Figure 5: Persistence contours with the y-axis the start month and the x-axis the months elapsed.
From: A unified nonlinear stochastic time series analysis for climate science
The persistence from the original Nino3 SST index, from 1870 to 2015 (a), and for comparison in (b) we show the associated results from equation 1. Note, even without the inclusion of the long-term forcing f(Ï„), the persistence calculated using the stochastic model is consistent with that from the original Nino3 index data. Therefore, the seasonal variability reflected in the persistence is well approximated solely by the monthly stability a(t) and the noise intensity N(t). In particular, regardless of the starting month, in both figures we see that the persistence drops sharply from March to May.
