Session Chair: Peter Murphy, National Oceanic and Atmospheric Administration (NOAA) Marine Debris Program

This session will advance the efforts for the detection, quantification and prioritization of marine debris in shoreline and nearshore environments, including techniques for application of emerging technologies such as UAS and automated post-processing.

Better understanding of the amount, composition and location of marine debris has important applications in improved understanding of sources, impact pathways, and optimal solutions both by prevention and removal.

Remote sensing, here meaning primarily aerial survey, has proven capacity to provide high-value products that aid in assessing the presence, concentration and composition of debris in multiple environments. The use of georeferenced photo survey of shorelines to quantify marine debris goes back decades, but is continuing to evolve; integrating new technologies and techniques to provide more capable and flexible products and tools. Likewise, the marine debris community is also working to integrate new platforms including UAS (Unmanned Aerial Systems) of different sizes and capabilities to answer multiple debris questions. Regardless of platform, surveys create huge amounts of data, in imagery, video, or other formats that must be analyzed in a frequently labor-intensive effort. Advancements in post processing techniques, including automated analysis have the capacity to improve the efficiency and accuracy of these efforts.

This session will bring together experts in aerial survey operations as well as post processing techniques to provide observations and lessons learned that can help guide the planning of future aerial surveys; identifying key questions and decisions that will result in the best mix of tools and techniques for the identified need.




Aerial remote sensing for shoreline surveillance: Lessons learned from mapping marine macro-debris throughout the main Hawaiian islands

presenting: Miguel Castrence (Resource Mapping Hawai’i, United States); authors: Miguel Castrence (Resource Mapping Hawai’i, United States)

Remote sensing of marine debris poses numerous technological and logistical challenges. Effective shoreline surveillance requires a consistent protocol that includes synoptic data coverage, high level of detail, rapid data collection and processing, and ease of analysis.

In the fall of 2015, we participated in the first attempt to systematically quantify, categorize and map marine macro-debris across the main Hawaiian islands. This project, led by the state of Hawai’i Department of Land and Natural Resources, was funded by the Ministry of the Environment of Japan via the North Pacific Marine Science Organization (PICES) as well as the Japan Gift Fund to the Pacific Coast states, administered by NOAA’s Marine Debris Program. This resulted in a unique collaboration between the Division of Aquatic Resources (DAR) of DLNR, NOAA’s Marine Debris Program, Resource Mapping Hawai’i (RMH) and the Hawai’i Coral Reef Initiative (HCRI). Our project involved manned-aircraft aerial surveys, photogrammetry, and GIS analysis to study a total of 1,223 miles of coastlines and locate 20,658 individual locations of debris.

We will share the lessons learned from this unique experience and provide recommendations for aerial remote sensing that can be applied to future studies in Hawai’i as well as other regions of the world.


Understanding the strengths and constraints of UAV data for monitoring marine debris.

presenting: Serena Lee (Griffith University, Australia); authors: Serena Lee (Griffith University, Australia), Dan Ware (Griffith University)

Unmanned aerial vehicle (UAV) drone surveys have the capacity to collect data over large spatial domains. Additionally, drones are able to collect data in locations difficult to access. Due to the potential benefit this technology provides, UAV surveys are being deployed to assist collect marine debris data. As UAVs are increasingly employed, it is important to understand the strengths and limitations of the data provided.

In this study we test the capacity of UAV’s to obtain reliable data with respect to marine debris. Two different drones were tested, the Phantom 4 Pro, and Sensefly eBee. By deploying the drones over known beach environments, with known quantity and type of marine debris, sensitivity to flight height (image resolution), debris shape, debris colour, debris size, sun angle, cloud cover, wind speed and background substrate, were evaluated. Additionally, the ability of the Parrot Sequoia multispectral sensor to detect marine debris was evaluated.

The study aims to provide guidance for researchers and coastal managers intending to use UAVs to capture marine debris data. By understanding the strengths and limitations of these data, UAV surveys may be designed to increase efficiency while minimising error. Additionally, the study provides methodology researchers can use to document the accuracy of subsequent UAV surveys. By employing a standardised approach and documenting data error, different UAV studies may be more easily compared. As a consequence, spatial and temporal trends may be better determined.


Use of Unmanned Aerial Vehicles and machine learning tools for efficient beach litter Monitoring

presenting: Cecilia Martin (King Abdullah University of Science and Technology (KAUST), Red Sea Research Center (RSRC), Thuwal, 23955-6900, Saudi Arabia, Saudi Arabia); authors: Cecilia Martin (King Abdullah University of Science and Technology (KAUST), Red Sea Research Center (RSRC), Thuwal, 23955-6900, Saudi Arabia, Saudi Arabia), Stephen Parkes (King Abdullah University of Science and Technology (KAUST), Water Desalination and Reuse Center (WDRC), Thuwal, 23955-6900, Saudi Arabia), Qiannan Zhang (King Abdullah University of Science and Technology (KAUST), Computer, Electrical and Mathematical Science and Engineering Division, Thuwal, 23955-6900, Saudi Arabia), Jie-Wei Chen (King Abdullah University of Science and Technology (KAUST), Computer, Electrical and Mathematical Science and Engineering Division, Thuwal, 23955-6900, Saudi Arabia), Xiangliang Zhang (King Abdullah University of Science and Technology (KAUST), Computer, Electrical and Mathematical Science and Engineering Division, Thuwal, 23955-6900, Saudi Arabia), Matthew F. McCabe (King Abdullah University of Science and Technology (KAUST), Water Desalination and Reuse Center (WDRC), Thuwal, 23955-6900, Saudi Arabia), Carlos M. Duarte (King Abdullah University of Science and Technology (KAUST), Red Sea Research Center (RSRC), Thuwal, 23955-6900, Saudi Arabia)

Beach cast litter assessment at a global scale is challenged by the use of low-efficiency methodologies and incomparable protocols that undermine high-scale data acquisition and estimation comparisons. The implementation of an objective, reproducible and time-saving approach is urged to systematically quantify loads of shore-deposited litter and help resolve the mass balance of marine anthropogenic debris, of which beach cast litter is a significant contributor. Here, we demonstrate the application of a likewise methodology coupling remote sensing and machine learning tools. Beach surveys were conducted along the Red Sea Saudi Arabian coast in March 2017 (n=19 beaches) and litter was recorded through image acquisition from an Unmanned Aerial Vehicle (UAV) allowing a beach coverage of 758 ± 90.7 m2 min-1 (mean ± SD). An automatic processing of the high volume of images is required and made possible by the development of a machine learning tool, where a histogram of oriented gradient (HOG) descriptor and a random forest classifier are employed for object detection and categorization, respectively. Standard visual-census beach surveys and a manual processing of drone images are used as efficiency and accuracy controls of the described methodology. Application of the proposed method resulted in 50x faster beach coverage compared to the visual-census approach and 80% accuracy of the machine learning tool in detecting litter. Random forest classification of debris types in ten categories was proportional to the classification obtained from the visual-census approach, further proving the method reliability. Machine learning tool improvements and further surveys will be performed in the coming months, culminating in the first high-scale study of beach litter in the Red Sea.


Classification of riverine floating debris based on true color images collected by a low cost drone system: Case study from the Citarum River, Indonesia

presenting: Elizabeth C. Atwood (RSS Remote Sensing Solutions GmbH, Germany); authors: Elizabeth C. Atwood (RSS Remote Sensing Solutions GmbH, Germany), Sabela Rodríguez Castaño (RSS Remote Sensing Solutions GmbH), Sarah Piehl (University Bayreuth, Dept. Animal Ecology I), Muhammad Reza Cordova (LIPI Indonesian Insitute of Sciences, Research Centre for Oceanography), Mathias Bochow (University Bayreuth, Dept. Animal Ecology I), Jonas Franke (RSS Remote Sensing Solutions GmbH), Sam Wouthuyzen (LIPI Indonesian Insitute of Sciences, Research Centre for Oceanography), Christian Laforsch (University Bayreuth, Dept. Animal Ecology I), Florian Siegert (RSS Remote Sensing Solutions GmbH)

Rivers have only recently begun to receive more attention as an important contributor to the plastic debris accumulating in our oceans, with an estimated 67% of annual global discharge coming from Asia. In Indonesia, there exists little to no rural municipal waste collection system. This leaves people with the options of either burning their waste or disposing of it in impromptu dump sites, often located near to rivers and thus “washed away” during the rainy season. Plastic debris is not only unsightly and a health hazard, it can contribute to stoppage of drainage canals and unnecessary flooding events. We present a low cost system capable of quantifying the amount of floating riverine debris. The monitoring device consisted of a commercial 3DR Solo drone mounted with a GoPro4 Silver camera and a low distortion lens. Seven locations along the Citarum River, Java, were surveyed. Images were post-processed to orthomosaics of at least 90 m² per site. A semi-automated hierarchical object-based classification scheme was developed using eCognition software to identify debris objects, which included not only plastic but also organic material such as bamboo and coconut hulls. Classifications were validated against a ground truth dataset created through evaluation of the original images by an independent observer. An overall accuracy of >88% was achieved for images with medium to high floating debris concentrations, regardless of water turbidity levels. Overall accuracy was reduced in the presence of abundant floating water plants and over clear water with relatively little debris. Our method can provide environmental groups or agencies with a low cost, relatively easy to operate system that allows quantification of floating debris, thus supporting monitoring activities aimed at flood control and focus of clean-up activities.


Drones for Debris: Utilizing Small Vessel-Launched UAS for Remote Coastal Surveys

presenting: Todd Van Epps (Oceans Unmanned, Inc., United States); authors: Matthew Pickett (Oceans Unmanned, Inc., United States), Todd Van Epps (Oceans Unmanned. Inc.)

While marine debris beach cleanups along accessible beaches and coastlines provide a great opportunity for easy success and community involvement, two of the greater challenges associated with marine debris detection and removal are at-sea identification and interception and shore-cast marine debris accumulated along remote, inaccessible shallow water habitats and beaches.

For the challenge of marine debris and detection along remote, rugged shorelines, small, consumer off-the-shelf Unmanned Aircraft Systems (UAS), or drones, provide a powerful tool to either pre-survey, or real-time survey these locations to target response and cleanup effort and reduce exposure to risky access thereby reducing costs and increasing safety. Stable vessel-launch and recovery capability are now possible with almost all small platforms, and combined with autonomous mapping software and online data processing programs, turn any small vessel and inexpensive drone into a powerful search and mapping combination. A variety of sensors, including dual payload capacity on small platforms, provide potential solutions for identifying a wide variety of debris across a broad spectrum of environments.

Over the past five years, we have tested a variety of platforms and payloads for a range of aerial data collection requirements, and in a broad variety of coastal and marine environments. Real world application of anomaly detection software has been utilized with varying results. While not specifically targeting marine debris, the tools and techniques tested, evaluated, and successfully developed are directly applicable to the detection of marine debris in shoreline and nearshore environments and are ready for implementation.


Debris Detection – Efforts & Lessons Learned

presenting: Peter Murphy (NOAA Marine Debris Program, United States); authors: Peter Murphy (NOAA Marine Debris Program, United States)

Marine debris is a global problem that impacts marine and coastal resources in many ways. One of the critical needs to better understand marine debris abundance and impacts is to be able to reliably detect, map and assess the debris. This data can inform impact assessments, but also prioritization and planning for removal. Many different tools and technologies have been used in efforts to detect debris – from satellites high above the earth down to in-situ cameras right near the waterline. Each has its own weaknesses and strengths. This talk will give an overview of the different detection technologies used by the marine debris community to detect, assess and map debris as well as lessons learned in those projects.