Media (photos, videos) of molids from the ACS and CCS, along with metadata, were compiled from several sources, including iNaturalist (www.inaturalist.org), public outreach campaigns, the authors’ private networks, and in collaboration with several ocean-focused organizations (e.g., Marine Education and Research Society, oceansunfish.org, Coastal Observation and Seabird Survey Team, JellyWatch, NOAA, and various museum collections). Crowd-sourced media comprised sightings from beach strandings, museum records, and live sightings from vessels, divers, snorkelers, and remotely operated vehicles. Most data (68%) originated from direct submissions to iNaturalist or the Marine Education and Research Society. Sighting metadata was verified with observers whenever possible (date, location/locality, molid size, and observation notes) except for records from iNaturalist, where observer-reported metadata was assumed correct. Each observation, including accompanying metadata, was reviewed before being included in the analysis. Only observations with an identifiable photo or video, and where date and location could be reasonably inferred, were considered. Duplicate observations, observations of captive fish, observations from other ecosystems, and observations without photo or video were excluded from the analysis. When observers did not provide GPS coordinates but provided a detailed description of the locality, coordinates were estimated within 10 km of landmarks referenced in the submission using Google Earth. Each record was also annotated with the observation type [at surface (seen from above water); at/near surface (seen from below water); at depth (seen from below water); stranded, on fishing boat (caught), in collection].

The size of each molid was subjectively gauged by a Molidae researcher (MN) as <1 m TL or >1 m TL based on 1) relative size to other objects in the images, 2) the aspect ratio (height to width) of dorsal and anal fins, which decreases with increasing molid size (Watanabe and Sato, 2008; Watanabe and Davenport, 2020), and 3) the degree of body bulkiness, which increases with molid size.

Molid species determination was based on all visible taxonomic characteristics following Sawai et al. (2018, 2020) and Sawai (2021) ( Figure 1 ). We holistically assessed all visible characters, because single characters are typically insufficient to determine to species. For example, the presence of a smooth band back-fold, or an indent in a smooth clavus margin (see Figure 1 ), does not, in isolation, identify M. tecta because M. mola and M. alexandrini may have faint back-folds (Sawai et al., 2018; Nyegaard et al., 2018b) and an indent in the clavus margin could be due to injury or malformation. Further, the presence or absence of morphological characters was considered relative to molid size, as nearly all characters develop with growth. Each molid observation was independently determined to the lowest possible taxon by two identifiers – a Molidae researcher with ten years of experience (MN) and a fish ecologist trained using the guide in Figure 1 (TML). Where determinations differed, a third Molidae researcher (TMT) independently identified the observation. Determinations differing in taxonomic level (e.g., M. mola versus Mola sp.) were resolved using the majority ID (2/3 of identifiers). Direct species- or genus-level conflicts (e.g., M. mola vs M. tecta) would have been resolved by reverting to the lowest agreed taxon (in this case, Mola sp.); however, no such conflicts occurred.

Spatial distributions of size and species data were plotted as hexagonal heatmaps in R (Villanueva and Chen, 2019), where a latitudinal boundary between the ACS and CCS was defined at 43°N (Auad et al., 2011). The frequencies of <1 m TL and >1 m TL molids, and M. tecta >1 m TL relative to all other molids >1 m TL in each current system were compared using Pearson’s chi-squared test for count data with Yates’ continuity correction ( α = 0.01 ) in R (version 4.3.1) using the function chisq.test().

Tissue samples from six stranded (dead) molids, identified as M. tecta from photos, were opportunistically collected by local scientists and members of the public (California: n=4, Alaska: n=2). Tissue was preserved in 70% ethanol, RNALater^® or dimethyl sulfoxide (DMSO) for transport to labs. All DNA extractions proceeded using Qiagen DNeasy Blood and Tissue Kits, following the manufacturer’s protocol. The Cytochrome c subunit 1 (CO1) locus was amplified and sequenced as described in Supplementary Table 1 . All sequences were submitted to GenBank [Accession numbers PQ636871 - PQ636876].

To verify the species identity of the sampled molids, the COI sequences were pooled with all CO1 sequences labeled Mola spp. or Ma. lanceolatus in NCBI (n=61). The sequences were aligned with Clustal W in Unipro UGENE v50.0 using default settings and trimmed to equal lengths (616 base pairs), except 7 shorter sequences (511 – 611 bp). Phylogenetic relationships were inferred using the PhyML (Maximum Likelihood; ML) analysis on the T-rex server (Boc et al., 2012; http://www.trex.uqam.ca). Following Nyegaard et al. (2018b) for Molidae COI model selection, the HKY85 model was used with optimized equilibrium frequencies, estimated tv/ts ratio, gamma distribution parameter, and four substitution rate categories, with the mean used for the middle of each rate class. Branch support was estimated with 1,000 bootstraps. The resulting phylogenetic tree was visualized in TreeViewer v2.2.0 (Bianchini and Sánchez-Baracaldo, 2024) as a rooted tree with Ma. lanceolatus as the outgroup.

To confirm the species identity of the sampled molids, we compared the tree topology with the analysis in Nyegaard et al. (2018b), which included an assessment of likely mislabeled Molidae sequences in NCBI. Finally, we determined which species group(s) our novel sequences clustered with.

Media from 1,213 molids (representing 1,000 observations) in the ACS and CCS were examined to determine species identities. Species determination by two independent identifiers resulted in 1,152 agreements (95.0%) and 61 disagreements (5.0%). All disagreements occurred due to differences in taxonomic level (e.g., M. mola vs. Mola sp.), with no direct, species-level conflicts (e.g., M. mola vs. M. tecta). Most disagreements (72.1%; n=44) occurred between determinations of M. mola versus Mola sp. Resolving determination disagreements with input from the third identifier resulted in an additional twenty-nine species-level identifications (all M. mola), twenty-six genus-level identifications (all Mola sp.), one family-level identification (Molidae), and three instances of unverifiable organisms (undetermined; removed from the dataset). In total, 471 specimens were identified to species level (423 M. mola, 45 M. tecta, 2 M. alexandrini, 1 R. laevis), 648 specimens identified as Mola sp., and 83 specimens identified as large-bodied Molidae (Mola spp. or Ma. lanceolatus). As our study focused on large-bodied molids, the small-bodied R. laevis was removed from the dataset.

The subset of observations for which a location and a size estimate were available (n=1,178 specimens) consisted of nearly equal numbers of molids <1 m TL (n=586) and >1 m TL (n=592) ( Figure 2A ). While most of the molids >1 m TL (62.3%; n=369) were determined to species, the vast majority (77.6%; n=455) of molids <1 m TL could only be determined to genus (Mola sp.; Figure 2A ) often due to insufficient development of taxonomic characters. All identified molids <1 m TL were determined to be M. mola. In both size groups, molids determined to family were consistent with Mola spp. but were insufficiently visible in the images to definitively exclude Ma. lanceolatus.

Across the dataset, most observation types were of molids basking or swimming at the sea surface (seen from above water) (59.7%, n=703) or stranded on beaches (25.7%, n=303). However, this differed between size categories; most >1 m TL molids were observed at the sea surface, while most <1 m TL molids were observed either at the sea surface or stranded on beaches ( Figures 2B, C ).

Across observation types, most molids <1 m TL could not be determined to species, except for specimens accessioned in museum collections (100% success) ( Figure 2B ). In contrast, most molids >1 m TL types could be determined to species across all observation types ( Figure 2C ).

The taxa- and size categories comprised different proportions of observation types ( Figures 2D, E ). Both Molidae <1 m TL and >1 m TL mainly comprised observations at the sea surface. The M. mola and Mola sp. <1 m TL observations were mainly of stranded specimens and specimens at the sea surface ( Figure 2D ). The M. Mola and Mola sp. >1 m TL observations were mainly from the the sea surface, with similar proportions of observations from other observation categories ( Figure 2E ). In contrast, relatively few M. tecta observations were made at the sea surface, with comparatively higher proportions of underwater observations, strandings, and fishing boat captures ( Figure 2E ).

Molids were observed from southern California to Kodiak Island in Alaska, with most sightings from the California coast for both size groups ( Figures 3A, B ). Mola mola and M. tecta observations had overlapping distributions from southern California to Alaska, with most sightings along the California coast for both species ( Figures 3C, D ). Mola alexandrini was only observed twice, and both times near the California-Mexico border ( Figure 3E ). A single R. laevis was observed in the southern California Current [iNaturalist observation 1472850 – removed from the final dataset]. Observed molids in the ACS comprised a significantly higher proportion of >1 m TL individuals compared with the CCS [X² (1, N = 1185) = 37.27, p< 0.01] ( Figure 3F ). Similar numbers of M. tecta (n=24 and n=21) were reported in both current systems; however, M. tecta (all >1 m TL) contributed a significantly higher proportion of total molids >1 m TL in the ACS (15.8%) than in the CCS (5.3%) [X² (1, N = 594) = 11.92, p< 0.001] ( Figure 3G ).

DNA from all six samples was successfully amplified and sequenced. The ML phylogenetic tree closely resembled that of Nyegaard et al. (2018), with four major clades with high branch support (95 – 100). Three of these clades consisted of NCBI sequences labeled either Ma. Lanceolatus, M. tecta, or M. mola, respectively ( Figure 4 ). The fourth was a mixed species cluster of NCBI sequences labeled M. mola and M. alexandrini. These findings were similar to those of Nyegaard et al. (2018b), where three “M. mola” sequences were found to be mislabeled M. alexandrini. Our mixed cluster included an additional three, newer “M. mola” sequences, with one of these (LC659949) being M. alexandrini (A. Yamada pers com Nov 2024). The remaining two sequences (OQ918272, PQ169543) are likely cases of mistaken identity, related to the recent redescription of M. alexandrini (Sawai et al., 2018), the historic taxonomic confusion in Molidae, and the lack of updated Molidae field guides. Regardless, all novel sequences from this study clustered with known M. tecta sequences, including the holotype (Nyegaard et al., 2018b), confirming the visual identification of the sampled fish.

Morphological species characteristics were sufficiently visible in most crowd-sourced media for species determination of molids >1 m TL, revealing two species (M. alexandrini and M. tecta) not previously reported from the Alaska or California Current Systems. Opportunistic genetic analyses corroborated the species identity of a subset of stranded M. tecta (n=6) identified visually from photographs.

Our holistic, media-based approach to molid species determination was remarkably consistent among identifiers (no direct determination conflicts; 95% were exact matches among two identifiers). However, as anticipated, our strategy was ineffective with molids <1 m TL (excluding R. laevis), as their diagnostic morphological characteristics had not yet developed or were not visible in crowd-sourced media. Therefore, a media-based approach to molid biogeography is currently most informative for larger molids and less so, without supportive methods, for smaller molids.

Our results suggest that molids exhibit ontogenetic shifts in habitat use in the Alaska and California Current Systems. Molids <1 m TL were frequently reported from the central California coast and comprised over half of all observations in the CCS. These findings align with the known seasonal influx of small molids (approx 30 – 50 cm TL; Thys pers obs; Gotshall, 1961) along the Californian coast, where they strand along beaches and are preyed upon by California sea lions (Thys, 1994). Conversely, larger molids (>1 m TL) dominated observations in the ACS. Larger molids have higher thermal inertia and utilize a broader depth range than smaller individuals (e.g., Nakamura and Sato, 2014), so the significantly higher proportion of larger molids in the colder ACS suggests these molids occupy a wider thermal range across geographic space, too. Our results are consistent with opportunistic spring and summer ship-board surveys in Queen Charlotte Sound, British Columbia, where reported molids were nearly all >1 m TL (Thys and Williams, 2013). These findings suggest that California’s southern and central coast may function as a nursery for small molids, with larger individuals being more broadly distributed in the northern CCS and ACS.

Our study revealed extensive geographic overlap of M. mola and M. tecta in the ACS and CCS. Overlapping distributions of large-bodied molid species have also been reported in other ocean ecosystems (e.g., Yoshita et al., 2009; Nyegaard et al., 2018a; also, see observations on www.inaturalist.org). However, niche partitioning between sympatric molid species is poorly understood (Arostegui et al., 2020), with recent research finding similar trophic niches and narrow diets for the close relatives M. mola and M. alexandrini in Taiwanese waters (Chang et al., 2024).

In our study, species composition analysis revealed that while M. mola was the most observed species within both current systems, the proportion of reported M. tecta (all >1 m TL) relative to other molids >1 m TL was significantly higher in the ACS compared with the CCS. This difference could potentially reflect differing thermal optima between the two species (e.g., Nyegaard et al., 2018a; Arostegui et al., 2020), or niche partitioning in habitat, diet, or foraging between the two species (e.g., Chang et al., 2024). However, little is known about the diet and horizontal and vertical movements of M. tecta. For example, while biotelemetry studies across several ocean ecosystems have reported seasonal latitudinal migrations and repeated deep-water excursions for M. mola (e.g., Dewar et al., 2010; Thys et al., 2015; review in Sousa et al., 2020), no research has been conducted on the movements of M. tecta. The E/V Nautilus’ ROV Hercules observation off Santa Barbara in July 2017 – included in this study – represents this species’ deepest confirmed record (71-74 m).

Gathering crowd-sourced media of molids is inexpensive (albeit time-consuming), however, the diverse means of data collection and variable data quality has implications for species determination. For molids <1 m TL, the species determination success was low for all observation types, except the ‘collection’ category, where detailed images were available. For >1 m TL molids, the species determination success was high and similar across observation types. Of these, the least successful observation type was ‘at sea surface, seen from above water’, mainly due to numerous images showing only a dorsal fin.

Pooling observations from different observation types could potentially lead to bias if behavioral differences between molid species lead to differing probabilities of human detection. For example, several observations in this study (n=5) confirm M. tecta bask at the ocean’s surface (like M. mola - Abe and Sekiguchi, 2012; Nakamura et al., 2015); however, it is unknown if M. tecta and M. mola spend similar time basking with equal likelihood of being observed by vessels. In our study, the smaller proportion of M. tecta observations from the sea surface compared with >1 m TL M. mola could indicate this may be the case. Similarly, while observations (n=4) confirm M. tecta solicit cleaner fish interactions on California reefs (like M. mola - Hobson, 1971; Vasco-Rodrigues and Cabrera, 2015), it is unknown if both species spend similar amounts of time at cleaning stations with equal likelihood of being observed by SCUBA divers. Any such differences could potentially introduce bias in a dataset like ours, originating from many sources (whale-watching vessels, SCUBA divers, beachcombers, etc.). This could further be exacerbated by the varying observational footprints in space and time, including the disparity in observational effort in the CCS (with large population centers) and the ACS (in general, sparsely populated, especially in the northern region).

In conclusion, the diverse means of data collection employed in this study and uncertainty in species-specific molid behaviors create challenges for directly comparing species commonality in the two current systems. Further research is needed to establish if M. tecta is proportionally more common among molids in the ACS than the CCS; two current systems that could be valuable comparative study areas to investigate niche partitioning between two morphologically similar species.

Mola tecta is considered a temperate species (Nyegaard et al., 2018b), with no current observations in tropical waters. While warm equatorial waters may form a thermal barrier for continuous M. tecta distribution between the Northeast and Southeast Pacific Oceans basins (i.e., between California and Peru), trans-equatorial movements could potentially occur through isothermal submergence as has been suggested for other species (e.g., Møller et al., 2003) or become more frequent during cool phases of climate oscillations (e.g., La Niña) when equatorial surface water temperatures drop (e.g., see ENSO-linked Humbolt squid range expansion - Zeidberg and Robison, 2007). While further research is needed to understand M. tecta’s distribution, the large number of confirmed M. tecta from the Northeast Pacific Ocean (n=45) provides compelling evidence for a species range extension in the CCS and ACS. Genetic research could help elucidate this species’ connectivity across its Pacific Ocean distribution.

Key aspects of M. tecta’s reproductive biology remain unknown, including the location of spawning grounds. It is worth noting that all sampled individuals in this study were female. Examination of the ovary of a 205 cm TL M. tecta stranded in Santa Barbara, California (US01 in Supplementary Table 1 ) revealed signs of secondary oocyte recruitment but no indications of prior spawning (Forsgren et al., 2020). The size at maturity is unknown for M. tecta, so it is unclear if this is unusual. It is also unclear whether the CCS functions as a nursery area for this species, like M. mola. All confirmed M. tecta observations were of relatively large fish (all estimated at >1 m TL; size estimates provided by observers ranged from 1.2–2.2 m), however, we cannot dismiss the possibility that smaller M. tecta occur in the Northeast Pacific Ocean.

The use of crowd-sourced observations to study large marine animals presents both exciting opportunities and notable limitations. Overall, we report that media crowdsourcing is a practical, noninvasive, complementary tool to genetic analysis to elucidate species-level molid biogeography, with potential to compile much larger datasets than genetic sampling alone. By mobilizing large numbers of community scientists (in this case, hundreds of observers), we were able to expand the temporal and spatial range of molid data, which would be cost-prohibitive and logistically challenging for traditional monitoring approaches. This approach can be particularly useful for understanding the distribution and habitat use of data-deficient species that are frequently observed by humans. However, key challenges include ensuring data consistency and accuracy (data quality varies), temporal and spatial observation biases (uneven observer density and interspecific behavioral differences), difficulty quantifying or standardizing effort across diverse data sources (e.g., divers, whale-watching vessels, beachcombing), and the loss of ‘absence data’ collected in traditional scientific surveys (which can be useful in predictive models of species range).

Media-based datasets and emerging technologies (e.g., artificial intelligence) have exciting promise for numerous applications, including molid species determinations and recognition of individuals; these technologies could be used to generate large datasets and investigate environmental perturbations on molid strandings, mortality, behavior and site fidelity (e.g., Nyegaard et al., 2023). However, the approach currently has limitations; our crowd-sourced data collection suggests that the central coast of California may function as a nursery for young molids. These small molids, in general, cannot be identified to species from crowd-sourced media alone because they haven’t yet developed clear diagnostic morphological characters. Developing a molid field guide (to include subadult molids) is vital to help identify key habitats for molids’ early life stages and assess conservation status (Thys et al., 2020; Phillips et al., 2023).

Finally, although we find that M. mola is the most frequently observed species in the ACS and CCS, we also report – for the first time – the occurrence of M. alexandrini (in the southern portion of the CCS) and M. tecta (from southern CCS to central ACS).