Default run infeasibilities in v4.3.5 q35_land_other(LAM_80)

[jansteinhauser]

Hello again!

I've just upgraded to MAgPIE v4.3.5, but sadly I can't seem to get it to run properly. Even with a clean install without any of my project-specific changes the default run stops due to an infeasibility:

WARNING: Infeasible solutions found (2020, 2025, 2030, 2035, 2040, 2045, 2050, 2055, 2060, 2070, 2080, 2090, 2100)!

The lst gives me the following:

An equation in the pre-triangular part of the model cannot be solved because the critical variable is at a bound.
Residual = 3.26475145E-02
Tolerance (Tol_Feas_Tria) = 1.00000000E-08
Equation q35_min_other(LAM_80) turned into simple lower bound on Variable vm_land(LAM_80,other). Bound = 0.120196
Equation q35_other_conversion(LAM_80) turned into simple lower bound on Variable v35_other(LAM_80,ac0). Bound = 0.110616
Variable vm_bv(LAM_80,other,nonforested) is forced to its lower bound = 0
Variable v35_other(LAM_80,acx) is forced to its upper bound = 0.0117809
Variable v35_other(LAM_80,ac25) is forced to its upper bound = 0.000181084
Variable v35_other(LAM_80,ac15) is forced to its upper bound = 0.030151
Variable v35_other(LAM_80,ac5) is forced to its upper bound = 0.000114557
Variable v35_other(LAM_80,ac0) is forced to its current value = 0.110616
Equation q35_bv_other(LAM_80,nonforested) acting as binding =L= has forced the above variables.
Equation q35_land_other(LAM_80) cannot be solved with respect to Variable vm_land(LAM_80,other) due to an active bound. Value = 0.120196  
Residual after last transformation = 0.032648 

R and R packages are on their respective newest versions, GAMS is v35.2.0.

R also gives the following warnings, but those don't seem to be actual issues as none of the files are used:

1: In copy_input(x = file2destination, sourcepath = "input", suffix = low_res,  :  File avl_cropland.cs3 seems to be missing!
2: In copy_input(x = file2destination, sourcepath = "input", suffix = low_res,  :  File avl_cropland_0.5.mz seems to be missing!
3: In copy_input(x = file2destination, sourcepath = "input", suffix = low_res,  :  File f38_historical_share.csv seems to be missing!
4: In copy_input(x = file2destination, sourcepath = "input", suffix = low_res,  :  File f38_regression_cap_share.csv seems to be missing!
5: In copy_input(x = file2destination, sourcepath = "input", suffix = low_res,  :  File f38_fac_req_fao.csv seems to be missing!

Is this a known issue of the current release or something unique to my default runs? Any suggestions what to fix where? Or should I simply revert to v4.3.4 - obviously not my preferred solution?

Thanks and all the best!
Jan

[abhimishr]

Possible solution here : #333 (reply in thread)

---

To replicate the issue, can you confirm if the following fails :

1. Fresh clone of 4.3.5 release
2. Using the default start script

We always make sure that the default runs work as expected before every release. Additionally, if you paste the config.log from your failing runs we can compare it to a known 4.3.5 default run and see what's going on/what changed.

I have a feeling that somehow the land protection policies are playing a role in your infeasibilities but this should not happen for a default run with freshly downloaded data.

[tscheypidi]

about the Miktex issue: this is usally an indication that some latex packages are not installed yet. Miktex has a setting which allows to install these missing packages automatically. Turning that on should solve the problem

[flohump]

Just a quick update on that: In the upcoming MAgPIE 4.4.0 release we introduce a shorter version of the validation.pdf file as default, which is about half in size and pages by including only aggregate crop groups such as cereals, sugar and oil. The shorter version also reduces the time for creating the PDF by more than 50% (rather 15-20 minutes instead of 45 minutes).

[devenazevedo]

Thanks both! Embarrassingly the Miktex issue turned out to just be from Miktex needing to be updated, and now that that is done the validation report compiles fine for me, ending up with 2927 pages. So @jansteinhauser this doesn't seem to be a windows issue. And Florian, that sounds like a great idea! As implied above, I often don't run the validation report due to the time required (on my computer it takes closer to 1.5 hours)

[jansteinhauser]

It's the little things sometimes. :)
Thanks for the update, good to know that it's working for you. I guess I'll have to keep digging for this then.
Thanks again!

[devenazevedo]

Haha indeed! No problem - let me know if I can help by testing anything on my windows machine in the future!

[jansteinhauser]

Since it came up in Wednesday's meeting, a quick update on the 4.4.0 RC: The default run does indeed solve without any infeasibilities on Windows. So, no problem there, it seems. (Still confused about my 51 page report, though. But that's another issue)

[flohump]

Good news and thanks for testing! Please note that the input data for rev4.65 on the RSE server is not yet final. There will be another update replacing the rev4.65 input files. Just to make you aware that you should re-download the data once 4.4.0 is released (Rscript start.R -> download data).