A waterfall plot is a cumulative spectral decay plot. It is an approach toward showing both time and frequency-domain behavior in a single graph. Primarily used for characterization of drivers.
Above waterfall display is from a Carver driver, testing performed without a baffle. Although the cavity resonance is pronounced in the regular magnitude/frequency plot, this resonances seems to decay fast. The 2.5 kHz resonance however, although hardy visible in the magnitude/frequency plots but present as a phase discontinuity in the phase plots, is definately visible in this waterfall display. There is also evidence of a long resonance at about 800 Hz. This showed also up as a dip in measurements with baffle. This resonance requires further research.
Above waterfall display is for the same Carver 60" driver but with a baffle of 32 inches. Note the extended time for decay all across the board. The resonances are present at the same locations. The low end of these waterfall displays are probably polluted with noise and measurement/calculation related floors. A non-decaying line is typically an indicator of spurious noise/signal. The above outdoor measurements had some wind interference affecting the lows.
The above plot is from a B&G RD-75 driver measured without baffle. The high end decay is seemingly different than the above Carver driver. Although not mentioned in the literature, there seems to be a resonance at approximately 2kHz, probably similar to the Carver 2.5 kHz resonance. Even the 700/800 resonance seems to be present.
The last waterfall plot in the series is from the same B&*G RD-75 driver with a 62 inch baffle. Similar remarks as with the above waterfall display. The data for these B&G plots were taken with a reduced "wind" factor. Hence the assumption that the low is more accurate displayed than the Carver measurements.