We combined herbarium material and observation records available on iNaturalist (https://www.inaturalist.org) to plan fieldwork in the municipality of Monchique, in the Algarve, and prepare the taxonomic treatment. A total of seven new gatherings of the putative new species were collected and deposited in public herbaria. Co-occurring angiosperms were recorded at all sites and identified using Flora iberica (Castroviejo et al. 1986–2021), adopting taxonomic updates from WFO (2026). A review of publications focused on the flora and vegetation of Serra de Monchique enabled the detection of two additional collections of the novelty. Both were first identified as L. spartea and were originally deposited at the herbarium of the University of Freiburg (Deil et al. 2008), which was integrated into the Staatliches Museum für Naturkunde Stuttgart in late 2024 (Thomas Ludemann and Stefan Abrahamczyk pers. comm.). A consultation of GBIF (2026) revealed no additional records of the new species from the Monchique Igneous Complex and adjoining areas, but one extra collection, identified as L. incarnata, was located through the virtual herbarium of Real Jardín Botánico, CSIC, Madrid (https://colecciones.rjb.csic.es). All herbarium material of the novelty was inspected as part of a near-comprehensive revision of L. subsect. Versicolores from the Algarve, building on a previously prepared dataset by Farminhão and Carapeto (2024). In total, we examined 143 specimens of L. subsect. Versicolores from the Algarve (Suppl. material 1) housed at ABH, ALGU, COI, LISE, LISI, LISU, MA, MACB, MAF, PO, and STU (acronyms following Thiers 2026). Scans of specimens housed at B, BR, L, P, W, and WAG were downloaded via GBIF (2026). Based on previous taxonomic treatments (Viano 1978; Sutton 1988; Sáez 2009; Fernández-Mazuecos et al. 2018a; Farminhão and Carapeto 2024), we selected 41 characters for the morphological study of L. subsect. Versicolores in the Algarve (Suppl. material 2). These include 34 quantitative characters, measured in 22 individuals of the new species (Suppl. material 3). Continuous characters were recorded to the nearest 0.1 mm using the “Measure” tool in ImageJ v.1.52d on previously acquired herbarium scans. Additionally, as in previous studies, seed morphology was inspected in detail using optical and scanning electron microscopy (SEM). Seeds were examined under an Emspira 3 digital microscope (Leica Microsystems) and photographed with Application Suite X (LAS X). For the SEM analysis, including the study of eight characters, seeds were gold-coated for micrographs to be taken with a Field Emission Scanning Electron Microscope (FE-SEM, GEMINI, SIGMA 300 VP, ZEISS). Synoptic tables were prepared to present the major morphological findings, highlighting the diagnostic characters of the new species.

Leaf tissue samples were collected from four populations of L. alfercensis and three of L. bimaculata, including one individual per population (Suppl. material 4). These samples represented the distribution ranges of the two species and included topotypic specimens, i.e. specimens collected at the type localities (Barranco da Fonte Santa for L. alfercensis, see below; and Faro, Gambelas for L. bimaculata). This sampling strategy was considered sufficient for our purposes, since we were interested in species-level phylogenetic relationships rather than population genetic parameters (cf. Eaton and Ree 2013; Andrews et al. 2016). Samples were preserved in silica gel until DNA extraction. A standard CTAB protocol (Doyle and Doyle 1987) was used to isolate total genomic DNA from all samples. The seven samples were included in a GBS library as part of a wider study of Iberian Antirrhineae (Alonso et al. in prep.). The GBS library preparation followed the procedure of Elshire et al. (2011) with modifications described in Escudero et al. (2014) and Fernández-Mazuecos et al. (2018b). In summary, we used 500 ng of DNA per sample, which were subjected to digestion using the PstI-HF restriction enzyme, and ligation of barcode and common adapters. Subsequently, 50 ng of each sample were pooled, and 50 µL of the pool were purified with AMPure XP magnetic beads (Beckman Coulter, CA, United States), eluting in 50 µL of milli-Q water. A MyCycler Thermal Cycler (BIO-RAD, CA, United States) was used to amplify DNA fragments for 19 PCR cycles starting from 35 ng of DNA. AMPure XP magnetic beads were used thereafter to purify the amplified library, and its concentration was quantified using a Qubit 3.0 Fluorometer (Invitrogen, CA, United States), with the 1X dsDNA HS Assay Kit. Quality control was conducted using a 2100 Bioanalyzer (Agilent, CA, United States). Finally, 30 µL of the genomic library were sent to Macrogen (Seoul, South Korea) for 150 bp paired-end sequencing in an Illumina NovaSeq X platform (Illumina, CA, United States).

GBS data for the remaining eight species of the Iberian clade of L. subsect. Versicolores, six representative species of the North African clade, including L. gharbensis Batt. & Pit., and the two species of L. subsect. Elegantes (Viano) D.A.Sutton (to be used as the outgroup) were obtained from Fernández-Mazuecos et al. (2018b). Individuals from Fernández-Mazuecos et al. (2018b) with a potential hybrid origin or low-quality sequencing results were excluded to facilitate phylogenetic inference, following that study. In total, GBS data for 88 individuals of L. subsect. Versicolores were used in the present study, including the newly generated data for four individuals of L. alfercensis and three of L. bimaculata (Suppl. material 4).

Assembly of GBS loci was performed using the ipyrad 0.9.107 pipeline (Eaton 2014; Eaton and Overcast 2020) on the CESGA supercomputing cluster (Santiago de Compostela, Spain). To ensure that the results were fully comparable with those of Fernández-Mazuecos et al. (2018b), we generally reproduced the procedures and parameters used in that study but implemented in a more recent version of the pipeline. Accordingly, we conducted de novo assembly, and only analysed the forward reads, thus treating the sequencing data as single-end. The latter approach provided higher resolution in preliminary analyses than analysing paired-end data. Given that assembly parameters are known to influence phylogenetic results, we generated assemblies using four different values of the clustering threshold (c = 0.84, 0.85, 0.87, and 0.92) that produced contrasting topologies in Fernández-Mazuecos et al. (2018b). For the minimum taxon coverage, we selected a low value of m = 4, following Fernández-Mazuecos et al. (2018b), who showed that this setting maximised the number of loci and consistently produced better phylogenetic resolution than higher values, albeit at the expense of a high proportion of missing data (cf. Rubin et al. 2012; Eaton et al. 2017). For the remaining assembly parameters, we also followed Fernández-Mazuecos et al. (2018b). Henceforth, the four assemblies are denoted as c84, c85, c87, and c92 according to the clustering threshold value. Demultiplexed sequences were deposited in the Sequence Read Archive (NCBI) under BioProject ID PRJNA1466108.

Phylogenetic analyses of the four datasets (generated using different c values) were conducted through concatenation-based and coalescent-based methods using full locus sequences. All analyses were run using the CIPRES Science Gateway (Miller et al. 2010). Concatenation-based analyses were conducted through maximum likelihood (ML), implemented in RAxML v.8.2.12 (Stamatakis 2014). We used the GTR substitution model with the CAT approximation of rate heterogeneity (Stamatakis 2006) during tree search, followed by evaluation and optimisation of the final tree under the GTR+GAMMA model. Statistical support of nodes was evaluated through non-parametric bootstrapping, with a number of replicates determined by the bootstopping criterion (Pattengale et al. 2010). Coalescent-based analyses were conducted using the SVDquartets method (Chifman and Kubatko 2014), implemented in PAUP v.4.0a (Swofford 2002). The multispecies coalescent model was selected, with samples assigned to their respective species, exhaustive quartet sampling, and the QFM algorithm for quartet assembly. For each matrix, 100 bootstrap replicates were conducted, following previous studies (Fernández-Mazuecos et al. 2018b, 2020), and results were summarised in a 50% majority-rule consensus tree.

All spatial data used in this study were compiled, processed, and analysed within a Geographic Information System (GIS) environment using ArcGIS Pro v.3.6.2. All data were projected to the study’s coordinate reference system (ETRS89/PT‑TM06) to ensure spatial consistency. Herbarium and observation records, available through iNaturalist (e.g. Suppl. material 5) and Flora-On (2026) were used to plot the distribution of the different taxa of L. subsect. Versicolores (viz. L. alfercensis, L. algarviana, L. bimaculata, L. spartea, and L. viscosa) in the Algarve. Linaria pedunculata (L.) Chaz., which is also present in the Algarve but belongs to the North African clade of L. subsect. Versicolores (Fernández-Mazuecos et al. 2013a), was not included in the analysis. After importing these datasets into ArcGIS Pro, the XY Table to Point tool was used to generate point features. For citizen science platform data, only records classified as reliable—following visual verification of associated photographs—and with a coordinate uncertainty smaller than 150 m were considered for this study. Subsequently, a few records with an original positional uncertainty greater than 150 m were also used, when the current presence of the species at the corresponding location was confirmed. Records falling outside the known ranges of each taxon (Flora-On 2026) were flagged and manually inspected by reviewing available photographs or, when necessary, by contacting the original observers. For L. algarviana and L. spartea, ‘geomorphogroups’ were defined as morphologically cohesive groups of individuals sharing the same geographical distribution over the same geological/edaphic units.

The base for the geological analysis was the Geological Map of the Algarve at 1:100,000 scale (Manuppella 1992), obtained from the National Laboratory for Energy and Geology (LNEG) via its Geoportal (https://geoportal.lneg.pt). All geological layers were georeferenced to the ETRS89/PT‑TM06 coordinate reference system. The lithostratigraphic designations used in this study follow the current nomenclature adopted by LNEG, which was preferred over the terminology present in the original printed map due to its alignment with the updated geological terminology used in Portugal. Administrative boundaries for the Algarve region were obtained from the Official Administrative Map of Portugal (CAOP), as published by the Directorate‑General for Territory (Direção-Geral do Território 2025).

A preliminary step in the geospatial analysis involved overlaying the compiled occurrence records of L. subsect. Versicolores with the geological map in order to identify the main geological formations associated with its occurrences. Based on this assessment, the formations considered most relevant were: (1) Mértola Formation: turbidites (shales and greywackes); (2) Silves sandstone; (3) Sands and gravels (Faro–Quarteira and Olhos de Água Formations). These units were subsequently digitised with higher spatial precision, using a maximum working scale of 1:2,000. All remaining geological units were digitised with lower precision (i.e. 1:30,000) and subsequently mapped as broader lithological categories based on their dominant composition: (1) shales and greywackes (other formations), (2) nepheline syenites, (3) limestones, marls, and other basic rocks, (4) fine sands and loose sandstones (Cacela Formation), (5) sands, sandstones, and gravels of Baixo Alentejo, (6) gravel beds and terraces, (7) aeolian sands, and (8) alluvium. All these combinations can be consulted in Suppl. material 6.

Small adjustments were made during the polygon digitisation process, particularly in areas occupied by water bodies (e.g. Ria Formosa, Arade Estuary) and other locations where the mapped geological boundaries were clearly misaligned with present‑day conditions observed in the field. All correction procedures were carried out using the World Imagery basemap (Esri 2024) to ensure accurate delineation of current landscape features.

A representative array of type specimens of Iberian and North African taxa of L. subsect. Versicolores was consulted via GBIF (2026) and the Global Plants Database (JSTOR 2000–2026), to test the morphological singularity of the new species and its endemicity to the Monchique Igneous Complex and adjoining areas, as supported by phylogenomic and geospatial analysis. The key to the Iberian clade of L. subsect. Versicolores by Fernández-Mazuecos et al. (2018a) was expanded to include L. alfercensis and L. bimaculata, and line drawings and photographic plates were prepared to illustrate the new species. Morphological terminology was standardised according to the Systematics Association Committee for Descriptive Biological Terminology (1962) and Beentje (2016). Vegetation is described in the habitat and ecology notes based on Rudner (2005), Deil et al. (2008), Neto et al. (2009), and Costa et al. (2012).

A risk of extinction assessment was prepared following Carapeto et al. (2020) and using the IUCN Red List guidelines (IUCN Standards and Petitions Committee 2024). Extent of Occurrence (EOO) and Area of Occupancy (AOO) were calculated using GeoCAT (Bachman et al. 2011).

Vegetative and floral diagnostic traits separating the new species from its morphologically and phylogenetically allied taxa are summarised in Table 1. Seed traits of Linaria alfercensis, L. algarviana, and L. bimaculata are presented in Table 2 and Fig. 1.

The sequencing of the GBS library yielded between 3.6 and 10.7 million reads for each of the seven newly sequenced individuals of L. alfercensis and L. bimaculata. The GC content was between 44 and 47%. Characteristics of the four assemblies are shown in Suppl. material 7.

Across concatenation-based RAxML analyses (Suppl. material 8A–D), the tree obtained from the c92 assembly (Fig. 2A) showed the best resolution within the Iberian clade of L. subsect. Versicolores, with each of the ten species (including L. alfercensis and L. bimaculata) supported as a monophyletic group with a bootstrap support (BS) = 100%, and all interspecific relationships supported by BS ≥ 83%. In L. subsect. Versicolores (BS = 100%), the North African clade (BS = 100%), and the Iberian clade (BS = 97%) were recovered as sister to each other. Within the Iberian clade, L. clementei was sister to the rest of the clade (BS = 99%), and a central–western Iberian subclade was recovered, including seven species (BS = 100%). Within the central–western Iberian subclade, an Algarvian subclade was obtained (BS = 100%), including L. alfercensis as sister to L. bimaculata plus L. algarviana (BS = 100%). The Algarvian subclade was recovered as sister (BS = 100%) to a clade including L. spartea and L. incarnata (BS = 83%). For the remaining assemblies (c84, c85, and c87), the same relationships were recovered, except for the sister group to the Algarvian subclade, which was unsupported (Suppl. material 8A–C).

All four coalescent-based SVDquartets analyses (Suppl. material 8E–H) recovered essentially the same topology, except for the uncertain position of L. clementei. The topology obtained from the analysis of the c92 assembly (Fig. 2B) was identical to that of the best-supported concatenation-based tree, except for the position of L. spartea, which was not sister to L. incarnata but formed a clade (BS = 83%) with L. viscosa plus L. onubensis. In all four analyses, the central–western Iberian subclade and the Algarvian subclade were supported with BS = 100%, and L. alfercensis was supported as sister to L. bimaculata plus L. algarviana, also with BS = 100% (Fig. 2B; Suppl. material 8E–H).

The Iberian clade of L. subsect. Versicolores is virtually absent from the basic rocks of the Meso-Cenozoic Algarve Basin, and from most of the Carboniferous shales and greywackes of the South Portuguese Zone (Fig. 3).

Most occurrences of L. algarviana and L. bimaculata overlap with the Plio-Pleistocene, rubefied coarse sands and gravels of the Faro–Quarteira Formation (FQF) (Fig. 3). Linaria algarviana (Fig. 4A, B) is mostly distributed in the Barlavento subregion of the Algarve, particularly west of Albufeira, where the FQF is interrupted by a relatively broad gap of limestones and other basic rocks that reaches the coast. A single subpopulation of L. algarviana occurs to the east of the ‘Albufeira Gap’, near Quarteira (Loulé), where it comes into contact with L. bimaculata (Fig. 4C), which locally exhibits a more coastal distribution. Besides being also present on the sands of the FQF, the “L. algarviana Costa Vicentina” geomorphogroup (Fig. 4A) occupies Quaternary gravel beds and aeolian sand deposits along the western coast of the Algarve, chiefly south of Ponta da Atalaia. It differs in having a more deeply coloured corolla and a typically more prostrate habit. The “L. algarviana Costa Central” geomorphogroup (Fig. 4B) occurs mostly on the FQF but occupies also a narrow stretch of the Triassic Silves Sandstone Line (TS) north of Portimão. To the east, L. bimaculata is predominantly distributed in the Sotavento subregion of the Algarve, from Quarteira to Pinheiro (Tavira). A single subpopulation occurs to the west of the Albufeira Gap, along the Galé–Arrifes coast. The easternmost limit of L. bimaculata is defined by limestones and the Cacela Formation, mostly composed of Miocene fine sands. A narrow band of limestones also delimits the distribution of L. bimaculata near Quarteira from the easternmost subpopulation of L. algarviana. The three eccentric occurrences of L. bimaculata (viz. Ponta da Piedade, Alcantarilha, Tunes), within the limestone belt (Fig. 3), coincide with anthropic sand deposits.

With a more hinterland distribution, L. alfercensis (Fig. 4D) occurs mostly on nepheline syenites along the eastern margin of the Monchique Igneous Complex (MIC), in Serra da Picota, with a single occurrence in the westernmost sector, near Marmelete (Fig. 3). Outside the MIC, it occurs along the Monchique river basin, a tributary of the Odelouca river, and in a single location on shales and greywackes to the southeast, near Pero Janeiro (Silves).

Plants ascribable to L. spartea s.l. (Fig. 4E–G, I, J) in the Algarve are distributed in four main geological settings (Fig. 3). In the northwest, plants with an erect habit, identified as L. spartea (Fig. 4E) occur in inland Cenozoic sandy deposits of the Sado Basin, and along the Seixe riverbed. Along the coast between Odeceixe and Ponta da Atalaia, the “L. cf. spartea var. expansa” geomorphogroup (Fig. 4F), with a decumbent habit, occupies mostly Quaternary aeolian sands. Separated by a ca 30 km mountainous gap of shales and greywackes, L. cf. spartea reappears, as the “L. cf. spartea Silves sandstone” geomorphogroup (Fig. 4G) along the TS, east of the Odelouca river, with a single occurrence in a neighbouring outcrop (Benaciate, Silves) of the FQF. Also along the TS, immediately east of the Odelouca river, we identified some putative hybrids between L. algarviana and L. cf. spartea> (Fig. 4H). To the northeast, the “L. cf. spartea Alcoutim–El Campillo” geomorphogroup (Fig. 4I) is predominantly distributed on the shales and greywackes of the Mértola Formation, with fewer records on the same rock types of other formations of the South Portuguese Zone, east to El Campillo (Huelva, Spain). Finally, the “L. cf. spartea Castro Marim–Isla Cristina” geomorphogroup (Fig. 4J) is present in small outcrops of the FQF near Azeda (Castro Marim), east to Isla Cristina (Huelva, Spain).

Linaria cf. viscosa occurs on Quaternary stabilised aeolian sand deposits west of the Guadiana mouth (Fig. 4K). To date, no taxa of the Iberian clade of L. subsect. Versicolores have been detected along the aeolian sand deposits from the barrier islands of Ria Formosa to Praia Verde (Castro Marim).

Morphological, phylogenomic, and geospatial data support the recognition of Linaria alfercensis as a new species within the newly recognised, fully supported Algarvian subclade of L. subsect. Versicolores, which is part of the wider central–western Iberian subclade. The Algarvian subclade also includes L. algarviana and L. bimaculata, with the three species being strictly endemic to the Algarve, in southern Portugal. The overall similarity of L. alfercensis in habit, inflorescence arrangement, and corolla colour to either L. spartea —with which it was previously confused (Deil et al. 2008)— or L. viscosa hypothetically results from convergent or parallel evolution within the central–western subclade. The phylogenetic topologies obtained here for the whole Iberian clade of L. subsect. Versicolores, now including ten species, are largely congruent with previous studies. As in Fernández-Mazuecos et al. (2018b), percentages of missing data in our GBS datasets were high (ca 89%), but they did not preclude the recovery of a highly-resolved tree. This agrees with previous studies showing that substantial levels of missing data in reduced-representation sequencing datasets do not necessarily compromise phylogenomic inference (Rubin et al. 2012; Eaton et al. 2017). Nevertheless, the exact position of L. clementei remains recalcitrant, possibly due to a hybridisation event (Fernández-Mazuecos et al. 2018b), which affects the delimitation of the Baetic subclade and deserves further investigation. The updated phylogenomic framework for the Iberian clade also confirms the validity of L. bimaculata as a well-supported species, sister to L. algarviana, as hypothesised by Farminhão and Carapeto (2024). More broadly, this study highlights the power of phylogenomic analyses based on genotyping-by-sequencing to disentangle recent plant radiations (Fernández-Mazuecos et al. 2018b, 2020; Otero et al. 2021). Additionally, it represents another example of the value of citizen science data, and particularly iNaturalist records, in accelerating taxonomic discoveries (Farminhão 2024; Mesaglio et al. 2025).

Our revision of citizen science records and herbarium specimens revealed remarkable morphological diversity among Algarvian populations of the Iberian clade of Linaria subsect. Versicolores surrounding those of L. alfercensis. Plants ascribable to “L. algarviana Costa Central”, with a more erect habit than typical L. algarviana, generally more light-coloured corollas and erect to erecto-patent corolla tubes (Fig. 4B) may fit the description of L. linogrisea Hoffmanns. & Link, nom. illeg. superfl. This name is based on material collected on the sandy fields of the Algarve at the end of the 18^th century (Hoffmannsegg and Link 1811) and is considered a synonym of L. incarnata (Viano 1978; Sutton 1988; Sáez 2009; Vigalondo et al. 2015). However, L. incarnata does not occur in the Algarve according to the most recent revision of the Iberian clade of L. subsect. Versicolores (Fernández-Mazuecos et al. 2018a), and the plants described as L. linogrisea by Hoffmannsegg and Link (1811) are now considered part of L. algarviana (Fernández-Mazuecos et al. 2018a; Farminhão and Carapeto 2024). Linaria algarviana s.str. (Fig. 4A) was described from Cabo de São Vicente (Chavannes 1833) and corresponds to our “L. algarviana Costa Vicentina” geomorphogroup. The slightly different “L. algarviana Costa Central” geomorphogroup identified as L. linogrisea by Hoffmannsegg and Link (1811) could represent a morphotype/ecotype of L. algarviana or a closely related taxon. Further research, including phylogenomics, is needed to clarify the taxonomic status of these populations.

Plants identified as a putative hybrid between L. algarviana and L. cf. spartea (i.e. LISI044803, Fig. 4H), recorded from a short stretch of the Silves Sandstone Line (SSL), near Arge (Portimão), and previously identified as a yellow morph of L. algarviana (Farminhão and Carapeto 2024), resemble L. alfercensis in corolla colour pattern, but have shorter corollas and more lax inflorescences, with the pedicels not appressed to the axis. They occur on a hill adjacent to the Odelouca river, which apparently acts as a possible dispersal corridor for L. alfercensis out of the Monchique Igneous Complex. This morphogroup should be targeted for future sampling along with populations of L. alfercensis from the Odelouca river.

Plants ascribable to L. spartea in the Algarve, attributed to four geomorphogroups in our geospatial analysis, should also be investigated within an integrative phylogenomic framework, to explore the potential occurrence of cryptic taxonomic diversity. Decumbent plants from the southwestern Portuguese coast agree with the type of L. spartea var. expansa (Samp.) Samp. described from Vila Nova de Milfontes, Baixo Alentejo (Sampaio 1946). Populations from northeastern Algarve may be identical to plants identified as L. spartea on similar geological formations east of the Guadiana in Spain. A sample of L. spartea from that area (Fuente de la Corcha, Huelva) revealed a candidate hybridisation event between L. spartea and L. onubensis (Fernández-Mazuecos et al. 2018b). Plants from the Plio-Pleistocene coarse sands of easternmost Algarve are similar to those found on similar geological formations in southwesternmost Spain, occupying apparently the same edaphic niche as L. bimaculata and L. algarviana to the west. Linaria spartea is widely distributed across the Iberian Peninsula and southwestern France and renowned for being highly polymorphic (Fernández-Mazuecos et al. 2018a), but eight geographically distant samples were recovered as a well-supported genetic cluster by Fernández-Mazuecos et al. (2018b) and again in the present study. Nevertheless, significant sampling gaps persist and we cannot rule out the occurrence of further phylogenetically distinct clusters, which may warrant taxonomic recognition if morphologically diagnosable.

Linaria viscosa was excluded from the flora of Portugal by Fernández-Mazuecos et al. (2018a), although it was subsequently reported immediately west of the Guadiana mouth, and east of Praia Verde (Castro Marim), by Farminhão and Carapeto (2024). These plants, occurring on Quaternary aeolian deposits on the coastal plain of the Guadiana mouth, are associated with Pinus pinaster Aiton and Retama monosperma (L.) Boiss. They occupy a similar habitat as genetically confirmed populations of L. viscosa from the Odiel and Guadalquivir river mouths, in southwestern Spain (Fernández-Mazuecos et al. 2018b). However, their morphology is somewhat intermediate between L. viscosa and L. spartea, and they occur in an area where the distribution ranges of the two species appear to meet. These populations may correspond to the name L. praecox Hoffmanns. & Link, probably described from Vila Real de Santo António (Hoffmannsegg and Link 1811; Farminhão and Carapeto 2024). Linaria praecox is currently considered a synonym of L. spartea (Sáez 2009; Farminhão and Carapeto 2024), although the intermediate morphology of these plants was already acknowledged by Sutton (1988). Other herbarium records of L. viscosa in the Algarve correspond to different geomorphogroups of L. cf. spartea in northeastern and northwestern Algarve. Their exact identification requires further studies.

Our analysis of seed morphology did not reveal prominent differences among the three species of the Algarvian subclade of L. subsect. Versicolores. The most conspicuous differential trait appears to be the height of the median papillae, which is shorter in L. algarviana (Fig. 1; Table 2). Seed characters present great diagnostic value for identifying some major clades of Linaria (Sutton 1988; Fernández-Mazuecos et al. 2013b). However, at lower taxonomic scales they are often less informative, owing to homoplasy (Fernández-Mazuecos et al. 2013b) or limited within-clade variation, as exemplified by the wingless, trigonous to reniform seeds with deep transverse ridges of the Iberian clade of L. subsect. Versicolores (Sáez 2009; Fernández-Mazuecos et al. 2018a).

Geospatial analysis of the Iberian clade of L. subsect. Versicolores confirms the calcifuge behaviour of this lineage in the Algarve, as more widely reported for the whole central–western Iberian subclade, which mostly occupies siliceous sandy substrates (Fernández-Mazuecos and Glover 2025). These include sandstone and coarse sand basins, igneous outcrops, and aeolian sand plains. Additionally, it reveals that some Miocene fine sand deposits (Cacela Formation), and turbidite formations (viz. Mira and Brejeira Formations), composed of shales and greywackes, are also virtually devoid of L. subsect. Versicolores. The exception to the latter is the Mértola Formation in northeastern Algarve, which concentrates most occurrences of L. cf. spartea in the region. The exact edaphic explanation for this pattern remains to be identified, but it is also evident in the neighbouring Portuguese region of Alentejo, where the same geological formations of the South Portuguese Zone overlap with an absence of records of L. spartea (Flora-On 2026; GBIF 2026). Overall, areas dominated by limestone, and other basic rocks, along with turbidites, seem to act as edaphic barriers within the clade. Conversely, Plio-Pleistocene coarse sand basins and nepheline syenitic outcrops act as habitat islands for L. subsect. Versicolores in the Algarve, highlighting some edaphic specialisation coupled with geographic isolation within this clade. We hypothesise that this distributional pattern, consistent with a model of edaphic archipelagos (de Luis et al. 2019), may also occur in other taxonomic groups, providing a new perspective for interpreting the biogeography of the Algarve within a broader southwestern Iberian framework of edaphically structured diversification associated with Mediterranean microrefugial dynamics.

Shifts in corolla colour from violet to yellow were also probably associated with speciation events within the Algarvian subclade, as inferred in other subclades of L. subsect. Versicolores (Fernández-Mazuecos et al. 2018b; Fernández-Mazuecos and Glover 2025). In total, five violet to yellow colour shifts are now suggested in the Iberian clade of L. subsect. Versicolores, in which a violet corolla is inferred as the ancestral state (Fernández-Mazuecos et al. 2018b). These transitions may have been associated with pollinator shifts, some of which could in turn have paralleled the inferred edaphic shifts, particularly considering that biotic interactions are reported to promote local adaptation to soil in plants (Dorey et al. 2024).

Linaria algarviana and L. bimaculata are endemic to the Plio-Pleistocene sand archipelago (PPSA) of the Algarve, mostly consisting of the Faro–Quarteira Formation (FQF). The Albufeira Gap, an edaphic barrier principally consisting of limestones, isolates most populations of L. algarviana in the west (Barlavento subregion of the Algarve) from most populations of L. bimaculata in the east (Sotavento subregion of the Algarve). Linaria algarviana, hypothetically originating in the west, was able to disperse to the Quarteira sector of the FQF, while L. bimaculata colonised the sand deposits on the top of the sea cliffs of Galé–Arrifes, immediately west of the Albufeira Gap. The Cacela Gap, another edaphic barrier consisting of limestones and Miocene fine sands (Terrinha et al. 2006), defines the easternmost limit of L. bimaculata. An eastward range expansion over the Albufeira Gap, or another major discontinuity within the PPSA, accompanied by a violet to yellow corolla colour shift, is a plausible scenario to describe the speciation event leading to this species pair. An even wider limestone gap seems to separate the two geomorphogroups of L. algarviana. In summary, limestone and fine sands appear to function as barriers isolating populations of L. subsect. Versicolores in the Algarve, potentially contributing to speciation. A potentially comparable edaphic archipelago scenario has been proposed for the Plio-Pleistocene inland sand ridges of the North American Coastal Plain, where geographically isolated sandy systems appear to have promoted diversification under strong edaphic niche conservatism in multiple angiosperm lineages (Schenk et al. 2018; Naranjo et al. 2023; Nevado et al. 2024). Accordingly, PPSAs may represent valuable systems for investigating continental edaphic diversification and identifying plant microrefugia.

Linaria algarviana has previously been reported from limestone substrates (Fernández-Mazuecos and Glover 2025). However, our geospatial analysis and field observations at a finer scale revealed that occurrences associated with carbonate rocks, actually overlap with mostly siliceous substrates: near Cabo de São Vicente, L. algarviana occurs on Quaternary sand and gravel deposits on top of limestone cliffs, harbouring other acidophilous species such as Rumex bucephalophorus L.; within the Barrocal limestone belt, smaller outcrops of FQF escape detection at larger scales. In turn, eccentric records of L. bimaculata, namely at Ponta da Piedade, Alcantarilha, and Tunes, in the Barlavento subregion, coincide with sand piles at disturbed sites, which suggests that these records may correspond to human-assisted introductions linked to construction sites. Therefore, this study underlines the importance of scale and ground truthing when interpreting the connection between geological maps and plant distributions.

Within the Algarvian clade, prior to the split between L. algarviana and L. bimaculata, L. alfercensis probably diverged after colonisation of the Monchique igneous island (MII) from the PPSA across the Brejeira Formation, a shale and greywacke barrier, where L. subsect. Versicolores is nearly absent. The presence of a mixture of different rock types, including nepheline syenite, along the Monchique and Odelouca riverbeds (Deil et al. 2008), offers an explanation for the presence of L. alfercensis outside the MII. Hydrochory along the subbasin draining Serra da Picota is a possible scenario to explain these outlying occurrences.

Within the MII, L. alfercensis is prominently absent from the Foia massif, with more humus-rich soils and a smaller extent of exposed bedrock compared to the Picota massif, but which otherwise shares the same geological context dominated by nepheline syenites (Malato-Beliz 1982). Nepheline syenites are a globally rare alkaline plutonic rock type, which generates thin soils with high Al, K, and Na concentrations, and an acidic upper sandy layer (Pereverzev et al. 2007; Ribeiro et al. 2007). Except for Brazil, where high levels of endemism are reported from syenitic outcrops in campo rupestre (Ribeiro et al. 2007), plant endemism on syenites is a little-studied phenomenon. In Flora iberica (Castroviejo et al. 1986–2021), syenites are only mentioned once, as the single substrate where Dianthus cintranus Boiss. & Reut. subsp. cintranus (Caryophyllaceae) can be found (Bernal et al. 1990), but this taxon actually occurs on other plutonic and volcanic rock types (Porto et al. 2026). In Mediterranean Europe, extremely few angiosperms have been reported to be mostly restricted to nepheline syenitic outcrops, with other examples being Armeria beirana subsp. monchiquensis (Bernis) Franco (Plumbaginaceae) (Vila-Viçosa and Arsénio 2021) and Taraxacum triforme Soest (Asteraceae) (Galán de Mera 2017), both endemic to the MII. However, chemical and physical properties of soils originating from the weathering of nepheline syenites, at least in temperate contexts, may not significantly differ from soils derived from other leucocratic plutonic rocks (i.e. poor in ferromagnesian minerals), such as granite. In this regard, the distribution of L. alfercensis may be interpreted as somewhat analogous to the occurrence of L. spartea on granitic massifs in the centre-west and northwest of the Iberian Peninsula. However, the particularly isolated position of the Monchique Igneous Complex, within a turbidite matrix, in an already described plant refugium area (Médail and Quézel 1997; Médail and Diadema 2009), suggests that this massif may represent an important edaphic island and overlooked endemism centre. Other narrow endemics of the MII, namely in the genus Sedum L. (Crassulaceae) and Arenaria Ruppius ex L. (Caryophyllaceae), await description (Farminhão et al. in prep.). Like similar small geological islands, this area escaped detection in large-scale screenings of plant diversity darkspots (Ondo et al. 2024).

Our study builds on recent literature on edapho-endemism (e.g. Voisin et al. 2026) by demonstrating that local geodiversity is underexplored in addressing the Linnaean shortfall, even in Europe’s well-studied flora. This is especially relevant when considering that most undescribed angiosperms are expected to be rare, narrow endemics, threatened with extinction (Brown et al. 2023). Indeed, L. alfercensis, preliminarily assessed as Endangered, requires immediate conservation attention, and should be included in a future addendum to the Red List of Vascular Plants of Mainland Portugal (Carapeto et al. 2020).

Geomorphogroups as first defined in this study are hypothesis-generating entities rather than implicit taxonomic units. This exploratory approach, of potential broad application, raised new questions on the ecology, taxonomy, and evolution of L. subsect. Versicolores in the Algarve and neighbouring southwestern Spain, which will be instrumental to guide additional sampling for an expanded phylogenomic backbone of this group, and further taxonomic and ecological investigations.