Re: #reproduciblemodels working group - getting organized #reproduciblemodels


 

Thanks Brandon for setting things up so nicely - I very much recommend
reading the repo readme that Brandon wrote if you haven't aready done
this.

We will adapt the scope of this working group as we go along, but let
me give you the motivation for forming it in the first place.

I am increasingly frustrated with the limitations of conventional LCA
formulations. In particular, the assumed linearity and independence of
parameters in the matrices makes little physical sense. Matrices have
a lot going for them, but I see no reason that we can't get the best
of both worlds: correlated, nonlinear sampling for individual product
systems, placed in a larger linear system. Until we come up with a
giant equilibrium model, this is probably a reasonable step forward.

Here are the kinds of things where I personally would like to use
these product system models:
- Parameterized LCIs in ecoinvent
- A model of auto transport which is adapted to load factors, driving
cycle and context, and which correlates fuel inputs with air emissions
- A model of electricity grid mixes which samples from time series
data, getting a much more accurate picture of uncertainty across the
various generation types

But this idea comes with real questions on transparency,
reproducibility, and ease of understanding. This is why I think we
need some standards, best practices, and checklists. I know that
Brandon has already done a lot here, though as far as I understand
this was focused on a specific form of LCI model, not computer models
for LCI (if that makes sense). So, for example, tests are great, but
rarely help transparency, as most unit tests are focused on small
parts of a model or library.

On Thu, 14 Mar 2019 at 11:08, Brandon Kuczenski <bkuczenski@ucsb.edu> wrote:

OK, I added a basic overview to the main README and added you all as contributors to the reproducibility repo. Please commit with wild abandon.

Regarding a call- I am free after 9am US pacific (I think 5pm CET) Thursday or Friday- not unfortunately Thursday night / Friday morning. Another possibility is to use github issues and comments for discussion. First person to open a new issue gets a cigar.

-Brandon
--
############################
Chris Mutel
Technology Assessment Group, LEA
Paul Scherrer Institut
OHSA D22
5232 Villigen PSI
Switzerland
http://chris.mutel.org
Telefon: +41 56 310 5787
############################

Join hackathon2019@bonsai.groups.io to automatically receive all group messages.