Simulation steps for Spatially independent, and spatially dependent generated datasets

[Travis-Neptune]

Hey there, I've been working on those requesets for a spatially independent and a spatially dependent set of simulations.

Spatially independent
This one is relatively straight forward, when set with a high nugget, and low de values for the up/down/euc partial sills this results in nice flat elevated semivariograms for flow unconnected, and euclidean distance.
However it does bring up something interesting that I had to scratch my head about for a while. Which I'll discuss below.

Spatially Dependent
This one has admittedly been a little trickier. I think partially because I don't know that I have a 100% full grasp of exactly what Jay is doing under the hood when creating/ combining these tailup/taildown/euclidean _params elements into the simulate step. I've held off fully diving into the bowels of those functions until we talked but if we want more detail I'd like to either do that or speak with the SSN2 squad about what exactly is going on in there. This is mainly due to the fact that the simulation discussion in the vingette is in my opinion pretty sparse. It gives an example, but honestly regardless of what I try in terms of reasonable alternate options for the xxx_type, de, or range parameters (ie the variogram type, partial sill, or range) I end up getting a scaled version of one of two shapes.

(I'm aiming to have a nice little setup of these figs for our chat)

Flow Unconnected
This torg I an get to make a

1. general uptick to roughly the range set in place, and then a minor drop to a pretty flat line out to the end off our Flow-un distance in the james
2. Totally flat line all the way from 0 to end flow-un distance
3. similar to 1 but with a little tick back up at the end

What I still haven't wrapped my head around yet here is the tick down that I'm seeing after reaching that rough range value, not sure if it's an artifact of the ssn object we're working with or an artifact of the simulate method

Euclid

1. totally flat line from 0 to euclid distance
2. nice initial uptick from lowish to our sill at a range roughly dictated by our range choices, then a gentle u shape out to just before the end of our euclid dist

Flow connected
This is where I pick up that head scratch discussion part.
So, this one try as I might, I've not been able to actually create a nice looking traditional semivariogram curve, or even a good horizontal line showing spatial independence. Instead, any combination of things I've tried has lead to some version of a very periodic looking semivariogram. Some with tighter periods, some with wider, but always some periodic shape.

While this still might have something to do with the simulation method that I'd like to research more, I think its more likely that this is due to the actual structure of the ssn object. In some literature I've heard these shapes described as often coming from a temporal seasonality component, but as we're generating our values without any sort of time component, my thought is that the structure isn't temporal, but an artifact of the actual physical layout of the stream network. Moreover, that the big dips in the semivarigram occur at distances where either it's very likely to have major confluences, or very unlikely to have them. I haven't dove in yet to figure out which of those hold true or whether that theory holds any water (pun medium levels of intended) yet as again I could imagine it could get a bit rabbit holey.

Let's discuss on our meeting today and I'll show you what I've been running into as far as all these different permutations of attempts.