Damage and recovery assessment of vessel grounding injuries on coral reef habitats by use of georeferenced landscape video mosaics

Vessel groundings are a major source of disturbance to coral reefs worldwide. Documenting the extent of damage caused by groundings is a crucial first step in the reef restoration process. Here, we describe the application of a novel survey methodology, landscape video mosaics, to assessment of the damage caused by vessel groundings. Video mosaics, created by merging thousands of video frames, combine quantitative and qualitative aspects of damage assessment and provide a georeferenced, landscape, high-resolution, spatially accurate permanent record of an injury. The scar in a Florida reef impacted by a 49-foot vessel, imaged in 2005 and 2006, covered an area of 150 m2 (total imaged area was >600 m2). The impacted coral community showed limited signs of coral recovery more than 3 years after the initial impact; the cover of corals was still significantly higher in the undamaged areas compared to the scar. However, seagrass colonization of the scar was observed. Finally, no evidence of further physical impacts was documented even when four hurricanes passed near the grounding site in 2005. The video mosaics developed in this study proved to be ideal tools to survey the grounding scars. Mosaics provide a means to collect information on the size of the damage area and the status and trends of the impacted biological communities and provide a permanent visual record of the damage, thereby expanding the quality and diversity of information that can be collected during field surveys. *Corresponding author: E-mail: [email protected] Acknowledgments Funding for this project was provided by the US Department of Defense (SERDP Program, award CS 1333 to R. P. Reid et al.), NOAA’s National Geodetic Survey (award NA06NOS4000184 to D. Lirman), and the Spanish Ministry of Education under the Ramon y Cajal Program (N. Gracias). We thank S. Viehman and G. Piniak for their guidance on selecting the grounding site in Biscayne National Park. This manuscript was improved based on the comments of Dr. Jaffe and two anonymous reviewers. Limnol. Oceanogr.: Methods 8, 2010, 88–97 © 2010, by the American Society of Limnology and Oceanography, Inc. LIMNOLOGY and OCEANOGRAPHY: METHODS DOI 10.4319/lom.2010.8.88 Lirman et al. Reef damage assessment with video mosaics 89 responsible for the damage and retain monetary recoveries that can be used directly for restoration (Precht and Robbart 2006; Shutler et al. 2006). To determine the proper amount of restoration required, a two-stage Natural Resource Damage Assessment (NRDA) is conducted to determine: (1) the “primary” actions needed to return the habitat to its original baseline structure and function and (2) the “compensatory” actions needed to compensate the public for the loss of resources and services until primary restoration is completed (Symons et al. 2006). Central to the NRDA process is the accurate and comprehensive quantification of the damage caused by a vessel on a benthic community. In this study, we describe the application of a novel methodology, landscape video mosaics, that is ideally suited for the quantification of damage caused by vessel groundings on coral reefs as well as subsequent recovery patterns. This methodology can, with limited time in the field, satisfy the crucial initial damage assessment needs that are required for the subsequent recovery of funds form responsible parties as well as establish a visual baseline of the damage against which future recovery can be ascertained. Accurately documenting patterns of physical damage (and subsequent recovery patterns) to benthic habitats can be especially challenging when the spatial extent of injuries exceeds tens of square meters. These large injuries are often too difficult to measure in situ by divers and too small or costly to be quantified effectively using aerial and satellite remote sensing tools. Moreover, in cases where immediate action is required to initiate recovery efforts and avoid secondary damage to the resources, damage assessment needs to be conducted quickly. Landscape mosaics capture data at a scale between diver observations and aerial imagery, thereby providing an ideal approach to assess grounding injuries because (1) the images are recorded close to the seabed (<2 m from the bottom), thus capturing detailed visual information; (2) the resulting mosaics cover large areas of the bottom at scales commensurate with the damage caused by large-vessel groundings; and (3) the imagery needed to document patterns can be collected quickly with an underwater video camera and, optionally, a surface GPS. Using this mosaic-based (or image-based) methodology, the dimensions of the injury caused by the 49foot cabin cruiser Evening Star in December 2002 in the waters of Biscayne National Park, Florida, as well as the condition of the affected benthic community, were documented in 2005. In addition to providing a method for measuring the extent of injuries, landscape mosaics create a spatially accurate map of the distribution and condition of benthic organisms so that patterns of recovery (or further damage) can be more easily assessed than by diver-based methods alone. Repeat mosaics Fig. 1. (A), Superficial damage to a coral colony caused by a ship grounding. (B), Severe reef framework damage caused by a large-vessel grounding. (C), Numbered tiles and painted disks used as ground control points (GPCs) for mosaic creation. (D), Diver version of the Shallow Water Positioning System (SWaPS) used to determine the location of GCPs. The unit integrates a GPS unit and a video camera to provide geotagged images of the bottom. Lirman et al. Reef damage assessment with video mosaics 90 taken over time at the same location can be used to measure changes to a study site without requiring extensive tagging of individual organisms. Gleason et al. (2007) exploited this advantage of mosaics to measure hurricane damage to Acropora palmata populations in the Florida Keys, and Gintert et al. (2009) showed how video mosaics can be used to document the impacts of bleaching on coral colonies in the Bahamas. In the present study, a second mosaic of the same grounding scar was constructed in 2006 to assess patterns of community succession and further damage caused by the passage of four hurricanes (Dennis, Katrina, Rita, and Wilma) during the summer of 2005 (Manzello et al. 2007). The ability to measure distances and benthic cover over time with just an underwater video camera and, optionally (or ideally), a surface GPS receiver makes mosaics an appealing tool for assessing damage and monitoring recovery of vessel grounding scars. As underwater landscape mosaics have not been used previously for this purpose, the overall objective of this effort is to test the utility of mosaics for the application of assessing grounding scars. Specific goals are (1) to show that landscape mosaics are capable of imaging large areas of the seabed efficiently; (2) to document an extension to the established mosaic method (Lirman et al. 2007) to take advantage of GPS input; and (3) to use the video mosaics to document status and trends of coral communities at a Florida reefgrounding site. Materials and procedures Data acquisition—On December 5, 2002, the 49-foot vessel Evening Star ran aground on a hardbottom community dominated by stony and soft corals within the waters of Biscayne National Park, Florida (25°23.332’ N, 80°09.874’ W, 3 m of depth). On May 23, 2005, and again on July 19, 2006, video data of the damaged and surrounding areas was collected using a Sony TRV900 DV camcorder placed in an underwater housing following the methods described by Lirman et al. (2007). The camera operator swam a lawnmower’s pattern of side-by-side strips followed by a similar pattern rotated 90 degrees. A bubble level taped to the back of the camera housing helped the diver keep the camera pointed in a nominally nadir angle. A digital depth gauge was used by the camera operator to keep a consistent depth during the surveys. The time required for a single diver to collect the video used for mosaic creation was <1 h in both years. During the 2006 survey, positional (GPS) information was obtained for the outline of the injury as well as 25 ground control points (GCPs) along the periphery of the scar using the diver platform of the Shallow Water Positioning System (SWaPS) (Fig. 1D). SWaPS consists of an integrated GPS and video system that collects video frames that are individually geotagged. A static GPS base station is established in the vicinity of SWaPS operations to track the detectable GPS satellites in synchrony with the mobile GPS receiver located in the SWaPS platform. Both receivers record the GPS L1 and L2 carrier phases and code ranges every second during operations. Both data files are postprocessed using the KINPOS program as described in Mader (1996). The position of the base station is accurately determined using OPUS, a GPS processing service created by the National Geodetic Survey (http://www.ngs.noaa.gov/OPUS). The SWaPS methodology has been previously used to document the position of objects underwater with submeter accuracy (Lirman et al. 2008). The GPS tracks recorded by the diver were used to demark the perimeter of the scar, and the area of the scar was then computed from the polygon delimited by the scar perimeter using linear distances between GCPs. Positions of the 25 GCPs, identified using numbered ceramic tiles and painted disks easily visible in the video (Fig. 1C), were captured by the SWaPS platform and used for mosaic creation. The deployment of the tiles used to establish the position of the GCPs as well as the SWaPS survey took a single operator <1 h. Mosaic creation—Three mosaics were created in total. One mosaic each from 2005 and 2006 used the algorithm described in detail by Gracias et al. (2003), Negahdaripour and Madjidi (2003), and Lirman et al. (2007), which is called in this study the “image-only method.” A third mosaic was created from the same raw video data acquired in 2006, but incorporating the SWaPS ground control points into the image registration algorithm. The differences between the image-only and the image-plus-GPC methods are summarized in Fig. 2 and outlined below. The two mosaics created with the image-only method were used to assess the status and trends of the benthic community between 2005 and 2006. The two mosaics created from the 2006 data were used to assess the spatial accuracy of the mosaics. Under the image-only method, the video is processed to estimate the image-to-image motion between pairs of sequential images. This information is used to recreate the camera trajectory. Subsequently, the estimated camera trajectory is refined by estimating motion between nonsequential but overlapping images. To create the final mosaic, contributions from all of the individual, registered frames are blended into a single image (Fig. 2). The image-only method, as described in Lirman et al. (2007), was used here with two improvements. First, the video was preprocessed to remove patterns of strong light intensity on the seabed caused by wave refraction by use of the method detailed by Gracias et al. (2008). Second, an improved blending method was used to create the final mosaic (Gracias et al. 2009). The mosaic creation method presented here assumes the imaged area is essentially flat. The robust image matching technique, however, is able to deal with departures from this assumption, up to the case where the average camera altitude is approximately twice the depth of variations of the sea-floor topography. Variations in altitude and pitch and roll are handled by the image-matching algorithm as changes in scale or planar-perspective projection. The image-plus-GCP method differs from the image-only method in the global optimization step (Fig. 2). Under the Lirman et al. Reef damage assessment with video mosaics 91 image-only method, the cost function that is minimized uses only the image-to-image registration points (Gracias et al. 2003). In contrast, under the image-plus-GCP method, the cost function to be minimized uses terms for both the image-to-image registration points and the image-to-GPS registration points (Ferrer et al. 2007). In both the image-only and image-plus-GCP algorithms, the image registration process estimates the 3D position and orientation of the camera for each image, thus accommodating for changes in altitude and pitch and roll. In addition, the image-plus-GCP algorithm georeferences the mosaic to a world coordinate system (Universal Transverse Mercator Zone 17N, in this case). Therefore, after the blending step, the mosaics created with GPS input are directly exportable to GIS software or Google Earth (Geotiff® and KMZ formats). The use of mosaics to survey the damage caused by groundings shifts the bulk of time needed to complete a diver-based classic damage assessment from the field to the lab. The time required to collect both the video (<1 h) and the GCPs (<1 h) in the field was minimal and easily achieved with one pair of divers. The processing time for the completion of the landscape mosaics ranged from 5 to 10 days. It is important to note, however, that most of the processing steps are automated and therefore require only minimal operator input, so the actual operator time required was only a few hours for each mosaic. More importantly, significant improvements to the mosaic algorithms have been made over the past 3 years, and total processing times for mosaics similar to those presented here are now 1–2 days. The processing time is roughly divided into the following portions: sunflickering removal (33% of the time [Gracias et al. 2008]), global matching (64%), optimization (1%), and blending (2%). For the 2006 mosaic, documenting the position of the GCPs from the geotagged video took approximately 3 h. Spatial accuracy of mosaics—To provide an independent method to evaluate the accuracy of the GPS locations for the GCPs obtained with SWaPS, as well as a way to assess the spatial accuracy of the video mosaics, the distance between adjacent GCPs was measured by divers using flexible underwater tapes. The distances measured by divers were compared to the same distances obtained independently from the GPS data as well as from the mosaics (Lirman et al. 2007). The accuracy of the distance measurements extracted from the GPS informaRemoval of Sun Flicker Image Matching Removal of Sun Flicker