Reference scores
Gallery
Curated example datasets scored under both the canonical
2015 weights and the experimental 2026-modern
weights. Each entry shows the side-by-side tiers, the Wallach
2024 calibration level, and a stable snapshot URL you can cite.
Münchenberg 2014
Source: Mirschel et al. (2018); the framework's canonical worked example (Christian Kersebaum's own classification).
Under 2015-kersebaum → Gold (adjusted 176.58).
Under 2026-modern → NA (the dataset
has no Sentinel-2 retrievals, no in-situ sensor record, no
SNP-level cultivar genetics — which is precisely what the v2
framework is designed to surface).
Selhausen / Merzenhausen 2015–2019
Source: Reichenau et al. (2020), ESSD 12, 2333–2364; TR32 / ICOS-D cropland record with companion CRNS soil-moisture series (Jakobi et al. 2018; Jakobi et al. 2022) and CLM5 evaluation (Boas et al. 2021). Dataset DOI 10.5880/TR32DB.39 (CC-BY-4.0).
Under 2015-kersebaum → Platinum
(adjusted 232.24; raw 222.70 × multiplier 1.0429).
Under 2026-modern → Silver
(composite raw 179.35); 7 of 7 Wallach 2024 calibration
groups supported. The Selhausen system carries Sentinel-2
coverage at 10 m and approximately 5-day revisit, a
cosmic-ray neutron sensor record alongside ICOS C1
eddy-covariance flux measurements, and a four-site
multi-year design — so the v2 framework rewards it
materially where Münchenberg's all-NA extension blocks pulled
its composite to NA.
Appiah et al. 2024 Nordic spring barley (HQ tier)
Source: Appiah, Bracho-Mujica, Svane, Styczen, Kersebaum, Rötter (2024), in silico Plants 6(2), diae010. Two-site Danish field trial (Taastrup + Ringsted, 2019), cultivar RGT Planet, APSIM Classic 7.9. Companion data deposit at 10.17894/ucph.f07f7109-e741-489f-a5d6-014e167b9a7f under open-access UCPH terms.
Under 2015-kersebaum → Gold
(adjusted 173.76). Under 2026-modern →
Copper (composite raw 126.66); 6 of 7
Wallach 2024 calibration groups supported — everything
except multi_year (single 2019 season). The
strongest published demonstration of the Kersebaum 2015
tier framework being used to render and defend a per-tier
APSIM model-skill comparison; the HQ tier is the
model-guided field-experiment exemplar that the v2 paper
§4 anticipates.
COSMOS-Europe CRNS network (Bogena et al. 2022)
Source: Bogena, Schrön, Jakobi et al. (2022), ESSD 14, 1125–1151. 66-station European cosmic-ray neutron soil-moisture network across 12 countries; 2011–2021, mean record 5.7 years per site, longest ~10 years. Deposit DOI 10.34731/x9s3-kr48 under CC-BY-4.0 via the Jülich TEODOOR database.
Under 2015-kersebaum → NA
(adjusted 55.61; no agronomic content). Under
2026-modern → NA
(composite raw 56.12); IoT block scores Gold
(51 pts). Only 2 of 7 Wallach calibration groups
(multi_site, multi_year) supported
— the soil-water group fails because the soil-properties
block is Silver, illustrating the framework's framing that
rich soil-water observations don't substitute for hydraulic
characterisation. The IoT-block contrast case for the modern
framework.
FLUXNET2015 cropland exemplar — DE-Geb (Gebesee) 2001–2014
Source: Pastorello et al. (2020), Scientific Data 7, 225. Anchored on the German ICOS Class 1 cropland flux tower DE-Geb (Gebesee, Thuringian Plain; rotation of winter wheat, sugar beet, rapeseed, barley and potatoes since 2001). Site DOI 10.18140/FLX/1440075 under FLUXNET2015 Tier 1 (CC-BY-4.0).
Under 2015-kersebaum → Silver
(adjusted 120.66). Under 2026-modern →
NA (composite raw 96.10, just under the
Copper threshold); IoT block scores Platinum
(59 pts — half-hourly EC + soil-water + soil-temperature
+ ET). 4 of 7 Wallach groups supported
(phenology, biomass,
final_values, multi_year);
soil_water / soil_n fail at the
single-tower scale, and multi_site fails on
single-site design. Demonstrates a strong IoT record running
into the framework's multi-site / cultivar limits.
Gallery datasets are curated during the v2.0 lit-review pass.
See docs/v2/lit-review/ in the repo for the working
shortlist; per-fixture tier justifications live in each fixture's
_meta.tierJustifications map.