Proper coupling with land-use model

[SongminYu]

Three tasks are included in this task

1. Soft-linking OPEN-PROM and MAgPIE until convergence ✅

• Task 7 is improved for this and now the two models run until convergence.
• Three parameters are added to config.json: max_iter, price_tol, and quant_tol.
• In each round, the gap between price/quantity timeseries from current and previous rounds are calculated as the maximum difference in percentage among all year&region pairs.

$$rel(r,t) = |x_k(r,t) - x_{k-1}(r,t)| / max(|x_{k-1}(r,t)|, floor)$$ $$delta_x = max(rel)$$

• When the price and quantity deviation are both below xxx_tol, the coupling is regarded as converged.
• In my test, it takes 4 rounds to reach convergence when price_tol and quant_tol are both set to be 1%.
• When the number of rounds reaches max_iter first, the coupling is stopped too.
• An interrupted coupling run can be restarted with existing_prom_run given, keeping all configs unchanged. But, a failed one cannot be restarted.

2. Coupling emulators of land-use models ✅

Update 2: problems solved

A quick summary of the coverage of MAgPIE and GLOBIOM (emulator):

Emission / quantity	MAgPIE	GLOBIOM non-EU	GLOBIOM EU
Agriculture CH₄ / N₂O (enteric / manure / rice / cropland soils)	✅ by gas	✅ by gas	❌ not covered — agriculture (a separate sector) is not in TOTAL_EMIS
Land-use change CO₂ (deforestation / afforestation / conversion)	✅	✅ by process	🔶 bundled in TOTAL_EMIS (CO₂eq), not process-resolved
Baseline standing-forest sink CO₂ (managed-forest net removal)	✅ (Indirect / Grassi)	❌ no baseline sink (only management increment)	✅ Forest Land remaining Forest Land
Peatland CH₄ / N₂O / CO₂	✅	❌ none	🔶 CO₂ only via Organic_soils; CH₄/N₂O undeterminable (no gas data)
Burning CH₄ / N₂O (savannah / agri-waste)	✅	✅ Savannah / Agricultural Waste Burning	🔶 undeterminable (no gas data)
Biomass supply	✅	⚠️ 5 discrete levels	⚠️ 5 discrete levels
Carbon price	✅ continuous	⚠️ 5 discrete levels	⚠️ 5 discrete levels

To fill in the gaps of coverage of GLOBIOM emulator, two files from PROM-MAgPIE converged-coupling under NPi are added to the input data preparation:

• eu_agriculture_ch4_n2o.csv
• noneu_stock_forest_sink_co2.csv

Besides, since GLOBIOM emulators provides results until 2070. The last 30 years are kept constant as 2070.

Comparison between PROM-MAgPIE converged-coupling under NPi and PROM with GLOBIOM emulator:

The deviations are significant, due to:

• Limited / unresolved GLOBIOM coverage
• Too few biomass-price scenarios → under-determined regression
• Carbon-price mismatch --> overcome by using same GHGScenario from GLOBIOM lookup table (GHG100), but the deviations are still big, so mainly the first two reasons above

Update 1: problems are identified and being solved

• Scope of emissions in GLOBIOM
  • EU27 + UK: No agriculture (CH4 and N2O), includes emissions from land use change + remaining land (CO2eq)
  • ROW: No remaining land emissions, includes agriculture and land use change (CH4, N2O, CO2)
  • Solution: Two missing parts are complemented by MAgPIE runs (converged coupling run with OPEN-PROM under NPi), the CO2eq from EU land use is regarded as pure CO2.
• Carbon price scenarios in GLOBIOM are far lower than our scenarios
  • Solution: Always map their highest, and we use NPi
• GLOBIOM models until 2070.
  • Solution: constant from 2070 to 2100 --> regression is done only for land use emissions, as agriculture emissions is not related to bioenergy demand. Besides, there are unit mistakes when running regression with GLOBIOM data. GLOBIOM uses EJ while OPEN-PROM uses Mtoe. It needs to be corrected in the regression step.
• Swiches for three modes of land-use coupling to be organized (in mrprom, OPEN-PROM, and postprom)
  • Three modes: current (exogenous bio-energy price and AFOLU emissions), soft-link MAgPIE, hard-link land-use emulators (GLOBIOM or MAgPIE)
  • The whole GLOBIOM emulator regression process should be moved to mrprom. It brings iBmswasEmisCoefGLOBIOM.csv and iBmswasSupplyCoefGLOBIOM.csv to the data folder.
  • Solution: logic needs to be smoothed, and new switches to be added in config. One more thing to be fixed: soft-coupling mode final variable names are not consistent with IAMC common definitions.

3. Daily run scipt is fixed ✅

• The script is scripts/tools/daily_run.bat.
• It needs correct paths, but only these. It contains all config information.
• When started, it installs mrprom and postprom main branch first, then run NPi with new data, then research mode for no-carbon-price, 1.5 degree, and 2 degree.

[SongminYu]

Updates:

• In reporting.mif, the AFOLU emissions are not classified in a clean way, e.g., AFOLU|Land|Agriculture should be moved/aggregated into AFOLU|Agriculture.