Abstract: Based on a statistics analysis of the experimental data of trigonometric leveling networks at four extra-long bridges such as the Humen bridge, etc., comparisons are made in forward and backward altitude differences, loop closures, the third-and the fourth-grade-leveling surveys difference limitations, and the altitude differences between trigonometric height survey and order leveling at common points. It is verified that in networks of rivers, using trigonometric height survey to replace the fourth-grade-leveling is completely feasible for the leveling measurement of crossing rivers, but when using it to replace the third-grade-leveling, the reliability is only about 75%.