Address hardcoded 2018 MECS in GHG methods

[catherinebirney]

cc:
Closes: #488

What changed? Why?

• Address the hard coded use of 2018 MECS data when we want the flexibility of using 2018 or 2022 in the GHG FBS methods
• The hardcoded methods relied on EPA_GHGI attribution tables, which needed to be replaced with UMD_GHGIA data
• Modified our approach by moving away from the EIA_MECS_Energy_CEDA_allocaition.yaml FBA - which called on CEDA functions and instead created FBS methods with use the flowsa logic/approach. This does modify how the EIA MECS data is used for attribution. The FBA calculated kg CO₂e weights, which were then used to allocate the GHG data. The FBS does not convert the MECS data to kg co2e, keeping data in original units - still used to proportionally attribute GHG.
• Moving to FBS also allows us to use the Cornerstone_2025 industry spec and makes it clear/easy to follow what data sources are being used for attribution, rather than the buried, hard coded sources.
• We estimate suppressed data and map the MECS sectors to target Cornerstone industry spec. We fully rely on the built in mapping logic that exists in flowsa to align data sets. Previously, CEDA had a specific list of relevant sectors.

Intentionally changed approach from CEDA:

• NEU petroleum product split (HGL / asphalt / remainder): CEDA merged all three into one FBA row (Petroleum, non energy) and nested a single combined EPA bucket. Now we keeps separate activity sets with product-specific MECS to BEA weight pairing.
• F01000 transport lubricants in NEU: We routes a PCE/price-scaled share of petroleum NEU to households (F01000) via neu_transport. CEDA approach was 0 on F01000 in T_3_14. Now +7.3 Mt (2018) and +3.5 Mt (2024) on F01000. CEDA did not assess households - our approach does.
• FBS mapping: Built in flowsa approach of estimating suppressed data and merging datasets on sectors is different than the lists of relevant industry codes identified in CEDA approach - results in some shifts of results between sectors.
• Using UMD_GHGIA data over EPA_GHGI data
• Using 2022 MECS within timeseries GHG methods.

Testing

Total FBS values for 2018 and 2024 are the same as we see on Main. There is a shift in results for T-3-11 and T-3-14.

Compared the results for the 2018 and 2024 GHG Cornerstone FBS

2018 national by NAICS-2 (nonzero diff)

Sector	Main Mt	cb_mecs Mt	Diff Mt	Pct diff
HH / F01000	1408.892	1416.226	+7.334	+0.52
Mfg 32	302.978	298.001	-4.977	-1.64
Transp 48	467.723	463.501	-4.222	-0.90
Construction	180.325	183.035	+2.710	+1.50
Gov GSLG*	405.439	403.655	-1.784	-0.44
Ag	1027.129	1028.149	+1.020	+0.10
Special S00	43.991	43.032	-0.959	-2.18
Mfg 33	74.966	75.810	+0.844	+1.13
Mining	530.969	531.774	+0.805	+0.15
Mfg 31	5.880	5.478	-0.402	-6.84
Transp 49	39.348	38.979	-0.369	-0.94

2018 T_3_11 by NAICS-2 (nonzero diff)

Sector	Main Mt	cb_mecs Mt	Diff Mt	Pct diff
Mfg 32	89.642	86.887	-2.755	-3.07
Construction	165.795	168.505	+2.710	+1.63
Mfg 33	10.123	8.934	-1.188	-11.74
Ag	383.136	384.155	+1.020	+0.27
Mining	188.819	189.624	+0.805	+0.43
Mfg 31	3.789	3.197	-0.592	-15.61
Utilities	0.296	0.296	+0.000	+0.08

2018 T_3_14 by NAICS-2 (nonzero diff)

Sector	Main Mt	cb_mecs Mt	Diff Mt	Pct diff
HH / F01000	0.000	7.334	+7.334	n/a
Transp 48	6.839	2.617	-4.222	-61.73
Mfg 32	68.481	66.259	-2.223	-3.25
Mfg 33	0.432	2.465	+2.033	+470.76
Gov GSLG*	3.312	1.528	-1.784	-53.86
Special S00	1.781	0.822	-0.959	-53.86
Transp 49	0.684	0.316	-0.369	-53.86
Mfg 31	0.000	0.190	+0.190	n/a

2024 national by NAICS-2 (nonzero diff)

Sector	Main Mt	cb_mecs Mt	Diff Mt	Pct diff
Mfg 33	75.240	94.546	+19.306	+25.66
Construction	164.357	153.067	-11.290	-6.87
Ag	992.957	988.709	-4.248	-0.43
HH / F01000	1304.340	1307.816	+3.476	+0.27
Mining	432.673	429.336	-3.337	-0.77
Transp 48	520.127	518.136	-1.990	-0.38
Gov GSLG*	414.719	413.877	-0.842	-0.20
Special S00	43.465	43.012	-0.453	-1.04
Mfg 31	5.422	5.088	-0.335	-6.18
Transp 49	39.217	39.043	-0.174	-0.44
Mfg 32	286.152	286.048	-0.104	-0.04
Utilities	1516.089	1516.082	-0.008	-0.00

2024 T_3_11 by NAICS-2 (nonzero diff)

Sector	Main Mt	cb_mecs Mt	Diff Mt	Pct diff
Mfg 32	71.416	91.606	+20.190	+28.27
Construction	147.087	135.797	-11.290	-7.68
Ag	389.614	385.366	-4.248	-1.09
Mining	171.291	167.939	-3.352	-1.96
Mfg 33	8.065	7.172	-0.893	-11.07
Mfg 31	3.019	2.619	-0.399	-13.23
Utilities	0.309	0.301	-0.008	-2.52

2024 T_3_14 by NAICS-2 (nonzero diff)

Sector	Main Mt	cb_mecs Mt	Diff Mt	Pct diff
Mfg 32	84.352	64.058	-20.294	-24.06
Mfg 33	0.523	20.721	+20.198	+3865.17
HH / F01000	0.000	3.476	+3.476	n/a
Transp 48	3.210	1.220	-1.990	-62.00
Gov GSLG*	1.554	0.712	-0.842	-54.19
Special S00	0.836	0.383	-0.453	-54.19
Transp 49	0.321	0.147	-0.174	-54.19
Mfg 31	0.000	0.064	+0.064	n/a
Mining	5.204	5.218	+0.015	+0.28

[bl-young]

Energy FBS national totals (2022 MECS) — follow-up fix incoming

While building the NEU activity-set pivot for this review, I found a pipeline bug that inflates 2022 MECS energy weights (e.g. Table 2.1 HGL national FBS ≈ 2,176 mmbbl vs MECS 31-33 control ≈ 870 mmbbl, ~2.5×).

Root cause: EIA_MECS_Energy is assigned NAICS_2022_Code for the 2022 survey (source_catalog.yaml). Energy methods target Cornerstone (target_naics_year: 2017), so map_to_sectors calls convert_naics_year (2022 → 2017). That path is correct for NAICS vintage conversion, but the final aggregation step reassigns group_id = df.index on every fan-out row after the Cornerstone crosswalk maps one MECS activity to multiple target sectors. Proportional BEA attribution then treats each fan-out row as an independent full group instead of splitting one group_total.

2018 MECS does not hit this path (source and target are both NAICS 2017, so convert_naics_year is skipped and group_id survives fan-out). National HGL stays ≈ 888 mmbbl.

Separate note: Change in disclosure for some sectors causes significant jumps between MECS years.

There are just a few cases, but data suppression, especially D (which is to avoid revealing confidential information) is different between 2018 and 2022. For example, **32531 has a +10.7 Mt jump in which aligns with 2022 MECS disclosing Table 2.1 NG feedstock for 325311 that 2018 suppresses (D → 0). We may want to consider alternate approaches to handle these cases (which are different than the * which reflect small amounts).

[bl-young]

Also turns out there are some extra blank columns in the 2022 Non-energy tables so the fuels were mis aligned