Seismic ground‐motion data provide valuable constraints on explosion characteristics, such as yield and height‐of‐burst/depth‐of‐burial (HOB/DOB). This study investigated a range of seismic amplitude features and their efficacy in minimizing errors associated with yield estimation. Using a set of explosion recordings from experiments conducted in alluvium and sedimentary rock geologies, we investigated the effectiveness of three different seismic feature types over a range of different frequencies. Using both velocity and displacement data, we investigated the zero‐to‐peak (ZTP) amplitude of the first arriving P wave, the peak‐to‐peak (PTP) amplitude of the first arriving P wave, and the amplitude over various time windows starting with the P wave. These three basic features were measured on both vertical component only data and the vertical–radial vector sum. In total, there were 56 different combinations investigated. Our seismic models combine the effects of scaled range and scaled HOB/DOB on the observed seismic features. The results show that the vertical–radial vector sum of the ZTP amplitudes measured on displacement seismograms most consistently produce models with the smallest difference between predicted amplitude values and observed amplitude values when accounting for both alluvium and sedimentary rock lithology. The improvement in fit to the data when incorporating the difference in source lithology is significant. If an alluvium lithology is assumed for a sedimentary rock lithology, the difference in yield could be up to times the appropriate yield in the far field.