Module 4: Dispersal Biogeography

Charles Darwin devoted Chapters 11 and 12 of On the Origin of Species(1859) to the problem of how species cross oceans. To test the feasibility of oceanic seed dispersal he conducted a deliberately low-tech but quantitatively rigorous series of experiments in glass jars at Down House. His method was unusually modern: carefully measured salt-water submersion times, controlled germination tests, and triplicated replicates (Darwin 1855, Gardeners’ Chronicle). The published results:

• Of 87 species of seeds immersed in sea water for 28 days, 64 still germinated.
• Mean floating time for viable seeds was 28.6 days; the record for the experiments was hazelnut (Corylus) at 90 days.
• Given the Gulf Stream at ~1.6 km/h, 28 days of floating implies transport of ~1000 km before loss of viability.

Darwin also examined the mud adhering to birds’ feet, demonstrating that a single partridge’s leg mud could yield 82 seedlings. He catalogued viable seed retention in the gut of herons, ducks, and fish. The message was unequivocal: even without atmospheric or driftwood transport, bird- and water-mediated dispersal could populate oceanic islands from continental sources over geologic time. Darwin’s experimental rigour set the standard for all subsequent dispersal biogeography.

van der Pijl’s (1972) Principles of Dispersal in Higher Plantsand Howe & Smallwood’s (1982 Annual Review of Ecology and Systematics) classification distinguishes five mechanisms:

• Anemochory (wind): plumed or winged seeds, spores, small insects, and dust are carried by boundary-layer eddies and, occasionally, jet-stream transport. The Saharan dust plume delivers ~60 Mt of mineral aerosol and microbial propagules to the Amazon basin each year (Yu et al. 2015 Geophys. Res. Lett.), supplying the nutrient-poor Amazonian soil with ~22 000 t of phosphorus annually. Dust-borne fungal spores and bacteria complete intercontinental jumps in a matter of days.
• Hydrochory (water): mangrove propagules, seagrass rhizomes, coconuts, driftwood rafts, and even floating pumice rafts support hydrochoric dispersal. The 1883 Krakatau eruption produced pumice rafts that drifted for months and dispersed at least 20 plant species to Christmas Island and Anak Krakatau. Mangroves (Rhizophora) have viable sea-water flotation periods exceeding a year.
• Zoochory (vertebrate): endozoochory in gut passage, epizoochory on fur and feathers, synzoochory as food caching. Migratory birds carry seeds and pathogens over thousands of kilometres each year. Viggiani Rocha et al. (2014) measured up to 480 seeds in a single wood thrush gut.
• Anthropochory (human-mediated): the dominant dispersal mechanism of the Anthropocene, ranging from deliberate crop and livestock movement to unintentional ballast-water transport. Hulme (2009 Journal of Applied Ecology) reviews the six pathways by which species cross borders under human transport.
• Ballochory / autochory: explosive self-dispersal (Impatiens, Hura crepitans), rarely producing distances over 10 m — irrelevant to biogeography but crucial to within-population dynamics.

Biogeographically relevant dispersal is dominated by wind, water, and vertebrate mediation. Within a given mechanism, the functional trait suite of a propagule (Raunkiær life-form, seed mass, appendages, buoyancy) determines its dispersal kernel shape (Thomson et al. 2011 Journal of Ecology).

A dispersal kernel \(f(r)\) is the probability density of a propagule landing at radial distance \(r\) from its parent. For most biogeographic purposes we treat it as isotropic and two-dimensional; the radial density is \(2\pi r f(r)\). Commonly fitted families:

• Gaussian: \(f(r) = \frac{1}{2\pi\sigma^{2}}\exp(-r^{2}/(2\sigma^{2}))\). Derives from Fickian diffusion of a cloud of propagules; tails fall as \(\exp(-r^{2})\) and are too thin to capture LDD.
• Exponential power (Clark 1998): \(f(r) \propto \exp(-(r/a)^{b})\) with shape parameter \(b\). For \(b < 1\) this is fat-tailed; \(b = 2\) recovers Gaussian.
• Log-Laplace (Austerlitz & Dick 2004): tails decay as \(r^{-b-1}\) (power-law). This is the canonical fat-tailed kernel for wind-dispersed tree seeds.
• Mixture kernels: sum of a local Gaussian and a long-distance Cauchy or power-law component, often the best fit to empirical seed-rain data with a tail that extends to unexpected distances (Nathan 2006 Science).

Fat tails matter for biogeography because the probability of colonising a distant island depends on the tail, not the mode. If the tail falls too fast (Gaussian), no island is reachable and biogeography should look like strict vicariance; if the tail is fat enough, even remote archipelagos receive occasional propagules and dispersal dominates. The distinction is quantitative.

The Hawaiian silversword alliance (Asteraceae: Madiinae) consists of ~30 species in three endemic Hawaiian genera (Argyroxiphium, Dubautia, Wilkesia). Baldwin & Sanderson (1998 PNAS) dated the radiation to ~5 Ma from nuclear ITS sequences; Landis et al. (2018) refined this with full-genome data. The alliance descends from a single colonising seed of a Californian tarweed (Madia-like ancestor) that reached Kauai ~5.2 Mya. That single propagule seeded a spectacular adaptive radiation spanning alpine rosette shrubs of Haleakala’s cinder cones, wet montane forest trees, desert cushions, and rainforest lianas — perhaps the most morphologically diverse angiosperm radiation on any island system.

Key observations:

• Single colonisation event (Baldwin & Sanderson 1998): silverswords are monophyletic with respect to their Californian relatives.
• Progression rule (Funk & Wagner 1995): older lineages occur on older islands; newer species on younger islands. Kauai (5.1 Ma) holds the sister to the clade-wide most recent common ancestor; the Big Island (< 0.5 Ma) holds mostly recently derived species.
• Hybridisation is extensive: Friar et al. (2007) documented recurrent inter-island hybrid swarms, complicating the strict progression rule.
• Trait disparity: leaf area spans 2.5 orders of magnitude, growth form spans rosette, liana, shrub and cushion. This is an order of magnitude greater morphological span than is found in all 300+ species of the continental parental tarweed clade.

The silversword case embodies the key insight of island radiation: a single chance colonisation event of an ecologically open island can seed extraordinary within-lineage diversification. The same logic explains Galápagos finches, Hawaiian honeycreepers, African rift-lake cichlids, and Caribbean anoles.

The Galápagos giant tortoises (Chelonoidis) provide a different LDD case: a vertebrate whose buoyancy and desiccation resistance allow ocean rafting. Caccone et al. (2002 PNAS) showed that all 13 extant Galápagos species descend from a single colonist of the Chaco tortoise lineage (C. chilensis sensu lato) that reached the archipelago ~2 Mya, crossing ~1000 km of open Pacific from the South American coast. Giant tortoises survive long flotation because they can exceed a month without food or water and have a shell volume that floats. The Galápagos founder carried the genetic bottleneck typical of colonisation: modern species have reduced genome-wide heterozygosity compared to the mainland source.

The same lineage reached the Mascarenes and Aldabra via independent rafting events. Austin & Arnold (2001) used mtDNA clocks to date the Aldabran giant tortoise (Aldabrachelys gigantea) colonisation at ~5 Mya from the Madagascar lineage. These parallel cases demonstrate that giant reptile LDD is not a rare anomaly but a repeatable biogeographic process, requiring only sufficient founder buoyancy and metabolic tolerance.

Madagascar separated from Africa ~160 Mya and from India ~90 Mya. Yet its endemic lemur fauna (Strepsirrhini: Lemuroidea) diverged from African strepsirrhines ~60 Mya, long after the island was biologically isolated. The oldest lemur ancestor therefore crossed the 400-km Mozambique Channel after the continental disjunction.

Ali & Huber (2010 Nature) resolved the apparent puzzle with palaeoceanographic reconstructions. They showed that the present strong east-to-west currents in the Mozambique Channel only developed in the middle Miocene (~20 Mya). Before that, during the Palaeocene and early Eocene, the Channel had an opposite westward-flowing surface current that would carry floating rafts from the African coast to Madagascar in ~25 days. This window of opportunity allowed the lemur ancestor — and also the ancestors of Malagasy tenrecs, euplerid carnivorans, and nesomyine rodents — to raft-colonise on driftwood or vegetation mats spawned by African tropical storms. After ~30 Mya the current reversed and Madagascar became the ecological island it is today.

The rafting model is testable: it predicts that each major Malagasy mammalian clade should have a single African sister and an Eocene-early Miocene divergence age. Poux et al. (2005) and Masters et al. (2006) provided molecular confirmation across multiple independent clades. Hippopotamuses reached Madagascar by a later, Miocene raft event — rather astonishing given their size, but consistent with the known buoyancy of floating mats and the occurrence of wild swimming behaviour in extant hippos. The Malagasy dwarf hippos went extinct within the last 1000 years during human colonisation.

Human colonisation of Remote Oceania between ~3500 BP and 1300 CE is a well-documented anthropogenic analogue of natural stepping-stone LDD. Austronesian-speaking Polynesians, sailing double-hulled canoes and using star compasses (Finney 1979; Irwin 1992), colonised successively more distant islands: Fiji, Samoa, Tonga (~3000 BP); Society, Cook, Marquesas (~1000–1500 BP); Hawaii, Easter Island, New Zealand (~800–1300 CE).

Each colonising voyage carried a deliberate cargo of economically useful plants and animals: Pacific rat (Rattus exulans), chicken (Gallus gallus), dog, pig, breadfruit, taro, banana, yam, sweet potato (the last providing strong evidence for a pre-Columbian contact with South America, Roullier et al. 2013 PNAS, genetically). Pigafetta (1522), a chronicler of Magellan’s voyage, documented the rat populations on Guam at European contact; these were already Pacific-rat populations introduced by earlier Polynesian colonisation, not European ship rats. The resulting “rat genetic barcode” across Polynesia provides an independent molecular tracer of human colonisation routes (Matisoo-Smith & Robins 2004 PNAS).

The archaeological and genetic picture shows that anthropochorous stepping-stone colonisation operates on time scales four orders of magnitude faster than natural LDD: Polynesians colonised the area from New Zealand to Easter Island (spanning 8000 km) in ~500 years, whereas natural stepping-stone plant radiations require millions of years.

Rosemary Gillespie (2004, Science; 2008, Philosophical Transactions B) developed network-theoretic models of island community assembly. In Gillespie’s framework, islands are nodes and dispersal events are directed edges whose weight is the expected number of colonisation events per unit time (product of donor propagule pressure, oceanic-current geometry, and recipient habitat availability). The resulting metacommunity network predicts both the expected richness and the turnover structure of island assemblages.

Applied to Hawaiian Tetragnatha spiders, Gillespie (2004) showed that the observed ecomorph radiation (colour-matched body forms on each island) is the signature of deterministic character convergence rather than accidental parallel evolution. Each island independently assembles the same four ecomorphs (maroon, green, small brown, large brown) starting from dispersal-limited colonisation pools. The network model quantifies the probability that each island’s assembly follows the same trajectory and identifies the geometric parameters (inter-island distance, habitat connectivity) that determine when parallel radiation is expected versus when divergent ecomorph sets arise.

Gillespie (2012 Evolution) extended the framework to arthropods more broadly, showing that per-clade predictability of ecomorph assembly scales inversely with dispersal ability: weakly dispersing spiders show tight parallel radiation; strongly dispersing insects show more idiosyncratic assembly. The same logic explains why Anolis lizards radiate into the same six “Greater Antillean ecomorphs” on Cuba, Hispaniola, Jamaica, and Puerto Rico (Losos et al. 1998 Science): the island network is dispersal-limited enough for independent convergence but connected enough for repeated colonisation.

Christen Raunkiær (1934) proposed a life-form classification of plants based on the height of perennating buds above the ground surface. His scheme — phanerophytes, chamaephytes, hemicryptophytes, geophytes, therophytes — implicitly tracks climate (harsh climates favour buds at or below the surface) and indirectly dispersal (therophytes produce abundant small seeds, phanerophytes produce heavy seeds or fruit rewards).

Dispersal syndromes are functional correlations between seed traits and dispersal vectors:

• Anemochorous: plumed or winged seeds, small mass (typically < 1 mg). Dandelions, pines, orchids.
• Hydrochorous: buoyant, often with oil-filled exocarp. Cocos, Rhizophora, seagrass rhizomes.
• Endozoochorous: fleshy fruit, hard-coated seeds surviving gut passage. Figs, cherries, mistletoes.
• Epizoochorous: barbed awns, hooks or burs. Goose-grass, burdock, and most grassland compositae.
• Ballochorous: explosive fruit, short-range. Impatiens, Hura.

The same syndrome may be achieved by convergent morphologies in distantly related taxa — the classic case of the exocarp of Cocos and the coriaceous pericarp of the unrelated West African Manicaria. Syndrome-based prediction of dispersal kernel shape is only moderately successful; empirical kernels show extensive within-syndrome variation (Thomson et al. 2011 Journal of Ecology).

A crucial distinction in dispersal biogeography is between jump dispersal (a single successful founder event across open ocean) and stepping-stone dispersal (a sequence of short-range movements across many intermediate islands). They produce very different phylogenetic signatures.

• Jump dispersal: long branches in the ancestral-area reconstruction, one lineage suddenly appearing in a new area with no intermediates. Typical of trans-oceanic colonisation by mammals and land reptiles.
• Stepping-stone dispersal: nested series of range expansions tracking adjacent landmasses or island chains. Produces a “progression rule” when islands have age ordering. Typical of flying arthropods, birds, and wind-dispersed plants in archipelagos.

In DEC ancestral-area modelling the distinction is encoded by whether the allowed range-transition matrix permits only adjacent-area transitions (stepping-stone) or allows any pair of areas (jump). Matzke’s (2014) +J extension of DEC is explicitly aimed at modelling jump-dispersal founder-event speciation on young island systems, where classical DEC under-predicts the frequency of sudden area shifts.

The compilation above illustrates the point emphasised throughout this module: dispersal capacity is a continuous quantitative trait, not a binary “can / cannot” category, and the biogeographic footprint of a clade reflects the entire shape of its kernel — especially the tail.

The modern consensus in dispersal biogeography is quantitative and hybrid: most clades show a mix of vicariance, jump dispersal, stepping-stone dispersal, and human-mediated transport. Open problems include:

• Rigorous inference of dispersal kernels from present-day phylogeography (Rousset 1997; Nathan 2012).
• Integration of oceanic-current palaeocirculation (Ali & Huber 2010) into ancestral-area models.
• Quantifying the rate of anthropogenic homogenisation of biota (McKinney & Lockwood 1999 TREE), and whether modern dispersal floods are already producing “New Pangaea” patterns.
• Predicting extinction debt from fragmentation under climate change using kernel-based metacommunity models (Thompson et al. 2020).
• Mechanistic explanation of kernel fat tails: are they the emergent property of hierarchical turbulent transport, of rare atmospheric jets, or of individual variation in propagule buoyancy/airborne time?