Despite the use of automation technology in the maritime industry, human errors are still the typical navigational risk factors in Maritime Autonomous Surface Ships with the third degree of autonomy, as defined by the International Maritime Organization. To analyse these human errors, a prediction model for human errors in the emergency disposal process is present. First, the risk factors are identified by analysing the emergency disposal behaviour process of a Shore Control Centre (SCC) under remote navigation mode. This is followed by the establishment of an event tree model of human errors using Technique for Human Error Rate Prediction (THERP). Furthers, a Bayesian Networks (BNs) model based on the THERP is proposed for the three stages: perception, decision, and execution. Subsequently, expert judgments based on the fuzzy theory are used to obtain the basic probability of root nodes and determine the conditional probability of each node in the BNs. Finally, the probabilities of human errors are calculated for the three stages, while the importance of human error factors is quantified with sensitivity analysis, which can provide flexible references for theoretical construction of the SCC and training of staff.