MS
Teams
(please
email
amgrad@uwaterloo.ca
for
the
meeting
link)
Candidate
Yazdan Babazadeh Maghsoodlo | Applied Mathematics, University of Waterloo
Title
The effect of different types of noise on early warning signals of climate tipping points
Abstract
The primary emphasis of our research is identifying the presence of tipping points in climate models. In so many dynamical systems, tipping points happen where slowly changing foreign conditions lead to a sudden shift to a new and sometimes very different state. When the system is close to the tipping points, its behavior can be described with a simpler dynamical model named” normal forms” that shows aspects of the new state of the system, like its stability. Indicators like increasing lag-1 auto-correlation and variance provide early warning signals (EWS) of the tipping point by measuring how dynamics slow down. But they do not reveal the nature of the new state. Here I have developed a deep learning algorithm that provides EWS in systems, using normal forms as training data. This algorithm can provide early warning signals more efficiently and also predict the type of bifurcation that the system is going to experience. Moreover, the climate data is mostly distributed on red noise, which is a subject that has not been studied so far. So, in this research, I aim to study the effect of using different types of noises in the time series, on the functionality and accuracy of the model. I used white noise and red noise with different redness values and studied the changes that happen.