City.sdst_v1_5 scenario matches baseline before the implementation year

[HughRunyan]

Purpose

City.sdst_v1_5 (the engine behind WasteMAP's /v1/site_emissions/sdst_v1_5)
models the baseline and the scenario as two independent landfills. Before the
scenario's implementation year nothing has changed, so the scenario landfill must
receive exactly the baseline waste — mass and composition — and emit identically
until then. For blank/custom sites it did not: a scenario that changed the waste
composition had the new composition wrongly back-dated onto deposits from before
the implementation year, so baseline and scenario diverged before the change was
even implemented.

Bug

• Component: SWEET_python/city_params.py, City.sdst_v1_5, the blank/custom-site branch (baseline_data=None).
• Reproduction: run a single custom site (no rmi_id) with baseline waste fractions ≠ scenario waste fractions (e.g. the /sdst "Custom Location" flow, where the UI exposes independent baseline and scenario composition columns).
• Expected: for every year before implement_year, scenario total emissions == baseline total emissions.
• Actual: they differed (~6% on the scenario series in a representative dump→landfill case), because the scenario composition was applied to pre-implementation deposits.

Root cause

_generated_waste_masses() splices the scenario total mass at implement_year
but multiplies by the scenario composition for every year. The two
TRACE-reconciled branches (real sites) already correct this:

waste_masses_df_scenario.loc[:implement_year-1, :] = waste_masses_df_baseline.loc[:implement_year-1, :]

(city_params.py:8236 and :8271), and advanced_dst.run_advanced_dst (:161) and
advanced_dst_city.run_advanced_dst_city (:296) both force baseline pre-implement.
Only the else (blank/custom-site) branch omitted the splice.

Fix

Add the same pre-implement splice the sibling branches already use, before the
open/close window is applied. It copies the baseline generated waste (mass +
composition) into the scenario frame for every year < implement_year.

Blast radius

• No-op whenever baseline and scenario composition match — so it only changes
  output for a composition-changing scenario on the custom-site path. Baseline
  output is never touched.
• Reachable from the live /sdst UI (custom locations send no rmi_id → the
  custom-site branch; the composition inputs have separate baseline/scenario
  columns).
• Breaks no existing test (the only active sdst_v1_5 tests use identical
  baseline/scenario composition).

Acceptance criteria

• For a blank/custom site with differing baseline/scenario composition, scenario emissions == baseline emissions for every year < implement_year.
• Composition-stable scenarios are byte-for-byte unchanged.
• Baseline emissions are unchanged.
• run_advanced_dst and City.sdst_v1_5 now produce identical output for equivalent single-site inputs including composition changes (verified by WasteMAP's compare_sdst_adst_parity.py).

Definition of Done

• Acceptance criteria met
• New regression test tests/test_sdst_v1_5_preimplement_composition.py (fails on the pre-fix engine, passes after)
• Full SWEET_python suite passes (92 passed)
• Changelog updated (changelog/2026-08.md)
• Reviewed & merged

model-output-change (not breaking-change): changes output values only for a
composition-changing scenario on the custom-site path; nothing stops running.

Paired with WasteMAP RMI/WasteMAP#747 (same branch name), which adds the
endpoint-level regression test and the sdst↔adst parity harness. Merge together.

🤖 Generated with Claude Code

[Copilot]

Pull request overview

Fixes a correctness bug in City.sdst_v1_5 for blank/custom sites (baseline_data=None) where scenario waste composition could be applied to pre-implementation-year deposits, causing baseline/scenario emissions to diverge before implement_year.

Changes:

• Splice baseline generated waste masses (mass + composition) into the scenario series for all years < implement_year on the blank/custom-site branch (matching the existing TRACE-reconciled branches’ behavior).
• Add a regression test covering composition-changing scenarios on the custom-site path to ensure pre-implement scenario emissions equal baseline emissions.
• Update August 2026 changelog and changelog index entry to document the model-output change.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated no comments.

File	Description
SWEET_python/city_params.py	Forces scenario generated waste masses to equal baseline before implement_year for blank/custom sites, preventing back-dated scenario composition.
tests/test_sdst_v1_5_preimplement_composition.py	New regression test ensuring scenario == baseline emissions pre-implement when only composition differs on custom-site path.
changelog/README.md	Updates the 2026-08 summary line to include this fix.
changelog/2026-08.md	Documents the fix under “Fixed” with impact/behavior notes and PR link.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.