Much of the geomorphic work of rivers occurs underwater. As a result, high resolutionquantification of geomorphic change in these submerged areas is important. Currently, to quantify thischange, multiple methods are required to get high resolution data for both the exposed and submergedareas. Remote sensing methods are often limited to the exposed areas due to the challenges imposedby the water, and those remote sensing methods for below the water surface require the collection ofextensive calibration data in-channel, which is time-consuming, labour-intensive, and sometimesprohibitive in dicult-to-access areas. Within this paper, we pioneer a novel approach for quantifyingabove- and below-water geomorphic change using Structure-from-Motion photogrammetry andinvestigate the implications of water surface elevations, refraction correction measures, and thespatial variability of topographic errors. We use two epochs of imagery from a site on the River Teme,Herefordshire, UK, collected using a remotely piloted aircraft system (RPAS) and processed usingStructure-from-Motion (SfM) photogrammetry. For the first time, we show that: (1) Quantification ofsubmerged geomorphic change to levels of accuracy commensurate with exposed areas is possiblewithout the need for calibration data or a dierent method from exposed areas; (2) there is minimaldierence in results produced by dierent refraction correction procedures using predominantlynadir imagery (small angle vs. multi-view), allowing users a choice of software packages/processingcomplexity; (3) improvements to our estimations of water surface elevations are critical for accuratetopographic estimation in submerged areas and can reduce mean elevation error by up to 73%;and (4) we can use machine learning, in the form of multiple linear regressions, and a Gaussian NaïveBayes classifier, based on the relationship between error and 11 independent variables, to generate ahigh resolution, spatially continuous model of geomorphic change in submerged areas, constrained byspatially variable error estimates. Our multiple regression model is capable of explaining up to 54%of magnitude and direction of topographic error, with accuracies of less than 0.04 m. With on-goingtesting and improvements, this machine learning approach has potential for routine application inspatially variable error estimation within the RPAS–SfM workflow.