Jason Fischer1, Robert Hunter2, Justin Chiotti1, Sophia Bonjour3, Matthew R. Acre3 1U.S. Fish and Wildlife Service, Alpena Fish and Wildlife Conservation Office - Detroit River Substation, 28403 Old North Gibraltar Rd. Gibraltar, MI, 48173, USA 2U.S. Geological Survey, Great Lakes Science Center, Lake Erie Biological Station, 380 Huron St., Huron, OH 44839, USA 3U.S. Geological Survey, Columbia Environmental Research Center, 4200 East New Haven Rd., Columbia, Missouri, 65201, USA Side-scan sonar has become a valuable tool for mapping benthic habitat, allowing habitat patches (e.g., substrate and large woody debris) to be mapped and classified with near complete coverage over large spatial extents. However, classifications derived from sonar imagery, like all remote sensing classifications, have inherent uncertainty. Ground-truthing can quantify classification accuracy, however, positional errors add additional uncertainties and reduce confidence in estimated classification accuracy. Thus, we propose an approach to account for positional errors when estimating classification accuracy and propagate uncertainty in patch classification. We provide a worked example of the process and its application to a habitat suitability index. We begin by accounting for the magnitude of positional error of geographic coordinates at each datum of a ground-truthing dataset, where we consider two options 1) sub-setting to retain ground-truthing data with a high degree of confidence they are within corresponding patches and 2) propagating uncertainty in ground-truthing locations as point clouds of possible locations. Multinomial models are then fit to the subset data or location realizations to estimate the accuracy of sonar classifications. Next, the estimated classification accuracies are used to reclassify patch types multiple times, propagating uncertainty in the composition of patch types. Lastly, the sets of reclassified patches are used to calculate multiple realizations of a habitat suitability index to estimate the median and probable area range of suitable habitat within the study extent and areas of interest. To provide a reproducible workflow for propagating classification uncertainties in side-scan sonar imagery and other remote sensing data, our approach is formalized as a set of functions available as an R package, groundTruther. This package provides end users with easy-to-use functions to develop habitat maps accounting for uncertainty to better inform management decisions.
