Ghost cavitation, which is a term describing that cavitation bubbles are generated acoustically, has been hypothesized to occur when the ghost reflected signals from many individual air guns beneath the sea surface produce a pressure that is close to zero in the water above the source array. Ghost cavitation is typically observed some milliseconds after the ghost reflection, and it may last for 5-15 ms, depending on the configuration of the source array. The cavitation process subsequently generates a weak high-frequency signal. To investigate this potential signal model and mechanism, we have performed a dedicated source experiment. We found that the distance between the source strings in a source array is a major factor that influences the amount and strength of the high-frequency signal. By increasing the separation distance from 6.5 to 8 m, we have observed a significant decrease in the high-frequency signal. Further, the amount of ghost cavitation can be reduced by increasing the distance between the guns. Also single sub-arrays may create ghost cavitation sound, of course weaker in signal strength compared with full arrays, in agreement with the model. Conventional air-gun modeling can be used to predict where ghost cavitation can occur. Therefore, in principle, a workflow could be developed to quantify grossly if and how much high-frequency signals could be generated by this mechanism, given the source array configuration, and further change the configuration to reduce to a very minimum the high-frequency signals, if deemed necessary. For an airgun array consisting of two subarrays separated by 6 m and fired at 9 m depth, we found that the high-frequency signals emitted between 1 and 10 kHz were of similar strength to the noise from conventional cargo ships, depending on their size and the vessels' speed.