converting growing volumes of reef imagery into useful scientific information.
Fluid Lensing can document changes across kilometers of reef while retaining centimeter-scale detail. PICOGRAM can examine underwater images at the colony level and produce consistent measurements with minimal manual intervention. Used with diver observations, satellite data and other environmental records, the technologies could create a more complete picture of how reefs respond to storms, warming waters, disease and restoration efforts.
That information has consequences beyond the reef itself. Coral reefs support fisheries and tourism, provide habitat for marine life and help reduce the wave energy reaching vulnerable coastlines.
"Better observations lead to better decisions," Chirayath said. "If we can determine where corals are declining, where they are surviving and whether restoration is working, managers can respond more quickly and make more informed use of the resources available to protect these essential ecosystems."
The PICOGRAM research was supported by NASA's Biodiversity and Ecological Forecasting Program. The Tumon Bay research received support from NASA's Biodiversity and Ecological Forecasting Program, NASA's Oceanography Program, the Pacific Islands Climate Adaptation Science Center, the National Geographic Society, the Aircraft Center for Earth Studies endowment, the G. Unger Vetlesen Foundation.
