Alexey-Ben meeting 2018-10-16

[ashiklom]

Goals from last week:

• Submit Frontiers paper (remote sensing and global change experiments) revisions -- done.
• Revise multivariate traits paper -- ~6 hours -- didn't work on.
• FoRTE
  • Background reading -- Read Biogeosciences modeling paper, but nothing else. Added lots of notes to paper outline. Need to do more reading this week.
  • Paper outline -- Added notes based on BGS paper.
• PEcAn
  • Test model runs at UMBS and SERC -- No model runs at UMBS/SERC yet, but have SIPNET and ED working at existing sites (had to do some additional debugging). Should be able to start doing runs at these sites this week.
• Hector
  • Run different scenarios. Plot inputs and outputs to understand what's going on -- done.
  • Parameter sensitivities for each vegetation parameter x each scenario -- not done, but the pieces are in place to do this.
• Other time use:
  • Ghassem paper review (~4 hours)

Goals for this week:

• GCAM community meeting -- ~8 hours (+/- depending on how relevant the conversations are)
• FoRTE
  • Background reading and paper outline -- ~8 hours
• PEcAn -- ~16 hours total
  • Short (several year) ED and SIPNET runs at UMBS and SERC
  • Make sure meteorology data download/workflow works
  • FIA site initial conditions data for ED
  • (Time permitting) PEcAn code hardening and improvements
  • Submit PEcAn runs via curl (make it easier to do an entirely script-based workflow)
  • Add more model Docker containers -- start with DALEC, MAESPA because they should be easy.
  • Test and harden remote execution workflow (e.g. for running models on PIC)
• Hector -- ~8 hours
  • Work towards calibration against historical data (AMERIFLUX, MODIS NPP)
  • Run against historical (observed) emissions -- already doable?
  • Extract NPP and respiration outputs (can use to calculate GPP?)
  • Write parameter calibration function and try to calibrate against observed global temperature data (e.g. from IRI Climate Data Library)
    • Touch base with Robert and Kalyn to see what they're doing. Try to avoid duplicated effort.
  • Obtain global NPP data from FLUXNET / MODIS -- need to find an efficient way to aggregate this data; it's probably huge! For MODIS, maybe AppEEARS?

[ashiklom]

The GCAM meeting DOE update from Gary Geernaert mentioned some upcoming workshops (slide 7 therein). There were a few on that list that it might be worthwhile for me to attend:

• "Disturbance and vegetation dynamics in ESMs" -- March 15-16 in Gaithersburg, MD
• Data assimilation in models -- April 9-10 in Rockville, MD
• Cyberinfrastructure -- April 30 in "Bolger" (Bolger Center in Potomac, MD?)

What do you think?

[ashiklom]

Also, bigger picture -- as I was watching Leon's presentation, it occurred to me that there was no direct mention of biodiversity or ecology anywhere. From a 5-10-20 year perspective, I could envision myself carving out a niche in bringing biodiversity into the context of integrated assessment. A few thoughts I jotted down while listening:

Why is it necessary? (a) Because we can't accurately model what's going to happen to the terrestrial carbon sink without considering biodiversity/ecological interactions. This is a conclusion that I expect will emerge out of the FoRTE modeling work (and already has substantial support from FASET, and more generally lots of literature on biodiversity-ecosystem productivity relationship over the last few decades). (b) Even if we didn't need to account for biodiversity to model carbon, there are significant socioeconomic consequences related to biodiversity and related ecological processes. Case in point -- ecosystem services value estimates are wildly uncertain, but IIRC even even conservative estimates place them in the trillions of dollars globally.

How is it possible to couple complex, ungainly ecological models with integrated assessment models? (a) By improving modeling cyberinfrastructure (PEcAn) (b) By interrogating the modeling process -- sensitivity analysis, uncertainty analysis, benchmarking, data assimilation. Using these tools to fully understand the trade-offs associated with increasing/decreasing model complexity, and then reducing the models down to a level of complexity tractable for integrated assessment while minimizing negative consequences. (Again, this fits PEcAn nicely).

Based on the DOE higher ups' presentations, I think these ideas also check off multiple DOE priorities, which is always good. Both terrestrial and marine biodiversity also feature prominently in the UN's Sustainable Development Goals (even independently from their known impacts on other goals related to climate change, public health, and economic well-being).

Time permitting, perhaps we could discuss? These ideas obviously aren't super well formed or anything, but I think it's useful to keep a long-term vision for my career in mind as we're starting to develop papers and plan projects in detail.

[ashiklom]

Another important discussion topic:

The more I dive into PEcAn for trying to do FoRTE-related runs, the more I see that there is at the very least room for, and possibly a need for, major revisions. There is a lot about the current system that is really fragile, poorly designed, and/or flat-out broken. I am eager to fix it, but reluctant to keep draining FoRTE funding to do so, because I know that I can stitch together the pieces of PEcAn with my own code (and which are working) to get the required runs done much faster (like I ended up doing for my dissertation).

Given the DOE's explicit interest in cyberinfrastructure, especially in the context of Earth System Modeling, what do you think about me trying to put together an LDRD proposal for, say, 100-200 hours (not sure about the scale of these projects) of my time to work on "PEcAn 2.0"? Would be interested to hear your thoughts.

[ashiklom]

A few more notes re: integrated assessment of biodiversity:

This paper uses the framework of "Human appropriation of net primary production" (HANPP) to effectively track the NPP consequences of "distant consumption" resulting from international trade/global supply chains. A few ideas: (a) Tracking stuff through supply chains seems like something GCAM should be good at exploring; (b) HANPP definition fundamentally depends on difference between "actual NPP" (after human land use) and "potential NPP" (what the un-transformed/unmanaged land NPP could have been). The main research objectives of FoRTE can be directly tied to the potential NPP calculation -- namely, if aging forests are more productive than we expect, then their potential NPP is higher (especially integrated through time), and therefore the HANPP impacts of many land use change activities will be much higher.

That same paper also notes that many ecosystem services can be tied directly to NPP, which provides a roadmap for a naive first-pass at the economic impacts of NPP decline (though we may have to think about how to separate the indirect climate effects of NPP decline -- i.e. Hector -- from more direct consequences of NPP decline, which are proxies for all the other services ecosystems provide besides climate mitigation).