Characterisation and quantification of morphometric traits

Morphological measurements of all specimens within the E. tietkensii species complex held at the Western Australian Herbarium were assessed. Preliminary examination of herbarium specimens indicated that some E. tietkensii specimens had been misidentified and were likely to belong, instead, to one of the putative taxa (confirmed with herbarium experts); these specimens were re-determined accordingly. Morphological characters of 67 specimens of E. tietkensii, 18 of E. sp. Hamersley Range, 14 of E. sp. Rudall River and 6 of E. sp. Calvert Range were measured or assessed quantitatively and qualitatively (Table 1).

Table 1.

Leaf surface trichomes were examined using compound and scanning electron microscopy (SEM) to determine their morphology and anatomy. No discernible differences were found between abaxial and adaxial trichomes; so, data from both leaf surfaces were combined. For light microscopy (LM), leaf samples of ∼5 × 5 mm were taken from herbarium specimens and placed on a glass slide with sufficient water to rehydrate. The sample was left to soften for 1 min to aid in removal of trichomes, which were scraped onto the slide with forceps and examined under a compound light microscope. For SEM, leaf samples of ∼2 × 2 mm were taken from herbarium specimens, coated with gold using a JEOL Smart Coater sputter coater, and imaged using a Philips 505 Scanning Electron Microscope at the Western Australian Herbarium.

Measurements of vegetative, floral and fruit characters were made by hand with a ruler to the nearest 0.5 mm. For all quantitative characters, three measurements per individual were averaged, with ratios calculated before averaging. For leaf characters, the three largest fully developed leaves on each specimen were chosen. Of the 23 characters measured or scored, only those that were assessable on most specimens were included to ensure the number of individuals retained in the final dataset was maximised. The final dataset contained seven morphological characters: L1, L4, L5, L9, L10, L11, and O3 (Table 1). Leaf length (L1) and ratio (shape) characters (L4 and L5) were included, while leaf width (L2) and distance to widest point (L3) were excluded so as to remove logical auto-correlations between the ratios and their base measurements. The same reasoning was used to include petiole width (L9) and ratio of petiole length:width (L10) while excluding petiole length (L8). The apex shapes (L6 and L7) were excluded because they were highly variable within individual on specimens and could not be adequately scored. While potentially informative, comparison of sepal size and indumentum across all specimens was problematic because sepals in members of the E. tietkensii complex are accrescent, enlarging significantly during and after flowering, and glabrescent. Consequently, sepal length, width and indumentum were excluded. Density of glandular and eglandular hairs on the ovary, another potentially useful character, was not able to be used because of an insufficient number of specimens with ovaries at a comparable phenological stage; indumentum density was found to vary greatly as the ovary matured. Using mostly vegetative characters meant that a higher proportion of herbarium specimens could be included in analyses; this approach also had the advantage that plant identification can be made in the absence of flowers.

The dataset was analysed using non-metric multi-dimensional scaling (nMDS), principal coordinate analysis (PCoA) ordinations, and unweighted pair-group method with arithmetic mean (UPGMA) classification in PRIMER (ver. 6.1, see  https://www.primer-e.com/; Clarke and Gorley 2006). Differences within and among recovered groups were analysed using ANOSIM, which generates an R statistic with a theoretical range from 0 (indicates a random distribution) to 1 (perfect separation of groups; Clarke and Gorley 2006).

Specimens were assigned to broad geographic groups (Fig. 2) to enable assessment of geographic structuring or partitioning within the widespread E. tietkensii complex.

Fig. 2.

Collection locations of all PERTH voucher specimens of Eremophila tietkensii, E. sp. Hamersley Range, E. sp. Rudall River and E. sp. Calvert Range in WA. Specimens were assigned to a priori groups numbered 1—7.

Molecular analysis

In total, 101 herbarium specimens were sampled for a ddRADSeq molecular analysis. Four specimens from the Western Australian Herbarium collections of E. tietkensii and the phrase-named taxa were excluded because of insufficient material or poor quality (Table 2). Additional non-vouchered silica-dried material of each species, except E. sp. Calvert Range, was provided by B. Buirchell for inclusion in the molecular analysis. As the purpose of this study was to determine species boundaries rather than produce a phylogeny or test evolutionary hypotheses or hierarchies, an outgroup was not included.

Table 2.

Genomic DNA was extracted using the Qiagen DNEasy Plant Minikit (Qiagen, Hilden, Germany) following the manufacturer‚s protocol modified in the following ways: at Step 2, 1.35 mL of Buffer AP1, in addition to 1 μL of dithiothreitol for every 1 mL of Buffer AP1, and 4 μL of RNase A was added to each tube containing 80 mg of plant material, vortexed, then samples were incubated for 30 min at 65°C; and at Step 11, 40 μL of Buffer AE heated to 65°C was added to each spin column filter for elution and then incubated at room temperature (15—25°C) for 10 min.

DNA concentration was quantified with a Qubit dsDNA BR Assay Kit and a Qubit 3.0 Fluorometer (Invitrogen, Carlsbad, CA, USA). Fragment lengths were checked with a LabChip GX Touch 24 Nucleic Acid Analyser using the HT DNA gDNA reagents (Perkin and Elmer, Waltham, MA, USA). Of the 109 samples, 86 were of suitable quality to be used for library preparation.

Library preparation followed the procedure in Severn-Ellis et al. (2020). DNA extracts were diluted to 300-ng concentration per sample with nuclease-free water and digested with 5 units of HpyCH4IV restriction enzyme (New England Biotechnologies (NEB), Ipswich, MA, USA), and 5 U of HinfI restriction enzyme (NEB) per sample and 3 μL of the master mix was added to each sample. Barcoded and common adapters designed to complement the pairs of restriction enzyme overhangs were prepared as described by Peterson et al. (2012). Ligated fragments were cleaned of excess, unligated adapters, and fragments within the range of 250—800 base pairs were selected using Ampure XP beads (Beckman Coulter, Brea, CA, USA) and enriched by polymerase chain reaction (PCR) amplification by using Phusion Hot-Start High-Fidelity Polymerase Master Mix (Thermo Fisher Scientific, Waltham, MA, USA).

Pooled libraries were cleaned (Ampure XP beads), quantified with Qubit and visualised and quality-checked using the LabChip GX Touch 24 with HT DNA HiSens Dual Protocol Reagents. Molarity of the final library was calculated and 50—100 μL of the ddRAD bead cleaned library was diluted to a final concentration of 10—20 nM by using normalisation buffer. The pooled ddRAD libraries were sequenced (2 × 150-bp reads) across two Illumina HiSeq X lanes at the Kinghorn Centre for Clinical Genomics (KCCG) Sequencing Laboratory in Darlinghurst, New South Wales, Australia.

Sequence data analysis

De novo assembly of RAD loci and single-nucleotide polymorphism (SNP) calling was performed following the bioinformatics workflow described in Severn-Ellis et al. (2020, see  https://github.com/ascheben/RAD_analysis_workflow/). Pooled data were de-multiplexed using Stacks (ver. 2.53, see  https://catchenlab.life.illinois.edu/stacks/) process_radtags (Catchen et al. 2013), with barcode rescue (-r), quality filtering (-q, -c) and RAD tag checks (—renz_1 HpyCH4IV —renz_2 HinfI). Low-quality reads were discarded and reads were trimmed (144 bp) with Trimmomatic (ver. 0.39, see  http://www.usadellab.org/cms/index.php?page=trimmomatic; Bolger et al. 2014). Quality checks using FastQC (ver. 0.11.9, see  https://www.bioinformatics.babraham.ac.uk/projects/fastqc/; Andrews 2010) were conducted, followed by a preliminary diversity assessment based on pairwise distances using Mashtree (ver. 0.37, see  https://metacpan.org/dist/Mashtree; Katz et al. 2019), whereafter individuals with extremely low genotyping rates were removed. De novo assembly of RAD loci and SNP calling was performed using Stacks (ver. 2.53; Catchen et al. 2013). For high levels of intraspecific divergence, it has been recommended that n = M in the Stacks software (Paris et al. 2017), and was considered appropriate for the present study. Stacks parameters of M-3, m-3, and n-3 were chosen following experimentation with different parameter settings (Paris et al. 2017) rendering the highest number of polymorphic loci in 80% of the populations studied.

Further filtering with VCFtools (ver. 0.1.16, see  https://vcftools.github.io/; Danecek et al. 2011) was executed to remove indels and retain only biallelic, high-quality SNPs. The minor allele frequency (MAF) filters rare alleles, and the threshold was set at 0.05. The filters that remove the most SNPs are usually depth (−minDP) and the missingness (max-missing) filters. For heterozygous samples, read depths ≥5 have been suggested to avoid undercalling of heterozygous genotypes (Maruki and Lynch 2017; Bilton et al. 2018). Depth and missingness filters were fine-tuned to find a balance between the quantity and quality of SNPs. Read depths of ≥3 (DP3) and ≥5 (DP5) were evaluated and a read depth of 5 (MAF = 0.05) was chosen. Missingness was maintained at 90% across samples (max-missing = 0.8). Finally, individuals with low genotyping rates (missingness >50%) based on the filtered SNPs were removed.

A principal component analysis (PCA) was conducted using the gdsfmt, SNPRelate, gridExtra, and ggrepel packages in R (RStudio, Inc., Boston, MA, USA) to visualise genetic diversity using detected SNPs and to assess whether samples formed clusters congruent with the results of the morphometric analyses.

Key morphometric differences among taxa

Examination of abaxial and adaxial leaf trichomes by using both SEM and LM showed that trichomes of Eremophila sp. Hamersley Range were substantially different from those of the other three taxa in the complex (Fig. 3a ). Trichomes in E. sp. Hamersley Range were distinctly septate with six to eight cells of approximately equal length, with the terminal cell with a rounded tip (Fig. 3a—b , 4f —i ). By contrast, E. tietkensii trichomes had four to five cells per trichome, with the terminal cell comprising half to one-third of the total length of the trichome and being attenuate at the apex (Fig. 3c—d , 4a—e ). Trichomes in E. sp. Rudall River (Fig. 3e—f , 4j— k ) and E. sp. Calvert Range (Fig. 3g—h , 4l —m ) were similar to those of E. tietkensii but were slightly longer (up to 350 μm in E. sp. Rudall River and 300 μm in E. sp. Calvert Range, compared with 250 μm in E. sp. Hamersley Range and E. tietkensii) and narrower (12—25 μm wide in E. sp. Rudall River and 10—22 μm wide in E. sp. Calvert Range, compared with 20—30 μm in E. sp. Hamersley Range and E. tietkensii), and tended to have slightly longer terminal cells. Overall, E. sp. Hamersley Range was the only taxon that could be distinguished on the basis of trichomes; the remaining three taxa differed only slightly and were highly variable.

Fig. 3.

Scanning electron microscopy (SEM) images of adaxial trichomes on leaves of Eremophila sp. Hamersley Range: (a) PERTH 03557464, (b) PERTH 09105972; E. tietkensii: (c) PERTH 03856275, (d) PERTH 03899918; E. sp. Rudall River: (e) PERTH 03878759, (f) PERTH 04201159; and E. sp. Calvert Range: (g) PERTH 03878570, (h) PERTH 07512821.

Fig. 4.

Trichomes under compound microscope at 400× of Eremophila tietkensii: (a) PERTH 08316899, (b) PERTH 03881202, (c) PERTH 06752519, (d) PERTH 03851281, (e) PERTH 07324545; E. sp. Hamersley Range: (f) PERTH 06653537, (g) PERTH 0653561, (h) PERTH 06017142, (i) PERTH 8521; E. sp. Rudall River: (j) PERTH 08305447, (k) PERTH 03878740; and E. sp. Calvert Range: (l) PERTH 07765886, (m) PERTH 07512821.

Eremophila sp. Calvert Range had a sericeous eglandular indumentum on the ovary that is distinct from that of the other three species (Fig. 5a ). Ovaries in Eremophila sp. Hamersley Range had a sparse to moderate covering of glandular and eglandular hairs (Fig. 5b ); those in E. sp. Rudall River mostly lacked eglandular hairs and had a moderate covering of glandular hairs, and, in E. tietkensii, ovaries mostly lacked eglandular hairs and had a moderate to dense covering of glandular hairs.

Fig. 5.

Ovary indumentum of (a) Eremophila sp. Calvert Range, showing sericeous eglandular trichomes; and (b) E. Hamersley Range, showing glandular and eglandular trichomes.

Morphometric groupings of taxa

Non-metric multidimensional scaling (nMDS) and UPGMA clustering showed that Eremophila sp. Hamersley Range and E. sp. Calvert Range are morphologically distinct from E. tietkensii and E. sp. Rudall River; however, each taxon is morphologically closer to E. tietkensii than to each other (Fig. 6a ). Whereas, overall, E. tietkensii and E. sp. Rudall River formed somewhat separate groups, some E. tietkensii individuals were morphologically more similar to E. sp. Rudall River than they were to other E. tietkensii specimens. Eremophila tietkensii showed high levels of morphological variability, with several individuals forming groups away from the rest of E. tietkensii when an 80% similarity threshold based on the UPGMA classification was applied (Fig. 6b ). Eremophila sp. Hamersley Range and E. sp. Calvert Range each clustered as one group both at the 70 and 80% similarity threshold (Fig. 6a ). Eremophila sp. Hamersley Range and E. sp. Calvert Range formed distinct groups in the UPGMA classification (Fig. 6b ). Eremophila sp. Rudall River and E. tietkensii formed weakly separated groups, with several individuals of E. tietkensii not being sufficiently distinguished from E. sp. Rudall River to join the larger E. tietkensii group (Fig. 6b ).

Fig. 6.

(a) Non-metric multi-dimensional scaling (nMDS) ordination of all individuals on the basis of selected traits (Table 1) overlain with groupings from cluster analysis (b); and (b) UPGMA classification of all individuals on the basis of selected traits (Table 1).

Pairwise ANOSIM tests between Eremophila sp. Hamersley Range and E. sp. Calvert Range and between E. sp. Calvert Range and E. sp. Rudall River returned an R statistic of 1, indicating a complete separation of groups on the basis of morphological characters (Table 3). Comparisons of Eremophila tietkensii and E. sp. Hamersley Range (R = 0.96, P < 0.001), and E. tietkensii and E. sp. Calvert Range (R = 0.88, P < 0.001), also showed high separation between groups (Table 3). The weakest separation of groups was between E. tietkensii and E. sp. Rudall River (R = 0.5, P < 0.001), as observed in the ordination (Fig. 6a ). When E. tietkensii and E. sp. Rudall River were analysed separately from the rest of the complex, the strength of separation of groups increased only slightly (R = 0.52, P < 0.001, Table 3).

Table 3.

When considering only Eremophila tietkensii and E. sp. Rudall River, the E. sp. Rudall River group was more similar to the larger E. tietkensii group at the 60% similarity threshold than five E. tietkensii individuals (PERTH 08317178, PERTH 08332215, PERTH 06752519, PERTH 03975320, PERTH 08316937; Fig. 7). At the 80% similarity threshold, two E. sp. Rudall River specimens were more similar to E. tietkensii than to each other. Individuals from the same geographic location did not cluster together, indicating that morphological characters were not correlated with geographical location (Fig. 7). Geographic Groups 1 and 3 clustered together at the 70% similarity threshold despite Group 3 being ∼1000 km east of Group 1.

Fig. 7.

Non-metric multi-dimensional scaling (nMDS) ordination of Eremophila tietkensii and E. sp. Rudall River individuals with pedicels present overlain with groupings from cluster analysis. Numbers indicate geographic areas of individuals as shown in Fig. 2.

The five Eremophila tietkensii individuals that clustered with E. sp. Rudall River had unusually short and wide leaves. One specimen (PERTH 08332215) occurred on the border of the E. sp. Rudall River population, and in the classification and ordination based on morphology was grouped with E. sp. Rudall River; it may have been misidentified initially. One specimen (PERTH 03975320) was the southernmost occurring E. tietkensii individual, ∼625 km from the nearest E. sp. Rudall River populations. Two specimens (PERTH 08316937 and PERTH 08317178) had a more compact habit than did the other three. They were collected 7 years apart and located ∼1 km from each other, but ∼500 km away from the nearest E. sp. Rudall River populations. One specimen (PERTH 06752519) occurred even closer to the E. tietkensii geographic Group 1, some ∼585 km from the nearest E. sp. Rudall River populations. These four specimens are unlikely to be misidentifications of E. sp. Rudall River; even if some specimens were reassigned from E. tietkensii to E. sp. Rudall River and vice versa, the morphological separation between these two entities was found to be weak.

No clear structure was found in the principal coordinate analysis (PCoA) performed on Eremophila tietkensii only. Axis 1 (PCo1) explained 54.9% of the variation and was correlated mainly with mean leaf and petiole length. Axis 2 (PCo2) explained 19.6% of the variation and was correlated with the distance to widest point:length ratio (Fig. 8). Despite the wide geographic range of the samples, there was no clear separation into geographically distinct morphotypes.

Fig. 8.

Principal coordinate analysis (PCoA) for six traits of all Eremophila tietkensii individuals. Numbers represent geographic groups as per Fig. 2.

Molecular analysis

Dried and carefully preserved herbarium material was found to be a generally excellent source of DNA material for ddRADSeq analysis. There was no clear correlation between the age of herbarium specimen and the quality or quantity of DNA suitable for use in the molecular analysis following demultiplexing and trimming (  Supplementary Fig. S1 (SB21005_AC.pdf)). The two oldest samples that passed filtering and quality checks and were used in the SNP analysis were collected in 1941.

A total of 1 851 822 978 sequence reads were generated, with 1 712 241 716 remaining following demultiplexing. After trimming, 1 511 865 102 reads remained. The total number of variant sites before filtering was 14 164 739. After removing samples with >90% SNPs missing, 67 of 78 individuals remained, and after filters were applied, 39 396 variant sites remained. Finally, when a single SNP from each ddRAD locus based on the distance between SNPs was randomly kept and samples with over 50% missing SNPs were removed, all 67 individuals remained, with a total of 7032 SNPs.

The first two components of the PCA based on SNPs explained 46.5% of the variation (Fig. 9). The broad geographic groupings within Eremophila tietkensii were recovered on the PCA ordination, except that E. sp. Rudall River sat between and mixed with E. tietkensii geographic Groups 2 and 3 and was not genetically divergent from them (Fig. 9). The E. tietkensii specimen from Exmouth (Group 4) clustered with geographic Groups 1 and 2 and did not appear to be genetically differentiated. Eremophila tietkensii Group 2 clustered with Groups 1 and 3, despite Group 3 being very geographically distant. Eremophila sp. Hamersley Range and E. sp. Calvert Range were strongly genetically segregated from each other and from the rest of E. tietkensii and E. sp. Rudall River, despite their close geographical proximity to all groups (Fig. 9).

Fig. 9.

Principal component analysis of detected SNPs of Eremophila tietkensii, E. sp. Hamersley Range, E. sp. Calvert Range, and E. sp. Rudall River. Inset shows positions of E. tietkensii and E. sp. Rudall River specimens.

Congruence between morphology and molecular patterns of variation

There is a high degree of congruence between patterns of morphological and molecular variation in the Eremophila tietkensii species complex. Morphological variation resolved three groups, comprising E. sp. Hamersley Range, E. sp. Calvert Range and E. tietkensii (including E. sp. Rudall River). The same three groups were strongly resolved in the SNP analysis. Eremophila sp. Rudall River could not be adequately segregated from E. tietkensii in either the morphological or molecular analysis and resolved between E. tietkensii geographic Groups 2 and 3 either genetically or geographically. While the morphometric analysis did not resolve geographic structuring in morphology in E. tietkensii, there was distinct geographical patterning between the a priori geographic groups on the basis of the molecular analysis (Fig. 9). The genetic patterning was not strong enough to indicate that E. tietkensii could be reasonably taxonomically split on the basis of geography; the spread of molecular variation across its range was slightly but not substantially greater than for the other taxa, indicating that gene flow, altough restricted, is sufficient to have prevented allopatric speciation. The lack of consistency in morphological types of E. tietkensii within geographic groups, in conjunction with the genetic structuring among geographic groups, suggests that differences in morphology may be the result of phenotypic plasticity. Detailed analysis of substrate, habitat and underlying geology may explain differences in phenotypic variation and drivers of species distributions, because these factors have been used to explain variation and distribution in other Eremophila species (Coates et al. 2014).

Eremophila sp. Hamersley Range was genetically highly divergent from E. tietkensii despite its close geographic proximity. Although this analysis was unable to estimate time since divergence our results suggest that these two taxa have been genetically isolated for a long period of time. A growing number of studies of evolutionary history and distribution patterns of Pilbara biota are showing similar patterns of divergence between species (e.g. Anderson et al. 2016). Since the formation of the Great Sandy and Gibson deserts <1 million years ago, the Pilbara region has been separated from similar landforms by extensive sand plains (Pepper et al. 2013b ). Species ranges contracted and expanded throughout this time, including those of Acacia spp. (Ladiges et al. 2006), Triodia spp. (Anderson et al. 2016), reptiles (Doughty et al. 2011; Pepper et al. 2013a ), beetles (Guthrie et al. 2010) and spiders (Durrant et al. 2010). Owing to the topographical heterogeneity of the Pilbara compared with surrounding low-lying plains, localised areas within the region were likely to be more climatically stable, acting as refugia for species that could not otherwise persist in the arid conditions outside of these environments (Byrne et al. 2016, 2017). These processes may have contributed to the high level of genetic divergence between E. sp. Hamersley Range and E. tietkensii.

The resolution of the Eremophila tietkensii complex provided here will allow the taxa to be included in future phylogenetic studies where reconstruction of patterns and timing of divergence among lineages can be inferred, providing insights into the evolutionary history of species in north-western Australia. These results will also assist in biological surveys, because these species can now be accurately discriminated, and population studies can be undertaken. We also note here the largely untapped potential of herbarium material for ddRADSeq molecular analyses for such studies. DNA extraction from herbarium material is often considered to be a supplement to field sampling (Beck and Semple 2015). We found that the quality of DNA obtained for molecular analyses was excellent in carefully stored specimens and not strongly correlated with specimen age. This observation is also consistent with those of recent similar studies elsewhere (Bakker et al. 2016; Forrest et al. 2019; Joyce et al. 2021). Overall, we have demonstrated that even without additional sampling, species delimitation can be achieved in an accurate, efficient and cost-effective manner by using ddRADseq to sample the genomes of herbarium specimens.

Taxonomic implications

The congruence between morphological and molecular patterns of variation, together with discriminatory characters, allows the recognition of Eremophila sp. Hamersley Range and E. sp. Calvert Range as morphologically and genetically distinct from E. tietkensii. These taxa occur in close geographic proximity and are likely to have the opportunity to interbreed. Their clear genetic and morphological separation indicates that they do not do so. This, in turn, indicates that they are likely to comprise independently evolving metapopulation lineages without significant gene exchange. Therefore, we believe that they should be recognised as distinct species under the unified species concept (de Queiroz 2007). By contrast, the similar morphology and lack of clear genetic separation between E. tietkensii and E. sp. Rudall River indicates that the latter cannot be recognised at species rank and should be absorbed back into E. tietkensii.

Accordingly, below we formally name and describe as new species Eremophila naaykensii A.L.Curtis & K.R.Thiele and E. hurteri A.L.Curtis & K.R.Thiele, and circumscribe E. tietkensii to include specimens currently phrase-named as E. sp. Rudall River.

Eremophila tietkensii F.Muell & Tate, Trans. Proc. & Rep. Roy. Soc. S. Australia 8: 109 (1890)

Type: Laura Vale, Northern Territory, [June] 1889, W. H. Tietkens s.n. (holo: MEL 82820).

Eremophila latrobei var. tietkensii (F.Muell. & Tate) Ewart & P.H. Jarrett [see  https://id.biodiversity.org.au/name/apni/114129/api/apni-format], Proc. Roy. Soc. Victoria 40: 87 (1928) [see  https://id.biodiversity.org.au/instance/apni/548054].

Eremophila pachomai Paczkowska & A.R.Chapman, W. Austral. Fl. Descr. Cat. 339 (2000), nom. inval. [manuscript name; no Latin description or diagnosis provided or referenced]

Eremophila sp. Rudall River (P. G. Wilson 10512), Western Australian Herbarium: L. J. Biggs & C. M. Parker, Nuytsia 23:504 (2013).

Rounded to flat-topped shrub 0.6—2(−3) m tall, aromatic. Young stems covered in a persistent, fine, grey to yellowish, appressed tomentum of simple hairs, obscurely tuberculate beneath the indumentum; older stems with grey to very pale grey, slightly fissured bark, at first with prominently raised and knob-like persistent leaf bases. Leaves scattered, pale greyish-green or grey, petiolate; petioles (2.5—)5—10(—16) mm long; lamina ovate to lanceolate, (19—)32—57(—91) × (4—)10—21(—32) mm, smooth; indumentum dense, very short, appressed, white to grey, velutinous, often matted-resinous, comprising simple, uniseriate hairs, the terminal cell much longer than the others and usually attenuate; margins entire; apex acute, attenuate, or mucronate. Flowers 2—4 per axil, pedicellate; pedicels (5—)10—14(—25) mm long, with indumentum as for stems. Sepals 5, imbricate, subequal, elliptic to oblanceolate, broadly acute to obtuse with a mucro, (7—)10—13(—18) × (1.5—)3—5(—7) mm, pinkish-purple to mauve, maroon or red, pubescent with +appressed, tangled hairs, the margins more densely so, enlarging after flowering and then glabrescent and with prominent veins. Corolla 22—28 mm long, pale blue, blue, pale lilac to pale mauve, white tinged lilac, mauve or pink; outer surface of lobes and tube with scattered eglandular hairs particularly near the margins, often almost glabrous; mid-inner tube with moderate density of eglandular hairs. Stamens 4, included; filaments with long eglandular hairs towards base, glabrous above; anthers glabrous. Ovary ovoid-oblong, densely glandular-puberulous with scattered or numerous longer eglandular hairs; style glabrous or with a few scattered, simple, spreading, short eglandular hairs. Fruit dry, woody, ovoid-conical, +beaked, ribbed, 6—7 × 3—4.5 mm; exocarp adhering to endocarp, glandular-puberulous but usually with some longer eglandular hairs, occasionally resinous; endocarp vertically ribbed, splitting into four segments towards apex.

Eremophila naaykensii A.L. Curtis & K. R. Thiele, sp. nov.

Eremophila sp. Hamersley Range (K. Walker KW 136) Western Australian Herbarium: L. J. Biggs & C. M. Parker, Nuytsia 23: 504 (2013).

Type: Hamersley Range (specifically Hancock Range) within mining tenement E-47/1329-I neighbouring Mining Area C, within Juna Downs Street, ∼103 km WNW of Newman townsite, 5 km E of Great Northern Highway, Western Australia, 21 Feb. 2018, C. van den Bergh CV Opp 18 (holo: PERTH 09105972!).

Rounded to obconical shrubs or small trees 1—2.5(—3.5) m tall, aromatic. Young stems clothed in a persistent, fine, grey to yellowish, appressed tomentum of simple hairs, obscurely tuberculate beneath the indumentum; older stems with grey to very pale grey, slightly fissured bark, at first with prominently raised and knob-like persistent leaf bases. Leaves scattered but tending to be clustered towards the stem apices, pale green or grey-blue, petiolate; petioles (6—)9—13(—18) mm long; lamina lanceolate, (37—)55—71.5(—89) × (5—)7.5—12(—15) mm, smooth; indumentum dense, very short, appressed, white to grey, velutinous, often matted-resinous, comprising simple, uniseriate hairs that are evenly septate, the terminal cell no longer than the others and with a bluntly rounded tip; margins entire; apex attenuate. Flowers (1)2—4 per axil, appearing clustered in the dense, terminal leaf clusters, pedicellate; pedicels (20—)28—33.5(—40) mm long and +sigmoidal, with indumentum as for stems. Sepals 5, imbricate, subequal, elliptic to oblanceolate, broadly acute to obtuse, sometimes mucronate, (7—)8—10(—14) mm × (2.5—)3—5(—6), yellowish, greenish, red or purple-tinged in flower (likely to be colouring further after anthesis), pubescent with +appressed, tangled hairs, the margins more densely so, enlarging after flowering and then glabrescent and with prominent veins. Corolla 20—28 mm long, cream, pale blue, lilac, yellow, pink or purple sometimes with spots on upper lobe, the throat and inside of tube pale yellow to cream; outer surface of lobes and tube with scattered eglandular hairs particularly near the margins, often almost glabrous; mid-inner tube with moderately dense eglandular hairs. Stamens 4, included; filaments with long eglandular hairs towards base, glabrous above; anthers glabrous. Ovary sparsely to moderate pubescent with glandular and eglandular hairs, ribbed; style with short, patent, eglandular hairs for most of its length. Mature fruits not seen.

Eremophila hurteri A.L.Curtis & K.R.Thiele sp. nov.

Type: base of Calvert Range (campsite), Calvert Range, WA, 7 August 2000, A. A. Burbidge 738 (holo: PERTH 07512821!).

Eremophila sp. Calvert Range (A. A. Burbidge 738) Western Australian Herbarium: L. J. Biggs & C. M. Parker, Nuytsia 23: 504 (2013).

Intricate flat-topped shrubs 1—1.5 m tall, aromatic. Young stems with indumentum of short, woolly, usually yellowish, sometimes grey, hairs, sometimes appearing sericeous, obscurely tuberculate beneath the indumentum; older stems grey to dark brown, scarcely fissured, often distinctly tuberculate, at first with prominently raised and knob-like persistent leaf bases. Leaves scattered, silvery, petiolate; petioles (7—)8—10(—11) mm long, decurrent; lamina lanceolate, (45—)50—69.5(—84.5) × (9—)9.5—12.5(—14) mm, finely strumose; indumentum dense, very short, white to grey, woolly, often matted-resinous, comprising simple, uniseriate hairs, the terminal cell much longer than the others and attenuate; margins entire; apex attenuate. Flowers 1 or 2 per axil, pedicellate; pedicels (4.5—)9—13(—15) mm long, straight to curved, with indumentum as for stems. Sepals 5, imbricate, subequal, elliptic to oblanceolate, broadly acute to obtuse, sometimes mucronulate, 7—9 × 2—3.5 mm in flower, yellow in bud, turning white or pink or mauve at anthesis, densely short-tomentose with +silky hairs, enlarging after flowering and then glabrescent and with prominent veins. Corolla 20—28 mm long, white to pale purple or mauve; outer surface of lobes and tube with scattered eglandular hairs particularly near the margins, often almost glabrous; mid-inner tube lanate with eglandular hairs. Stamens 4, included; filaments with woolly eglandular hairs towards base, glabrous above; anthers glabrous. Ovary densely sericeous with yellow, simple, eglandular hairs; style with sparse, long spreading, eglandular hairs for most of its length. Mature fruits not seen.