Event Navigation
NOAA Seminar Series: Spatial Resolution Impacts on Remotely Sensed Product Uncertainty and Representativeness
August 28, 2025 1:00 pm - 2:00 pm EDT
Title: Spatial Resolution Impacts on Remotely Sensed Product Uncertainty and Representativeness
Abstract: Satellite remote sensing, while providing broad geographic coverage, faces limitations in spatial resolution for detailed benthic mapping, particularly in coastal regions such as Puerto Rico. Small unmanned aerial systems (UAS) offer a promising solution due to their ability to capture high-resolution imagery with flexibility. This study examines the impact of spatial resolution and classifier training strategies on the accuracy and consistency of benthic habitat classifications derived from drone-based imagery. It determines the optimal airborne sampling parameters ” balancing effective spatial resolution and flight parameters ” when using UAS for marine habitat mapping. Using high-resolution RGB orthomosaics (0.036 m/pixel) collected via UAS and upscaled to coarser resolutions (0.5 m to 10 m), we assessed the classification performance of coral, sand, seagrass, and substrate using Support Vector Machine (SVM) classifiers under four case-study scenarios. Spatial metrics (total area, patch count) and accuracy assessment indicators (self-transition and Critical Success Index) were applied to quantify classification degradation across scales and scenes. Results show fine-scale features, especially coral and seagrass, rapidly degrade beyond 1 meter, while more homogeneous classes, such as sand and substrate, remain relatively stable.
Presenter(s): Biajani Gonzalez, CESSRST II Graduate Fellow
Google Meet :
https://meet.google.com/ugm-keyg-pgrAbstract: Satellite remote sensing, while providing broad geographic coverage, faces limitations in spatial resolution for detailed benthic mapping, particularly in coastal regions such as Puerto Rico. Small unmanned aerial systems (UAS) offer a promising solution due to their ability to capture high-resolution imagery with flexibility. This study examines the impact of spatial resolution and classifier training strategies on the accuracy and consistency of benthic habitat classifications derived from drone-based imagery. It determines the optimal airborne sampling parameters ” balancing effective spatial resolution and flight parameters ” when using UAS for marine habitat mapping. Using high-resolution RGB orthomosaics (0.036 m/pixel) collected via UAS and upscaled to coarser resolutions (0.5 m to 10 m), we assessed the classification performance of coral, sand, seagrass, and substrate using Support Vector Machine (SVM) classifiers under four case-study scenarios. Spatial metrics (total area, patch count) and accuracy assessment indicators (self-transition and Critical Success Index) were applied to quantify classification degradation across scales and scenes. Results show fine-scale features, especially coral and seagrass, rapidly degrade beyond 1 meter, while more homogeneous classes, such as sand and substrate, remain relatively stable.