EWSmethods: an R package to predict tipping points

Output

One peer reviewed publication (O’Brien et al. 2023), a dedicated website, and stable release on CRAN.

This R package aims to provide tipping point indicators for univariate and multivariate time series via a simple interface.

It supports simple ‘early warning signals’ and machine learning models such as EWSNet to allow real-time monitoring of ecosystems, fisheries, climate systems etc risk of sudden collapse.

These sudden changes can occur through a number of different mechanisms (Figure 1), but EWSmethods provides different methods/indicators to classify dynamics.

Figure 1: Three forms of system dynamics detectable using EWSmethods.

Specifically, it builds on the seminal earlywarnings R package to pull upon multiple data sources and the wider stability literature - e.g. Pimm 1984 & Donohue et al. 2013.

I wrote and maintain this package so please feel free to email me questions directly. Further tutorials can be found at https://duncanobrien.github.io/EWSmethods/.

Figure 1: Three forms of system dynamics detectable using EWSmethods.