Abstract: For long baseline underwater acoustic localization scenarios, a method based on the horizontal dilution of precision (HDOP) is proposed for selecting array elements. This method avoids the participation of array elements with large time-delay estimation errors in the solving process of localization, and adjusts the selection strategy timely based on the localization results. First, the combination of array elements with the smallest HDOP are selected to calculate the target location, and then the Kalman filter is used to predict the target location. If the prediction value of the Kalman filter differs significantly from the calculated value, the combination of array elements with the second-smallest HDOP is selected to resolve the problem. This process is repeated until the difference between the calculated value and the predicted value is less than a certain threshold. The optimal combination of array elements and the target localization results are then obtained. The processing results of the data measured on the lake show that using the intersection and Gaussian iteration solution reduce the localization error of the proposed method by 51.0% and 22.2% respectively, compared to the fixed element selection method. This method effectively improves the accuracy of long-baseline underwater acoustic localization.