Objective 9 of 9 · a quantity proxy, not an LCA

Embodied-carbon proxy

carbon = partitions + envelope + slab. Interior partition length × 2.7 m × 25 kgCO2e/m², boundary perimeter × 2.7 m × 90 kgCO2e/m², floor area × 60 kgCO2e/m². For a fixed programme only the partition term varies, so the proxy rewards layouts that need less internal wall.

What this is not. No operational energy, foundations, roof, services, finishes, windows or end-of-life. Coefficients are conservative mid-range cradle-to-gate figures for stud partitions, masonry cavity walls and a 150 mm RC slab. Compare plans of the same programme only. See docs/CARBON.md.

How much can layout move it?

Across each task's union Pareto front, the spread between the lowest- and highest-carbon plan, next to the fixed envelope + slab term the layout cannot change:

Task	Tier	Floor area (m²)	Fixed envelope+slab (kg)	Front min (kg)	Front max (kg)	Layout swing	Partition share at min
large-001	large	132	21,528	25,578	27,266	6.6%	15.8%
large-002	large	140	20,064	24,924	26,713	7.2%	19.5%
large-003	large	113	18,672	22,587	23,903	5.8%	17.3%
large-004	large	121	20,109	23,754	25,070	5.5%	15.3%
large-005	large	130	22,137	26,727	28,192	5.5%	17.2%
large-006	large	108	19,116	22,086	23,605	6.9%	13.4%
large-007	large	113	17,928	21,776	23,193	6.5%	17.7%
large-008	large	121	21,081	24,659	26,346	6.8%	14.5%
medium-001	medium	116	18,108	21,314	23,103	8.4%	15.0%
medium-002	medium	82	14,397	16,321	17,198	5.4%	11.8%
medium-003	medium	111	18,051	21,089	22,337	5.9%	14.4%
medium-004	medium	71	12,537	14,326	14,900	4.0%	12.5%
medium-005	medium	84	15,489	17,750	18,830	6.1%	12.7%
medium-006	medium	94	15,861	18,527	20,080	8.4%	14.4%
medium-007	medium	77	13,368	15,224	16,304	7.1%	12.2%
medium-008	medium	81	13,593	15,551	16,664	7.2%	12.6%
realistic-2bed	realistic	81	13,593	15,854	16,799	6.0%	14.3%
realistic-3bed-house	realistic	106	17,052	20,832	21,035	1.0%	18.1%
realistic-small-office	realistic	111	17,079	20,387	21,635	6.1%	16.2%
realistic-studio	realistic	39	8,415	9,394	9,968	6.1%	10.4%
small-001	small	60	11,133	12,314	13,327	8.2%	9.6%
small-002	small	67	12,009	13,325	14,237	6.8%	9.9%
small-003	small	60	11,103	12,352	12,959	4.9%	10.1%
small-004	small	63	11,556	12,771	13,446	5.3%	9.5%
small-005	small	64	11,601	12,850	13,862	7.9%	9.7%
small-006	small	70	12,462	14,015	14,690	4.8%	11.1%
small-007	small	57	10,953	12,134	12,877	6.1%	9.7%
small-008	small	50	10,017	11,131	11,435	2.7%	10.0%

The honest message: layout choices move this proxy by single-digit percentages of the three-element total; structure and envelope choices move it by tens of percent. The objective is still worth keeping because it is the only one that penalises gratuitous partitioning, which every other objective is indifferent to or rewards.

Why a partition-length proxy is defensible

Partition quantity is a real, additive material take-off; the coefficient just prices it. Anyone can substitute their own factor from the raw partitionLength stored with every plan.
It is monotone and cannot be gamed by the optimiser except by doing the architecturally sensible thing: aligning walls and avoiding slivers.
It stays out of the way: it is one of five headline objectives, normalised so that the fixed term does not dominate the hypervolume.