The overall objective of this project was to demonstrate the utility of underwater image mosaics for coral reef monitoring. The problem of efficiently mapping and monitoring coral reef resources has relevance to the Department of Defense (DoD) for several reasons. First, at least 46 US military facilities have adjacent coral reef sites. Second, federal policy mandates that DoD characterize, assess, and monitor underwater benthic communities at these sites to ensure that DoD operations do not lead to natural resource degradation. Third, coral reef ecosystems worldwide are presently threatened by increasing levels of both human and natural disturbance. Thus, monitoring efforts that can efficiently provide data that will help distinguish between reef degradation that can be directly attributed to DoD activities versus those that that are correlated with region-wide decline are of primary concern.
A previous Strategic Environmental Research and Development Program (SERDP) project (RC-1333) led by the University of Miami’s Roesenstiel School of Marine and Atmospheric Research (UM/RSMAS) researched the creation and use of underwater image mosaics. The result was a suite of image processing algorithms, software, and best-practices that together enabled new capability for mapping and monitoring coral reef resources. The project projected that use of meso-scale, 2-D, mosaicked images of reef plots could circumvent the limitations of current state-of-the-art methods in coral reef monitoring (i.e., diver transects, photo-quadrats, strip mosaics), while simultaneously maintaining the strengths of a diver-based approach. Testing this premise was the overall objective of this project.
In order to demonstrate the capabilities of the underwater landscape mosaic technology, the approach was to determine potential end users’ specific applications and needs that could benefit from these new capabilities. Five field demonstrations were conducted to test 18 performance objectives that had been identified in response to the needs assessment. Most of the performance objectives had more than one metric to test, so there were a total of 57 metrics that were assessed during the project. Of the 18 performance objectives, 9 were considered completely successful, 9 were a partial success, and 0 were failures. Of the 57 metrics, 45 were considered completely successful, 12 were a partial success, and 0 were failures. Eight of the 12 metrics that were partial successes were technically not complete successes due to the way the performance tests were designed, but for practical purposes were still quite acceptable.
A mosaic is a single large image composed of many smaller overlapping images, each covering a small portion of the total area. Individual underwater images are taken close (~1-2 m) to the seabed; they have high spatial resolution and minimal water column attenuation. A mosaic of such underwater images enables a high-resolution “landscape view” of the seabed. The RSMAS team has developed techniques to construct spatially accurate mosaics covering areas up to 20×20 m with millimeter-scale resolution. First-generation mosaics (ca. 2004) were created with video images only and provided millimeter-scale resolution. In 2007, a second-generation system with sub-millimeter scale resolution was developed by integrating a high-resolution still camera with the original video acquisition system. This demonstration used the second-generation system.
The innovative aspect of the current mosaic technology is that the images provide both landscape-level maps and high resolution (sub-millimeter) images of individual coral colonies. Users can, moreover, collect imagery at both landscape and colony-levels for areas of several hundred square meters in under an hour of in-water dive time, creating mosaic products that provide increased information on coral colony health and small scale competitive interactions. Landscape mosaics address several limitations of traditional, diver-based, coral reef monitoring techniques:
The overall goal was to demonstrate that landscape mosaics extend traditional methods of coral reef monitoring by providing new capabilities, while simultaneously retaining the strengths of diver-based methods. Four field demonstrations were designed to test the mosaic capabilities relative to other techniques currently used for coral reef assessment. In addition, a fifth demonstration was conducted under controlled conditions in a pool using man-made targets of known size in order to assess the absolute spatial accuracy of the mosaics (as opposed to the field demonstrations, which assessed mosaic accuracy relative to diver-based measurements).
Four field demonstrations and one pool demonstration were completed to evaluate 57 metrics organized under 18 performance objectives.
Long-term Monitoring Demonstration - The goal of this demonstration was to assess the potential benefits of using landscape mosaic technology in a long-term coral reef monitoring program. In particular, the project assessed the effectiveness of using mosaics to extract 1) colony-based metrics of coral reef condition, and 2) the metrics needed to map and monitor large-scale reef plots for change-detection purposes. In addition, the ease of use of the mosaicing technology in terms of data collection was evaluated. The results showed that measurements of colony size and percent cover made by divers in the water were not significantly different than those made from mosaic image analysis. Mosaic imagery was also capable of providing the same information as from hand-mapping reef areas. In addition, mosaic imagery and the process of mosaic analysis was found to be as consistent as using multiple diver observations. Finally, nonscientific divers were trained in mosaic image acquisition and acquired usable data. In terms of cost, there was little difference in mosaicing and diver methods of measuring coral colony sizes. However, given the same cost per unit effort (4 days of sampling, 2 divers), the project estimated that divers would be able to map 62m2 of reef resources using hand-mapping techniques as compared to ~3,800m2 using landscape mosaics.
Endangered Species Demonstration - The goal of this demonstration was to evaluate the utility of mosaics for monitoring populations of threatened corals, particularly the species Acropora palmata. The technology was evaluated in its ability to replicate diver metrics of 1) coral location and abundance, 2) coral colony size, and 3) colony condition. Mosaic imaging technology was able to replicate diver assessments of coral colony counts, location information, colony size estimates, and provide mosaic analysts with the information to accurately assess colony health information such as % live cover and colony type. When comparing the cost of assessing coral colonies for the above metrics image mosaicing was less expensive than traditional diver methods.
Grounding Demonstration - The goal of this demonstration was to evaluate the utility of mosaics for assessing damage to reefs caused by vessel groundings. The project evaluated the utility of using mosaics to measure 1) large areas of damage, 2) long linear-distances, 3) multiple methods of damage assessment, and 4) reef health. A fifth performance objective was devised to assess whether new users can extract data from image mosaics. No significant differences were found in measures of long-linear distances between divers and mosaics. Global Positioning System (GPS) information was found to be less accurate than either divers or mosaics for the purpose of damage assessment. Measures of reef health agreed with mosaic-derived indices with the exception of categories such as sand and gorgonian cover that varied greatly between observers and methods. Novice analysts were able to derive estimates of coral colony sizes and percent cover of major categories that were indistinguishable from diver estimates. When comparing the cost of assessing reef damage, GPS methods were the least expensive followed by mosaic imaging. Diver –based assessment of reef damage was the most expensive method tested. The GPS method, although inexpensive was also the least accurate and most variable of the three methods tested. The mosaic is the most cost-effective method of measuring reef damage due to the accurate results and the increased ecological information provided over both diver-based and GPS methods.
Traditional Metrics Demonstration - The goal of this demonstration was to evaluate the utility of mosaics for coral reef monitoring efforts traditionally performed using diver-transect surveys. The performance objectives of this demonstration examined if 1) mosaics could replicate ecological information extracted from diver surveys, 2) mosaics could estimate metrics obtained through multiple diver methods of reef health assessment, and 3) novice users can be trained to create image mosaics using a manual. In cases where the exact diver-based transect was replicated directly on a mosaic image, there was no significant difference found in estimating coral reef health parameters. However, some differences in methods were detected based on differences in the areas sampled by various diver transects and the variability of the reef itself. Novice users were trained to use mosaic software and create mosaic image data that was indistinguishable from those created by expert analysts. When comparing diver and mosaic methods of estimating ecological metrics, the project found that single-variable diver methods of estimating coral health are less costly than mosaic surveys. However, if end-users are interested in estimating more than one health parameter in a given survey, such as coral cover, coral size frequency, and species diversity, mosaic imaging is less expensive since all of these metrics can be obtained from a single mosaic survey. In addition, the ability to measure multiple variables at a later date, without advance planning, is a distinct advantage of mosaic imaging over diver surveys.
Absolute Accuracy Demonstration - The absolute accuracy demonstration was designed to evaluate the accuracy and precision of size measurements made from mosaic image analysis and diver surveys. Unlike the previous performance objectives, the success criteria in this demonstration are based on the known sizes of objects and not the performance relative to diver surveys. This demonstration evaluated 1) absolute accuracy of mosaic and diver size measurements, 2) precision of multiple mosaic and diver size measurements, 3) the precision of multiple mosaic analysts and diver size measurements, and 4) the bias of pool and field derived mosaic imagery. The average bias of measuring targets of known size between 5 and 120cm was approximately 1cm for both diver and mosaic methods. The same was true for estimating the projected length of inclined targets. No differences were observed when comparing results over multiple mosaics or when using multiple mosaic analysts. In addition, the measurement bias of objects placed in a pool was not significantly different than the bias measured in field mosaics. Thus mosaics were found to be highly accurate methods of estimating coral colonies on the cm scale and these results were found to be repeatable over different images and using different observers to carry out the analysis.
This project addressed an important question, “when are mosaics superior to traditional methods (as opposed to equaling performance of diver methods)?”
Considering both performance and cost, the project concluded mosaics are a superior approach:
One intended end-user community includes the marine/coral reef ecologists with the Navy’s Scientific Diving Services. Transfer of this technology to that group has completed and they have executed several field surveys already. The University of Miami continues to partner with other federal, state, local, and private organizations to expand the pool of users of this technology. Current UM/RSMAS partners include: the National Oceanic and Atmospheric Administration (NOAA) (Restoration and Southeast and Pacific Fisheries Science Centers), Biscayne National Park, The Nature Conservancy, New England Aquarium, American Museum of Natural History, Scripps Institute of Oceanography, U. North Carolina Wilmington, Coral Restoration Foundation, and Dial Cordy, Inc.
The eight partial successes that would be acceptable for practical purposes were: (1) Diver and mosaic analyst disagreed on definition of a colony. (2) All colony types agreed between diver and mosaic except for one attached fragment labeled as a lose fragment. (3) GPS and mosaic agreed; the difference with the divers may be explained by different areas sampled. (4) Some errors were not exactly 0, but were very small. (5) Divers were better for only the small, flat targets, mosaics were better otherwise. (6) Some errors were not exactly 0, but were very small. (7) One of the divers differed from 0, but mosaics were successful. (7) The GPS differed from 0 but mosaics were successful.