Article by Eric Worrell
Poor quality climate data and oversimplified model assumptions make it impossible to accurately locate major climate events in the future, the Technical University of Munich said.
Not the day after tomorrow: Why we can’t predict the timing of climate tipping points
Munich Industrial University
August 2, 2024A study published in scientific progress Revealing that current uncertainties are too great to accurately predict the exact timing of overturning of key Earth system components such as the Atlantic Meridional Overturning Circulation (AMOC), polar ice caps or tropical rainforests.
…
First, forecasting relies on assumptions about underlying physical mechanisms, as well as assumptions about future human behavior, to extrapolate past data into the future. These assumptions can be oversimplified and lead to significant errors.
Second, there are few long-term, direct observations of the climate system, and the data may not adequately represent the Earth system components in question. Third, historical climate data are incomplete.
…
To illustrate their findings, the authors examined the AMOC, an important system of ocean currents. Previous historical data projections suggested that the collapse could occur between 2025 and 2095. New research shows that the uncertainty is so great that these predictions are unreliable.
…Learn more: https://phys.org/news/2024-08-day-tomorrow-climate.html
Research cited in the article;
Uncertainty is too great to predict the timing of the collapse of major components of the Earth system based on historical data
Maya Ben-Yami, Andreas Mohr, Sebastian Badiani and Nicolas Boles
scientific progress
August 2, 2024
Volume 10 Issue 31DOI: 10.1126/sciadv.adl4841
Abstract
One way to warn of impending critical transitions in Earth system components is to use observations to detect decreases in system stability. It has also been suggested to extrapolate this stability change into the future and predict tipping times. Here, we argue that the uncertainties involved are too high to robustly predict pouring times. We raise concerns about (i) the modeling assumptions used to extrapolate historical results into the future, (ii) the representativeness of single Earth system component time series, and (iii) uncertainties in the observational datasets used and the impact of preprocessing, focusing on non-stationary observation coverage and gap filling. We explore these uncertainties generally, using the Atlantic meridional overturning circulation as an example. We believe that even assuming that a given Earth system component is close to a tipping point, the uncertainty is too great to reliably estimate tipping points by extrapolating historical information.
Learn more: https://www.science.org/doi/10.1126/sciadv.adl4841
One thing that really caught my attention is how sensitive the model is to changes in model input or data processing.
…The most basic assumption in all these methods is that the system in question can tip over for a given force. However, not all systems can be tilted, and since these methods assume tilt, they are prone to false positives. We now apply these three methods to time series produced by linear models without any bifurcation but with the addition of an average trend and with red noise forcing to increase the correlation strength (see Materials and Methods). The first and third methods described above can predict the toppling situation of this system (Figure 1). For such a linear system, the ideal method would give infinite tipping time (i.e., no tipping time). However, the MLE method always predicts finite pouring times, whereas the AC(1) extrapolation method only gives infinite pouring times in about a quarter of the cases. The results of a regression method based on generalized least squares (GLS) designed to account for non-stationary correlated noise did not show a significant decrease in system stability. Still, it provides limited pour time for about half of the cases. …
Learn more: https://www.science.org/doi/10.1126/sciadv.adl4841
Scientists caution that the error could be that models underestimate the risk of tipping points.
But these models are clearly not suitable for guiding public policy.
Trying to infer behavior from a model that contains degrees of freedom that no one knows about, based on incomplete and unreliable data over too short a period of time, a model based on such noisy data, switching to a different statistical method can lead to very different results , that is, this is not science, but superstition. Ouija boards can rival the quality of data provided by such models.
related
