Allow choice of initial period y₀ (e.g. 2025) for Transport

[khaeru]

Currently message_ix_models.model.transport can only produce valid results on scenarios with y₀=2020. Much of the input data and input data processing logic is fixed to this period/year.

To resolve,

1. Identify code changes necessary to support usage with y₀>2020, for instance y₀=2025. This may include:
   • Building on a scenario that already has y₀>2020.
   • Building on a scenario with y₀=2020, such that the transport solution still meets the criteria when y₀ is shifted forward by 1 or more periods, as is common in some workflows.
2. Identify additional data needs or data generation necessary for (1)
3. Implement the above changes.

Until completed, the issue thread can collect info on (1) and (2); and also known issues using .model.transport with y₀>2020 and workarounds.

Note

This was originally opened as iiasa/message_data#561 on 2024-07-04.
On GitHub, “a private repository issue cannot be transferred to a public repository,” so this is a duplicate.

[khaeru]

@yiyi1991 @r-aneeque and I discussed and indentified one issue today in a draft scenario under preparation:

The symptoms:

• In a scenario with y₀=2035, CO₂ emissions remain high from F ROAD technologies (FR_*, f road *) in future periods even with high emission prices.
• The model was only choosing to build and use FR_ICE_M and FR_FCH, and not f road electr which has the lowest emissions.

The cause:

• Currently, the file data/transport/R12/act-non_ldv.csv is used to fix initial activity (bound_activity_{lo,up}) of some transport technologies.
  • This file contains data for period 2020 only.
  • For f road electr in particular, this ensured that there was some small, non-zero activity in the initial period.
  • After this, the ACT (and thus CAP, CAP_NEW) of this technology could grow, as constrained by growth_{activity,capacity,new_capacity}_up derived from data/transport/constraint-dynamic.csv.
• When .model.transport was built on a scenario with y₀=2020 and then y₀ was advanced to 2035, these bound_activity_{lo,up} constraints no longer had any effect.
• The model thus would not choose to build and operate new vintages of this technology.

A workaround we discussed:

• Take a corresponding, solved scenario with y₀=2020.
• Note the future values of ACT for at least (t=f road electr, y_a=yₓ).
• Copy these entries as bound_activity_lo for the same technology and future period.

In the longer run, the file act-non_ldv.csv and its transformation to bound_activity_{lo,up} will likely be removed. This will happen as bring the representation of non-LDV passenger and freight modes/technologies in line with the LDV mode/technologies. In this representation, the historical_new_capacity and growth_new_capacity_{lo,up} are populated such that new sales (CAP_NEW) in y₀ match empirical data and sufficient stock (CAP) exists to satisfy y₀ demand; this ACT is constrained indirectly.

@yiyi1991 is going to attempt this workaround, and asked that I make this note for your reference, @macflo8.

[yiyi1991]

Take a corresponding, solved scenario with y₀=2020.

Tried this but since f road electr was never picked in 2020, still nothing allowed bound_activity_lo to constrain on.
As a quick solution I duplicated their bound_activity_up (their <- what was prepared in data/transport/R12/act-non_ldv.csv) as bound_activity_lo (locking the y₀). Then the ACT in the following years does show up now.

Let me know if this does not make sense.

(I took a quick look at the function bound_activity() and bound_activity_lo() in model.transport. freight. It seems the act-non_ldv.csv is not designed to serve as a floor showing historical values, only the ceiling. So let me know if the quick solution would be unacceptable.)