Arctic permafrost coasts make up about one third of the global coastline and are likely to witness some of the most dramatic changes linked to changing environmental conditions in the 21st century. Increasing sea level, warming sea temperatures, longer open water season and increasing open-water area all bear the potential to increase the frequency of storm surges impacting the coasts and inducing coastal erosion. The consequences are obvious to the human eye in the form of threats to community and industry infrastructure, especially on ice-rich coasts like the southern Laptev Sea and the southern Beaufort Sea but they are also palpable in the impact on sediment and nutrient pathways in the nearshore zone. Coastal erosion delivers annually just as much particulate organic carbon to the ocean as rivers do if not more. These large quantities of carbon are released from the coast throughout the thaw season, contrary to arctic rivers which unleash most of their sediment laden during the spring discharge peak. The current knowledge on carbon release from coastal erosion is scarce, despite the availability of these quantitative estimations at the global level. Coastal erosion, ground ice contents, coastal geomorphology vary greatly spatially and make it difficult to estimate the amount of organic carbon being released locally as well as its impact on the nearshore environment. I this study, we show results from two studies conducted in the southern Laptev Sea, along the shore of the Bykovsky Peninsula, and in the southern Beaufort Sea, along the Yukon Coastal Plain and show the challenges associated with the computation of carbon budgets for coastal areas. We selected these two areas because they are both ice-rich, yet of a very different stratigraphic nature, one being made of syngenetic ground ice (Laptev Sea), the other of epigenetic ice (Beaufort Sea). We emphasize the need to consider the whole coastal stratigraphic column, including its subaqueous part, in the computation, as well as the role of coastal thermokarst, the inclusion of ground ice and difficulty in finding adequate geospatial datasets to perform envelope calculations.