Evaluating the complementarity of acoustic and satellite remote sensing for seagrass landscape mapping

Accurate maps representing seagrass spatial distribution are essential components for effective monitoring and management of coastal vegetated habitats. Satellite and acoustic remote sensing provide valuable spatial data for seagrass mapping, though few studies have evaluated the complementarity of these methods. In this study, the complementarity of seagrass mapping was assessed through comparison of acoustic and satellite remote-sensing data sets. QuickBird (R) satellite imagery representing the seagrass landscape of the Richibucto estuary, New Brunswick, Canada, was classified through an object-based procedure and evaluated against a single-beam sonar data set. Acoustic percentage cover values were classified into binary presence/absence format through the application of a decision threshold, allowing comparison with satellite data using the error matrix and derived metrics. Though the binary satellite classification resulted in relatively high accuracy compared with independent ground reference data, agreement between satellite and acoustic data sets was limited. Local differences in seagrass prevalence and patchiness affected classification accuracy, highlighting the potential for under-or overestimating seagrass cover when applying bay-scale classification to areas with different landscape structure. These results emphasize the importance of landscape context in seagrass mapping. Satellite and acoustic remote sensing were seen to fundamentally differ in their depiction of the landscape. Comparison of multiple remote-sensing methods allowed for assessment of complementarity as well as ecologically relevant insight to seagrass spatial dynamics, with implications for mapping and monitoring of seagrass habitats.