For purposes of design of new structures or safety assessment of existing structures, extreme effects need to be assessed, for example through calculation of characteristic values of high quantile. Indeed, these values make it possible to determinate stresses for extreme events (extreme weather impacts like temperatures or rain, but also extreme traffic events). This paper focuses on modeling and estimating tail parameters of bridge traffic load effects by generalized Pareto distribution (GPD) to obtain these characteristic values. Its main advantage is that it makes use of all relevant data on the high load effect induced by the passage of traffic, not just the daily maxima used for generalized extreme value (GEV) distribution. However, two critical problems of parameter estimation and threshold selection complicate the use of GPD in practice. A Kolmogorov-Smirnov (KS) statistic based threshold selection method is used to choose the optimal threshold, and the maximum likelihood estimator is adopted to estimate the parameters in this study. Results show that the GPD method is a theoretically well supported technique for fitting a parametric distribution to the tail of an unknown underlying distribution.