With the exploitation of marine oil and gas resources, steel catenary risers (SCR) have been widely used worldwide, which has promoted the development of automatic pipeline welding technology. Root-pass welding is critical in the pipeline welding process, since the resulting backside weld geometry largely affects the corrosion resistance and fatigue properties of the pipeline. STT (surface tension transfer) welding has been one of the main methods of root-pass welding for pipeline, in which the short-circuiting period, peak current period and base current period can be identified clearly. A signal sensing method based on the change in arc voltage during the peak current period (ΔU¯) was proposed to characterize the backside weld pool geometry with the help of heat amount inputted to the pipe in 1 s (J). Unlike pulsed gas metal arc welding (GMAW-P), the peak arc force of the short-circuiting transfer was not large enough or was not with sufficient time to excite the weld pool oscillation with a large amplitude, further to affect the backside weld height and width to a great extent. Thus, J, which has a great influence on the weld pool volume (that will determine the backside weld pool geometry to some extent when full penetration occurs) in the low current range, was introduced and was included to corresponding models to characterize the backside weld pool geometry jointly with ΔU¯. The results showed that the models were accurate under the welding parameters and conditions in the present work, which could be applied in the subsequent prediction and control of backside weld geometry in the STT welding.