2.5.2 Early warning signals: What can we learn from social-ecological models?

Social-ecological models often consist of existing classical ecological models coupled to a human system where the population size, the behaviour of individuals in the population or both are represented as state variables (Figure 2.5.1). In most cases, the rate of harvesting or pollution is a function of the state of the human system and the evolution of this state is determined by a number of feedbacks, some of which are described below. Social norms work to either reinforce dominant behaviours or encourage sustainable behaviour through incentives or sanctions imposed on defectors. Rarity-based conservation occurs when public support for conservation increases as the natural system approaches collapse. Conservation cost represents the effort, financial or otherwise, required to interact with the ecological system in a sustainable manner.

Many models in the literature, including generalised resource models (Sigdel et al., 2019; Bieg et al., 2017; Lade et al., 2013), forest-cover models (Bauch et al., 2016; Innes et al., 2013), a grassland model (Thampi et al., 2019) and a fishery model (Horan et al., 2011) have directly compared traditional ecological models to their coupled social-ecological counterparts. In all cases, the addition of a coupled social system leads to more alternative stable states, and in turn a greater number of tipping points, which are not present in the uncoupled model. 

The increased propensity for these coupled systems to abruptly transition motivates the necessity of tools that can give sufficient warning to these tipping events so that actions can be taken to mitigate potential catastrophes. These tools, known as early warning signals, typically look at statistical signatures in time series data which exhibit significant trends as a tipping point is approached (Dakos et al., 2012). The ambiguity in the transitions that early warning signals herald, paired with a muting of the strength of these signals, provide a unique challenge in the prediction of tipping points that may occur in social-ecological systems. However, there has been some work done in the modelling literature comparing the strength of early warning signals between the time series of state or auxiliary variables in social-ecological models. These studies have found early warning signals in the social time series data to be the only reliable indicators of the system approaching a tipping point (Bauch et al., 2016; Richter & Dakos, 2015; Lade et al., 2013). These data range from fraction of conservationists to average profits by resource harvesters and catch per unit effort. This suggests great potential for the monitoring of ecological resilience through analysing socio-economic data, which fortunately is much easier to gather and is already more frequently generated than ecological data (Hicks et al., 2016).

Economic time series allow for straightforward monitoring of profits tied to resource extraction and the use of early warning signals on previous financial tipping points (Figure 2.5.2) shows promise for use of this data social-ecological systems. This is especially pertinent as financial tipping points will be exacerbated in the future both by climate change risks and mitigation (see Chapter 2.3.6 on financial tipping points). A caveat is warranted, though. Financial systems do not act like other social systems as the constituent actors in the system are themselves trying to predict its future, and act based on those predictions, potentially affecting predictability.

Sentiment analysis of social media data – for example the number of tweets in a given area raising concern over an exploited resource – can give estimates of the fraction of conservationists that have stakes in the social-ecological system. Additionally, citizen science generates not only ecological data, but social metadata through the number of engaged users monitoring specific areas. Using existing infrastructure such as CitSci.org, we have the ability to use this data as a proxy for trends in conservationists (Wang et al., 2015). This approach allows for the deployment of generic real-time monitoring of ecological systems with existing data without requiring extensive knowledge or models of the system. Social data over longer timescales may also provide valuable resilience indicators, as seen in archeological data using variance in settlement size as a reliable indicator for societal collapse under environmental forcing (Spielmann et al., 2016).