As a caveat, the statistical data for New Zealand NG does not contain any extra cruft. Just a few fields contribute to the sample space. This means that the stochastic approximations I make do not have as great a bearing on the results as a larger sample would, and deterministic effects consequently have a greater effect. However, the prospect of a "natural gas cliff" has some potentially huge ramifications; the New Zealand case study has proved the best example yet. So, if we model correctly, the cliff will show up -- whether we have determinism or not. I used mean time constants of approximately 6 years for the fallow, build, and maturation phases and also 6 years for the 1/e "half-life" extraction time constant in computing the results. This generated the red curve in the following chart.
But clearly, right around the year 1996, you can see production (in billion of cubic feet) starting to ramp back up. To get this to work out, in the context of the oil shock model, I have to add a strong linear extraction rate increase (see the dotted green line).Assuming no future discoveries, this shortens the half-life to just over a year by the time 2020 rolls around. Like the case of petroleum (i.e. UK North Sea oil), extraction rates have to increase to meet the needs of demand. Since New Zealanders can't get the gas from anywhere else, they basically have to follow the cliff down.