Geoscience Reference
In-Depth Information
are ot en of interest. Recurrence analysis can be used to detect dif erent types
of such transitions. Applying moving windows along the main diagonal of
the recurrence plot, we divide it into sub-recurrence plots and calculate the
recurrence measures of these sub-plots. In our example we choose a moving
window length of 150 and an overlap of 20%
w = 150;
We then calculate the recurrence rate and transitivity coei cient within these
moving windows (Fig. 5.28).
Movie
5.5
w = 150;
Trans = zeros(length(R)-w,1);
RR = zeros(length(R)-w,1);
for i = 1:w/5:length(R)-w
subR = R(i:i+w,i:i+w);
RR(i) = mean(subR(:));
subA = A(i:i+w,i:i+w);
numTripl = sum(sum(subA * subA));
numClosTria = trace(subA * subA * subA);
Trans(i) = numClosTria/numTripl;
end
plot(t(round(w/2) + (1:w/5:length(RR))), RR(1:w/5:end),...
t(round(w/2) + (1:w/5:length(RR))), Trans(1:w/5:end))
xlabel('Time')
legend('recurrence rate','transitivity coeff',4)
h e results suggest slight changes in the dynamics with respect to recurrence
probability (due to the visible amplitude variations in the time series) and
regularity. For a reliable interpretation of the variations in the recurrence
Fig. 5.28 Recurrence rate and transitivity coei cient for the synthetic data in Fig. 5.26, using
a moving window of 150 data points and an overlap of 20%.
Search WWH ::




Custom Search