Abstract. Tidal marshes, mangrove forests, and seagrass meadows are important global carbon (C) sinks, commonly referred to as coastal blue carbon. However, these ecosystems are rapidly declining with little understanding of what drives the magnitude and variability of C associated with them, making strategic and effective management of blue C stocks challenging. In this study, our aims were threefold: (1) identify ecological, geomorphological, and anthropogenic variables associated with C stock variability in blue C ecosystems; (2) create a predictive model of blue C stocks; and, (3) map regional blue C stock magnitude and variability. We had the unique opportunity of using a high-spatial-density C stock dataset from 96 blue C ecosystems across the state of Victoria, Australia, integrated with spatially explicit environmental data to reach these aims. We used an information theoretic approach to create, average, validate, and select the best general linear mixed effects model for predicting C stocks across the state. Ecological drivers (i.e. ecosystem type or dominant species/ecological vegetation class) best explained variability in C stocks, relative to geomorphological and anthropogenic drivers. Of the geomorphological variables, distance to coast, distance to freshwater, and slope best explained C stock variability. Anthropogenic variables were of least importance. We estimated over 2.31 million Mg C stored in the top 30 cm of sediment in coastal blue C ecosystems in Victoria, 88 % of which was contained within four major coastal areas due to the extent of blue C ecosystems ($\sim$ 87 % of total blue C ecosystem area). Regionally, these data can inform conservation management, paired with assessment of other ecosystem services, by enabling identification of hotspots for protection and key locations for restoration efforts. Globally, these methods can be applied to identify relationships between environmental drivers and C stocks to produce predictive C stock models at scales relevant for resource management.