Statistical Approach to Model the Deep Draft Ships’ Squat in the St. Lawrence Waterway

Abstract

In shallow waterways such as the St. Lawrence River, an accurate prediction of the squat is important to ensure a balance between the security and the efficiency of traffic. The Canadian Coast Guard is now studying the squat phenomenon and considering to reassess the actual underkeel clearance standards of the St. Lawrence Waterway. Hence, a field campaign was conducted with 12 deep draft ship sailings, during which the maximal squat was measured with on-the-fly global positioning system. All the variables that may influence the squat (speed, draught, water level, etc.) were also measured. Twenty of the empirical models that are used in practice to predict the squat were tested and the Canadian Coast Guard recommended to either optimize these models or develop new models. Therefore, statistical approaches to model the squat of deep draft ships that navigate on the St. Lawrence Waterway are proposed in this paper. The Eryuzlu model, which is presently used by the Canadian Coast Guard, was optimized by modeling its errors with a stepwise regression. New models were also developed with the regression tree technique. The performance of the statistical models was better than 10 empirical models that are considered the most suitable to predict the maximal squat in the St. Lawrence Waterway. The models built by regression tree gave the best predictions.