gcam-core
gcam-core copied to clipboard
JGCRI
Time Steps & Time of Departure Changes in GCAM 5.4 for MAC
Hi There! I have two questions:
- if it is possible to reduce the time steps simulation from 5 years down to 1?
- And if it is possible to manually change the time of departure of the simulation? for instance right now I believe the base year is 2020; but I was wondering if I could change that and for example making the simulation begin in 2019? Thanks a lot in advance! Filo
Thanks Paul for your answer! Do you think it would be possible if I re-compile the C++ source code? I saw that by doing that I would be able to "add new features or otherwise customize the model" - do you know exactly the type of customization that I could do? or if I could find material that actually talk about this? Thanks a lot again
@filofossati check this paper https://iopscience.iop.org/article/10.1088/1748-9326/ac2965/meta ; "For this study, we modify GCAM v5.1 (Calvin et al 2019) from 5 years to annual time steps and incorporate adaptive expectation into the model."
First off, the final calibration year is 2015, not 2020. One can re-set the final calibration year to a different year, to run in hindcast mode, by modifying the objects under the "Time constants" header in gcamdata/constants.R. The comments within that file can guide you. But there's no way to run 2016 or any later year as the necessary data aren't available for those years. Second, assigning annual timesteps is possible in theory, but if you try to run every single year to 2100, the XML file sizes will be massive and will probably fail to write or properly read in. When this has been done in the past, people typically generate the XML files only to 2030 or so, in order to not run into the file size issue, but then there will be issues of exogenously specified share-weights in years that are not model time periods, which crashes the model. E.g. there are a bunch of technologies with share-weights assigned in the year 2050, and these assignments are scattered in a bunch of files in the inst/extdata/ subfolders, with the string "shrwt" in the filename. There's no real easy way around this; just have to find all of the instances where a year that you don't want to create is specified, and modify the assumption, making sure also that there isn't an interpolation rule ("interp" in the filename) expecting to inherit a value in the given year. So yeah, unless you've got lots of time on your hands, I second what Paul wrote initially!
Thanks a lot for your answer @pkyle! I'll dig into it and let you know if I have additional questions.
I don't think that any of these changes are possible.
On Thu, May 19, 2022 at 5:33 PM Filo @.***> wrote:
Hi There! I have two questions:
- if it is possible to reduce the time steps simulation from 5 years down to 1?
- And if it is possible to manually change the time of departure of the simulation? for instance right now I believe the base year is 2020; but I was wondering if I could change that and for example making the simulation begin in 2019? Thanks a lot in advance! Filo
— Reply to this email directly, view it on GitHub https://github.com/JGCRI/gcam-core/issues/219, or unsubscribe https://github.com/notifications/unsubscribe-auth/AI3X5EDBOQZSJD556BVJSVDVK2XRFANCNFSM5WNQRMPQ . You are receiving this because you are subscribed to this thread.Message ID: @.***>
Its quite a bit of work to modify the last base-year, it requires updating quite a bit of code and data. We're actually working on that now. Also, 2020 is not a good choice for a GCAM base-year, GCAM can;t simulate the unusually dynamics in 2020 properly. See the presentation in the annual meeting about these two things. But as Page mentioned, its very doable to select an earlier base year. Finally, note that moving to 1 year time step does not gain you annual dynamics. There are system behaviors that are not part of GCAM yet that would be missing. So you would not gain as much as you might think.
