Data acquisition

I used expert-vetted sources for the ranges of California marine fishes originally published by Horn and Allen in 1978 [9] (the dataset was obtained through personal communication with L. Allen). Their range data were largely based on the first edition of Miller and Lea [44] and were published during a cold decadal oscillation regime [27]. In their study, species were assigned as present or absent within latitudinal bands of 1-degree intervals. For example, a hypothetical species ranging from 32.0°–34.9° N would be considered present in the 32, 33, and 34-degree bands. There are ten latitudinal bands within California borders (from 32–42 degrees). This dataset served as the historical baseline for my study (referred to herein as the 1978 dataset). Note that the authors published an update to their analysis in 2006 [12] with the addition of 22 species known to have dispersed to California from Mexico, but ranges for preexisting species were identical to the 1970s dataset.

The modern dataset (referred to herein as the 2022 dataset after the time of compilation) was based on the second edition of the Guide to the Coastal Marine Fishes of California published in 2020 [34] as well as a checklist of species from Alaska to Baja California published in 2021 [45]. These two sources were nearly identical. Following Horn and Allen [9], I assigned species to 1° latitudinal bands to allow compatibility with the 1978 dataset. I used the “validate.names” function in the R package rfishbase [46] to validate and standardize species names and ensure consistency across sources. I noted when species’ northern or southern latitudinal limits differed between the 2022 and 1978 datasets (S1 Table in S1 File). I also noted whether species were new to California since the 1978 dataset (S2 Table in S1 File), including species that were described since the original study. A small number of species had unclear range information; these cases are detailed in S3 Table in S1 File. Some of these had to be removed from downstream analyses (S3 Table in S1 File). Human-introduced or invasive species were not considered in this study.

While the original datasets (1978, 2006) recorded latitudinal range boundaries outside of California (i.e., further north than 42.0°N or further south than 32.5°N), I chose to only record limits as “north of CA” or “south of CA” as appropriate. This is because guides were usually not as detailed for ranges outside of California (e.g., stating the northern range boundary as “Alaska”, or southern range boundary as “Peru”, which could span many potential latitudes).

The 1978 study excluded deep-sea fishes. Still, I noted the ranges of deep-sea fishes based on modern sources and included them in biogeographic analyses. While many deep-sea fishes are cosmopolitan, others have clear biogeographic affinities [28, 43, 47]. Deep-sea fishes were generally defined in this study as those belonging to taxonomic groups that were excluded in the 1978 dataset (such as Myctophiformes or Stomiiformes). Due to high latitude emergence, the usual cut-off of 200 m depth to define a deep-sea species [48] is difficult to apply on a local scale [49]. Biogeographic results for deep-sea fishes are reported separately from inshore fishes (S1 Fig in S1 File). Only inshore fishes were compared to the 1978 dataset, and reported if new to California waters (S2 Table in S1 File).

Approach to historical biogeography

I developed a novel application of historical biogeography methods to uncover biogeographic boundaries when their location is uncertain. This approach could be applicable to other biogeographic questions and requires (a) delimiting arbitrary boundaries for a “region” that subsumes putative biogeographic breakpoints of interest, and (b) identifying external regions that represent potential regions-of-origin and dispersal routes. The pattern of boundary-crossing inferred by the ancestral range reconstruction [40] can be used to assign a region-of-origin category to species. The relative proportion of these species categories mapped over a finer spatial scale within the artificial region can then be used to detect patterns of turnover in biogeographic affiliation. In my case, I determined whether species found within California were descended from a lineage that originally entered California via Oregon (USA) or Baja California (Mexico) (Fig 1). Then, the relative proportion of northern-origin and southern-origin species within each latitudinal band in California was visualized. If Point Conception is the site of turnover of southern-origin to northern-origin communities as expected, then the proportion of these species categories should “flip” from southern-biased to northern-biased beyond the 34° latitudinal band.

Normally it is not advisable to use political boundaries for these models, because historical biogeography pertains to deep-time processes that predate (and are agnostic to) human activities. Therefore, dispersals inferred by the model would refer to lineages crossing an arbitrary boundary that is not biogeographically relevant. In this case, I am purposely setting up artificial boundaries for lineages to cross (one boundary to the north and one to the south), and using this information to infer the direction of dispersal for species within the boundaries. When regional boundaries are inappropriate or incorrect, the number of inferred dispersals should be inflated, because there is no true barrier to dispersal. We can exploit this property to gain information about species within the region of interest. Lineage dispersal and colonization must have happened for species to occur in waters today recognized as Californian, regardless of the political boundary existing or not. Whether lineages crossed the southern or northern artificial boundary is used as evidence to code species as deriving from a northern or southern lineage. These species-specific codes can then be visualized on a finer-grain scale (i.e., 1° latitudinal bands) to visualize patterns of turnover within the arbitrary area of interest, thereby revealing features that are biogeographically relevant.

To model the evolution of ranges, I used published time-calibrated molecular phylogenies for Actinopterygii [36] and Chondrichthyes [37]. These phylogenies were constructed from a supermatrix of sequences mined from GenBank. Excellent taxonomic sampling is critical for this approach for two reasons. First, region-of-origin can only be inferred for species sampled in the phylogeny without making tenuous assumptions about biogeography based on taxonomy. Second, broad sampling of higher taxa is needed to capture dispersal events that occurred in deep time. The ray-finned fish phylogeny used here contains 11,638 species; the cartilaginous fish phylogeny contains 610 species. These phylogenies sampled 73.8% of ray-finned fish and 83.5% of cartilaginous fish species of California. Agnathans were not included in these analyses, but range shifts are still reported in S1 and S2 Tables in S1 File.

Biogeographic transitions were modelled among five “regions”. The temperate East Pacific was broken into three regions: (a) North of the Oregon-California border (equal to the “Cold Temperate Northeast Pacific” of [50] exclusive of California; (b) within California, and (c) South of the California-Mexico border (equal to the “Warm Temperate Northeast Pacific” province exclusive of California). Species were assigned to the “North” and “South” regions based on the checklist by Love et al. [45] (spanning Alaska to Baja California, Mexico). Since the “Warm Temperate Northeast Pacific” province includes the Gulf of California, I also assigned species in this region based on information from the STRI Shorefishes of the Tropical Eastern Pacific database [51]. In total, 784 species were assigned to the “North” region (741 and 43 species of ray-finned and cartilaginous fishes respectively), 902 species were assigned to California (836 and 66), and 1,377 species were assigned to the “South” region (1,270 and 107).

The final two regions were “marine but outside the temperate East Pacific”, and “freshwater”. These regions allowed me to model dispersal into the marine temperate East Pacific itself, as a precursor to dispersal into California. I used a dispersal rate modifier matrix that limited transitions from freshwater into the other four marine states by multiplying the base rate by 0.05. This approach was used in past studies of broad-scale fish biogeography [7, 48, 52, 53]. The freshwater state was broadly defined to include brackish or diadromous states. Habitat for each species was based on past compilations [48, 52] with some updates based on Love and Passarelli [34]. I set the maximum number of areas that a species could co-occur in to five. I removed area combinations from the state space that allowed species to simultaneously occur in the North and South regions but not reside in California. This disjunct state was not observed in any living species. I allowed direct dispersal from the “outside” state into California, but I downweighed the probability of these transitions by 0.05 (as with marine-freshwater transitions). This could be imagined if a lineage crossed the open ocean from the Central Pacific to colonize the East Pacific, which occurs rarely [54].

Biogeographic models were fit using the BioGeoBEARS R package v. 1.1.2 [40]. Note that only the BAYAREA class of models [39] are appropriate in this case. This is because these models allow cladogenetic inheritance of ranges spanning multiple regions, which is important because few species are endemic to California itself. In contrast, the DEC class of models [38] would force a widespread range to be broken up into individual regions during cladogenesis, followed by anagenetic dispersal to achieve the widespread range again, inflating the number of dispersals (these issues are discussed in detail in [48]). The fit of the BAYAREA and BAYAREA+J (with founder event speciation) models were compared using Akaike weights. To incorporate uncertainty, I performed biogeographic stochastic mapping [55] to produce 100 independent simulations (“stochastic maps”) of biogeographic history that are possible under the best-fit model. These can be thought of as 100 different boundary-crossing arrangements that can plausibly explain the origins of the California marine ichthyofauna.

To infer the region-of-origin for each California species, I used a custom R script (available in the Dryad repository associated with this study) to “walk backwards” from the tip to the branch where dispersal into California occurred (Fig 1). The source region was obtained for this dispersal event (either the “North” region, “South” region, or “other”). I repeated this sequence for all species and across all 100 stochastic maps. The region-of-origin for a species was considered to be North or South if >70 stochastic maps inferred that region as the source of the ancestral lineage of that species. Of 617 California species, 75.8% could be assigned to a source region this way. Species with the region-of-origin inferred to be North or South in similar frequency across stochastic maps were considered to be of unclear origin (20.2% of species). For the trailing ~4% of species, the source region was inferred to be “North” plus “other”, or “South” plus “other”, in similar frequencies. Nearly all of these were euryhaline or diadromous species, and the “other” potential source region was “freshwater”. These were assigned to North or South categories respectively.

I generated plots of species richness and north and south range terminals within each of ten latitudinal bands, with species separated by inferred region-of-origin (North, South, or unclear). I also plotted the difference in richness between northern- and southern-origin species within each latitudinal band. I plotted these variables separately using the 1978 and the 2022 range datasets of inshore marine fishes. For additional context, species were coded as demersal or pelagic based on [48] since these life habits can influence dispersal [21].

Range shifts of California fishes

The ranges of 151 species of California marine fishes have changed since the 1978 dataset (S1 Table in S1 File). The two most common types of range shifts were: northern range limit extended northward (84 species or 55.6% of range changes); and southern range limit extended southward (43 species or 28.5% of range changes). The majority of range changes involved a species growing its range either northward, southward or in both directions (129 species or 85.4% of range changes).

Two types of range changes are consistent with effects of anthropogenic climate warming: a northward expansion of the northern range limit (caused by dispersal), and a northward expansion of the southern range limit (caused by migration or extirpation). While the former was common, the latter was only seen in 5 species (S1 Table in S1 File). Extensions of the southern range limit further southward were the second most common type. These are not readily explained by climactic changes, since the period 1977–1999 was a warm PDO regime [26, 27] which should drive northward dispersal. Range shifts other than the predicted northward direction have been documented in the East Pacific and other marine ecosystems [56, 57]. These could be explained by factors such as biotic interactions or idiosyncratic clade-specific responses to warming. Another caveat is that our knowledge of local marine fishes has increased since the 1970s, which could explain some range updates. However, this factor should not drive range changes in any particular direction, whereas shifts driven by warming should generally occur northward.

In addition to range shifts of existing California species, 110 species are new to California waters (S2 Table in S1 File). Of these, 18 species were newly described since the 1978 study, so their occurrence cannot be explained by recent dispersal (S2 Table in S1 File). Among the remaining species, 69 species dispersed from Baja California, Mexico (south-to-north dispersal), 15 species dispersed from Oregon, USA (north-to-south dispersal), with 8 species of unclear origin based on their range (either California endemics or spanning both northern and southern borders). These counts do not include invasive or introduced species through anthropogenic means, which were discarded from this study.

The greater number of new species added from northward than southward dispersals is consistent with a role of warming climate, either anthropogenic climate change or the recent warm PDO cycle. One complication is that species richness in Mexico is greater than that of Oregon, because richness increases towards the equator (i.e. the latitudinal diversity gradient). Therefore, we would expect the number of northward dispersals to be greater than southward dispersals even under a null model of random dispersal. I performed a post-hoc analysis where I simulated a null distribution of ratios of northward to southward dispersals. The null distribution was made assuming that species have dispersed into California at random since 1978, drawing from the combined pool of species North of California and South of California (details in S1 Note in S1 File). The observed ratio of 4.6 dispersers from Mexico for every 1 disperser from Oregon was well outside the 95% quantile interval of simulated ratios (median simulated ratio = 2.5; S2 Fig in S1 File). These results support the assertion that recent dispersals are influenced by climate warming, even given the confounding influence of the latitudinal diversity gradient.

Due to new species, richness of the California inshore marine fauna has increased 22.3% overall since the 1970s from 494 to 604 species (now including 531 ray-finned fishes, 66 cartilaginous fishes, and 7 jawless fishes). Similar increases in richness primarily due to northward range expansions have been documented in other temperate marine ecosystems [58–60]. Richness increased in all latitudinal bands, but especially in the 32° and 33° bands (including the Mexican border, San Diego, Orange and Los Angeles counties) in which over 75 new species were added to the fauna. This suggests that many of the new species are not dispersing beyond southern California and are affected by the same biogeographic barrier as the pre-1978 fauna (see below).

Like all studies comparing only two time periods, these results should be interpreted with caution [56]. It is unlikely that new species were gradually and consistently added over the 44 years between datasets (1978 and 2022). Many of these new species are known to have entered California during extreme warm events [32, 41]. Likewise, it should not be assumed that the historical dataset [9] represents the “normal” or “unmodified” state for California [61]. Fishes in 1978 faced anthropogenic threats as they do today. Cycles of northward and southward dispersals are a persistent feature of the California marine fauna as documented in the fossil and historical record [29–31]; therefore, it is difficult to identify any single snapshot of California’s marine fauna from the past as the “normal” state.

It should be noted that most subtropical species that entered California during the recent warm period from 2014–2018 were not sighted again in 2019 [32]. Whether these new additions will persist in California over the long term is unclear. Breeding populations of at least one new species have been documented [62]. Hubbs [30] noted that subtropical species that entered during the 1853–1860 warm period persisted through the 1880s and were probably breeding within California. With the frequency of extreme warm events predicted to increase in the future [42, 63], and given the overall warming trend in California [64], it seems plausible that conditions will eventually remain suitable for subtropical species to establish in California.

Marine biogeography of California

The BAYAREA+J model had a better fit than the BAYAREA model in both ray-finned fishes and cartilaginous fishes (Akaike weight = 1). By “walking backwards” from each tip of the tree to the branch where colonization of California waters occurred, I determined that 188 species of inshore California ray-finned fishes are derived from a northern lineage, 205 species are derived from a southern lineage, and 49 species were of unclear origin (S4 Table in S1 File). Uncertainty was much greater for cartilaginous fishes (2 species from a northern lineage, 29 species from a southern lineage, and 22 species of unclear origin; S5 Table in S1 File). This could be due to the lower richness and smaller size of the cartilaginous fish phylogeny (610 tips versus >11,000 in the ray-finned fish phylogeny) which affects power, and also by the widespread ranges of many species (i.e. spanning most of the East Pacific). Interestingly, while demersal fishes were of northern and southern-origin in similar proportions, pelagic fishes were overwhelmingly of southern origin (Fig 2). Of deep-sea fishes (S1 Fig in S1 File), roughly half had unclear affinities, which is unsurprising given their widespread distributions.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Species richness in each latitudinal band by region-of-origin as inferred from historical biogeography.

Range data refer to the 1978 dataset of Horn and Allen [9] or the 2020’s data assembled in this study based on [34, 45]. Only inshore marine fishes are shown; for deep-sea fishes see S1 Fig in S1 File. Region-of-origin for individual species are given in S4 and S5 Tables in S1 File.

https://doi.org/10.1371/journal.pone.0291776.g002

Examples of northern-origin lineages include: rockfishes (Sebastidae), sculpins and poachers (Cottoidei), gunnels and pricklebacks (Zoarcoidei), and Bathyraja skates (S4, S5 Tables in S1 File). Southern-origin lineages include: many sharks and batoids, surfperches (Embiotocidae), damselfishes (Pomacentridae), scorpionfishes (Scorpaenidae), basses (Serranidae), and jacks (Carangidae). These examples help show how recent speciation and dispersal can erode the signal of events deeper in time. Specifically, surfperches are most diverse today in central and northern California [65], yet they were found to be derived from a southern-origin lineage in my analyses (S4 Table in S1 File). This is because the family Embiotocidae is nested within the larger clade Ovalentaria, which is predominately tropical. Similarly, richness of Sebastes rockfishes peaks near Los Angeles in southern California (33° band) [66], yet the lineage was confidently inferred to have colonized California from the northern United States (S4 Table in S1 File). This suggests the modern distribution of rockfishes is due to southward dispersal and speciation within California after their initial colonization. Overall, the novel application of BioGeoBEARS used here helped untangle the influence of deep-time events (colonization of a region by higher taxa) from more recent events (subsequent dispersal of individual species within a region) on the biogeography of California.

Fishes in the 32° and 33° bands were mostly of southern origin (derived from lineages in Baja California) (Figs 2 and 3). Fishes in the 34° band were of northern- and southern- origins in similar proportions with a slight bias towards northern-origin species. Fishes in the bands from 35°–42° were mostly of northern origin (derived from lineages in Oregon). These patterns were true of both 1978 and 2022 ichthyofaunas. However, the difference between northern and southern California faunas has become more pronounced since the 1970s; that is, southern bands are even more biased towards southern-origin species, and northern bands are even more biased towards northern-origin species, than they were in 1978 (Figs 2 and 3). This is due to the addition of new species from Baja California (i.e. tropicalization), especially those that do not disperse beyond 33.99° N, with a lesser role from species dispersing southward from northern latitudes (i.e. borealization [57]). Note that deep-sea fishes also showed a south-to-north turnover (S1 Fig in S1 File), but the transition was more gradual than inshore fishes and there was no obvious biogeographic boundary.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3.

(A) Difference in richness of northern- and southern-origin species within each latitudinal band, using range data from the 1978s and 2020s. Bands are colored by the region-of-origin with the highest richness (blue = southern origin, orange = northern origin). (B) Difference in total species richness in each latitudinal band between the 1978 and 2022 dataset. Only inshore marine species are shown.

https://doi.org/10.1371/journal.pone.0291776.g003

If range shifts persist under future warm conditions in California, these results suggest this will have the biogeographic consequence of widening the disparity in faunal composition between communities north and south of Point Conception. It seems that many subtropical species new to California are limited by the same biogeographic barrier as core California species (Figs 3 and 4). This is unnerving for future survival of species under climate change. If species are prevented from dispersing northward, they may be trapped into unfavorable thermal conditions [67, 68]. It is unclear whether continued warming will change oceanographic conditions enough to weaken this barrier in the future.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. Count of species’ range terminals (northernmost and southernmost records) within each latitudinal band, colored by region-of-origin as inferred from historical biogeography.

Only inshore marine species with a range terminal within California are shown.

https://doi.org/10.1371/journal.pone.0291776.g004

The locations of northern and southern species range terminals have been traditionally used as evidence for biogeographic boundaries within California [9]. Northern range terminals of southern-origin species peaked in the 33° band (near Los Angeles; Fig 4). Southern range terminals of northern-origin species peak at 32° N (near San Diego and the Mexican border). These patterns were also reported by Horn and Allen [9, 12] but have become more dramatic in the present day due to range shifts. These results provide support to the suggestion that biogeographic barriers in California are more influential on ranges of southern-origin species than ranges of northern-origin species [9], and this continues to be true for new species entering California. This is consistent with the idea that the southward flowing California Current inhibits recruitment north of Point Conception [21, 69].

Role of point conception

Traditionally, it is believed that Point Conception is the barrier with strongest influence on California marine fishes, representing the site of turnover between southern and northern fish faunas of the temperate East Pacific. In favor with this view is the roughly balanced proportion of northern- and southern-origin species within the 34° band (Fig 3). This band spans 34.0°–34.99°N, with Point Conception near the center at 34.4° N. The balanced proportions in this band could be explained if southern-origin species outnumber northern-origin species below Point Conception, but the reverse is true above Point Conception. Therefore, the pattern of turnover of northern-origin and southern-origin lineages, as inferred from historical biogeography, is in support of the Point Conception barrier.

Still, an observation against the importance of Point Conception is that southern range terminals concentrate within the 33° band (Fig 4). If Point Conception was most influential on species’ ranges, we would instead expect southern range terminals to concentrate within the 34° band instead. The observed pattern of range terminals suggests that the most obvious changes in faunal composition occur well outside the vicinity of Point Conception. This is consistent with phylogeographic studies across a variety of marine organisms, which tend to show that intraspecific genetic groupings split near the Los Angeles area [16–18, 21, 70]. In particular, the site of divergence appears to be Santa Monica Bay (33.7–34.0° N), which is contained by the Palos Verdes peninsula to the south and Point Dume to the north. While Pleistocene sea level change has been implicated as the driver of these genetic splits [17, 70], present-day factors must be operating as well, given that many new species entering California during warming events have not dispersed beyond the 33° band (Figs 3 and 4). One possibility is that sharp transitions in substrate between soft bottoms and rocky reefs and canyons in this area discourages dispersal [70, 71].

How do we reconcile the fact that historical biogeography gives a different answer about the relevance of Point Conception than range terminals and intraspecific phylogeography? One answer is that the approach used in this study considers the role of deep-time events on community assembly, specifically the dispersal of ancestral lineages. Subsequent dispersal of individual species away from their center of origin can erode the signal of earlier events [5]. Further, it is worth considering that the California marine fauna is largely dispersal-assembled rather than speciation-assembled. Specifically, faunas on either side of Point Conception are distinct because they are two distantly related groups that have dispersed southward and northward, respectively, and meet in California. This is different from a scenario where the faunas are distinct because they are the product of repeated allopatric speciation events (Burton 1998). The latter scenario may be ongoing in the Los Angeles region, but the conditions that favor genetic structuring in this area may only date to the Pleistocene [17, 18, 25]. This is long after major fish lineages initially established in California, a process in which Point Conception appears to have been more influential. To sum, these sources of evidence operate on different time scales and can be expected to give signals of different events on biogeographic regionalization.