Alexey-Ben meeting 2019-07-08

[ashiklom]

Updates from this week

• FoRTE
  • Some progress on manuscript draft, but still a long ways to go.
• Hector
  • Addressed comments from biome PR (jgcri/hector#317)
  • Productive conversations with Corinne on Wednesay (7/3)

This week

• FoRTE
  • Continue work on manuscript. The path forward for introduction and methods is clear and straightforward, but we should discuss what still needs to be done on the results.
• Hector
  • Background reading + progress on paper outline
• LDRD / PEcAn
  • Progress on paper outline

[ashiklom]

Upcoming deadlines

(An additional section that may be useful for guiding priorities)

• FoRTE paper (intro, methods, results draft; discussion outline) to co-authors -- July 23?
• AGU Abstract -- July 25
• Ecological Applications revision -- August 11
• PNNL Post-grad symposium presentation -- August 14 (presenting; draft August 7?)

[bpbond]

1. Why do some runs die? What are the common characteristics, and why in terms of both ecology (potentially) but model code?
2. What combination of structure & params gets us to observational data? E.g. a box in NPP-LAI space, with each run (last X years) plotted as a point and colored as in Figure 2.
3. Composition/diversity - for example, select runs with a common set of parameters (e.g. from median of draws), highlight those runs in 2 above, and then make a new Figure 3 based on those.

[bpbond]

This week:

• Tighten intro plus ensure logical flow (see PR)
• Entry point: question 2 above, produce figure(s) and discuss
• Rough draft of submodel descriptions

[ashiklom]

End-of-week update

• Tighten intro plus ensure logical flow (see PR) -- Made significant progress on this through these changes. There are still a few sentences to fill in and lots of places where the language should be polished, but I feel pretty good about the overall structure and flow of ideas.
• question 2 above, produce figure(s) and discuss -- The figure is done (and in the current rendered version). No discussion yet.
• Rough draft of submodel descriptions -- The most involved of these, the multiple-scatter RTM, is close to finished. I still need to write up the exact ways in which it's different from the two-stream, but that shouldn't be too much more work since I've already looked at that model in detail elsewhere (including an R implementation that I used for my AGU 2019 presentation).
• BONUS: Why do some runs die? -- I could have sworn this was one of my goals for this week, but I remembered wrong and turns out it wasn't. I spent most of today doing exploratory analysis around this question and was able to figure out that most of the runs I thought were collapsing immediately were actually growing, but very slowly and only just barely, which is a significantly different problem. I was further able to narrow this down to a relatively small subset of variables and processes, but my formatted outputs were missing some output variables necessary to explore this further. So I put together a script that loads all of the output variables from my existing ED runs that should finish over the weekend so I can return to this problem on Monday.
  • As part of this exercise, I also drafted some code that looks at runs with the same sets of parameters, which will be useful for (3).

[bpbond]

Hey, that sounds like really solid progress! Nice. Let me know what/when you'd like me to look at something, provide feedback, or anything else. Really looking forward to seeing the no. 2 figure above.