EWSNet. The Early Warning Signal Network

The EWSNet is built to serve as an indicator in varied domains where a tipping is suspected. It comprises of the fully convolutional blocks (FCN) and the long-short term memory (LSTM) blocks that learn temporal dependencies in the time series. It has been trained using simulated time-series data from nine different models, including biological, ecological, and climate models displaying sudden, smooth, and no transitions.

To process your time series and visualize the predictions of EWSNet, enter the time series values (separated by comma, blankspace or newline) in the textbox below and click Process.
You may also use the buttons provided below to load predefined time series obtained from natural systems, known to exhibit a catastrophic transition.