The seismic signatures of calving events, i.e., calving icequakes, offer an opportunity to examine calving variability with greater precision than is available with other methods. Here using observations from Yahtse Glacier, Alaska, we describe methods to detect, locate, and characterize calving icequakes. We combine these icequake records with a coincident, manually generated record of observed calving events to develop and validate a statistical model through which we can infer iceberg sizes from the properties of calving icequakes. We find that the icequake duration is the single most significant predictor of an iceberg's size. We then apply this model to 18 months of seismic recordings and find elevated iceberg calving flux during the summer and fall and a pronounced lull in calving during midwinter. Calving flux is sensitive to semidiurnal tidal stage. Large calving events are tens of percent more likely during falling and low tides than during rising and high tides, consistent with a view that deeper water has a stabilizing influence on glacier termini. Multiple factors affect the occurrence of mechanical fractures that ultimately lead to iceberg calving. At Yahtse Glacier, seismology allows us to demonstrate that variations in the rate of submarine melt are a dominant control on iceberg calving rates at seasonal timescales. On hourly to daily timescales, tidal modulation of the normal stress against the glacier terminus reveals the nonlinear glacier response to changes in the near‐terminus stress field.