A study examines the reasons behind the reactivation of the world’s tallest geyser.
Image credit: Mara H. Reed.
Edited by Atholl J. Anderson, The Australian National University, Canberra, Australia, and accepted by Editorial Board Member James O'Connell September 15, 2016 (received for review July 27, 2016)
Significance
The colonization of Remote Oceania between ∼3400 and 800 B.P.—from multiple origin points and across hundreds or thousands of kilometers of open ocean—represents some of the most impressive population dispersals on Earth. For decades, scholars have sought to explain when and how these occurred. Here we show that seafaring simulation techniques coupled with sophisticated analyses of climatic data, including the role of El Niño Southern Oscillation events and island distribution on ancient voyaging, are critical comparative tools for understanding the variables—culturally, technologically, and environmentally—that structured movement across the world’s largest ocean. These data also pinpoint likely departure points for ancient Pacific Islanders that in some cases support or negate current archaeological and other lines of evidence.
Abstract
The prehistoric colonization of islands in Remote Oceania that began ∼3400 B.P. represents what was arguably the most expansive and ambitious maritime dispersal of humans across any of the world’s seas or oceans. Though archaeological evidence has provided a relatively clear picture of when many of the major island groups were colonized, there is still considerable debate as to where these settlers originated from and their strategies/trajectories used to reach habitable land that other datasets (genetic, linguistic) are also still trying to resolve. To address these issues, we have harnessed the power of high-resolution climatic and oceanographic datasets in multiple seafaring simulation platforms to examine major pulses of colonization in the region. Our analysis, which takes into consideration currents, land distribution, wind periodicity, the influence of El Niño Southern Oscillation (ENSO) events, and “shortest-hop” trajectories, demonstrate that (i) seasonal and semiannual climatic changes were highly influential in structuring ancient Pacific voyaging; (ii) western Micronesia was likely settled from somewhere around the Maluku (Molucca) Islands; (iii) Samoa was the most probable staging area for the colonization of East Polynesia; and (iv) although there are major differences in success rates depending on time of year and the occurrence of ENSO events, settlement of Hawai’i and New Zealand is possible from the Marquesas or Society Islands, the same being the case for settlement of Easter Island from Mangareva or the Marquesas.
Population dispersals that led to the colonization of islands in Remote Oceania, which began ∼3400 B.P. from multiple origin points across millions of square kilometers of open ocean, are considered to be some of the greatest feats of seafaring in human prehistory. Though archaeological, linguistic, and genetic data have provided substantial evidence for when peoples first arrived to many of the region’s islands and archipelagoes (1⇓⇓–4), important questions still remain as to where these early colonists originated from and the strategies and trajectories used to reach what are arguably the most distant and isolated patches of land on earth.
Several models have been developed using combinations of these datasets to explain how movements into the Pacific were structured. In terms of Austronesian speakers who first settled Near Oceania and that eventually dispersed into non-Micronesian Remote Oceania, four models are often cited (5). The “Express Train to Polynesia” model suggests that peoples moved quickly from Taiwan into West Polynesia ∼3400 B.P. with minimal or no contact with other native groups. Though not explicitly addressed in this model, the populations venturing southward seem to have splintered off from somewhere in Island Southeast Asia between ∼3400 and 2900 B.P., moving east into western Micronesia (the Marianas, Palau, and perhaps Yap). An alternative model known as the “Bismarck Archipelago Indigenous Inhabitants” argues that the development of the Lapita Cultural Complex was an in situ process and not related to major population dispersal. The “Slow Boat to the Bismarcks” proposes that peoples established interaction networks and honed their voyaging skills between islands from eastern Indonesia to the Bismarcks and Solomons ∼6000–3500 B.P., and then ∼3100 B.P. (possibly centuries earlier for W Micronesia) these Lapita groups moved quickly into Remote Oceania. The last “Voyaging Corridor Triple I” (intrusion, integration, and innovation) model follows the “Slow Boat” option, but concedes that some aspects of the Lapita cultural complex could have been introduced by indigenous groups already living in Near Oceania. An additional scenario is also presented for how Lapita peoples colonized Remote Oceania—the “Mobile Founding Migrant” model—that it was a quick but unstable movement that eventually reached Samoa and Tonga, but went no further (5). Nearly 1,000 y later, much of central (Chuuk, Pohnpei, Kosrae, and various atolls) and eastern (e.g., Marshalls and Kiribati) Micronesia was settled from Island Melanesia and/or West Polynesia ∼2000 B.P. After another 1,000 y, between ∼1200 and 800 B.P., the nodes of the Polynesian triangle (Hawai’i, New Zealand, and Easter Island) were settled, as were many of the islands in between, though questions still remain as to the processes involved technologically, culturally, and environmentally that influenced these dispersals.
An additional analytical tool that has proven extremely useful in examining ancient seafaring techniques and capabilities is computer simulation of voyaging. Past research in the Pacific and elsewhere (6⇓⇓⇓⇓⇓⇓–13) have used a number of environmental, cultural, and biological variables integrated into varying types of simulation platforms to identify likely departure and arrival points of colonists as well as seasonal or annual variations of climatic regimes (e.g., El Niño) that may have influenced those journeys, and modeled them successfully against known historical voyages (14).
In this study, we have conducted the most detailed and sophisticated seafaring simulation project in the Pacific to date, which is also a systematic, region-wide evaluation of the potential role of ENSO-related wind and rainfall variability in colonization of Remote Oceania. Our analysis takes advantage of new, high-resolution databases of wind, currents, precipitation, and land distribution in the region to estimate seafaring and “shortest-hop” trajectories from likely origin points and seasonality of departure for peoples who colonized five major regions of the Pacific: (i) western Micronesia (Palau and the Marianas) ∼3400–3000 B.P.; (ii) central and eastern Micronesia (Chuuk, Pohnpei, Kosrae, and the Marshall Islands) ∼2000 B.P.; (iii) the Marquesas and Society Islands; (iv) Hawai’i and New Zealand; and (v) Easter Island between ∼1200 and 800 B.P. (Tables S1–S5 and SI Data and Methods). Our results suggest that climatic variability was paramount in decision-making, with precipitation variability a potential driver in colonization ventures. Our spell displacement vs. easting distance analysis indicates how efficient downwind sailing could be during relatively short and common voyages between islands and archipelagos throughout the region.
View this table:
Table S1.
Simulated downwind voyages arriving at Palau from distinct origin sites
View this table:
Table S2.
Simulated downwind voyages arriving at the Mariana Islands from distinct origin sites
View this table:
Table S3.
Simulated downwind voyages arriving at central and eastern Micronesia from distinct origin sites
View this table:
Table S4.
Simulated directed voyages: First semester
View this table:
Table S5.
Simulated directed voyages: Second semester
SI Data and Methods
Computer Simulations of Voyaging with the Seascape Model.
For western Micronesia (Tables S1 and S2), simulations of downwind sailing are conducted from nine origin points: Taiwan; Luzon (northern Philippines); Samar (central Philippines); southeast of Mindanao (southern Philippines); Halmahera (Maluku Islands); western, central, and eastern New Guinea; and the Bismarck Archipelago. Downwind simulations are also conducted between Palau and the Marianas in both directions. For origins from central and eastern Micronesia (Table S3), the four departure points used in downwind sailing simulations include the western and eastern Solomon Islands, Vanuatu, and Fiji. The following voyages were evaluated by both downwind and directed sailing experiments: (i) from Samoa to the Marquesas and to the Society Islands; (ii) from the Marquesas and Society Islands to Hawai’i and to Easter Island; and (iii) from the Society Islands to New Zealand (Tables S4 and S5).
Land Distribution and Shortest-Hop Trajectories.
The change in spatial resolution of the land distribution data are performed to facilitate processing and done via linear interpolation. To make sure the procedure did not generate spurious results, we compare the courser (∼1.8 km) set against coastlines from the version 2.3.4 of the Global Self-Consistent, Hierarchical, High-Resolution Geography Database (31). The analysis indicates no land areas are lost or created during the interpolation of the original GEBCO data from ∼900 to ∼1,800-m resolution.
There are indications that between ∼3000 and 1500 B.P., relative sea level in at least some areas within the region of interest was ∼1 m higher than today (32). To evaluate the effect of sea level change on our results, short-hop trajectories are also determined, in addition to the current state, for conditions in which sea level is raised by 1 and 2 m relative to present day. Under the assumption that small islands are harder to locate and reach, additional paths are generated based on maps from which islands smaller than ∼1.8 × 1.8 km and then 3.6 × 3.6 km are removed. Trajectories are also generated for maps that combine the effects of sea level change and minimum island size cutoff. Finally, to test trajectory robustness, short-hop paths based on the second and third closest islands to the east are calculated for all sea level and land distribution configurations described above. The selection of three sea levels; three island-size cutoff maps; and closest, second closest, and third closest paths results in 27 shortest-hop trajectories being generated for each departure point. Many trajectories started from the same point, but established under different controlling parameters, overlap, which we take as an indication of the method’s robustness (Fig. 1).
If no restrictions are applied, shortest-hop paths initiated at the Philippines and Maluku Islands result in trajectories that move southeast to New Guinea, the Solomons, and then into islands in East Polynesia. To evaluate crossings into Micronesia from starting points in the Philippines and Maluku Islands, the New Guinea and the Solomon groups are removed from the analysis by setting their land bins to ocean. In addition, Mindanao and the Malukus are removed for the central Philippine departures, and the Maluku Islands removed for the southern Philippines departures.
To evaluate the potential path from the central portion of the northern New Guinea coast to the Marianas, we generate a directed shortest-hop trajectory, which contains a destination point in addition to the initial departure one. Individual “hops” along this type of trajectory are given by the shortest crossing from the present position that at the same time decreases the distance to the destination point.
Wind and Precipitation.
The analysis focuses on seasonal as well as the interannual variability associated with ENSO cycles that have a periodicity ranging from ∼2–7 y and are comprised of an El Niño, or warm phase, and a La Niña, or cold phase. Given the temporal scales of interests, wind data are averaged into daily, monthly seasonal, and annual values with precipitation averaged into seasonal and annual values. Wind and precipitation analysis are conducted at the original spatial resolution of the datasets: 1 × 1° for wind and 2.5 × 2.5° for precipitation.
The adopted present-day data provide significantly higher spatial and temporal resolution than available proxies, and very likely, a better representation of variability than paleo models do for the period of interest. Also, there are many indications that the present-day atmospheric patterns over the low-latitude Pacific, including the variability associated with ENSO, had been established by 3000 B.P. (33⇓–35).
Data and Methods
Computer Simulations of Voyaging.
The analysis is conducted with the Seascape model and based on monthly current and wind data from the US Navy Marine Climatic Atlas of the World Version 1.1 (14, 15). The model randomly selects frequency-weighted wind and current data from the Marine Climatic Atlas. Wind and current forces are allowed to operate on vessels for a period of 24 h and then a new selection is made. Because the precise vessel type used in the region at the time is unknown, generalized speeds for small vessels under various wind and current forces are used (11). Vessels are either sailed before the wind (downwind sailing) or given a heading toward target islands (directed sailing). In the directed-sailing simulations, vessels are provided with minimum headway from 1.852 to 5.556 km/h [1–3 knots (kn)]—in other words, approaching the target island by a minimum of 44.5–133.3 km (24–72 nautical miles) a day regardless of wind direction (11). Maximum voyage duration is set to 200 d, reflecting survival limits of known drift voyages (11). One hundred voyages are simulated from each point of origin beginning in all 12 months of the year. Success is achieved after coming within detection distance of 10 nautical miles (18.5 km) from the islands (16) (Figs. S1 and S2).
Fig. S1.
Paths of successful crossings for directed voyaging simulations between (Upper) Samoa and the Marquesas and (Lower) Samoa and the Society Islands. Both results refer to trips started under April conditions.
Fig. S2.
Paths of successful crossings for directed voyaging simulations between (Top) Society Islands and Easter Island with start under February conditions; (Center) Marquesas and Easter Island for start under August conditions; and (Bottom) Mangareva and Easter Island for start under January conditions.
The necessary minimum headway speed is likely unrealistic given average conditions, particularly for crossings from the Marquesas Islands to Hawai’i, the Society Islands to Hawai’i, and the Society Islands to New Zealand. The crossings from Samoa to the Marquesas and Societies are likely feasible given sufficient windward sailing efficiency, although this would be greater than the generalized speeds for the vessels used in the downwind sailing simulations. It should be noted that there is some debate as to whether windward sailing capabilities were limited or absent during the era of colonization (17, 18); this would apply to the directed simulations to Easter Island from the Marquesas and Society Islands, but there is also the problem of the target being very small. Maintaining a consistent heading over the distance would be extremely difficult.
Environmental Data and Analysis.
Land distribution comes from a global gridded bathymetry data set (19) with a resolution of 0.0083° (∼900 m at the equator) interpolated to 0.017° (∼1.8 km at the equator) before our analysis (SI Data and Methods). These data are used to determine shortest-hop trajectories, defined as trajectories that would be formed if eastward displacement took place by a sequence of crossings in which each individual crossing always reaches the closest island to the east of the departure island. We analyze trajectories starting from the easternmost limits of the central and southern Philippines, Maluku Islands, and Solomon Islands. To account for uncertainties in the methodology and potential changes in sea level, 27 shortest-hop trajectories are generated for each departure point (SI Data and Methods and Fig. 1). To produce trajectories into Micronesia, some landmasses to the south of departure points in the Philippines and Maluku Islands are removed from these experiments (SI Data and Methods).
Fig. 1.
All shortest-hop paths and mean westerly wind conditions. (Upper) Filled red lines: all shortest-hop paths with starts from central and southern Philippines, Maluku, and Solomon departure areas; dashed red line: present-day sea level and all island sizes directed shortest-hop path from New Guinea to Marianas. Gray shade areas with westerly winds on the green line: annual average; dashed yellow line: all years June–November average; blue dashed line: El Niño June–November average. See Table S6 for island abbreviations. (Lower) Same as Upper with gray shade showing areas with westerly winds on the dashed yellow line, all years December–February average; blue dashed line: El Niño December–February average.
View this table:
Table S6.
Island and island group name abbreviations used in maps
Wind data come from the ERA-Interim Reanalysis (20) with 1 × 1° spatial resolution and one estimate every 6 h. Monthly gridded precipitation values based on satellite and rain gauge measurements with 2.5 × 2.5° resolution come from the GPC Project (21). Wind and precipitation are analyzed from January 1, 1979, to December 31, 2012. The use of a present-day wind and rain data are based on the assumption that current, observation-based values offer the best available representation of atmospheric conditions in the area at ∼3000 B.P. (SI Data and Methods).
Assuming periods of constant westerly (toward the east) winds could play a role in voyaging strategy, we determine duration and location of westerly spells, defined as periods during which daily averaged winds flow continuously to the east. Spells are used to generate potential spell eastward displacement P, given by
where d is spell duration, U is average westerly wind speed during spells of duration d and e, and a downwind sailing efficiency of 0.5, based on published values (7). Our choice of e results in vessel speeds from 2.25 to 4 kn (4.6–9 km/h), well within the speed range of traditional Pacific watercraft.
Results
Colonization of Micronesia.
Though easterly (toward the west) winds dominate the annually averaged flow over the southern and central Philippines and most of Micronesia, the area in the vicinity of the Maluku Islands exhibits mean annual westerly winds. Large portions of Micronesia exhibit average westerly winds during the Northern Hemisphere (NH) summer and fall, and this area of seasonal westerlies expands significantly during El Niño (Fig. 1). A strategy of downwind sailing during the relatively common 3- and 5-d westerly spells—which are more frequent under El Niño conditions—would be capable of providing the easting distance required by all shortest-hop crossings in Micronesia (Fig. 2 and Figs. S3 and S4). La Niña is associated with easterly wind anomalies that hinder eastward displacement (Fig. 3).
Fig. 2.
Crossing distance and along-path spell displacement and frequency. (Top) Shortest-hop path from Maluku (Mal) with present-day sea level and all islands considered. (Center) Eastward distance for individual crossings above 50 km (red X’s at the longitude of the arrival island) and average potential eastward displacement of vessels sailing downwind with 50% efficiency for westerly spells lasting 3, 5, and 7 d (filled lines). (Bottom) Average annual frequency of 3-, 5-, and 7-d westerly spells (same color scheme as Center). Potential displacement and frequency represent along-path averages.
Fig. 3.
ENSO annual anomalies. Wind (green vectors) and precipitation (colors, mm/d) deviations from the long-term annual mean during El Niño (Upper) and La Niña (Lower) years.
Fig. S3.
Same as Fig. 3 but for northern crossings starting from the Solomons (Sol). (Top) Shortest-hop path with present-day sea level and all islands considered. (Center) Eastward distance for individual crossings above 50 km (red X’s at the longitude of the arrival island) and average potential eastward displacement of vessels sailing downwind with 50% efficiency for spells lasting 3, 5, and 7 d (filled lines). (Bottom) Average annual frequency of 3-, 5-, and 7-d spells (same color scheme as the center). Potential displacement and frequency represent along-path averages.
Fig. S4.
Average duration of eastward spells in days during (Upper Left) all SH summers (December–February); (Lower Left) El Niño SH summers; (Upper Right) all NH autumns (September–November); and (Lower Right) El Niño NH autumns. Note the difference in color scale between Left and Right. In all panels, red lines represent shortest-hop trajectories (Fig. 1).
Crossings to Palau.
For shortest-hop trajectories departing from the Philippines and Maluku Islands, movement into Micronesia requires crossings to Palau and Yap. Paths departing from the Maluku Islands can arrive at Palau through crossings that do not exceed 360 km, significantly shorter than the longest crossings required to reach Palau from the southern (∼657 km) or central (∼735 km) Philippines. Many crossings departing from the Maluku Islands take place in an area with average annual westerly winds (Fig. 1).
Simulations are in good agreement with wind and shortest-hop results. No downwind voyage from Taiwan, the northern and central Philippines, or the Bismarck Archipelago reaches Palau. Successful crossings to Palau take place from the southern Philippines, the Maluku Islands, western New Guinea, and the Marianas (Table S1). Average crossing times of ∼1 mo are simulated for the first two areas, but Maluku Island crossings have a larger probability of success (5–35%) than those from the Philippines (1–11%). Crossings from New Guinea are faster (15 d), but occur only for simulations started in July and have much smaller success rates (1%). With the exception of the Mariana departures, feasible downwind voyages are simulated for NH summer and fall, with crossings under September conditions showing the highest success rates. Trips departing the Marianas to Palau between December and April have success rates ranging from 1% to 5%, the highest value coming from February departures (Table S1).
Drier conditions associated with El Niño over the Philippines, the Maluku Islands, and New Guinea could act as an incentive for movement away from these areas (Fig. 3 and Figs. S5 and S6). El Niño wind anomalies favor eastward movement out of the Philippines and northeastward displacement from the Maluku Islands, particularly during the already favorable conditions of the NH summer (Fig. 3 and Fig. S6).
Fig. S5.
ENSO seasonal anomalies. Wind (green vectors) and precipitation (colors, mm/d) deviations from the long-term December–January–February mean under El Niño (Upper) and La Niña (Lower) conditions.
Fig. S6.
ENSO seasonal anomalies. Wind (green vectors) and precipitation (colors, mm/day). (Upper) deviations from the long-term June–July–August mean under El Niño conditions. (Lower) Deviations from the long-term September–October–November mean under La Niña conditions.
Crossings to the Mariana Islands.
The majority of shortest-hop trajectories departing from the Philippines and the Maluku Islands move east along Micronesia, but when one or two bin-sized islands are ignored (SI Data and Methods), some pathways include crossings to the Marianas (Fig. 1). Simulated downwind voyages from Taiwan and the Philippines are not able to reach the Marianas, but August–October departures result in successful crossings from Palau, the Maluku Islands, New Guinea, and the Bismarcks. Central New Guinea voyages initiated in September show the highest probability of success (25%). Bismarck September crossings also show relatively high probability of occurrence (13%) and have an average duration of 93 d, ∼10% shorter than average central New Guinea crossings (Table S2). La Niña is associated with drier conditions as well as easterly and southeasterly wind anomalies that could have played a part in movement from New Guinea and the Bismarcks toward the Marianas (Fig. 3 and Fig. S6). Crossings from Palau to the Marianas exhibit the smallest probability of success (1–2%) but have the shortest durations (52 d on average).
Crossings to central and eastern Micronesia.
Central and eastern Micronesia are reached by simulated downwind voyages starting from the eastern and western Solomon Islands, Vanuatu, and Fiji, although eastern Solomon and Vanuatu departures result in success rates of only 1% (Table S3). Successful voyages from the western Solomons are observed between July and October. The largest success rate (24%) occurs in July crossings that arrive in the region around Chuuk after an average trip of 84 d. August is the month with the most successful crossings from Fiji (14%), which have an average duration of 124 d and also arrive in the region near Chuuk. Both source areas become drier during El Niño, when wind changes also facilitate movement to the north and east (Fig. 3 and Fig. S6).
Colonization of Polynesia.
Easterly flow dominates the annually averaged mean wind over West Polynesia, but during the southern hemisphere (SH) summer (December to February), westerly winds are found over a significant area within the region. The area of mean westerlies expands during El Niño summers, reaching as far east as the northern Cook Islands (Fig. 1). Areas with longer westerly spells also expand eastward during El Niño conditions (Fig. 1 and Fig. S4). In addition to the westerly anomalies around the equator, El Niño conditions are also characterized by southerly wind anomalies further south (Fig. 3). Much of West Polynesia becomes drier during El Niño years, with some increase in precipitation observed over the western portions of East Polynesia. La Niña causes a decrease in precipitation over most of Polynesia in addition to easterly wind anomalies near the equator and northerly anomalies at higher southern latitudes.
Crossings to Samoa.
Shortest-hop departures from the East Solomon Islands generate three general trajectories that reach Samoa: a northern route through Tikopia, Rotuma, and Tuvalu, another that from Rotuma reaches Fiji, and a third that moves through Vanuatu, Fiji, and then Tonga. The wind regime would favor eastward displacement along the northern routes (Fig. 1 and SI Data and Methods).
The eastward distance required in all shortest-hop crossings toward Samoa from the Solomon Islands can be covered by downwind sailing trips making use of 3- and 5-d westerly spells (Fig. 2). Shortest-hop trajectories toward Samoa that start in Palau, move east through Micronesia, and then southeast along Kiribati and Tuvalu contain fewer challenging crossings (long distance compared with wind easting) than trajectories that reach Samoa through West Polynesia (Fig. 2 and Fig. S3), though there is no archaeological evidence to support this advance.
Crossings to the Marquesas and Society Islands.
In all shortest-hop trajectories, movement from West Polynesia into East Polynesia occurs through Samoa (Fig. 1). Several components of our analyses point to the relative difficulty of moving eastward from that area. All shortest-hop trajectories encounter above-average easting distance for crossings from Samoa toward East Polynesia (Fig. 2 and Fig. S3). For present-day sea level, required crossings span ∼400 km, and though this is only 11% larger than the longest Micronesia crossing and ∼50% shorter than the Tikopia–Rotuma hop in West Polynesia, these take place in a region where westerly winds are less frequent than in areas to the west (Fig. 1 and Fig. S4). Downwind sailing simulations did not generate successful crossings in East Polynesia where viable trips only occurred in directed sailing experiments with minimum headway speeds from 1 to 3 kn. However, it seems unlikely that such speeds could be maintained over long distances, largely to windward, or that the target islands could be reached so directly. Under a minimum headway speed of 1 kn, the Marquesas and the Society Islands could be reached from Samoa in trips started during several months. Crossings to the Marquesas generally show higher success rates. For both target island groups, voyages started in April show the largest probability of success and shortest duration (Tables S4 and S5 and Fig. S1). El Niño conditions decrease precipitation over Samoa and make eastward travel easier, particularly toward the Marquesas (Fig. 3 and Figs. S5 and S6).
Crossings to Hawai’i, Easter Island, and New Zealand.
Hawai’i can be reached by directed voyages from both the Marquesas and Society Islands with most successful crossings taking place during the NH fall. Trips from the Marquesas are ∼60% faster, but require more efficient windward sailing capabilities (Tables S4 and S5). La Niña conditions result in drier conditions for the Society Islands. No large change in precipitation is observed over the Marquesas under La Niña conditions, but movement to the northwest toward Hawai’i, especially during the NH fall and in the area north of the equator, is easier than usual due to wind anomalies (Fig. 3 and Fig. S6).
Simulations indicate that Mangareva is a more likely source area for the settlement of Easter Island than the Marquesas. Mangareva crossings have more viable start months (three instead of two), higher probability of success (92–45% vs. 57–6%), and shorter average duration (65–47 d compared with 109–92 d) (Tables S4 and S5). Both crossings occur in an area that experiences northwesterly wind anomalies during El Niño SH winters, which would favor movement toward Easter Island. Directed voyages from the Society Islands to Easter Island are only successful for January departures with a 13% probability of occurrence. El Niño wind anomalies are small and do not clearly facilitate the Society Islands to Easter Island voyage (Fig. 3 and Figs. S5 and S6).
Simulations show that voyages from the Society Islands to New Zealand can occur throughout the year, with a larger probability of success and shorter travel times for trips started between November and March (Tables S4 and S5). During La Niña years, the area near the Society Islands is drier and wind anomalies favor movement to the southwest toward New Zealand (Fig. 3 and Figs. S5 and S6). Paleoclimate reconstructions point to decadal scale wind anomalies between AD 1100 and 1300 that would have facilitated downwind sailing voyages from East Polynesia to New Zealand (22).
Conclusions
Wind/distance and shortest-hop analyses coupled with the Seascape simulations demonstrate the significance that seasonal and interannual climatic variability would have had on choices made by ancient voyagers in Remote Oceania. Changes to precipitation could have acted as motivation for migration and knowledge of how winds change with the seasons and events like ENSO would significantly increase the probability of successful trips (Fig. 4). The general El Niño pattern of drier conditions over western portions of Micronesia and Polynesia associated with winds that favor eastward displacement appears to be particularly relevant to the colonization process (Figs. 1 and 3).
Fig. 4.
Synthesis of results. Filled and dashed arrows refer to crossings that, according to simulations, are viable under downwind and directed sailing, respectively. Numbered notes provide additional information for particular voyages. Seasons (e.g., NH summer) refer to period of the year when environmental conditions facilitate crossing. Colors refer to aspects of ENSO-related anomalies that favor particular crossings (e.g., dark green is used for crossings where, under El Niño conditions, departure areas are drier and wind anomalies help displacement toward target). The open arrow (4) reflects the fact that this is a general trend and not a particular crossing. Arrows from Society Islands and Marquesas that point out of the map (7) refer to crossings toward Easter Island, which is not shown.
Earlier efforts have recognized downwind sailing as a viable voyaging/island discovery strategy in the Lapita region (23). Here we demonstrate that this is also the case for Micronesia and go on to show that over a large expanse of Oceania, despite average easterly wind conditions, downwind sailing during frequent westerly wind spells is a viable way to travel east (Fig. 2 and Fig. S3).
The identified discontinuities in both land distribution and westerly wind occurrence to the east of Samoa are in agreement with models that point to environmental factors as influential in preventing eastward movement of Lapita settlers into Central and East Polynesia for nearly 2000 y (12). The lack of successful simulated downwind crossings from Samoa into East Polynesia offer support to the suggestion that this pause in the colonization process was related to the development of technological advances that would allow more efficient windward voyaging (24), though the degree to which this was a viable strategy has been debated (17, 18).
There is considerable debate as to when and how the Mariana Islands were settled prehistorically, with some scholars suggesting that this occurred ∼3500 B.P. from the northern Philippines (1) and others arguing for a more southerly origin (25, 26). Our results point to the Solomon Islands and New Guinea as the more likely sources of voyages to the Mariana Islands, supporting the latter hypothesis. The present analysis also suggests a strong likelihood that other areas in West Micronesia were settled from the Maluku Islands or somewhere in the near vicinity and not from the northern Philippines. Yap remains an enigma archaeologically and linguistically, although it has been suggested that the original colonists may have come from the Bismarck Archipelago (27); our simulations do not discount this scenario, with northwestward movements to and near the Marianas quite possible. Still, in terms of specific islands and crossings, Palau and Samoa appear to have been central nodes for movement into Micronesia and East Polynesia, respectively. The similarity of several artifact types (Conus shell ornaments, Tridacna shell adzes) between late Lapita forms and those found in Central Micronesia, along with the settlement of Kapingamarangi and Nukuoro (Polynesian outliers) further south (13), also suggest a movement from somewhere in island Melanesia, perhaps the Solomons, as evidenced by simulated voyages that are successful, particularly between July and October (11–24%).
In terms of settlement models, our analysis suggests that the Voyaging Corridor Triple I may have some merit given the nearly contemporaneous settlement of West Micronesia and Island Melanesia/West Polynesia between ∼3400 and 3100 B.P. that would have likely required the refining of sailing skills through widespread interaction networks and increased recognition of climatic variations shared between groups. This model is also supported by human genetic diversity observed in Near Oceania (28), genetic data on rats that suggest Rattus exulans was an intrusive aspect of Lapita colonization (29), and a high degree of linguistic heterogeneity (27). The Mobile Founding Migrant models (5) that proposed a rapid movement into Remote Oceania also seems reasonable given the timing of dispersal and supporting genetic data (29), though it seems clear that the late expansion into East Polynesia was also the result of factors such as land distribution, technological advances in seafaring, and climate (12). There are indications that post-AD 1000, more favorable winds (22) and changes to long-term mean precipitation, with drier conditions in the west in the increase rainfall over East Polynesia (30) could have promoted eastward colonization. It has also been suggested that the punctuated temporal pattern of Pacific colonization could be associated to millennial variability of ENSO frequency and intensity and the changes to ease of eastward movement these generate (24).
Overall, these results provide critical comparative data for examining how Pacific Islanders structured voyaging ventures and the role that environmental variables likely played in decision-making in their efforts to settle the world’s most expansive ocean. The evidence, from a multitude of perspectives, points to the need for cultural recognition of seasonal and interannual climatic fluctuations and the development of new technologies and seafaring strategies to ensure successful voyages. The evidence also helps to target new areas of research (e.g., the Solomons and Yap) focused on determining how and when these ancient seafarers reached some of the most remote places on earth.
Acknowledgments
We thank Patrick V. Kirch for comments on the manuscript prior to submission as well as the editor and two anonymous reviewers for their corrections and comments. A.M. was partially supported by the Unesp International Visiting Scholar Grant.
Footnotes
- ↵1To whom correspondence should be addressed. Email: montenegro.8{at}osu.edu.
Author contributions: Á.M., R.T.C., and S.M.F. designed research; Á.M. and R.T.C. performed research; Á.M. and R.T.C. analyzed data; and Á.M., R.T.C., and S.M.F. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission. A.J.A. is a Guest Editor invited by the Editorial Board.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1612426113/-/DCSupplemental.
References
Simulations of prehistoric colonization of Oceania
Article Classifications
- Social Sciences
- Anthropology
- Physical Sciences
- Environmental Sciences
Sign up for Article Alerts
Jump to section
You May Also be Interested in
A study quantifies the racially disparate impacts of COVID-19 on life expectancy in the United States.
Image credit: Pixabay/PIRO4D.
A study examines links between climate change and rainfall-induced flood damage in the United States.
Image credit: Pixabay/Hans.
We must learn from the current pandemic to prepare better resource-deployment strategies, governance directives, and policy responses capable of addressing multiple crises.
Image credit: Dave Cutler (artist).
Coal mines like those shown here, as well as groundwater use and even the sheer weight of enormous cities, can cause the ground to sink in destructive ways.
Image credit: Science Source/US Geological Survey.