mdietze
commented
9 years ago @kevinwolz, spatially explicit is overkill here -- you don't need to know where every tree in the forest is across the whole field since you've got a simple repeating pattern. you just need the model to be aware of the LOCAL pattern, which is much easier (for truly spatially explicit ED is the wrong model)
For dealing with local spatial pattern there's different degrees of complexity you might want here:
1) radiative transfer: there are probably RTMs that can deal with this sort of arrangement without having to resort to a fully implemented 3D ray tracing model (@serbinsh any suggestions?) in which case you could just incorporate that logic into the ED RTM. If not, these systems are simple enough that you could probably work it out. What you want is a way of estimating the light each tree gets without having to do a spatially-explicit calculation for a whole volume of voxels
2) soils: option 1 is to have homogeneous soils and just weight water and nutrient uptake by distance. Option 2, analogous to the 3d RTM is to do 3d soil voxels as well
If you do the 3d voxel route for everything the key would be to set up your patch-level topology to be wrapping (ie a torous) so your explicit calculation is limited to just one pattern repeat
I'd guesstimate that the first approach would be a couple months, while the second voxel approach is a year or so
DanielNScott
fabeit
dlebauer
kevinwolz
ghost
Hi all, My name is Kevin Wolz, and I'm a PhD student at UIUC. I've worked sporadically with a few of you in the past (hey there @mdietze, @ryankelly-uiuc, @serbinsh, @dlebauer), but I'm pretty new to ED and this group.
I come to you from the world of agroforestry. Specifically, my work focuses on mixed-species agroforestry systems, where multiple plant functional types are integrated into a a single system (e.g. canopy trees, shrubs, and ground cover). I utilize several field sites here at Illinois to understand how the transition from corn/soy to food-producing woody systems will effect biogeochemical processes. However, since there are an infinite number of possible agroforestry designs, crops, orientations, etc., field sites are not enough. Models are critical.
A really great review of modeling in mixed-species ag systems (including agroforesty systems) can be found here: http://link.springer.com/article/10.1051%2Fagro%3A2007057. Honestly, this article is quite depressing. This article concludes with: "Today, it is barely feasible to simulate multi-species systems and, due to the absence of efficient models, it is difficult to understand the effects of the different factors that interact within those systems." As developers of the preeminent model that does exactly this, I imagine you would take serious offense to this conclusion.
The truth is that most of the agroforesty models included in that review, as well as most others, have been all but abandoned because they were either overly simplistic (i.e just dealt with competition for light) or were created for very specific systems. The state of the modeling science in that community is pathetic, at best. Most attempts to model agroforesrty systems have started from the agronomy side of things and then tried to layer in ecological complexity. This is not only difficult, but, I think, a complete waste of time since ED already has most of that ecological complexity!
WE NEED ED!
I think ED can become a major asset for the agroforestry community, and my vision is to adapt ED for use in these types of systems. While there are many minor tweaks that would help enable ED for use in agroforestry, there is one critical need: making ED spatially explicit. What I mean by this is that individuals in each PFT need to be arranged in a specific spatial pattern in the landscape. To make this a bit clearer (I hope), the image below shows example systems that we are studying here at Illinois. They represent the "typical" agroforestry system: parallel rows of one or more woody species separated by an "alley" that can be used for row crops, grazing, etc.
So, I'm throwing this out there to this community because, while I am extremely excited about this, I neither have the time nor the skill set to make this happen. I honestly don't even know how monstrous the task I'm proposing is. Is this the scope of a late night's work? A master's thesis? An entire career? Any and all ideas, suggestions, condemnations, or anything else would be deeply appreciated. All of us working in this field are just spinning our wheels on the modeling side of things. I just wanted to break down the wall and open communication with those of you on the other side...
Looking forward to discussing and collaborating!