Functional landscapes · web edition

Interactive edition · refreshed 4 August 2026

How much habitat does a landscape need before its species can stay?

Bergman and colleagues put a number on it for the 102 species that depend on ancient oaks and semi-natural grassland across 65,504 km² of southern Sweden. This edition rebuilds their result as something you can interrogate: every fitted curve is redrawn, every species links out to GBIF and iNaturalist, and every conservation status has been restated against the red lists that are actually in force in 2026 — not the ones the paper cites.

Karl-Olof Bergman · Leif Andersson · Markus Franzén · Victor Johansson · Lars Westerberg
IFM Biology, Conservation Ecology Group, Linköping University, 581 83 Linköping, Sweden · Pro Natura, Halna 27, 545 93 Töreboda, Sweden

Landsc Ecol (2026) 41:101 · doi:10.1007/s10980-026-02373-4 · received 4 July 2025, accepted 1 May 2026, published 22 May 2026 · Open Access CC BY 4.0

About this edition. This is an independent web edition, not published, reviewed or endorsed by the authors of the article or by Landscape Ecology. The science, the data and the figures are theirs, reused under CC BY 4.0. The reconstruction of the fitted curves, the 2026 status update, the measurements taken off the figures, and every claim in what to be careful with are this edition's — and so is any error in them. Cite the article, not this page: Bergman K-O, Andersson L, Franzén M, Johansson V, Westerberg L (2026) Landscape Ecology 41(6):101, doi:10.1007/s10980-026-02373-4.
25 of the 102 species are nationally threatened today — the paper's own baseline said 16 102 species modelled 86 with a significant habitat response — the paper's count, matched by our own reading of Fig. 3 245 fitted models rebuilt 81 / 102 with a photograph

The refresh

The paper's conservation statuses were out of date on the day it appeared

Fig. 3 of the paper marks each species with its Swedish Red List category, citing SLU Artdatabanken 2020. But the next Swedish Red List — Rödlistade arter i Sverige 2025 — was published on 24 March 2026, and the paper went online on 22 May 2026. Two months separate them. Restating the 102 species against the list now in force changes the picture substantially.

16 → 25 Nationally threatened (CR, EN, VU) A rise of 9 species, over half again as many
11 Species moved to a worse category Against 5 moved to a better one
2 Now CR (PRE) — possibly regionally extinct Both are grassland butterflies, and both sit near the top of the 15 km threshold ranking — ranks 1 and 3 of 81
24 Species whose name is unsettled Where the paper, GBIF, iNaturalist and the national list disagree

Where the species went

to a worse category to a better one unchanged

Ribbon thickness is the number of species. 99 of the 102 species are drawn — a species needs a category in the LC–CR range on both lists to have a ribbon. The 3 left out are Buellia violaceofusca (absent from the 2025 list), Clavaria zollingeri (absent from the 2025 list), Microglossum olivaceum s.lat. (NA on the 2025 list). The two species that gained a possibly regionally extinct qualifier keep their base category of CR and are drawn. This is why the 2020 column adds up to 15 threatened species where the tile above says 16: the missing one is Clavaria zollingeri, VU in 2020 and not on the 2025 list under that name.

The burnet moths moved as a block

All four Zygaena burnet moths in the study went from NT to VU: Z. filipendulae, Z. lonicerae, Z. osterodensis and Z. viciae. So did two blues, Cupido minimus and Phengaris arion, and the copper Lycaena hippothoe; the blue Cyaniris semiargus went straight from LC to VU.

That is eight of the 11 uplistings in one guild: day-flying Lepidoptera of semi-natural grassland. The other three are fungi.

It would be neat if the habitat data pointed the same way. It does not. Ranked by their grassland thresholds these are middling to undemanding species: 6 of the 7 with a 5 km threshold sit below the 250 ha median of the 80 grassland species, and 6 of 8 sit below the 518 ha median at 15 km. Zygaena filipendulae's 15 km figure is one of the negative values flagged below, so it is not a habitat requirement at all. Only Phengaris arion, at rank 10 of 81, is among the demanding ones.

So the red-list signal and the habitat-amount signal are not saying the same thing about this guild. Whatever is pushing these moths and butterflies into the threatened categories, this study's habitat-amount models are not where the explanation lies — which is itself worth knowing, and is what the authors' own caveats about connectivity, nectar resources and climate point at.

Two of the most demanding grassland species may already be gone. Of the 81 grassland species, Plebejus argyrognomon has the highest 15 km threshold and Melitaea britomartis the third highest — 31,865 and 3,903 hectares. Both are now CR (PRE): critically endangered, possibly regionally extinct in Sweden. Read that pairing carefully rather than as a finding: P. argyrognomon's figure is larger than a 15 km cell can physically hold, so its rank says as much about extrapolating a model fitted where GBIF holds two Swedish records as it does about the species' needs. (Those two are GBIF's holdings, not the paper's sample — the study drew on Swedish sources GBIF only partly mirrors.)

Uplisted since the paper's baseline (11)

Species20202025Group
Piptoporus quercinus tungtickaENCRFungus
Cupido minimus mindre blåvingeNTVUInsect
Cyaniris semiargus ängsblåvingeLCVUInsect
Hygrocybe ingrata rodnande lutvaxingLCVUFungus
Hygrocybe ovina sepiavaxingLCVUFungus
Lycaena hippothoe violettkantad guldvingeNTVUInsect
Phengaris arion svartfläckig blåvingeNTVUInsect
Zygaena filipendulae sexfläckig bastardsvärmareNTVUInsect
Zygaena lonicerae bredbrämad bastardsvärmareNTVUInsect
Zygaena osterodensis smalsprötad bastardsvärmareNTVUInsect
Zygaena viciae mindre bastardsvärmareNTVUInsect

Downlisted (5)

Species20202025Group
Centaurium erythraea var. erythraea flockarunNTLCVascular plant
Gentianella campestris bredgentianaENVUVascular plant
Gymnopus fusipes räfflad nagelskivlingNTLCFungus
Hemaris tityus svävflugedagsvärmareNTLCInsect
Microglossum olivaceum s.lat. olivjordtunga agg.NTNAFungus

Microglossum olivaceum s.lat. is not a real improvement: the 2025 list treats the aggregate as NA, not applicable, and assesses its segregates individually instead.

Nationally the threatened share held flat. Here it rose by half again.

The cleanest way to compare is the share of assessed taxa in a threatened category, because it holds the denominator steady on both sides. Across Sweden as a whole that share did not move: 2,249 of 21,740 assessed species were threatened in 2020 (10.3%) and 2,373 of 23,103 in 2025 (10.3%).

Among the 102 species of this study the same share went from 15.7% to 24.5% — 16 species to 25. Sweden's biodiversity picture stayed where it was; the species of ancient oaks and semi-natural grassland moved.

Two caveats, both real. Nine species out of 102 is a small number, and the two national editions do not assess an identical set of taxa (21,740 species in 2020 against 23,103 in 2025), so this is an indication rather than a formal test. And one counter-current is worth naming: the number of these species red-listed at all — NT and worse — actually fell, from 56 to 53. Almost none of that fall is good news. 6 species left the list: three were genuinely downlisted to LC (Centaurium erythraea var. erythraea, Gymnopus fusipes, Hemaris tityus); Microglossum olivaceum s.lat. is NA only because the 2025 list assesses its segregates instead; and Buellia violaceofusca and Clavaria zollingeri are absent under the paper's names, their material carried as Lecanographa amylacea (NT) and, on one reading, Clavaria amethystina (VU). Against that, 3 species entered the list from LC and went straight to VU. So of the 25 threatened today, 22 were already red-listed in 2020 and moved deeper in; the other 3 arrived from outside it.

The global picture is nearly empty — which is the point

Of the 102 species, only 17 carry a global assessment on the IUCN Red List version 2026-1 (released 9 July 2026, 175,909 species assessed worldwide, 49,505 of them threatened). 7 are globally threatened, and every one of them is a fungus — Clavaria zollingeri, Entoloma griseocyaneum, Entoloma prunuloides, Hapalopilus croceus, Hygrocybe citrinovirens, Hygrocybe ingrata, Piptoporus quercinus — a legacy of the first wave of fungal assessments rather than of these taxa being better known than the rest.

No butterfly, no burnet moth, no grassland plant in this study has a global IUCN assessment. For a reader outside Sweden that is the honest headline: the global red list has almost nothing to say about the species this paper is about, and the national list carries the whole weight of the evidence.

The study

Two habitats, two scales, one question

The study area is 65,504 km² of boreo-nemoral southern Sweden, of which 54,908 km² is land. About 70% is forest, mostly production spruce and pine; arable fields take 14%, and semi-natural grassland just 3.4%. The west coast, Öland, Gotland and the far south were left out as too different in climate and bedrock.

Two habitat resources were mapped. Large oaks, from the Swedish national species observation system's Trädportalen project: 40,326 records of oaks over 100 cm in diameter at breast height and 3,626 over 150 cm, covering 42,152 km² of the study area. And semi-natural grassland, from the national meadow and pasture inventory TUVA plus the higher-conservation-value class of the agricultural subsidy database.

Against that, 5,700,341 species records were assembled, of which 4,898,150 survived a 5 km precision filter. 143,504 of them belong to the 102 focal species; the remaining 4.8 million are the background used to weight each grid cell by how hard anyone had actually looked there.

Occurrence was modelled per grid cell with a binomial GLM, habitat amount as the only predictor, weighted by Ruete's ignorance score. The habitat threshold is the habitat amount at which the fitted probability of occurrence reaches 50% — an LD50 analogy, and, as the authors say, a conventional rather than an ecologically derived cut.

Fig. 1 — The extent of the study area in southern Sweden.
Fig. 1. The extent of the study area in southern Sweden. Bergman et al. 2026, Landsc Ecol (2026) 41:101, CC BY 4.0.
21species tied to old oaks 11 epiphytic lichens, 9 fungi, 1 beetle — each modelled against both the ≥100 cm and the ≥150 cm oak class
81species tied to semi-natural grassland 41 insects, 32 vascular plants, 8 fungi
86 / 102 significant positive habitat response 84%, the paper's own count. Reading the colours of Fig. 3 independently also gives 86 — with the caveat in what to be careful with
0.77 vs 0.67 mean AUC, oak vs grassland models Oak-dependent species respond far more cleanly than grassland ones
Fig. 2 — Probability of occurrence for Cliostomum corrugatum (black lines) and Grifola frondosa (grey lines) in relatio
Fig. 2. Probability of occurrence for Cliostomum corrugatum (black lines) and Grifola frondosa (grey lines) in relation to the number of oaks above 100 cm in diameter within 2500 ha (5 × 5 km grid). The horizontal line at 50% probability shows an interval representing 5–95% of the bootstrapped values. Bergman et al. 2026, Landsc Ecol (2026) 41:101, CC BY 4.0.

What a threshold looks like

Fig. 2 of the paper shows two oak species. The lichen Cliostomum corrugatum reaches 50% occurrence probability at 46 oaks over 100 cm dbh within a 5 × 5 km cell; the polypore Grifola frondosa needs 135. The lichen's bootstrap interval is much tighter, meaning its presences and absences track tree numbers more consistently.

Those two curves are the whole method in miniature. The explorer below rebuilds them — and the other 243 — from the published numbers, so you can put any species on the same axes.

The case for the method

What holds up

This edition spends a section on what to be careful with, so it owes the same space to why the result is worth having. Three things carry it: the thresholds agree with independent earlier estimates, the choice of scale rests on published evidence rather than convenience, and the authors tested their own sensitivity to several of the choices a sceptic would query first.

Three independent estimates, same order of magnitude

The paper's headline figure is roughly 400 oaks over 1 m dbh within 25 km² for many oak-dependent species. Earlier work over smaller regions, on different taxa and by different methods, lands in the same broad range — though "the same range" is the honest claim, not "the same number": two of the three differ from their Table S2 counterparts by factors of about 2 and 5.

  • Beetle species richness plateaued at about 250 oaks ≥ 100 cm dbh within 16 km² (Bergman et al. 2012).
  • At least 8 ancient oaks ≥ 160 cm dbh within 12.5 km² for 50% predicted occurrence of the red-listed lichen Ramalina baltica (Paltto et al. 2010) — but note that is occupancy of an individual tree, not of a landscape cell, so it is not directly comparable with a Table S2 threshold. Set beside it, Table S2's 3.3 oaks ≥ 150 cm within 25 km² for another oak lichen, Cliostomum corrugatum, is at least the same order of magnitude.
  • 11% grassland and deciduous forest within 78.5 km² for burnet moths (Bergman et al. 2004), against Table S2's 4.5% grassland within 25 km² for Zygaena viciae — a factor of about 2.4, and the earlier figure counts deciduous forest as well as grassland.

Convergence from independent data is the strongest argument that these thresholds are not artefacts of one dataset. It is also why the two Discussion percentages that do not reconcile matter so little to the overall picture and so much to the individual number.

The two grid sizes come from a literature review

Supplementary Table S1 reviews 16 published scale-of-response estimates for the relevant taxa in northern Europe, restricted to scales above 100 ha so that the question is the landscape rather than the site. The answers range from site level for grassland plants (Dainese et al. 2015) to a 64 km radius for oak-dependent lichens (Ranius et al. 2008), but 5–15 km radii dominate, which is where the paper's 5 × 5 km and 15 × 15 km grids come from.

GroupHabitat measure ScaleSource
Oak beetlesOak dominated woodland key habitats1 kmFranc et al. (2007)
Oak beetlesAmount of dead oak wood12-40 kmFranc et al. (2007)
Oak beetlesNumber of large/hollow oaks2.3 kmBergman et al. (2012)
Elater ferrugineus (click beetle)Number of large/hollow oaks4.7 kmMusa et al. (2013)
LichensNumber of large/hollow oaks64 kmRanius et al. (2008)
FungiNumber of large/hollow oaks8 kmRanius et al. (2008)
Wood-inhabiting fungiTemperate broadleaved forests1 km/10 km*Paltto et al. (2006)
Vascular plantsTemperate broadleaved forests5 kmPaltto et al. (2006)
BryophytesTemperate broadleaved forests5 kmPaltto et al. (2006)
LichensTemperate broadleaved forests10 kmPaltto et al. (2006)
MollusksTemperate broadleaved forests10 kmGötmark et al. (2008)
13 species of butterfliesSeminatural grasslands10-30 kmBergman et al. (2018)
ButterfliesSeminatural grasslands5 kmBergman et al. (2004)
Vascular plantsSeminatural grasslandsSite levelDainese et al. (2015)
ButterfliesSeminatural grasslands10X10 km (grid)Dainese et al. (2015)
Vascular plantsSeminatural grasslandsNo scale responseLindborg & Eriksson (2004)

Two things to hold in mind when you pick a scale in the controls above. A grid cell is not a radius: a 15 × 15 km cell is 225 km², a 15 km radius is 707 km². And both grids are applied to every species, so for a lichen whose literature scale is 64 km even the larger grid is small.

The authors tested the choices a sceptic asks about

Each threshold carries 1,000 bootstrap resamples, with the 5th and 95th percentiles drawn as the intervals in Fig. 3. The survey-effort weighting was repeated at ignorance thresholds of 10, 20 and 50 — the rank order of species thresholds held, and you can switch between all three in the explorer above. Thresholds are published at three occupancy levels, not one. Model discrimination is reported per model as AUC. And the map cut-offs were checked against alternatives, which the authors state yielded similar spatial patterns.

The paper is also explicit about its own limits: occurrence is a proxy for viability, P = 0.50 is conventional rather than ecological, fixed non-overlapping cells ignore habitat just over a boundary, and the thresholds are minimum targets that say nothing about extinction debt. Most of what this edition adds in what to be careful with extends those admissions rather than contradicting them.

Interactive · the centrepiece

Threshold explorer

Supplementary Table S2 publishes each model's threshold at three occupancy levels — 25%, 50% and 75%. Because the models are logit-linear, any two of those fix the fitted curve exactly. Every curve below is therefore the paper's own model, recovered rather than approximated.

Habitat
Grid cell
Ignorance threshold
Backdrop

Pick up to three species — the list is ordered by threshold, most demanding first. A tag marks any model that is not significant at p<0.05.

Reading the ignorance control. Each grid cell is weighted by Ruete's ignorance score, I = t / (observations + t), where t is the number of background observations at which absence has a 50% chance of being real. The paper sets t to 20, so a cell with 20 background observations scores 0.5, a well-surveyed cell approaches 0 and an unsurveyed one approaches 1. Table S2 also publishes t = 10 and 50, which is what this control switches between. Across the three settings the rank order of species is stable, which is the reassuring result; individual thresholds do move, and for some species a good deal — compare Zygaena viciae at 5 km across the three.

One thing to hold in mind, discussed further in what to be careful with: the paper states that I itself is used as the regression weight, and I is largest where survey effort is lowest.

Interactive · Fig. 3 rebuilt

Every species, ranked by how much habitat it needs

The paper's Fig. 3 as a live chart: sortable, and every row opens the full species record. The interval is the span from 25% to 75% occurrence probability — not a confidence interval. Marker darkness is the significance of the fitted model.

Habitat
Grid cell

p<0.05 p<0.1 not significant no marker in Fig. 3 to read ↓ falling model Shape carries the class, and the category code is printed beside every row, so nothing here depends on colour.

The original Fig. 3, for comparison

The published figure. The chart above is built from Supplementary Table S2, which is the numerical source behind this figure; the significance colours above were read off these bitmaps. Note that Fig. 3 uses Aurantiporus croceus, Buglossoporus quercinus and Inoderma byssaceum where Table S2 uses Hapalopilus croceus, Piptoporus quercinus and Arthonia byssacea.

Fig. 3a–b — The habitat amount required for 50% probability of occupancy and 5/95 percentile of bootstrap replicates for d
Fig. 3a–b. The habitat amount required for 50% probability of occupancy and 5/95 percentile of bootstrap replicates for different habitats and spatial scales. (a) oaks > 1 m dbh, (b) oaks > 1.5 m dbh. The rug plot at the bottom of each panel indicates observed amounts of habitat. Swedish Red List categories (SLU Artdatabanken 2020) are indicated by asterisks: NT=*, VU=**, EN=***, CR=**** Bergman et al. 2026, Landsc Ecol (2026) 41:101, CC BY 4.0.
Fig. 3c — The habitat amount required for 50% probability of occupancy and 5/95 percentile of bootstrap replicates for s
Fig. 3c. The habitat amount required for 50% probability of occupancy and 5/95 percentile of bootstrap replicates for semi-natural grasslands (ha). The rug plot at the bottom of each panel indicates observed amounts of habitat. Swedish Red List categories (SLU Artdatabanken 2020) are indicated by asterisks: NT=*, VU=**, EN=***, CR=**** Bergman et al. 2026, Landsc Ecol (2026) 41:101, CC BY 4.0.

Interactive · Fig. 4 generalised

Is my landscape functional?

The paper's conservation value pyramid rests on nestedness: if a landscape holds enough habitat for a demanding species, it holds enough for every less demanding one. Fig. 4 shows four species. Here is the same ladder for all of them — set a habitat amount and read off which thresholds it meets.

Habitat
Grid cell
These are floors, not certificates. The authors' own conclusion is that thresholds "should be interpreted as minimum targets, and complementary consideration of habitat connectivity and climate change is recommended". Habitat amount is the only predictor in these models, so this ladder cannot see either.

The paper names its own exceptions. For a poor disperser like Osmoderma eremita, distance between patches may matter as much as their total number. Several grassland butterflies need larval host plants and nectar in adjacent habitats, so grassland area alone overstates what is usable. And a landscape that clears a threshold today may still be shedding species through extinction debt, while one below it may hold them for decades inside a window of conservation opportunity. Meeting a threshold is a necessary condition for persistence, not a sufficient one.

Fig. 4 — A conservation value pyramid for species dependent on oaks >150 cm dbh within 15x15 km. If the landscape fulfi
Fig. 4. A conservation value pyramid for species dependent on oaks >150 cm dbh within 15x15 km. If the landscape fulfils the habitat requirements for a species that requires a high number of oaks >150 cm dbh the landscape for all other species with lower demands is considered functional. Species name in bold indicates the species in the photo. Photo credits: Karl-Olof Bergman except Grifola frondosa by Pethan CC BY-SA 3.0 Bergman et al. 2026, Landsc Ecol (2026) 41:101, CC BY 4.0.

Species ladder, most demanding first

Six species

What the numbers look like on the ground

Six of the 102, chosen because each one shows something the aggregate hides. Every figure in these cards is read straight from Table S2, the Swedish Red List 2025 and live GBIF record counts.

4 of the 102 species have no Swedish record in GBIF since 2015. They are Buellia violaceofusca, Melitaea britomartis, Piptoporus quercinus, Plebejus argyrognomon. Three of the four are the study's most habitat-demanding species; the fourth is the lichen that has since been folded into another taxon. The study's occurrence window closed in 2014, so it could not have seen this.
Plebejus argyrognomon

Plebejus argyrognomon

Reverdin's Blue · kronärtsblåvinge · Insect

CR (PRE)
Threshold
31,865.3 ha per 15 × 15 km cell — larger than the 22,500 ha cell
Model, 15 × 15 km
AUC 0.67 · UNRESOLVED
Swedish records
2 total · 2 in 1980–2014 · 0 since 2015

The species the study ranks as needing more grassland than any other is also the species it can say least about. Its 5 km model has an AUC of exactly 0.50 and a falling fitted slope; its 15 km threshold asks for more grassland than a 15 km cell contains. And GBIF holds two Swedish records of it, ever, both before 2014.

GBIF ↗iNaturalist ↗Wikipedia ↗

(c) Mirko Tomasi, some rights reserved (CC BY-NC), uploaded by Mirko Tomasi · CC BY-NC 4.0 · original ↗

Melitaea britomartis

Melitaea britomartis

Assmann's Fritillary · veronikanätfjäril · Insect

CR (PRE)
Threshold
3,903.4 ha per 15 × 15 km cell
Model, 15 × 15 km
AUC 0.60 · n.s.
Swedish records
236 total · 17 in 1980–2014 · 0 since 2015

The other CR (PRE) butterfly. GBIF holds 236 Swedish records, 17 of them inside the study's 1980–2014 window and none since 2015. GBIF is not the paper's dataset — the study drew on Swedish sources GBIF only partly mirrors — but a species this sparsely represented in the public record leaves very little with which to check a threshold against.

GBIF ↗iNaturalist ↗Wikipedia ↗

(c) Айдън Асанов, some rights reserved (CC BY-NC), uploaded by Айдън Асанов · CC BY-NC 4.0 · original ↗

Piptoporus quercinus

Piptoporus quercinus

Oak polypore · tungticka · Fungus

CR IUCN VU was EN in 2020
Threshold
36.0 oaks per 5 × 5 km cell
Model, 5 × 5 km
AUC 0.75 · p<0.05
Swedish records
39 total · 31 in 1980–2014 · 0 since 2015

The only oak species to be uplisted, from EN to CR, and one of the seven species in the study that are globally threatened (Vulnerable). The paper's own Fig. 3 already calls it Buglossoporus quercinus; only Table S2 still says Piptoporus. Zero Swedish records since 2015.

GBIF ↗iNaturalist ↗

(c) Ryan Patrick, some rights reserved (CC BY-NC), uploaded by Ryan Patrick · CC BY-NC 4.0 · original ↗

Hapalopilus croceus

Hapalopilus croceus

Orange Polypore · saffransticka · Fungus

CR IUCN VU
Threshold
42.8 oaks per 5 × 5 km cell
Model, 5 × 5 km
AUC 0.75 · p<0.05
Swedish records
892 total · 367 in 1980–2014 · 448 since 2015

The apex of the four-species pyramid the paper draws in Fig. 4 — though Table S2 puts five other oak species above it at this scale: Perenniporia medulla-panis, Gymnopus fusipes, Bactrospora corticola, Chaenotheca hispidula and Arthonia byssacea. CR in Sweden, Vulnerable globally (Dahlberg 2019, criteria A2c+3c+4c). Unlike the butterflies, it is being found more often, not less — 448 of its 892 Swedish records postdate the study's window.

GBIF ↗iNaturalist ↗

(c) Katja Schulz, some rights reserved (CC BY), uploaded by Katja Schulz · CC BY 4.0 · original ↗

Osmoderma eremita

Osmoderma eremita

Hermit beetle · läderbagge · Insect

VU IUCN NT
Threshold
12.5 oaks per 5 × 5 km cell
Model, 5 × 5 km
AUC 0.80 · p<0.05
Swedish records
4,341 total · 2,107 in 1980–2014 · 2,050 since 2015

The best-known species in the study and the one with the most tractable target: a dozen oaks over 150 cm within 25 km². Near Threatened globally, Vulnerable in Sweden, and modelled here with an AUC of 0.80 — among the most reliable relationships in the whole analysis.

GBIF ↗iNaturalist ↗Wikipedia ↗

(c) Stanislav Snäll, some rights reserved (CC BY) · CC BY 4.0 · original ↗

Gentianella campestris

Gentianella campestris

field gentian · bredgentiana · Vascular plant

VU was EN in 2020
Threshold
323.5 ha per 5 × 5 km cell
Model, 5 × 5 km
AUC 0.67 · p<0.05
Swedish records
32,330 total · 14,259 in 1980–2014 · 12,733 since 2015

The clearest case of the two signals disagreeing. The red list moved it the right way, EN down to VU. The habitat maths did not: its threshold is 12.9% of a 5 km cell, in a study area that is 3.4% semi-natural grassland. The listing improved; the landscape it needs did not appear.

GBIF ↗iNaturalist ↗Wikipedia ↗

(c) Wolfgang Jauch, some rights reserved (CC BY), uploaded by Wolfgang Jauch · CC BY 4.0 · original ↗

Record counts are live GBIF totals for the accepted taxon and are not the paper's dataset — the paper used 143,504 records of these species drawn from Swedish sources, filtered to 5 km precision. GBIF counts are shown because they are the figures a reader can click through and check today. Photographs from iNaturalist under the licence shown with each.

All 102 species

The species list, linked out to the world

Every species links to its GBIF taxon page, its GBIF occurrence search and its iNaturalist page — the two systems a reader anywhere can use, rather than a national portal. Click any row for the full record: fitted curves, thresholds at all three occupancy levels, taxonomy, record counts and photograph credit.

oaks ≥ 100 cm oaks ≥ 150 cm semi-natural grassland The group name is printed beside every dot, so nothing here rests on colour.

Thresholds are the P = 0.50 value at ignorance 20, for the species' own habitat — for the 21 oak species the ≥150 cm class is shown, since it is the scarcer resource. AUC is the highest of all that species' models, which for an oak species spans both diameter classes as well as both scales — so it need not be the model behind the threshold in the same row. The per-model AUCs are in the species record. GBIF record counts are live totals for the accepted taxon, not the paper's dataset, and for a handful of species GBIF's backbone pools two taxa — those are flagged in the record.

Where

Functional areas, and the places that fall short

Two things are mapped in each figure. On the left, how many species' thresholds a grid cell actually meets. On the right, the habitat mismatch: places where sensitive species are recorded even though the habitat sits below their threshold — a signature of extinction debt, and the paper's best candidates for restoration.

Fig. 5 — Functional areas with enough habitat to harbour a certain number of species (a, c) and areas with habitat mism
Fig. 5. Functional areas with enough habitat to harbour a certain number of species (a, c) and areas with habitat mismatch (b, d), for oak-dependent species based on oaks > 100 cm dbh. The number of species, seen in the legend (of a and b), for the different habitat amount classes correspond to habitat thresholds in Fig. 3. The habitat mismatch maps are based on the distribution of species requiring > 327 oaks at 15 km and > 122 oaks at 5 km, representing moderately sensitive species. As an example, red areas (in b) lack at least 300 oaks to be functional for the species occurring there, while green areas have a surplus of up to 1600 oaks. Grey areas lack tree data. The values for functional areas and habitat mismatch thresholds were chosen to illustrate the general pattern; alternative threshold values yielded similar spatial patterns Bergman et al. 2026, Landsc Ecol (2026) 41:101, CC BY 4.0.
Fig. 6 — Functional areas with enough habitat to harbour a certain number of species (a, c) and areas with habitat mism
Fig. 6. Functional areas with enough habitat to harbour a certain number of species (a, c) and areas with habitat mismatch (b, d) for oak-dependent species based on oaks > 150 cm dbh. The number of species, seen in the legend (of a and b), for the different habitat amount classes correspond to habitat thresholds in Fig. 3. The habitat mismatch maps are based on the distribution of species requiring > 60 oaks at 15 km and species requiring > 22 oaks at 5 km, representing moderately sensitive species. As an example, red areas (in b) lack at least 50 oaks for being functional for the species occurring there, while green areas have a surplus of oaks. Grey areas lack tree data. The values for functional areas and habitat mismatch thresholds were chosen to illustrate the general pattern; alternative threshold values yielded similar spatial patterns Bergman et al. 2026, Landsc Ecol (2026) 41:101, CC BY 4.0.
Fig. 7 — Functional areas with enough habitat to harbour a certain number of species (a, c) and areas with habitat mism
Fig. 7. Functional areas with enough habitat to harbour a certain number of species (a, c) and areas with habitat mismatch (b, d) for semi-natural grassland dependent species. The number of species for the different habitat amount classes correspond to habitat thresholds in Fig. 3. The habitat mismatch maps are based on the distribution of species requiring > 370 ha at 15 km and > 131 ha at 5 km, representing moderately sensitive species. As an example, red areas (in b) lack at least 450 ha for being functional for the species occurring there, while green areas have a surplus of grasslands. Grey areas lack grassland data. The values for functional areas and habitat mismatch thresholds were chosen to illustrate the general pattern; alternative threshold values yielded similar spatial patterns Bergman et al. 2026, Landsc Ecol (2026) 41:101, CC BY 4.0.

North-east strong, south and west thin

For oaks the pattern is consistent at both scales: large functional areas in the north-east, and regions in the south and west with too few oaks to reach even the lowest threshold in the study.

200–300 oaks short

Across much of the south and west the negative mismatch for oak species runs 200–300 trees, and in places more than 300 — that is how many oaks are missing from landscapes where the species are still being recorded.

200–450 hectares short

For semi-natural grassland the equivalent deficit is 200–450 ha, over 450 in places. Functional grassland regions sit mainly in the north and the south-east.

Why mismatch is the useful map. A cell that is already functional needs holding; a cell 30 oaks short can be lifted over the line. The paper's argument for cost-effectiveness is exactly this — restoration that carries a landscape across a threshold buys far more than restoration that leaves it well below one.

Read this before using a number

What to be careful with

Rebuilding a paper from its own supplement surfaces things the printed version does not dwell on. None of this undoes the study. All of it changes how a particular number should be used.

1 24 of the 245 thresholds are negative numbers, for two different reasons

At P = 0.50, ignorance 20, 24 species-by-habitat-by-scale combinations return a negative habitat amount. Reconstructing each fitted curve shows these are not one phenomenon but two:

  • 19 rising curves that already sit above 50% occurrence probability at zero habitat. Extrapolating one backwards to find where it crosses 0.5 lands below zero. Taken literally, Table S2 asks for −5,187 hectares of grassland for Dactylorhiza maculata at 15 km.
  • 5 falling curves, where occurrence decreases as habitat increases. These never reach 50% at any positive habitat amount, and at zero habitat they sit at 3–17%, not above half. Melitaea britomartis, Plebejus argyrognomon, Satyrium pruni, Pyrgus alveus and Glaucopsyche alexis at their respective scales.

Neither is a habitat requirement, and neither should be read as a target. The paper notes the analogous problem for the P ≈ 0.2 inflection point that Rueda et al. propose, but not the negative values at P = 0.50. In the explorer above both are drawn as they actually are.

2 7 models run the wrong way, and two of them look fine

The paper reports significant positive relationships. But 7 of the 245 reconstructed curves have a negative slope: occurrence falls as habitat rises. Five of them surface as the negative thresholds above, where at least the number is obviously unusable.

The other two do not. Pulsatilla vulgaris at 15 km returns 22,116 ha and Thymus serpyllum at 15 km 1,049 ha — positive, plausible-looking figures that Table S2 presents like any other threshold. Both come from curves that say more grassland means less chance of finding the species, and both have an AUC below 0.55. A practitioner reading either number off the table would be setting a target derived from a relationship pointing the opposite way. Every chart on this page marks a falling model; the species table flags it too.

3 The paper never says which species failed

The paper reports that 86 of 102 species showed a significant positive relationship, and leaves the other 16 unnamed. Table S2 carries no p-values, so the only route to them is the colour of each marker in Fig. 3. Reading those colours here gives 86 species reaching p<0.05 and these 16 reaching it nowhere:

Clavaria zollingeri, Entoloma griseocyaneum, Hygrocybe ingrata, Hygrocybe ovina, Lasiommata megera, Melitaea britomartis, Microglossum atropurpureum, Microglossum olivaceum s.lat., Nymphalis antiopa, Parnassius apollo, Plebejus argyrognomon, Pulsatilla vulgaris, Pyrgus alveus, Satyrium pruni, Thymus serpyllum, Zygaena minos

Take that list as indicative, not as the paper's own. The count agrees with the paper's 86, but that is not the validation it looks like: two earlier and definitely wrong versions of this measurement also returned exactly 86, once from row positions that were twelve rows out at the bottom of the panel. A total can be right while the rows behind it are not. What the current version has going for it is structural: rows are seated on the axis ticks, the panel frame and gridlines are excluded before anything is read, and a row with no plotted marker is reported as unread rather than guessed — 28 of the 246 rows fall in that class, because their threshold lies outside the plotted axis. What it is reliable enough for is a warning flag, and the warning is worth having: for a species with no significant response the study establishes no threshold, yet Table S2 still prints a number against it. One of them, Plebejus argyrognomon, carries the highest 15 km grassland threshold of all 81 grassland species and has an AUC of 0.50 at the 5 km scale — exactly a coin flip.

4 Two percentages in the Discussion do not reconcile

The Discussion states a threshold of 7.9% grassland within 25 km² for Zygaena viciae and 8.7% for Gentianella campestris. Table S2 gives 111.6 ha for Z. viciae at the 5 × 5 km scale, which is 4.5% of the 2,500 ha cell, and 323.5 ha for G. campestris, which is 12.9%.

Neither figure can be recovered from any cell of Table S2 — not at any of the three ignorance levels, not at any of the three occupancy levels, and not at the 15 km scale. The two sets of numbers describe the same two species and disagree. The supplement is the auditable source, so this report uses it, and flags the discrepancy rather than choosing quietly.

5 Some grassland targets are larger than the landscape

Semi-natural grassland covers 3.4% of the study area. A 5 × 5 km cell is 2,500 ha, so an average cell holds about 85 ha of it; a 15 × 15 km cell about 765 ha. Against that, 91 of the 161 grassland thresholds sit above the study area's own mean grassland density.

The paper sees this and says so — grassland species show "threshold values at the top end of the available area of grasslands". But 2 thresholds go further and exceed the total area of the grid cell they are defined on: Plebejus argyrognomon at 31,865 ha within a 22,500 ha cell, and Aporia crataegi at 4,339 ha within a 2,500 ha cell — 142% and 174% of the cell respectively. No cell can contain that much grassland, so these are extrapolations past the edge of the predictor rather than habitat targets. At P = 0.75 six thresholds do the same.

6 Discriminatory power is thin for grassland species

Mean AUC is 0.767 for the oak models and 0.673 for grassland — the paper's own figures, reproduced here from the supplement. But 13 of the 245 models fall below 0.55, and 6 below 0.50, meaning they discriminate presence from absence worse than chance. The weakest is Glaucopsyche alexis at 0.40.

A threshold from a model that cannot separate presence from absence is a number without a claim behind it. The species table marks every non-significant model, and the ladder greys them out.

7 Four authorities, 24 disagreements about names

This report deliberately uses GBIF for taxonomy and iNaturalist for vernacular names and photographs, because a reader in Brazil or Japan can use both. That choice exposes how unsettled these names are. For 24 of the 102 species, at least one of the paper, GBIF's backbone, iNaturalist and the Swedish Red List 2025 uses a different name — and GBIF is the outlier as often as the paper is.

GBIF's backbone still calls Phengaris arion a synonym of Maculinea arion, Favonius quercus a synonym of Quercusia quercus, and Satyrium pruni a synonym of Fixsenia pruni — all older generic arrangements. In two cases it is simply wrong: it points Schismatomma pericleum at Lecanactis abietina and Sclerophora coniophaea at Sclerophora pallida, which are different species. Occurrence counts for those two therefore pool two taxa, and are marked accordingly.

The paper is not consistent with itself either: Table S2 says Hapalopilus croceus, Piptoporus quercinus and Arthonia byssacea where Fig. 3 says Aurantiporus croceus, Buglossoporus quercinus and Inoderma byssaceum.

Name in the paperDiffers inWhat is going on
Arthonia byssaceaGBIF, iNaturalist, Swedish Red List 2025Superseded. Now Inoderma byssaceum, the name the paper's own Fig. 3 uses and the name under which the Swedish Red List 2025 assesses it. Only Supplementary Table S2 retains the Arthonia placement.
Boloria euphrosyneGBIFGBIF's backbone points this at Clossiana euphrosyne, an older arrangement. Boloria euphrosyne is current.
Buellia violaceofusca—No longer a separate taxon. Ertz et al. (2018) showed that Buellia violaceofusca and Lecanographa amylacea share one fungal partner and differ only in their photobiont. The Swedish Red List 2025 carries no Buellia violaceofusca at all; the material is assessed as Lecanographa amylacea (NT). The paper knowingly kept the two apart as separate functional units because they associate with oaks of different ages — a defensible choice for modelling, but it means two of its 21 oak species are now one listed taxon.
Centaurium erythraea var. erythraeaiNaturalistRank only. Current floras treat this as Centaurium erythraea subsp. erythraea; the Swedish Red List 2025 assesses it as Centaurium erythraea (LC). Nothing about the circumscription has changed.
Clavaria zollingeriiNaturalistUnresolved. No Clavaria zollingeri appears anywhere on the Swedish Red List 2025, which instead carries Clavaria amethystina (violett fingersvamp, VU). Some treatments synonymise the two; others hold C. zollingeri to be the tropical species and European material to be C. amethystina. GBIF accepts C. zollingeri and gives it a global IUCN category of VU. This report leaves the national category blank rather than transferring a category across a contested boundary.
Favonius quercusGBIFGBIF's backbone points this at Quercusia quercus, an older arrangement. Favonius quercus is current and is what iNaturalist and the Swedish Red List 2025 use.
Hapalopilus croceusiNaturalist, Swedish Red List 2025The paper is inconsistent with itself. Supplementary Table S2 calls this species Hapalopilus croceus; the Discussion, Fig. 3 and Fig. 4 all call it Aurantiporus croceus. GBIF's backbone currently accepts Hapalopilus croceus; the Swedish Red List 2025 uses Aurantiporus croceus. Both names are in current use and refer to the same fungus, the saffron polypore.
Haploporus tuberculosusGBIFGBIF's backbone is behind here: it points Haploporus tuberculosus at Pachykytospora tuberculosa, which is the older arrangement. The Swedish Red List 2025 and current polypore treatments use Haploporus tuberculosus, the name the paper uses.
Hygrocybe ingrataiNaturalist, Swedish Red List 2025Superseded. Now Neohygrocybe ingrata, the name the Swedish Red List 2025 uses.
Hygrocybe ovinaGBIF, iNaturalist, Swedish Red List 2025Superseded. Now Neohygrocybe ovina, the name the Swedish Red List 2025 uses.
Inonotus dryadeusGBIF, iNaturalist, Swedish Red List 2025Superseded. GBIF treats Inonotus dryadeus as a synonym of Pseudoinonotus dryadeus; the Swedish Red List 2025 assesses it under Pseudoinonotus.
Lasiommata maeraGBIFGBIF marks its own record for this name DOUBTFUL, which is a defect in the backbone rather than a real nomenclatural problem: Lasiommata maera is a well-established species.
Lasiommata megeraGBIFGBIF marks its own record for this name DOUBTFUL, which is a defect in the backbone rather than a real nomenclatural problem: Lasiommata megera is a well-established species.
Lecanographa amylacea—Accepted and unchanged, but see Buellia violaceofusca: on the current Swedish list this name now covers both of the paper's two functional units.
Melitaea britomartisGBIFGBIF's backbone points this at Mellicta britomartis; Mellicta is now generally treated as a subgenus of Melitaea. The paper's name is current.
Microglossum atropurpureumiNaturalistDifferent name in: iNaturalist.
Microglossum olivaceum s.lat.iNaturalistThe paper models an aggregate. The Swedish Red List 2025 treats the aggregate as NA (not applicable) and assesses its segregates separately, so the NT the paper cites for 2020 has no 2025 counterpart at aggregate level.
Phellinus robustusGBIF, iNaturalist, Swedish Red List 2025Superseded. Now Fomitiporia robusta; the Swedish Red List 2025 assesses it under that name.
Phengaris arionGBIFGBIF's backbone still calls this Maculinea arion. Phengaris is the accepted genus following Fric et al. (2007), and is what iNaturalist and the Swedish Red List 2025 use. The paper's name is the current one.
Piptoporus quercinusGBIF, iNaturalist, Swedish Red List 2025Superseded. GBIF treats Piptoporus quercinus as a homotypic synonym of Buglossoporus quercinus (Schrad.) Kotl. & Pouzar, which is the name the paper's own Fig. 3 uses. Buglossoporus pulvinus (Pers.) Donk, seen in some British treatments, is a heterotypic synonym of the same species.
Platanthera bifolia subsp. bifoliaiNaturalistAccepted at subspecies rank; iNaturalist resolves the query to the parent species Platanthera bifolia, so the observation count shown covers the species, not the subspecies alone.
Satyrium pruniGBIFGBIF's backbone points this at Fixsenia pruni. Satyrium pruni is current and is what iNaturalist and the Swedish Red List 2025 use.
Schismatomma pericleumGBIFGBIF's backbone is wrong here: it points Schismatomma pericleum at Lecanactis abietina, a different lichen. The Swedish Red List 2025 assesses Schismatomma pericleum in its own right, and the occurrence counts shown here inherit GBIF's error — treat them with caution.
Sclerophora coniophaeaGBIFGBIF's backbone points Sclerophora coniophaea at Sclerophora pallida. These are treated as distinct species by the Swedish Red List 2025, which assesses S. coniophaea separately. The occurrence counts shown here follow GBIF and therefore pool the two.

8 Two of the 21 oak species are now one listed taxon

The paper models Buellia violaceofusca and Lecanographa amylacea separately, noting openly that Ertz et al. (2018) found them conspecific on the fungal partner and differing only in photobiont, and arguing they remain distinct functional units because they occupy oaks of different ages. That is a reasonable modelling decision.

On the current Swedish Red List it is no longer available: Buellia violaceofusca does not appear on the 2025 list at all — not even as LC, NE or NA — and the material is assessed as Lecanographa amylacea (NT). Anyone translating the paper's 21 oak species into red-list terms today gets 20 taxa.

9 The survey-effort weight, as printed, favours the cells nobody looked at

Occurrence records are weighted by Ruete's ignorance score. The paper gives it as I = threshold / (observations + threshold) with the threshold set to 20, and states plainly that the model was fitted "with habitat amount as the explanatory variable and I as weight".

Read literally that inverts the intent. I is a measure of ignorance: it approaches 1 in a cell with no background observations and 0 in a thoroughly surveyed one. Using it directly as the regression weight therefore gives the most influence to the cells whose absences are least trustworthy, and the least to the cells that were actually searched — the opposite of what weighting by survey effort is for. Weighting by 1 − I would do what the method section describes.

This report cannot tell from the outside which was done. The published thresholds are internally consistent and the AUC values are respectable, which is easier to reconcile with an effort-weighted fit than with an ignorance-weighted one, so the most likely explanation is a description slip rather than a modelling error. But it is the single sentence on which every number in Table S2 depends, and as written it says the wrong thing. Anyone rebuilding this analysis should settle it against the code before trusting the direction of the weighting.

Smaller things worth knowing

Table S2 is missing a row. Phengaris arion is reported at the 15 km scale only. Fig. 3c draws it at both. The 5 km model exists; its numbers were not published.

Fig. 3's bootstrap intervals are not reproduced here. The paper draws 5–95% bootstrap percentiles from 1,000 resamples. Those values are not in the supplement, so no chart in this report claims to show them. The intervals here are the span from P = 0.25 to P = 0.75, which is a different quantity and is labelled as such everywhere it appears.

One rounding disagreement. The Discussion gives Cliostomum corrugatum's threshold as "four oaks >150 cm dbh"; Table S2 gives 3.28, which rounds to three.

No corrections or errata exist. Checked on 4 August 2026: Crossref's update-relationship query for this DOI returns zero results, and its Crossmark record carries only the received, accepted and first-online dates and the competing-interests declaration. Nothing in this report is working around a published correction.

The paper's data are not deposited. The data-availability statement reads, in full: "The datasets generated and/or analysed during the current study are available from the corresponding author on reasonable request." No repository, no accession number, no dataset DOI. Everything in this report therefore comes from the published article and its three supplements, plus public APIs — which is also why it can be audited line by line.

The 2020 red-list column is measured off the figure too. The paper gives no table of red-list categories: they appear only as asterisks printed beside each species name in Fig. 3, where NT = *, VU = **, EN = *** and CR = ****. Every "2020" category on this page was read from those glyphs, twice — once from the full label and once from a narrow strip covering only the line ends, so a clipped name cannot lose an asterisk. Two consequences follow. A species with no asterisk is recorded here as LC, but the legend does not distinguish LC from NA, NE or DD, so that column is really "not in an NT–CR category". And the whole 2020 baseline is an image measurement, like the significance classes — not a figure the paper tabulates.

The significance classes are measured, not published. Table S2 carries no p-values, so the significance of each model was read off the colour of its marker in Fig. 3. The method is described in Methods below. 28 of 246 rows have no marker plotted at all — their threshold lies outside the axis — and are reported as unread. Six more sit between two reference greys and are flagged low-confidence. Any single row may still be misread.

Provenance

How this edition was made

Sources, and only these

The published article Landsc Ecol (2026) 41:101 — correctly cited as Landscape Ecology 41(6):101, ISSN 1572-9761, an article number rather than a page range — and its three supplements: Table S1 (spatial scales from the literature), Table S2 (AUC and habitat thresholds at three ignorance levels × three occupancy levels — 245 rows, the backbone of this report), and Fig. S1 (survey effort).

Conservation status from two places and no others. Nationally, Rödlistade arter i Sverige 2025 (24 March 2026, SLU Artdatabanken, Swedish University of Agricultural Sciences, Uppsala) via the open dataset Rodlistearbete_2025.csv — Red List work 2025: all red-listed taxa plus all taxa in the categories LC, NE and NA, DOI 10.5878/ffrv-6x57, CC0 1.0. Globally, the IUCN Red List version 2026-1 as republished in GBIF's IUCN checklist.

Taxonomy from the GBIF backbone (doi:10.15468/39omei); vernacular names, observation counts and photographs from iNaturalist. Both were queried on 4 August 2026, and every response is cached alongside the pipeline, so each figure can be traced to the payload it came from. Both are moving targets: a reader checking a record count or a name opinion later may find it changed. Neither Dyntaxa nor Artfakta is used as a taxonomic or linking authority: this edition is written for readers outside Sweden, and every species link resolves to a system with global coverage.

Recovering the fitted curves

Table S2 gives, for every model, the habitat amount at which occurrence probability reaches 0.25, 0.50 and 0.75. A binomial GLM with a logit link is logit(p) = a + b·x, so any two of those points determine a and b exactly. This report uses b = 2·ln3 / (x₇₅ − x₂₅) and centres the curve on x₅₀.

The reconstruction is checkable, because the third point is redundant. Estimating the slope from x₂₅→x₅₀ and from x₅₀→x₇₅ independently gives two numbers that should be identical. Across all 245 models the median disagreement is 2.4 × 10⁻⁶ and the worst is 2.5 × 10⁻³ — consistent with nothing worse than the two-decimal rounding of the published table. The curves in this report are the paper's models, not a fit to its summary statistics.

Checks the build must pass

The pipeline asserts rather than hopes. It fails if the species count is not 102, if the habitat groups are not 21 / 21 / 81, if the two oak dbh classes do not carry the same species, if any species appears in both habitat systems, or if any species × habitat lacks both spatial scales — with one documented exception, Phengaris arion at 5 km, asserted by name so that a future re-extraction silently dropping a different row fails the build.

Every red-list read passes keep_default_na=False. The Swedish list uses the literal string NA for not applicable, and the default behaviour of the CSV reader is to turn that into a missing value — which converts an assessed species into an unassessed one.

Two independent reproductions of the paper's own statistics were used as validation: mean AUC of 0.767 (oak, range 0.58–0.91) and 0.673 (grassland, range 0.40–0.90) against the paper's 0.77 and 0.67 with identical ranges. Reading significance off Fig. 3 gives 86 of 102, which is what the paper states — but that agreement is reported as weak evidence, not strong: two earlier versions of the measurement, both since shown to be wrong, returned 86 as well.

Reading significance off Fig. 3

Table S2 publishes no p-values, but Fig. 3 encodes them in the colour of each marker: light grey n.s., mid grey p<0.1, black p<0.05. The figure bitmaps were taken from the PDF at native resolution, each panel rectangle located from its rules, and the marker core read as the 8th percentile of non-background ink in a band around each row. The three colour classes were then found by 1-D k-means over the observed inks of each source bitmap — not hard-coded — and each row assigned to its nearest class.

Row positions come from the axis tick marks, detected one per species, and the build fails unless it finds exactly 21, 21 and 81 of them at regular spacing. An earlier version divided each panel's height into equal bands instead, on the assumption that a ggplot discrete axis is evenly expanded. It is not: on Fig. 3c that guess sat 1.6 rows too high at the top and 12 rows too high at the bottom, so most of the panel was being read against the wrong species — and it happened to return exactly the paper's 86, which is why the count alone is not a validation. The corrected geometry was checked by overlaying the detected row centres on the figure and confirming that each line passes through both its label and its marker.

The panel letters and the colour legend are printed inside the 5 km panels of Fig. 3a and 3b in black, and would be read as markers; those rectangles are blanked first, which is safe because both panels are sorted by descending threshold, so the affected rows have their markers far to the right.

Images and licences

The article is Open Access under CC BY 4.0, so its figures are reproduced here with attribution, downsampled and re-encoded to keep the page to a workable size. Species photographs come from iNaturalist, each centre-cropped to a square and resized to 168 px — an adaptation, which is why NoDerivatives photographs are excluded rather than embedded. Every one carries its photographer, the name of its licence, a link to the licence deed and a link to the original. 81 of the 102 species have one; the remaining 21 are either unphotographed on iNaturalist or have an all-rights-reserved lead photo, which is not redistributable and was therefore not embedded.

The page makes no external requests. All CSS, JavaScript, data, figures and photographs are inlined, so it works offline and from a local file.

Swedish Red List notation, for readers outside Sweden

The categories are the IUCN ones — LC least concern, NT near threatened, VU vulnerable, EN endangered, CR critically endangered, RE regionally extinct, DD data deficient, NA not applicable, NE not evaluated — applied at national scale.

Three national annotations appear in the categories on this page. A ° after a category means the outcome of the criteria was adjusted — most often downgraded because regular immigration from a neighbouring country reduces the risk of the species disappearing from Sweden, occasionally upgraded where the Swedish population is a sink dependent on immigration that is expected to decline. A following *, as in the NT°* and VU°* that appear here, marks an adjustment made for a reason other than immigration. CR (PRE) means critically endangered and possibly regionally extinct: not confirmed recently enough to be sure it is still present.

The 2025 list assessed 23,103 species; 5,217 are red-listed and 2,373 threatened. It also assesses lower taxa — the open dataset carries 1,727 of them (1,348 småarter plus 379 subspecies and varieties), where SLU's own summary of the publication gives 1,687. This edition quotes the dataset figure, because the dataset is what everything else here is derived from.

Reproducing this report

Six scripts, run in order. 10_species parses Table S2 into 245 model rows and 102 species records and runs the structural assertions. 20_enrich resolves every species against the GBIF backbone and iNaturalist and caches each raw API payload, so any printed value can be traced to the response it came from. 30_fig3_significance reads the significance classes off Fig. 3. 40_redlist joins the Swedish Red List 2025 on four name variants per species. 50_assets fetches and optimises photographs and figures. 60_build_report writes this page.

References

Bergman K-O, Andersson L, Franzén M, Johansson V, Westerberg L (2026) Identifying functional landscapes in Sweden for semi-natural grasslands and old-growth oaks (Quercus robur) based on habitat thresholds. Landscape Ecology 41:101. doi:10.1007/s10980-026-02373-4

SLU Artdatabanken (2026) Rödlistade arter i Sverige 2025 [Red-listed species in Sweden 2025]. SLU Artdatabanken rapporterar 37. SLU Artdatabanken, Swedish University of Agricultural Sciences, Uppsala. 150 pp. Published 24 March 2026. ISSN 2003-5373; ISBN 978-91-87853-90-6 (electronic), 978-91-87853-89-0 (print).

SLU Artdatabanken (2020) Rödlistade arter i Sverige 2020 [Red-listed species in Sweden 2020]. SLU Artdatabanken, Swedish University of Agricultural Sciences, Uppsala. Published 22 April 2020. ISBN 978-91-87853-55-5 (electronic), 978-91-87853-54-8 (print). The edition Bergman et al. cite as "SLU Artdatabanken 2020".

SLU Artdatabanken (2026) Red List work 2025: all red-listed taxa as well as all taxa in the categories LC, NE and NA. Dataset, version 1, CC0 1.0, 80,611 rows × 18 columns. doi:10.5878/ffrv-6x57 Deposited with the Swedish National Data Service; the file used here matches the published dimensions exactly.

Dahlberg A (2019) Hapalopilus croceus. The IUCN Red List of Threatened Species 2019: e.T58521209A58521216. Assessed 29 March 2019 as Vulnerable (A2c+3c+4c); reviewer Irmgard Krisai-Greilhuber. The assessment the paper cites for the species it calls Aurantiporus croceus; IUCN files it under the synonym.

IUCN (2026) The IUCN Red List of Threatened Species, version 2026-1, released 9 July 2026. iucnredlist.org

GBIF Secretariat (2026) GBIF Backbone Taxonomy. Checklist dataset doi:10.15468/39omei, accessed via the GBIF species and occurrence APIs on 4 August 2026. The backbone is rebuilt several times a year and occurrence counts change daily, so both the name opinions and the record counts on this page are a snapshot of that date.

iNaturalist (2026) iNaturalist taxon and observation data, accessed via the iNaturalist API. inaturalist.org

Ertz D, Guzow-Krzemińska B, Thor G, Łubek A, Kukwa M (2018) Photobiont switching causes changes in the reproduction strategy and phenotypic dimorphism in the Arthoniomycetes. Scientific Reports 8:4952. The source for Buellia violaceofusca and Lecanographa amylacea sharing one fungal partner.

Fahrig L (2013) Rethinking patch size and isolation effects: the habitat amount hypothesis. Journal of Biogeography 40:1649–1663.

Kuussaari M, Bommarco R, Heikkinen RK, et al. (2009) Extinction debt: a challenge for biodiversity conservation. Trends in Ecology & Evolution 24:564–571.

Poiani KA, Richter BD, Anderson MG, Richter HE (2000) Biodiversity conservation at multiple scales: functional sites, landscapes, and networks. BioScience 50:133–146.

Rueda M, Hawkins BA, Morales-Castilla I, Vidanes RM, Ferrero M, Rodríguez MÁ (2013) Does fragmentation increase extinction thresholds? A European-wide test with seven forest birds. Global Ecology and Biogeography. The LD50 analogy the habitat threshold is defined by.

Ruete A (2015) Displaying bias in sampling effort of data accessed from biodiversity databases using ignorance maps. Biodiversity Data Journal 3:e5361. The ignorance score used as the survey-effort weight.

Fric Z, Wahlberg N, Pech P, Zrzavý J (2007) Phylogeny and classification of the Phengaris–Maculinea clade (Lepidoptera: Lycaenidae). Systematic Entomology 32:558–567.

The article's full reference list — 90-odd works — is in the published paper. Listed here are the sources this edition relies on directly.