LOVE paper method correspondence

Theuerkauf and Couwenberg (2024) describe analytical LOVEr and optimized LOVEoptim using a simulated landscape. The exact 1,097,851-byte article PDF (SHA-256 3336b5b6450b89d97b6e52f18b42090333091bf355a3f4715676215e2b18cc3e) was acquired from the publisher PDF. It is byte-identical to the formerly retained mirror copy. The article prints copyright 2024 by Martin Theuerkauf and John Couwenberg and the CC BY 4.0 license on page 1. The public source packet under data/love/paper/ binds the PDF and its separately native-qualified text derivative. The table producer derives 151 numeric cells from printed page 6, including the wrapped Tilia header; the retained table is the genuine source-backed producer output, checked against all 151 formerly transcribed constants. Ordinary configured-owner reference/replay proof remains a separate output-custody requirement. This establishes public article identity and printed constants only. The paper's data availability statement points to DISQOVER functions, without identifying a deposited simulation input bundle. No supplementary simulation bundle was identified in the accessible article. The local source identity receipt establishes the exact later DISQOVER 0.9.14 archive used by our R oracle; the article accessed DISQOVER in February 2024, whereas that archive was packaged in January 2026. Archive parity therefore does not establish the exact software version behind Figure 2.

The retained published parameter tables are generated by bijux_pollenomics_dev.source_completion.love.paper_tables from the exact publisher PDF, its qualified text derivative and their source manifest. Its 121 map-unit fractions from Table 1 and 30 fall-speed, PPE and PPE SD constants from Table 2 reproduce all 151 previously transcribed numeric cells on page 6. The retained source metadata now identifies the public article and the byte-authenticated extraction, rather than a private mirror transcription. Each map-unit row sums to one, and every printed PPE SD is 10% of its PPE. Table 2 does not state a fall-speed unit, so the transcription does not assign one. These constants resolve a bounded paper input; they do not identify the spatial extent of each map unit, the pond placements or the random realizations behind Figure 2.

The source-parameter comparison is a fresh execution of the maintained verifier against that genuine public-source table and the pinned packaged DISQOVER LOVEexampleParams.json. All 30 fall-speed, PPE and PPE SD values agree exactly by row position, with zero residuals. The refreshed report changes only its table input receipt; its comparison rows and scope are unchanged. Nine labels agree exactly; the paper prints Grasses where the package prints Gras. This is a numeric source correspondence with an explicit label difference, not evidence that the 2024 simulation ran the later pinned package version. The pinned package example has only local and regional pollen rows and one 36-ring layout reaching 100,000 distance units. Those inputs do not encode the article's eight artificial ponds or their realized pollen/draws. Check the producer output byte for byte against the governed JSON from the repository root with python3 docs/report/method-certification/love-paper-correspondence/verify.py --repository . --check. Use --output artifacts/method-certification/love-paper-source-parameter-replay.json instead when an isolated replay file is needed; the output must be absent. The configured comparison adapter bijux_pollenomics_dev.source_completion.love.parameter_correspondence invokes this same fixed verifier. Separate genuine reference/replay calls for both configured owners are required before output-ownership acceptance. Neither these constants nor their agreement supplies a Figure 2 simulation replay.

The same pinned DISQOVER 0.9.14 archive's disqover/man/LOVEexampleRings.Rd (member SHA-256 43c8f7166a88d35b609d8772bab2c5830dc1e76f97ff5d4e9d6fa2a5fd0fc5ad) describes its example as simulated from BÜK300. The retained 2024 paper describes its eight-site Figure 2 simulation as using BÜK200. We retain those source descriptions as a difference in stated map basis; the example parameter match does not establish that either archive example contains the paper's Figure 2 landscape or random runs.

Paper method or assumption	Current executable evidence	Scope of correspondence
Small-site pollen deposition combines local and regional contributions.	Optimized forward objective and analytical subtraction.	Source equations have bounded R/Python parity; no ecological truth is inferred.
Analytical LOVE assumes comparable total distance-weighted abundance among small sites and local vegetation homogeneous with the regional composition.	Analytical ring sensitivity retains negative and above-one estimates rather than imposing cover constraints.	These are paper-stated limitations, not assumptions validated for Nordic basins.
LOVEoptim fits ring-specific local cover by log-Euclidean pollen mismatch using RcppDE::DEoptim.	Common objective and solution-quality study uses installed RcppDE and the Python optimizer on identical declared objectives.	Candidate evaluations agree; achieved objective quality is bounded, while search paths and stopping events differ.
Regional and local count/PPE perturbations are repeated for uncertainty.	Paired LOVE draw study holds explicit local count/PPE draws equal across R and Python; source compatibility records the optimized wrapper's sampled-but-unused local PPE.	The analytical supplied-draw calculation is checked. The article's general uncertainty description does not certify coverage, and the optimized source's local PPE behavior is narrower than that description implies.
Ring land-cover fractions can represent lakes, wetlands and terrestrial area separately.	Forward input profiles name source and revised fraction rules; source dimensional refusals and revised ring/taxon tests are explicit.	No paper landscape ring raster or site-specific geometry has been admitted.
Paper simulation uses BÜK200 map units, one regional lake, eight artificial ponds (diameters about 10, 50, 100 and 200 m), a 100 km source extent, LSM unstable dispersal and 100 model runs.	Dispersal and parity map covers the LSM kernel and core LOVE equations.	The published Tables 1 and 2 give map-unit compositions and ten taxon parameters, but the precise vector map extraction, pond placements, ring-area arrays, realized pollen inputs and random draws needed for Figure 2 are not supplied by the article.
Paper evaluates true-versus-reconstructed ring cover with RMSE and discusses a 100–500 m range for its simulated landscape.	Ring sensitivity reports source-compatible numerical behavior only.	The paper's Figure 2 RMSE and preferred radius cannot be reproduced or transferred to Nordic sites without its spatial truth and ring inputs.

The accessible article supports an assumption and equation mapping. The original R reference, paired local draws and optimizer comparison provide actual bounded numerical tests on declared inputs. Reproducing the paper's eight-site Figure 2 requires the author's exact landscape extraction, site and ring geometry, pollen simulation inputs or code, and stochastic settings. The article tables alone cannot determine those numerical outputs; digitizing the figure would not establish a simulation replay. This paper-example obligation remains open with that exact source input requirement.