singular model fit

Thanks for answering. 

I have added points at shorter distances to my dataset. Now I can observe a hole effect in the variogram (periodicity due to elevation coming from 1 mountain I suppose). I can fit a variogram on residuals from a linear regression nicely and even fit a gaussian variogram in ln_depth. However, if I try to fit other models the sofware wont converge. 

How can I get around this issue? Does it make sense to limit the range or psill? I can send the new dataset if requested... 

Thanks for your help.

Javier