Skip to content

Tagged Petr Kuneš LRA comparison

The bounded comparison executes two pinned R sources against the same LandClim II author example counts and parameters: the author's 30 October 2016 LRA-REVEALS-release.R and Petr Kuneš's LRA v0.1.2 source at commit d29075692ed730e3934bc461eb9081c404aef59b. The tagged source tarball is retained outside the product repository with SHA-256 3292d4726e3cb0204d6f8326c5ce9208aecf3e064c180846adab3be7fa03a353. Its REVEALS.R, KPf.R and DESCRIPTION SHA-256 values are respectively 50353523cd1e1cb85b3fc572d55320771bdc4bf6b03b7b98ba7fa6c3b321d72d, bebe75561016cfe12138d9bf49326da760d475afa762e3c15ad7150d51938247, and 7828cfa0220637cabb029b93d447da41cbf6215327088179569e7b157d38ee90. The author source SHA-256 is 7675b6e615cddba328bad506692b3a6a31e00d9273cef5ad4641a1737fddfd87. The pinned zipfR 0.6-70 source tarball is retained privately with SHA-256 0c318569cf5dd13ab5ea22f7cf797a2daeda09d8332d89ba54fd2d7c643ecf96. The tagged package's DESCRIPTION says GPL-2 while the GitHub root shows an MIT license; its source is not copied into this product repository.

The taxon comparison has all 31 cover and 31 standard-error values for each of four cases at the author 225 BP midpoint window: one 150 m lake, a 150/10 m lake pair, all six sites with positive source counts, and the full nine-site grid 512 input (six active, three absent at this window). Both functions use the same author 31-taxon PPE/covariance and GPM-neutral parameters, wind 3 m/s and region 100,000 m. The same R seed is set before each call. All 124 matched cover pairs and 124 matched SE pairs have zero difference at the CSV's retained precision; source site counts also agree. The tagged KPf function body is identical to the 2016 author body's parsed R expression; the REVEALS bodies differ. This is a version comparison within a shared code lineage, not an independent model derivation or a corrected uncertainty certification.

The R replay also retains the full grid's 1,000 original-site bootstrap indices (SHA-256 1515f0b680e066bfc1810b0ce691e70dea56856076bf9a149c70d00f4efb9a45) and verifies that both R versions sampled identical indices. The direct native comparison feeds those exact zero-based indices and the same full nine-site, 31-taxon, 100–350 BP input into the historical native profile. Against the tagged R output, all 31 cover values pass the predeclared 1e-11 absolute/1e-10 relative tolerance (maximum residual 1.21958e-13); all 31 SE values pass 5e-8 absolute/1e-9 relative tolerance (maximum residual 2.86438e-14). Six sites have source support in that window. The native report SHA-256 is 273c1b129551e6392ca926988bf594a66c7de9f435b299832e24b771731d8995.

The isolated one-window calculation does not replay the full 25-window R random-number stream from the two-grid author/native result. That prior result's 225 BP values must not be compared to these outputs as if their bootstrap draws had the same history. The current comparison also does not test the LSM transport branch, chronological or basin uncertainty, source missing-taxon semantics, or Nordic empirical reconstruction. The reviewed conditional draw profile is a different method with a restricted positive-count subcomposition.

The source replay script parses only the author's KPf and REVEALS function definitions, avoiding the archived script's old installation side effects. It loads the tagged functions from their pinned source files and requires zipfR 0.6.70 and LRA 0.1.2. Reproduce into artifacts/ using the private source roots:

mkdir -p artifacts/jobs/petrkunes-r-comparison/r-library
R CMD INSTALL --library=artifacts/jobs/petrkunes-r-comparison/r-library \
  /path/to/bijux-pollenomics-data/lra/dependencies/zipfR_0.6-70.tar.gz
OMP_NUM_THREADS=1 OPENBLAS_NUM_THREADS=1 MKL_NUM_THREADS=1 Rscript \
  docs/report/method-certification/petrkunes-r-comparison/replay.R \
  /path/to/landclim/author-reference \
  /path/to/bijux-pollenomics-data/lra/petrkunes-v0.1.2/LRA-d29075692ed730e3934bc461eb9081c404aef59b \
  artifacts/jobs/petrkunes-r-comparison/r-library \
  artifacts/jobs/petrkunes-r-comparison/replay
cmp docs/report/method-certification/petrkunes-r-comparison/summary.csv \
  artifacts/jobs/petrkunes-r-comparison/replay/summary.csv
cmp docs/report/method-certification/petrkunes-r-comparison/taxa.csv \
  artifacts/jobs/petrkunes-r-comparison/replay/taxa.csv
cmp docs/report/method-certification/petrkunes-r-comparison/bootstrap-full-grid.csv \
  artifacts/jobs/petrkunes-r-comparison/replay/bootstrap-full-grid.csv
PYTHONDONTWRITEBYTECODE=1 OPENBLAS_NUM_THREADS=1 OMP_NUM_THREADS=1 \
  MKL_NUM_THREADS=1 .venv/bin/python \
  docs/report/method-certification/petrkunes-r-comparison/native_comparison.py \
  --repository . --private-root /path/to/landclim/author-reference \
  --output artifacts/jobs/petrkunes-r-comparison/native-replay
cmp docs/report/method-certification/petrkunes-r-comparison/native-report.json \
  artifacts/jobs/petrkunes-r-comparison/native-replay/native-report.json

The governed summary SHA-256 is 5f7957d2d318518e96959a286b74a9a183a9b23655aef3216d61dfd38137541a; the taxon comparison SHA-256 is b37211fda3a692d1b903501a14323ab14820950d434f8a2953b9ea25a269b915.