Dirty ice areas of patchy and discontinuous debris typically occur above the limit of continuous supraglacialdebris cover, and are important both as regions of high melt rates and as the locus of the identified increase indebris cover. Dirty ice ablation is spatially highly variable, and debris thicknesses straddle the effective thicknessthreshold separating enhanced and reduced ablation. The melt response is therefore very sensitive to small changesin debris thickness and extent in this zone. Despite this no previous study has sought to directly quantify ablationdynamics over an area of dirty ice. This may be due to the unsuitability of traditional point ablation stakes whichcannot capture the likely spatial variation in ablation. The Østrem curve which relates debris thickness to ablationhas been well-established by plot based field experiments, although only recently have models been able toreplicate the peak of the curve. However, modelling has suggested that a similar peak in ablation can be achievedby altering the percentage debris cover, although as yet this has not been substantiated with field measurements.Indeed, the existence of very thin, uniform debris layers is questionable, and a discontinuous debris cover is muchmore likely.We present a novel approach to determine distributed ablation, based on photogrammetric processing of re-peat UAV (Unmanned Aerial Vehicle) imagery, involving correcting the second digital elevation model for thehorizontal glacier velocity field, downslope ice movement and the emergence velocity. A high-resolution (approx.0.04 m x 0.04 m) spatially continuous ablation map was created for a one-month period over a 2.7 ha dirty-ice areaon Miage Glacier (Italy). It provides unique insights into ablation beneath continuous and discontinuous debris.We identify several processes that affect surface melt at the local scale, including melt foci along sub-debrisdrainage routes, stream enlargement, and sliding of large clasts. We are able to quantify the ablation for a givenpercentage debris cover and identify the peak in ablation associated with moderate percentage covers. Evolutionof the debris is investigated by detecting change in percentage debris cover between images. Methods are alsodeveloped to link ground-truthed data on debris albedo and clast depth to the UAV imagery, allowing spatialextrapolation of these variables in preparation for future modelling.