Repository navigation
Dam release follow-up: flood-pool smoke results, full-population measurement, non-dam levers - #52
Merged
Merged
Conversation
…fts the base recipe; journal Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_015TepsZAS2LkFC5EmvuZrz9
…ho-180 no-dam twins
dam_release_full_s7_s{42,43}: dam_release_full_nopool_s{42,43} with
release_head.per_dam_l2 0.0 (the S7base recipe) and dam_floor carry
(mass-conserving; allowed without a flood pool). Joint training.
dam_release_full_nodam_r180_s{42,43}: the same experiment block (rho 180,
warmup 30, testing warmup 5) with use_reservoirs false and no dam table,
ranges or release head, so each pair differs by the dam model alone. The
earlier no-dam full runs used rho 90 and are not twins of these arms.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015TepsZAS2LkFC5EmvuZrz9
…, Phase A) Honest runoff multipliers (fitted on training years) move the population median by +0.0005 to +0.0008: the model is volume-unbiased on 1982-1995 (pred/obs 1.004) and the test-period excess is a drought/drift in Q'. gamma = 0 is worth about +0.002 (+0.006 with learned p, q; one seed); a two-seed ensemble +0.001. A training-year-weighted blend of three Q' products (UH 0.50, DIST 0.25, LSTM 0.25) gives +0.0259 [+0.0221, +0.0309] at the population median (KGE about flat; paired KGE -0.004); with gamma = 0 variants +0.0297. Pre-registration in prereg.txt; report in report.txt. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_015TepsZAS2LkFC5EmvuZrz9
Dammed gauges +0.0112 / +0.0066 paired NSE (seeds 42 / 43); population median +0.0093 / -0.0016, about +0.004 on average, inside seed noise. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_015TepsZAS2LkFC5EmvuZrz9
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_015TepsZAS2LkFC5EmvuZrz9
This was referenced Oct 1, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Follow-up to #51: results and records produced after it merged. No engine code changes; configs, analysis scripts, journal entries and findings only.
experiments/reservoir/smoke_v2/flood_gate.py(the pass bar fixed before results).per_dam_l2finding. The per-dam L2 at 1e-3 had been holding every per-dam parameter near its init; with it off the base recipe gains about 50 % more on the smoke set.config/experiments/dam_release_full_{s7,nodam_r180}_s{42,43}.yaml): dammed gauges +0.0112 / +0.0066 paired median NSE; population median +0.0093 / -0.0016 (about +0.004 on average, within seed noise).experiments/reservoir/full_run/paired_full_run_full_s7_s{42,43}.*.experiments/levers/): honest runoff multipliers +0.0005; gamma = 0 about +0.002; a training-year-weighted blend of three Q' products +0.026 at the population median.research/journal/2026-09.md), research-status and findings updates.Test plan
src/changes since Learned dam release: per-dam calibration, additive dam row, rule curve, water accounting #51 (configs and scripts only);ddrs planpasses on every new config🤖 Generated with Claude Code
https://claude.ai/code/session_015TepsZAS2LkFC5EmvuZrz9