"Abstract

One way to warn of forthcoming critical transitions in Earth system components is using observations to detect declining system stability. It has also been suggested to extrapolate such stability changes into the future and predict tipping times. Here, we argue that the involved uncertainties are too high to robustly predict tipping times. We raise concerns regarding (i) the modeling assumptions underlying any extrapolation of historical results into the future, (ii) the representativeness of individual Earth system component time series, and (iii) the impact of uncertainties and preprocessing of used observational datasets, with focus on nonstationary observational coverage and gap filling. We explore these uncertainties in general and specifically for the example of the Atlantic Meridional Overturning Circulation. We argue that even under the assumption that a given Earth system component has an approaching tipping point, the uncertainties are too large to reliably estimate tipping times by extrapolating historical information."

...

"… The most fundamental assumption made in all these methods is that the system in question can undergo tipping for a given forcing. However, not all systems can undergo tipping, and as the methods assume tipping, they are susceptible to false positives. We now apply the three methods to time series generated by a linear model without any bifurcation but with an added mean trend, forced with red noise that increases in correlation strength (see Materials and Methods). The first and third method above predict tipping for this system (Fig. 1). For such a linear system, the ideal method would give the tipping time as infinite (i.e., no tipping time). However, the MLE method always predicts a finite tipping time, and the AC(1) extrapolation only gives an infinite tipping time for about a quarter of the cases. The generalized least squares (GLS)–based regression method is designed to account for nonstationary correlated noise, and its results do not indicate a notable decrease in system stability. Despite that, it still gives a finite tipping time for about half of the cases. …"

https://www.science.org/doi/10.1126/sciadv.adl4841

So really the models aren't fit for purpose.