Periodogram with significant peak at 0 Hz

Hi,

I'm analysing a time series and trying to make spectra, but either with periodogram or fft functions I still get the most significant peak at 0, which should be incorrect.

The time series comes from the CAmodelForced_edited model, that follows:

function dxy = diffxy(t, xy)

%

%split xy into variables in our equations

%

A = xy(1);

C = xy(2);

%

% define the derivatives of these variables from equations

%

%time t is in kyr

om=0.31416;

k=25;

S=100;

I=0.004;

beta=0.0025;

Adot = I-k*(1+beta*cos(om*t))*A*C;

Cdot = k*(1+beta*cos(om*t))*A*C-S*C;

%i

%return the derivatives in dxy in the right order

%

dxy = [Adot; Cdot];

and the script I am using is this:

options = odeset('RelTol',1e-12);

% integrate the unedited CAmodelForced function

[t,xy] = ode45(@CAmodelForced_edited,[0,2500],[2.0,0.0004],options);

A = xy(:,1); % first column

C = xy(:,2); % second column

A= A - mean(A);

Adt= detrend(A);

L=length(Adt);

Fs=length(Adt)/2500; %sampling rate in kyrs

[PxxA,f] = pwelch(Adt,hann(L),[],[],Fs);

magnitude = abs(PxxA);

figure(1)

plot(f,magnitude)

xlabel('Frequency')

ylabel('Power')

title('Power Spectral Density Estimate')

I am not an expert and I would really appreciate some help!

Thank you,

Arianna

[Robert]

Do you know what frequencies you expect to see? It is easy for frequencies much slower than your sampling rate to appear as a bias when plotting on a linear scale. Try viewing the results with frequency on a log scale.

set(gca, 'XScale', 'log)