Abstract
Coastal habitats (e.g., salt marshes, tidal flats, seagrass meadows, and mangrove forests) in the Texas Coastal Bend support diverse biological communities and provide critical ecosystem services. However, these ecosystems are sensitive to anthropogenic stressors, climate change, and acute weather events, which can threaten long-term resiliency. Rapid detection of habitat cover is critical for increasing our understanding of local systems and our ability to conserve natural resources. Traditional field monitoring techniques are time and labor-intensive and are often carried out with coarse spatiotemporal resolution. Additionally, commonly available pre-existing products and approaches have limitations due to infrequent observations and /or coarse spatial resolution, making it difficult to detect sudden shifts in habitat coverage. The goal of this study was to create an open-source method derived from freely available repeat datasets for evaluating coastal habitat change over time in the Texas Coastal Bend, an under-studied region. Our work combined ground-truthing data collected from May to August 2022 across two Texas barrier islands with a database of multi-platform remotely sensed data. We used this data to create a habitat classification model that relied on geo-informatics and extreme gradient boosting algorithms. Our highest performing algorithm resulted in an 83.6% classification accuracy across novel testing data withheld via nested spatiotemporal cross-validation procedures, with 6 out of 8 habitat classes estimated at > 80% accuracy. These results provide for accurate, rapid habitat assessments and change, as well as provide resource managers with a tool to inform conservation decision-making.