Data questions

  1. What are the most relevant and appropriate datasets for early warning of social tipping points? As outlined in this chapter, social tipping points are more complex than environmental tipping points due to the interacting relationships between climate parameters and social responses. Given this complexity, there is a need to identify relevant data sources that can be used to detect and anticipate tipping points. Moving forward, it would also be useful to explore datasets that can predict endogenous social tipping, as opposed to predictable events stemming from primarily environmental issues. Recent advances in remote sensing and Earth observation, machine learning and deep learning, and increasing social data from social networks all offer an unprecedented opportunity to understand early warning signals for social tipping points. In this chapter we outlined a handful of use cases, but additional research is needed to fully unpack the potential of these emerging datasets. Once datasets are identified, ensuring that these are accessible and usable for analysis is highly important. For instance, data from social media which could be used for detecting tipping points are often only available at a cost, rendering them inaccessible. Moving forward, it will be important to consider sharing platforms to ensure access to critically important datasets.
  2. What are the characteristics of datasets that can render them more (or less) useful for detecting social tipping points? A key practical question for tipping point analysis is whether there are specific characteristics that make datasets more appropriate for detection of critical transitions. Early warning of tipping points ultimately depends on reliable, high-frequency data. For example, in an analysis of data requirements for early warning of food security tipping points, Krishnamurthy, Choularton and Kareiva (2020 highlighted the importance of temporal resolution over spatial resolution in order to detect autocorrelation or flickering in coupled climate-food systems. A long historical database (with at least 30 years of data) is also preferred as it can help determine climatology and anomalies that could lead to tipping. However, research has shown that even limited datasets such as SMAP soil moisture (available since 2015) can provide transformative opportunities for detecting food security transitions (Krishnamurthy et al., 2022).
  3. Which early warning signals (autocorrelation, variance, skewness, threshold exceedance) are more meaningful for different applications? Identifying the most useful metrics and statistics for early warning of tipping points translates to actionable information. For instance, recent work has shown that increased autocorrelation and variance can detect transitions in managed vegetation systems (Fernandez-Gimenez et al., 2017). In food security applications, too, autocorrelation is the key metric used to detect a transition in food security states, with the rolling average statistic indicating the direction of the transition (Krishnamurthy et al., 2022). Such insights can help leverage resources in a timely fashion to avert negative effects associated with social systems that exhibit tipping points. 
Bezos Earth Fund University of Exeter logo
Earth Commission Systems Change Lab logo Systemiq logo
Global Tipping Points logo
Share this content