The airborne transient electromagnetics method has been widely used in geophysical exploration in recent years, but it still faces challenges in balancing the accuracy and efficiency of electromagnetic data interpretation. As far as forward modelling is concerned, the higher the dimension of the geophysical model, the higher the accuracy of data interpretation, but, correspondingly, the more computing resources need to be consumed, which will greatly reduce investigation efficiency and practicability. In this paper, the spectral element method is first introduced for solving the 2.5-dimensional forward modelling of the airborne transient electromagnetic system, which has a smaller computing scale than three-dimensional modelling and is closer to the actual geological structure than either one-or two-dimensional modelling. In the forward algorithm, a non-uniform quadrilateral structured mesh is adopted to simplify the computing scale, and the Talbot algorithm rather than Gaver-Stefest algorithm is applied to the inverse Laplace transform to improve the numerical precision of this conversion. Moreover, we use parallel computing technology to improve the algorithm efficiency while keeping satisfactory accuracy. The study shows that, whether a low-resistivity or high-resistivity layered geophysical model, the numerical solutions of the proposed spectral element method forward algorithm agree well with the analytical solutions of the corresponding models; furthermore, the key factors affecting the accuracy of the numerical solution are analysed by experiments. Finally, we successfully applied it to the 2.5-dimensional geoelectric model simulation.