{
  "schema_version": "sambosjon-direct-pollen-protocol.v1",
  "response": "Observed Neotoma pollen fractions on the existing 16 exact-label closed axis; native NISP and outside-axis mass remain visible",
  "target_site_id": "neotoma:site:3398",
  "target_site_name": "Lake Sämbosjön",
  "target_collection_unit_id": "neotoma:collection-unit:4262",
  "target_collection_handle": "SAMBOTOT",
  "target_dataset_id": "neotoma:dataset:4403",
  "target_aliases_withheld": ["Lake Sämbosjön", "SAMBOTOT"],
  "nearby_name_join_refused": "LandClim SAM is 3.316459 km from Neotoma site 3398; the retained site-identity review does not establish a common physical core or basin, so SAM is excluded as a source join rather than asserted as an alias",
  "training_site_exclusion": "Exclude the entire physical site, all target cores, aliases, dataset rows and dependent pollen products from every training and model-selection input",
  "training_population": "The 81 previously selected DK/FI/NO physical sites and three previously selected Swedish lake sites, deduplicated by Neotoma site ID",
  "age_model": "Sämbosjön source-default comparable calibrated-radiocarbon chronology neotoma:chronology:4262:30017; separate radiocarbon claims are context only",
  "bin_width_cal_bp": 500,
  "candidate_bin_ids": [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22],
  "methods": ["training_mean", "nearest_site", "inverse_distance_squared"],
  "primary_method": "inverse_distance_squared",
  "pre_score_predictions_sha256": "97a2b45ce51e2e42c851e64ba5b91ea721e887fac6c1e92e16f3938b428f4e65",
  "primary_metric": "equal-bin mean absolute error on the closed pollen fractions",
  "secondary_metric": "equal-bin mean Hellinger distance on the same bins",
  "support_rule": "Predict a target bin only when at least two distinct training physical sites have positive closed-axis NISP in that bin; retain every refusal",
  "preparation_read_set": "The shared Neotoma prerequisite screen parses the retained observation partitions, including site 3398 target rows, for source validation. Its returned training profiles select DK/FI/NO only; the three Swedish profiles are separately fixed to ABB/BUS/BOKE. No Sämbosjön taxon rows or count composition are selected, aggregated or passed into prepared training, prediction or method choice.",
  "prediction_read_set": ["prepared training profiles", "target site coordinate, sample and selected age-claim IDs, and candidate bin IDs", "this protocol"],
  "scoring_read_set": ["frozen predictions and their SHA-256", "source-locked Neotoma observation parts and sample-age claims", "this protocol"],
  "target_count_use": "Target native and closed-axis totals and positive-bin support were inspected for feasibility before protocol declaration. Shared source preparation parses target observation rows but does not select or aggregate them into predictor inputs. Per-taxon target composition and scores were first interpreted after a prediction freeze; subsequent protocol annotation corrections do not change predictions or method choice.",
  "claim_scope": "A newly evaluated physical site supplying descriptive observed-pollen development evidence, with prior aggregate-count exposure and later aggregate-score exposure during a protocol annotation correction. No independent-confirmation, untouched-lockbox, LRA, measured-vegetation, basin-buffer or broad Swedish-generalization claim."
}