How does WAFO select a good threshold through the MRLplot?

[Francisco Fonseca]

I'm doing a POT analysis to a set of significant wave heights data over a 14 years duration for my Master Thesis. When selecting my threshold I've read Coles (2001) and some bibliography about the subject and I don't think I understand how the WAFO routine "reslife" works. 

From what I've understood from Coles, the MRL should be linear for the ideal threshold. My question is: should it be linear to the left or to the right of the selected threshold?

Currently I have something like this:

[HS,time]=decluster(dat_HS,0,Tmin);

dat_HSpot1=[time HS];     

optionsdisp.method='MPS';       

subplot(2,2,3);

%STEP 2: Excess mean over threshold

dat_HSpot1sort=sort(dat_HSpot1(:,2));

mrl=reslife(dat_HSpot1sort);   

plot(mrl,'b');

for io=numel(mrl.data):-1:1

    CI=mrl.dataCI(io:end,:);

    if ~(max(CI(:,1))<=mrl.data(io) && mrl.data(io)<=min(CI(:,2)))

        thres_mrl=mrl.args(io);   % Retira threshold minima por MRL

        break;

    end

end

vline(thres_mrl,'g');

[Francisco Fonseca]

Thank you in advance for the attention and all replies are deeply appreciated.

Francisco

[pab]

The MRL should be linear to the right from the threshold you select.