We mainly relied on herbarium materials from 15 herbaria (A, BR, CHR, CONC, GOET, E, M, MA, MO, MSB, NY, SI, UPOS, and WS; codes according to Thiers, 2020) to study 45 species (49 taxa) belonging to subg. Psyllophorae (Supplementary Table 1): 22 species plus four varieties from sect. Junciformes (all except C. archeri Boott, C. boelckeiana Barros, and C. moorei G. A. Wheeler); all 7 species of sect. Psyllophorae; and 17 included in sect. Schoenoxiphium (all except C. acocksii C. Archer, C. chermezonii Luceño and Martín-Bravo, C. gordon-grayae Luceño, Márq.-Corro and Sánchez-Villegas, and C. sciocapensis Luceño, Márq.-Corro and Sánchez-Villegas; see Luceño et al., 2021 for more information). Our dataset comprises 87% of all the extant diversity known for subg. Psyllophorae (following Roalson et al., 2021 with some modifications; 53 species) including all main lineages and representing the geographical and morphological variability of the subgenus. Carex phalaroides s.l. constitutes a complex species much in need of detailed study; so we opted for a synthetic view and considered it as a single species (herein C. gibertii G. A. Wheeler, C. moesta Kunth, and C. hypsipedos C. B. Clarke are treated as C. phalaroides). We used as outgroup one or two representatives of each of the other five Carex subgenera (Villaverde et al., 2020): C. hypolytroides Ridl. and C. siderosticta Hance (subg. Siderostictae Waterway), C. canescens L. and C. gibba Wahlenb. (subg. Vignea), C. flava L. and C. dissitiflora Franch. (subg. Carex), C. arctogena Harry Sm. (subg. Euthyceras Peterm.), and C. meridensis (Steyerm.) J. R. Starr (subg. Uncinia Pers.).

We used four DNA regions, the nrDNA ITS and ETS, and the ptDNA matK and rps16. These markers were selected because they have been used successfully in previous studies in Carex in general and in subg. Psyllophorae in particular (Gehrke et al., 2010; Jiménez-Mejías et al., 2016a; Villaverde et al., 2017b; Márquez-Corro et al., 2020). DNA extraction and sequence amplification followed Jiménez-Mejías et al. (2016a), except for rps16 amplification that was performed as indicated in Shaw et al. (2005). All PCR products were sequenced by Macrogen (Madrid, Spain). Sequence chromatograms were manually edited using Geneious v. 11.0.2 (Biomatters Ltd., Auckland, New Zealand).

Four independent matrices were compiled, each one containing sequences for one of the DNA regions (ITS, ETS, matK, or rps16). The sequences were automatically aligned with Muscle v.3.8.425 (Edgar, 2004) and alignments were manually corrected in the flanking regions of indels. Informative indels were coded as a binary character using SeqState v.1.4.1. according to Simmons and Ochoterena (2000) simple coding method. A complete multiaccession matrix (105 accessions, 23.26% missing data) was built concatenating all markers.

We performed maximum likelihood (ML) and Bayesian Inference (BI) phylogenetic analyses using RAxML v.8.2 (Stamatakis, 2014) and MrBayes v.3.2 (Ronquist et al., 2012), as implemented in CIPRES Science Gateway (Miller et al., 2010), with parameters detailed in Supplementary Data 1.

We inferred divergence times in subg. Psyllophorae with a Bayesian analysis implemented in BEAST v.1.10 (Suchard et al., 2018) establishing three primary calibration points based on reliable Carex fossils (Jiménez-Mejías et al., 2016b): C. colwellensis (Eocene: 38.0-33.9 mya) for the crown node of the genus, C. marchica (Early Miocene: 23.0-16.0 mya) and C. hartauensis (late Oligocene: 28.1-23 mya) for the stem node of subg. Vignea and Carex, respectively (Table 2). We also used a secondary calibration point for the crown node of subg. Psyllophora (24.41 mya; Martín-Bravo et al., 2019). Prior node age distributions and analyses settings are detailed in Supplementary Data 1.

Diversification rates for subg. Psyllophorae were estimated on the dated singleton tree using the speciation-extinction model implemented in BAMM (BAMMtools package; Rabosky et al., 2014) using R v.3.6 (R Development Core Team, 2019). This analysis was performed in order to estimate the diversification rates by the effective sample size of the log-likelihood and number of shift events along branches using reversible jump Markov chain Monte Carlo (rjMCMC) to infer sample mixtures of distinct evolutionary rate dynamics across the branches. Four chains of 1.000.000 generations, saving trees each 1,000 (see Supplementary Data 1 for more information) were performed.

Occurrences belonging to sect. Junciformes (Supplementary Table 2) were downloaded from GBIF (https://www.gbif.org/), iDiGBio (https://www.idigbio.org), and the ALA database (https://www.ala.org.au/). In addition, we manually georeferenced 154 vouchers from 14 herbaria (A, CONC, E, LIL, M, MICH, MO, MSB, NY, SI, TRIER, UPOS, US, and WS), as well as 115 records obtained from literature sources (Barros, 1948, 1950, 1957; Wheeler, 1988, 1989, 1998; Wheeler and Muñoz-Schick, 1990; Wheeler and Guaglianone, 2003; Wheeler and Beck, 2011; Jiménez-Mejías et al., 2020). We cleaned the resulting database by removing duplicates and unreliable records after visual inspection, and finally, 616 occurrences were recovered (Supplementary Table 2).

We retrieved 37 bioclimatic and ecophysiological variables from WorldClim2 (Fick and Hijmans, 2017) and Envirem (Title and Bemmels, 2018) to characterize the bioclimatic niche of our study group. Our variable selection procedure (Supplementary Data 1) retained five bioclimatic variables that contributed more than the rest to the variability of our data, and have easy biological interpretation for our species: Annual mean temperature (AMT), temperature annual range (TAR), mean temperature of the driest quarter (MTDQ), annual precipitation (AP), and terrain roughness index (TRI). These variables illustrate species-climatic limits regarding water availability (AP), temperature optimum (AMT), and limits (TAR, MTDQ), as well as microclimatic determinants driven by topography (TRI). We conducted a principal component analysis (PCA) of the retained variables using the package ggplot2 (Wickham, 2006) and the function prcomp in R for visualizing the environmental space of sect. Junciformes.

We performed ancestral state reconstruction of ecological preferences for sect. Junciformes using the dated pruned singleton tree and the mean value for each retained variable in each species as implemented in the package phytools in R (Revell, 2012; Figure 2A). We previously assessed two models of continuous trait evolution [Brownian motion (BM) and Ornstein-Uhlenbeck (OU)] and selected the best using corrected Akaike Information Criterion (AICc) (fitContinuous function in geiger package; Harmon et al., 2008).

In order to study fine-level ecological differences within sect. Junciformes, we obtained and plotted response curves for each retained variable for the main sister lineages (Figure 2B) and single species (Supplementary Figures 1, 2). To characterize the environmental space for a whole lineage, we merged all occurrences of the corresponding species. To further evaluate the Phylogenetic Niche Conservatism hypothesis between sister lineages from SA and NZ within the Aciculares-clade (see Results), we also performed tests of similarity and equivalence (Warren et al., 2008; Broennimann et al., 2012) implemented in the package ecospat (Di Cola et al., 2017). We also measured the niche overlap between these sister clades with Schoener's D index (Schoener, 1968).

Ancestral area reconstruction (AAR) was performed using the package BioGeoBEARS (Matzke, 2014a) in R. We performed two different biogeographic reconstructions focusing on two different evolutionary-geographic scales. First, we performed a large-scale analysis based on our complete sampling for the whole subg. Psyllophorae, coding species distribution according to six biogeographical regions (Figure 3A): North America, SA, Eastern Palearctic, Western Palearctic, Afrotropical region, and Australasia (including NZ). For species in subg. Psyllophorae, we obtained distributions from the World Checklist of Selected Plant Families (WCSP) (Govaerts et al., 2020), while for each outgroup tip, the coded distribution was obtained from the ancestral area of the correspondent subgenus as inferred by Martín-Bravo et al. (2019). Second, we focused only on sect. Junciformes to explore in detail its biogeography, delimiting subregions within the Neotropics and NZ, based on their current (Supplementary Figure 3) and inferred present potential distribution (Supplementary Figure 4; see Supplementary Data 1 for methodological details). As a result, we considered seven areas (Figure 3C): Northern Andes, Central Andes, Northern Patagonia (N Patagonia), S Patagonia, Atlantic South America, Falklands, and NZ. As input data, we used the pruned calibrated tree and a file recording the presence/absence of species in each area.

Both datasets were analyzed under three different models: Dispersal-Extinction-Cladogenesis (DEC; Ree and Smith, 2008), a likelihood version of the Dispersal-Vicariance analysis (DIVA-like; Ronquist, 1997), and a likelihood interpretation of BayArea model (Landis et al., 2013) implemented in BioGeoBEARS (BAYAREA-like; Matzke, 2014a), and each of these in combination with the founder-event speciation (+J model; Matzke, 2014b). The fit of the different models was tested using AICc (Burnham and Anderson, 2002).

Subgenus Psyllophorae was recovered in a strongly supported monophyletic group (Supplementary Figure 5) including sects. Psyllophorae, Schoenoxiphium, and Junciformes, which also formed well-supported clades. Section Psyllophorae (Clade A) was sister to Clade B which in turn contained sects. Junciformes and Schoenoxiphium as sister groups (see Supplementary Data 2 for a detailed description of phylogenetic results).

Divergence times obtained with BEAST are shown in Supplementary Figure 6. The crown age of subg. Psyllophorae coincided with the early Miocene (mean = 22.55 mya, 95% highest probability density (HPD) = 20.76–24.44 mya). The most recent common ancestors (MRCA) of sect. Psyllophorae (Clade A; mean = 18.40 mya, 95% HPD = 15.01-21.56 mya) and sects. Schoenoxiphium and Junciformes (crown node of Clade B; mean = 20.04 mya, 95% HPD = 17.25-22.84 mya) were inferred to have diversified just after the subgenus as a whole, still in the early Miocene. Interestingly, both sections from the Southern Hemisphere were originated almost at the same time in the early Miocene (Clade B1: sect. Schoenoxiphium, mean = 16.10 mya, 95% HPD = 12.5-19.47, and Clade B2: sect. Junciformes, mean = 16.05 mya, 95% HPD = 12.86-19.17 mya). Within sect. Junciformes, we found two main clades diversified in late Miocene (Aciculares-clade, mean = 8.60 mya, 95% HPD = 5.6–11.89 mya; core Junciformes, mean = 10.25 mya, 95% HPD = 7.89–12.82 mya). On the one hand, Junciformes-clade (including the monospecific lineage of C. argentina) seems to have diversified during the middle Miocene (mean = 11.66 mya, 95% HPD = 9.04–14.38 mya). The MRCA of core Junciformes was dated to the late Miocene (mean = 10.25 mya, 95% HPD = 7.89–12.82 mya), and in turn, the Lineages A and B derived from this ancestor were also inferred to have differentiated in the late Miocene (Lineage A: mean = 7.71 mya, 95% HPD = 5.6-10 mya; Lineage B: mean = 8.46 mya, 95% HPD = 6.14-10.9 mya). On the other hand, within Aciculares-clade, the diversification of SA-clade was during the late Miocene (mean = 6.44 mya, 95% HPD = 4.06–8.95 mya), whereas NZ-clade probably occurred a bit later during early Pliocene (mean = 4.87 mya, 95% HPD = 2.79–7.28 mya).

Bayesian Analysis of Macroevolutionary Mixtures (BAMM) analyses estimated a similar and constant diversification pattern along the three sections (Psyllophorae, Schoenoxiphium, and Junciformes; Figure 3B) of subg. Psyllophorae. In addition, it did not detect significant shifts in diversification rate.

The PCA (Supplementary Figure 7) of the variables using mean values for all occurrences per species of sect. Junciformes displayed a clear trend toward variables related to temperature along the first principal component, which explained 42.1% of the total variance. Precipitation (AP) and terrain roughness (TRI) were negatively correlated with this axis and displayed lower values than temperature ones, while they contributed positively to the second principal component that showed 32.4% of the total variance explained. Therefore, the cumulative proportion of the variance explained by both principal components was 74.5%.

The best evolutionary model for bioclimatic reconstruction over sect. Junciformes phylogenetic tree according to the corrected AICc value was the Ornstein–Uhlenbeck model (BM = 112.10, log-likelihood = −53.76; OU = −34.99, log-likelihood = 20.76). The bioclimatic niche evolution of the five selected variables on the obtained phylogeny is shown in Figure 2A, as well as density functions exhibiting the frequency of the distribution of the bioclimatic variables in Figure 2B. In general, similar values across internal nodes of the phylogeny were retrieved for each independent variable (Figure 2A). This is also depicted by the mostly overlapping and relatively narrow peaks for the different sister clades (Aciculares-SA vs. NZ; Junciformes lineages A vs. B; see also Supplementary Figures 1, 2) in the response curves (Figure 2B). The deep nodes along the phylogeny displayed medium values on their bioclimatic preferences, while tips revealed the existence of more extreme values (Figure 2A).

Despite the general scenario of relatively uniform bioclimatic values across sect. Junciformes, some degree of heterogeneity was observed for certain variables in particular lineages. Thus, TAR was the variable that showed the highest environmental heterogeneity along the phylogeny, with contrasting bioclimatic preferences for Junciformes-clade (C. andina Phil., C. molinae Phil., C. patagonica Speg., and C. setifolia Kunze). The monospecific lineage of C. phalaroides showed high values for AMT in comparison to low-medium values in the rest of the species of sect. Junciformes. In addition, closely related species of Aciculares NZ-clade presented contrasting bioclimatic preferences across different variables, such as AP and TRI, as well as TAR and TRI for some species belonging to Junciformes-clade (C. setifolia and C. transandina). Finally, the high values displayed in the irregularity of terrain (TRI) were associated with species that inhabit a wider altitude range across NZ and SA. In addition, C. setifolia revealed higher values of MTDQ in comparison to medium-lower ones for the other tips. Otherwise, C. transandina G. A. Wheeler displayed contrasting values for AP and TRI, more similar to species from Aciculares NZ-clade than Junciformes Lineage B to which it is phylogenetically more related (Figure 2A and Supplementary Figure 2).

The PCA comparing the e-space for NZ and SA clades (Aciculares-clade) revealed a large overlap between both groups within both principal components (67% explained of the total variance; Supplementary Figure 8). This niche overlapping was also revealed by Schoener's D index (D = 0.523). In addition, pairwise statistical comparison for similarity and equivalence tests suggested that these clades were also significantly more similar and equivalent than expected by chance (p < 0.01).

Dispersal-vicariance analysis (DIVA-like model) yielded lower corrected AICc values than DEC and BAYAREA-like model (76.84, 81.15, and 111.8, respectively), with ΔAICc > 2 (Burnham and Anderson, 2002). DIVA-like and DEC models resulted in biologically congruent reconstructions. However, the BAYAREA-like model displayed scarce uncertainty for node inferences although this model is more limited in the number of evolutionary scenarios considered (i.e., vicariance is not considered). In order to simplify the results, we mainly focused on the DIVA-like model (Figure 3A). Results from DEC and BAYAREA-like models are presented in Supplementary Figures 9, 10, respectively. The stem node of subg. Psyllophorae (Oligocene) was restricted to the Palearctic, from where it dispersed to SA (crown node widespread in the Western Palearctic and SA, early Miocene). The differentiation of sect. Psyllophorae took place in the Western Palearctic from the early Miocene. For the clade embracing sects. Schoenoxiphium and Junciformes, the ancestor was recovered as widespread in the Afrotropical region and SA during the early Miocene. The differentiation of each of these two sections implied the constriction of their crown node to one of the two landmasses. While the arrival to SA necessarily implied a LDD event from the Old World, the early Miocene colonization of the Afrotropical region remains obscure, since we cannot rule out the dispersal from SA or dispersal/vicariance from the Western Palearctic. Finally, a LDD event to SW Pacific from SA across the Mio-Pliocene boundary led to the more recent diversification of the NZ-clade (Figure 3A). Within-area lineage diversification from the middle Miocene onwards was inferred for the remaining subclades within the three sections commented above (Figure 3A and Supplementary Figures 9, 10).

Regarding the biogeographic reconstruction for sect. Junciformes, the DEC model yielded lower AICc values than the DIVA-like and BAYAREA-like models (141.2, 142.4, and 143.3, respectively), although with ΔAICc <2 between DEC and DIVA-like models. Thus, we decided to focus on the DEC model (Figure 3C). Results for DIVA-like and BAYAREA-like models are presented in Supplementary Figures 11, 12, respectively. A widespread ancestral area including Patagonia as a whole and NZ was inferred for the crown node of sect. Junciformes at the late early Miocene, although uncertainty was present under the three biogeographic models tested with ranges probabilities below 0.10 (Figure 3C and Supplementary Figures 11, 12). DIVA-like model inferred N Patagonia, while BAYAREA-like model inferred both N-S Patagonia and Central Andes.

In the Aciculares-clade and the DEC model, a widespread area in N Patagonia and NZ was inferred for the MRCA (late Miocene). On the one hand, a LDD event from N Patagonia gave rise to the NZ-clade at the Mio-Pliocene boundary. On the other hand, from the late Miocene onwards, the diversification of SA-clade took place within N-S Patagonia, with at least two later Plio-Pleistocene independent range expansions (to the Falklands -C. caduca Boott- and Northern Andes -C. via-incaica Jim. Mejías and Roalson-). Remarkably, a similar scenario of Patagonian diversification was inferred for the Junciformes-clade, although starting earlier (from the Middle Miocene) than in the Aciculares-clade, and restricted to the N Patagonia. Subsequent expansions took place southwards to the S Patagonia (Pleistocene; Lineage B), and northwards to the Central Andes (from the Late Miocene; C. setifolia). The isolated lineages of C. camptoglochin, C. vallis-pulchrae, and C. phalaroides underwent unprecedented range expansions within the section, including multiple extra-Patagonian regions.

We preferred models not accounting for the J parameter for the AAR analyses as these results should be treated with caution (Ree and Sanmartín, 2018).

Our sanger-based phylogenetic findings suggest that subg. Psyllophorae, as delimited by Roalson et al. (2021), is a well-supported monophyletic group (Supplementary Figure 5), but not including C. baldensis L. and C. curvula All. The placement of these two species in this subgenus needs further study and it was considered provisional by Roalson et al. (2021) as they were placed on a very short branch in the phylogenomic reconstructions (Villaverde et al., 2020). Three well-supported major clades are identified within subg. Psyllophorae, each one representing a section (Psyllophorae, Schoenoxiphium, and Junciformes), unlike previous phylogenies where lower taxonomic and molecular sampling yielded less robust reconstructions (Jiménez-Mejías et al., 2016a; Martín-Bravo et al., 2019). The prolonged isolated evolution of some species (Supplementary Figure 6) has yielded a remarkable morphological differentiation within sections (e.g., C. distachya, C. camptoglochin, and C. phalaroides), which has traditionally caused taxonomic problems in their sectional delimitation (Wheeler and Guaglianone, 2003; Luceño, 2008; Silveira and Longui-Wagner, 2012). The monophyly of sect. Schoenoxiphium is well-supported (Clade B1; Supplementary Figure 5), which agrees with previous studies focusing on that group (Luceño et al., 2021, and references therein) despite resolution problems reported for its inner phylogenetic structure. Meanwhile, C. camptoglochin and C. phalaroides are nested within sect. Junciformes clade, supporting their treatment within that section (Roalson et al., 2021). Section Aciculares as conceived by Wheeler (1989) is not monophyletic due to the position of C. transandina and C. vallis-pulchrae with respect to the members of sect. Aciculares (Supplementary Figure 5). As currently circumscribed, sect. Junciformes is primarily characterized by the presence of inflorescences formed by a single androgynous spike, usually dense, and with a staminate tip concealed by the female part. The only exception is C. phalaroides, with multispicate inflorescences, which seems a reversion to the ancestral state of the genus (Roalson et al., 2021). In addition, C. camptoglochin, which displays reflexed utricles with a protruding rachilla that may play an epizoochorous role (Villaverde et al., 2017a), has been traditionally included in sect. Leucoglochin Dumort. primarily because of this distinctive morphological character (Wheeler and Guaglianone, 2003).

Subgenus Psyllophorae is inferred as one of the oldest lineages of Carex, dated to the early Miocene (22.55 mya), with previous studies retrieving even slightly older ages around the Mio-Oligocene boundary (24.4 mya; Martín-Bravo et al., 2019). In congruence with the origin and early diversification of Carex in the Eastern Palearctic (Martín-Bravo et al., 2019), our results point to an ancient migration from the Northern to the Southern Hemisphere, specifically from the Palearctic to SA (Figure 3A). Thus, although the biogeographic reconstruction inferred the Western Palearctic and SA continent as the ambiguous ancestral area for the crown node of subg. Psyllophorae, the immediate parent node was inferred to be exclusively distributed in the Northern Hemisphere (Eastern Palearctic or Western Palearctic). In a previous reconstruction using a representative sampling for the whole genus, the Western Palearctic was revealed as the origin of subg. Psyllophorae (Martín-Bravo et al., 2019).

Allopatric differentiation is suggested for the three main sectional lineages of subg. Psyllophorae originating in different regions (sect. Psyllophorae in the Western Palearctic, sect. Schoenoxiphium in the Afrotropical region, and sect. Junciformes in SA) and diversifying almost exclusively within each one (except sect. Junciformes, which also colonizes the SW Pacific for more than 15 million years (Figure 3A). A remarkably synchronous timing of the origin was inferred for these three main lineages (16–18 mya).

As expected, a primary (tectonic) Gondwanan disjunction has to be ruled out for the Southern Hemisphere lineages (sects. Schoenoxiphium and Junciformes) because their ages long postdate the split of the southern supercontinent (c. 16 mya vs. 135–105 mya; McLoughlin, 2001), unlike what is found in other, much older groups of plants (Araucariaceae, Biffin et al., 2010; Dicksonia, Noben et al., 2017). It is unclear whether the Afrotropical region was colonized from SA by LDD or from the Western Palearctic (Figure 3A). In the latter case, since this colonization occurred in the early Miocene (c. 20 mya) the disjunction is probably too old to fit a climate-driven Rand-Flora vicariance (Mairal et al., 2017), while LDD may not be ruled out (Míguez et al., 2017). Moreover, the origin of sect. Schoenoxiphium seems to be in central Southern Africa, which also dismisses the Rand-Flora pattern (Márquez-Corro et al., 2020). An additional, more recent (probably Pleistocene) colonization of the Afrotropical region from the Western Palearctic by LDD can be identified within sect. Psyllophora involving C. peregrina Link (Figure 3A). Remarkably, this constitutes the only recent out-of-continent dispersal within the subgenus.

The diversification of sect. Junciformes within SA started close to their origin in the middle Miocene and was characterized by relatively constant cladogenesis (especially in the Junciformes-clade), while diversification rate decreased over time (Figures 3B,C). The group remained confined and diversified in N Patagonia for several million years (Figure 3C). This long-permanence and within-area diversification is known in other groups at comparative evolutionary levels (Mathiasen and Premoli, 2010; Otero et al., 2019; Benítez-Benítez et al., 2021). The persistence of the group in N Patagonia and lack of evidence of early presence at lower latitudes suggest that sect. Junciformes colonized the Southern Cone from the Northern Hemisphere by direct LDD from the Western Palearctic (Figures 3A,C). Contrastingly, most cold-adapted Neotropical plant groups used the American cordillera as a corridor into SA (Zemlak et al., 2008; Xu et al., 2009; He and Sun, 2017). It would not be until the latest stages of the orogeny that the Andes served as a natural passageway for multiple colonizations in sect. Junciformes, including northward migrations from Patagonia (C. camptoglochin, C. setifolia, C. vallis-pulchrae, and C. via-incaica). This could have been favored by the concurrent, Pliocene cooling of the low-latitude Andes (Roberts et al., 2017). Both, the early permanence of sect. Junciformes in N Patagonia and the posterior colonizations within SA entailed no strong ecological innovations but a relative niche conservatism (Figures 2, 3C). Our bioclimatic reconstructions suggest that such ecological conservation is favored by selection (Figure 2A, OU model; Blomberg et al., 2020). Accordingly, the changing environment in SA (Rabassa, 2008) instead of inducing niche shifts in sect. Junciformes entailed the range expansion as similar suitable niches became available. By contrast the entrance of C. phalaroides in multiple areas out of Patagonia implied striking changes in bioclimatic preferences, especially regarding AMT (Figure 2A). This is expectable since this species inhabits radically different, much more temperate habitats (Atlantic Forest, Pampas, montane tropical forest) in comparison to the other species belonging to sect. Junciformes.

In the late Pliocene and during the Pleistocene, several groups (Junciformes-clade Lineage B and Aciculares-clade SA-clade) underwent in situ diversification either in N or S Patagonia. In addition, these clades present changes in bioclimatic preferences at very shallow levels, especially in the Junciformes-clade Lineage B (Figure 2). These could be interpreted as a result of local adaptation in glacial refugia, as reported in other plant genera in different regions around the world (Tremetsberger et al., 2009; Premoli et al., 2010; Kremer, 2016), as well as in other Carex groups (Benítez-Benítez et al., 2018). Many plant groups were able to survive Patagonian glaciations in high latitude refugia (see references in Table 1) since ice sheets were generally not extensive after the Greatest Patagonian Glaciation (1–1.2 mya; Rabassa, 2008). During the Pleistocene, major Patagonian rivers have been invoked as barriers for allopatric speciation in plants (Jakob et al., 2009; Sede et al., 2012; Cosacov et al., 2013), but these do not seem to have acted as barriers in Carex given its striking capacity for LDD (Villaverde et al., 2017a). In other plant groups, glacial survival in Patagonia seems to be conditioned by the aridification and establishment of the steppe (Jakob et al., 2009; Cosacov et al., 2010; Sede et al., 2012). In sect. Junciformes, the steppe may have played a certain role as a barrier since these species hardly penetrate this biome (Supplementary Figure 3).

Section Junciformes colonized SW Pacific (NZ and probably also SE Australia and Tasmania) from SA (Aciculares-NZ clade; Figure 3A), a biogeographic pattern also reported in other families of plants (Von Hagen and Kadereit, 2001; Meudt and Simpson, 2006; Otero et al., 2019). Whether the colonization took place by direct LDD or stepping stone is unclear, despite the absence of sect. Junciformes in circum-Antarctic archipelagos seem to point to a direct LDD as the most plausible explanation. In any case, inferred divergence ages (late Miocene-Pliocene; Supplementary Figure 6) also rule out the role of Antarctica as the last tundra remnants of this region are reported to disappear around the Middle Miocene (Lewis et al., 2008; but see Barrett, 2013). New Zealand species inhabit alpine habitats which became available with the orogeny of the Southern Alps starting in the Pliocene (Heenan and McGlone, 2013), which mostly matches the crown node age inferred for this group (Figure 3C and Supplementary Figure 6). Interestingly, the colonization of NZ entailed a remarkable long-term niche conservatism within Aciculares-clade for different variables (AMT, TAR, MTDQ; Figure 2), which indicates that the lineage was somehow preadapted to the newly colonized habitats. Nonetheless, a certain degree of local adaptation is revealed by the somewhat different values for AP and TRI (Figure 2).

The phylogenetic reconstruction of subg. Psyllophorae shows relatively synchronous cladogenesis through time and an unbalanced topology, with a relatively poor lineage (sect. Psyllophorae) sister to a much more diversified group (sects. Schoenoxiphium-Junciformes; Figure 3B). This has sometimes resulted in a ladderized rather than bifurcating topology (Crisp and Cook, 2005; Vargas and Zardoya, 2014), with frequently long branches and deep nodes originating old monotypic and/or species-poor lineages. Their stem nodes date as old as the early-middle Miocene. Interestingly, this pattern has been found in the three sections: Psyllophorae (C. distachya Desf., c. 18 mya; C. illegitima Ces., 12 mya); Schoenoxiphium (C. multispiculata Luceño and Martín-Bravo - C. lancea (Thunb.) Baill., c. 11 mya), and Junciformes (C. camptoglochin, c. 15 mya; C. vallis-pulchrae, c. 14 mya; C. phalaroides, c. 14 mya).

Diversification rates seem to progressively decrease in subg. Psyllophorae (Figure 3B), which agrees with the previous findings in Martín-Bravo et al. (2019) using a larger sampling of Carex species but a smaller set of sect. Junciformes taxa. As stated before, sect. Junciformes is one of the only three Neotropical Carex groups with more than 15 species (see Introduction). This lack of significant increase in diversification rates points to a process of disparification rather than true evolutionary radiation (Simões et al., 2016). Thus, its current relatively high diversity would be simply the result of the progressive accumulation of changes during its long evolutionary history.

The allopatric distribution of the three sections in different landmasses (Figure 1) could have imposed differential intrinsic/extrinsic factors influencing their diversification. The limited ecological diversification revealed by our bioclimatic niche analysis in sect. Junciformes (Figure 2B) provides insights into the intrinsic conditionants of its evolution. A relatively conserved niche in this lineage, perhaps constrained by its own ecological limits, may have prevented a significant increase in diversification rates. It means that the ecologically suitable space could have been early filled. In particular, the case of the NZ-clade is remarkable, since the group hardly diversified after colonizing NZ, also involving limited ecological (Figure 2) and morphological change (Hamlin, 1962; Edgar, 1970), which strongly contrasts with other spectacular cases of morphological/ecological adaptive radiations within NZ (Wagstaff et al., 2002; Glenny, 2004; Meudt and Simpson, 2006). Furthermore, other Carex groups (sects. Spirostachyae and Uncinia; Roalson et al., 2021) underwent remarkable radiations in NZ, but they could have arrived earlier to the archipelago than NZ-clade of sect. Junciformes (Martín-Bravo et al., 2019), perhaps preventing the establishment of its species through high density blocking (Slingsby and Verboom, 2006; Pender et al., 2021).

Further studies are needed to characterize in detail the niche of the other sections of subg. Psyllophorae. On the one hand, sect. Schoenoxiphium was inferred to have originated in the Drakensberg range in South Africa, where its current center of diversity is still located (Márquez-Corro et al., 2020). In this area, species distributions are often overlapping, but with frequent turnover along various ecological gradients (elevation, wetness, forest to grassland; Luceño et al., 2021). Thus, speciation in this lineage appears to have been sympatric and driven by ecological factors. On the other hand, sect. Psyllophorae combines widely distributed species (C. distachya and C. pulicaris) which are partly sympatric with several allopatric narrowly distributed species (Jiménez-Mejías and Luceño, 2011). Interestingly, while it is the poorest diversified section within subg. Psyllophorae, its species display remarkably different ecological preferences, ranging from Mediterranean shrublands to mountain bogs (Borges et al., 2008; Luceño, 2008; Silva et al., 2010; Verdcourt, 2010; Gehrke, 2011).