Dredging and related activities are common across nearshore marine environments, potentially threatening nearby ecosystems. Regulatory frameworks are essential for minimizing environmental impacts, yet rely heavily on a sound understanding of how ecosystems will respond to environmental stressors, and the thresholds that delineate benign and harmful conditions. Here, we use novel statistical approaches and mathematical tools to account for uncertainty in deriving in situ dose–response relationships and thresholds for the environmental management of dredging, based on estimates of the probability of non-zero mortality of corals during a dredging campaign at Barrow Island, Western Australia. Using modified receiver operating characteristic curves, we derive thresholds with explicit Type I and Type II errors rates, across the full range of primary stress pathways and exposure dimensions (intensity, frequency and duration). Monitoring coral health and mortality can be expensive and water quality indicators are often used to supplement direct receptor monitoring during dredging management. We found strong relationships between coral mortality and a range of water quality exposure metrics, lending support to the use of water quality metrics as management tools for protecting corals during dredging. Metrics based on sediment deposition were more statistically powerful than those based on either light or turbidity, but may be more difficult to implement in practice. Thresholds reflecting aversion to a false sense of security in environmental protection or aversion to the costs of false alarms varied substantially for all exposure metrics examined. Synthesis and applications. Strategies for managing environmental harm under uncertainty are critical to achieving an informed risk-weighted balance between environmental protection outcomes and development costs. Our study demonstrates the complexities of how communities respond to variable environmental exposure across a range of pressures in time and space, and the value of integrating probabilistic approaches in environmental management to account for that complexity and its associated uncertainty. Coral mortality was strongly related to a range of water quality exposure metrics associated with dredging activities, and regulatory thresholds based on water quality can provide a solid and cost-effective foundation for protecting corals during dredging. The probabilistic dose–response relationships and thresholds presented here are the first to be derived from in situ data using dredging related coral mortality and represent a step forward in integrating formal decision science approaches into environmental management.