Problem in using user-defined covariance base class (CovModel)

[himansh78]

I am trying to generate a user-defined spatial random field in 2D (Earthquake slip realisations in my case) using the CovModel base class for Von-Karmen Auto-correlation function (link ). Unfortunately I am stuck at an error message and it is getting difficult to get past it. It would be a great help if you could help me get some insights to solve the issue.

Here is the Covmodel I created-

class VK(gs.CovModel):

    def cor(self, h):
        h = np.array(h, dtype=np.double)
        nu = 0.8
        res= (4*np.pi*nu*(h**nu)*sps.kv(nu, h)) /sps.kv(nu,h.min() )
        return res

I can plot spectral density using the covmodel but cannot create the model using SRF method.

model = VK(dim=2)
slip_srf = gs.SRF(model,seed=10)

I get a TypeError when I run the above commands.- TypeError: 'float' object cannot be interpreted as an integer
I get the same error, even when I create a gaussian covariance model (as given in the tutorial itself).
Thanks a lot for your time. Looking forward to your response.

[MuellerSeb]

Hey there,
thanks for using GSTools.

I see two problems here:

1. the integer error comes from emcee but we fixed this in GSTools v1.3.2. so upgrading GSTools should solve this. (See: Calling gs.Gaussian() with dim=3 leads to TypeError: 'float' object cannot be interpreted as an integer #183)
2. the covariance model given with your link seems to be wrong, since they are dividing by K(v,0), which is infinit and the model would result in a constantly zero one.
   
   The cited reference has a similar definition:
   
   But they use another function with a finite value to normalize the function.
   In the end, the described model is exatly the matern model with a slightly modified rescaling factor. So I would recommend using this predefined model.

Hope that helps!
Cheers, sebastian

[himansh78]

Thanks . that helps.