Conserving and Restoring Critical Habitat

Seagrasses provide a wide range of ecosystem services, such as nursery habitat for fisheries species, and reducing erosion. The development of drones, or Unoccupied Aircraft Systems (UAS), equipped with high resolution cameras offers new opportunities to improve the efficiency and accuracy of mapping and monitoring seagrass beds over large areas at a low cost compared to historic on-ground methods. So far, UAS have primarily been used in areas with high water clarity and extensive, dense seagrass, providing clear imagery and accurate information on the area of seagrass beds. However in many regions, such as the northern Gulf of Mexico from Mobile Bay west, seagrass occurs as sparse, patchy beds in relatively turbid waters, presenting challenges for the application of UAS methods. We are developing and ground-truthing methods for classifying and quantifying seagrass from UAV imagery in turbid coastal waters. We use semi-supervised classification of UAS imagery to identify the area and extent of apparent seagrass. We then generate stratified random points within each substrate class (seagrass, possible seagrass, bare substrate) to identify georeferenced ground truth points at which quadrats will be placed to quantify the presence and coverage of seagrass in each substrate class. If successful, this research will provide a framework for more efficient and accurate mapping and monitoring of seagrass in turbid coastal waters.

The integration of drones, or small unoccupied aircraft systems (sUAS), into environmental restoration has ushered in a new era of data collection and analysis. Equipped with advanced technologies like real-time kinematic (RTK) GPS and multispectral cameras, drones offer a cost-effective and efficient means of monitoring restoration projects. This study explored the application of drones and open-source workflows for living shoreline monitoring. By leveraging RTK GPS, we significantly improved the accuracy of drone-captured imagery, eliminating the need for ground control points (GCPs) and reducing project costs. Results showed improved accuracy when imagery was collected with onboard RTK GPS (0.05 m mean horizontal accuracy) as opposed to imagery processed with ~ 1 GCP/ha (0.43 m mean horizontal accuracy). Additionally, multispectral cameras enabled us to assess vegetation health and monitor changes in plant species composition, providing valuable insights into the success of restoration efforts. Our open-source workflows, including OpenDroneMap and QGIS, facilitated efficient data processing and analysis, making this approach accessible to a wider range of users. These findings highlight the potential of drone technology and open-source tools to revolutionize living shoreline restoration monitoring. By providing high-quality data at a fraction of the cost of traditional methods, drones offer a valuable tool for restoration ecologists, resource managers, and other stakeholders. As drone technology and open-source software continue to advance, we can expect even greater benefits for environmental restoration efforts. This study highlights how the continued advancement of these workflows creates opportunities to adapt monitoring methods and metrics of success to the multifaceted, often site-specific, nature of living shoreline restoration.

The monitoring of ecosystem parameters is a central component of coastal conservation and restoration projects, allowing managers to create project-specific designs and measure progress towards project goals. However, financial and workforce limitations often limit managers' capacity for the intensive fine-scale measurements required for such monitoring. The recent introduction of the unmanned aerial system (UAS) has allowed coastal researchers to capture data at both fine-scale temporal and large-scale spatial resolutions. For example, it is possible to produce fine-scale digital elevation models (DEMs) using UAS based LiDAR or Structure-from-Motion (SfM) techniques. These DEMs have increased in popularity due to their usefulness in informing modeling and planning efforts for marsh sustainability in relation to climate change and sea-level rise. Previous studies have investigated the accuracy of LiDAR-derived salt marsh DEMs and tested correction methods to reduce observed error. However, LiDAR technology can be expensive and require significant technical expertise. These requirements can limit its accessibility for coastal monitoring. Structure-from-Motion technology presents a more accessible alternative to LiDAR. However, there have been limited studies on the accuracy of and potential correction methods for DEMs produced using SfM techniques. This study investigates the accuracy of, and potential correction methods for, salt marsh DEMs developed using UAS and SfM techniques. The accuracy of these DEMs will be evaluated using real-time kinematic (RTK) elevation and vegetation field measurements. Principal Component Analysis (PCA) will be used to determine which vegetation characteristic variables are influential in DEM error and how those variables are correlated. Potential correction methods will be explored using vegetation species-specific correction factors. Evaluation and potential improvement of SfM derived DEMs will increase their usefulness in coastal management.

Salt marshes are imperiled yet critically important coastal habitats and are subject to extensive restoration efforts. There are many approaches to mapping salt marsh shorelines, but it is not clear how the underlying uncertainties associated with each mapping approach affect estimates of shoreline change. Therefore, researchers and restoration managers cannot identify methods that will produce adequate shoreline change estimates for their specific needs. Furthermore, mapping technologies and data repositories have rapidly improved, but the underlying data uncertainties have been poorly described.
We aim to provide improved methods for mapping salt marsh edge of vegetation shorelines for shoreline change analysis. First, we compare mapping approaches using in-field and publicly available data by characterizing their underlying errors. Next, we model scenarios of marsh shoreline change to examine how shoreline change estimates would differ by mapping approach, under scenarios of low, moderate, or high erosion. Therefore, our methods will specify the tradeoffs between mapping approach and quality of shoreline change estimates, helping end-users identify approaches most suitable for their local erosional environment.