The existent ecological risk assessment methods for polycyclic aromatic hydrocarbons (PAHs) treat each compound as directly independent of each other, without considering the possible correlations between them. To study the multi-dimensional joint distribution of PAHs with correlations, based on the principle of considering multivariate correlations, a Copula approach with adaptability of marginal distributions and a multivariate dependency model used for constructing the model of PAHs exposure concentrations based on Copula functions, then ecological risk assessments were carried out. The types of ecological risk sources of PAHs were analyzed and evaluated based on the results of the Copula model, enhancing the effectiveness and feasibility. Taking the surface sediments of Taihu Lake as examples and comparing with existing assessment methods, the Copula model shows that it takes into account the correlations between variables and makes the risk evaluation results more accurate compared to the risk quotient (RQ) method. And the results of model evaluation can directly reflect the types of risk sources and the degree of risk of each sample point. The results of model evaluation are more comprehensive, more effective and more feasible.