Design and re-analysis of offshore structures requires the joint estimation of extreme values for a set of environmental variables, representing so-called long-term and short-term characteristics of the environment, subject to sources of systematic variation including directionality and seasonality. Estimation is complicated by numerous sources of uncertainty, typically including limited sample size and the specification of a number of analysis parameters (such as thresholds for peaks over threshold analysis). In this work, we present a model to estimate joint extremal characteristics of the ocean environment incorporating non-stationary marginal and conditional extreme value analysis, and thorough uncertainty quantification, within a Bayesian framework. The model is used to quantify the joint directional–seasonal structure of extremes waves, winds and currents at a location in the Danish sector of the North Sea.