Apr 7, 2016 - extinction dynamics influenced by changes in island characteristics ... Maximum (LGM) island area, isolation, elevation and climate on.
Letter
doi:10.1038/nature17443
Late Quaternary climate change shapes island biodiversity Patrick Weigelt1,2, Manuel Jonas Steinbauer3,4, Juliano Sarmento Cabral1,5 & Holger Kreft1
Island biogeographical models consider islands either as geologically static with biodiversity resulting from ecologically neutral immigration–extinction dynamics1, or as geologically dynamic with biodiversity resulting from immigration–speciation– extinction dynamics influenced by changes in island characteristics over millions of years2. Present climate and spatial arrangement of islands, however, are rather exceptional compared to most of the Late Quaternary, which is characterized by recurrent cooler and drier glacial periods. These climatic oscillations over short geological timescales strongly affected sea levels3,4 and caused massive changes in island area, isolation and connectivity5, orders of magnitude faster than the geological processes of island formation, subsidence and erosion considered in island theory2,6. Consequences of these oscillations for present biodiversity remain unassessed5,7. Here we analyse the effects of present and Last Glacial Maximum (LGM) island area, isolation, elevation and climate on key components of angiosperm diversity on islands worldwide. We find that post-LGM changes in island characteristics, especially in area, have left a strong imprint on present diversity of endemic species. Specifically, the number and proportion of endemic species today is significantly higher on islands that were larger during the LGM. Native species richness, in turn, is mostly determined by present island characteristics. We conclude that an appreciation of Late Quaternary environmental change is essential to understand patterns of island endemism and its underlying evolutionary dynamics. The Quaternary is characterized by marked climatic oscillations between glacials and interglacials that had worldwide consequences for species distributions and diversity8,9. Warm interglacials, like the Holocene (past 11,500 years), can be regarded as short-term anomalies embedded in much longer, cooler and drier glacial periods3. For a Hypothesis 1
most of the Late Quaternary, massive amounts of water accumulated in continental ice sheets. This lowered sea levels considerably4 and influenced size, spatial arrangement and isolation of landmasses. In this context, the estimated 122 m sea level rise from the LGM (~21,000 years before present) to the early Holocene was particularly drastic and rapid3 (Fig. 1a). Late Quaternary sea level lowstands had far-reaching consequences for island geography (Fig. 2). Many islands were much larger and less isolated5,10. Neighbouring islands today separated by shallow waters formed single landmasses, for example, Fuerteventura and Lanzarote made up the larger Mahan11 (Fig. 2d), and currently submerged seamounts emerged and created stepping-stones for species dispersal10–12. Moreover, many islands were repeatedly connected and disconnected to the mainland13,14. Such profound changes in island geography should have strongly affected immigration, speciation and extinction, the core processes determining the assembly and diversity of island biota5,7. Present island diversity thus probably carries an imprint of past physical and bioclimatic conditions. While the effects of Quaternary climate change on biodiversity are increasingly well-recognized for mainlands8,9,15, they are under represented in island biogeography theory5,7. The equilibrium theory of island biogeography1, perhaps the most influential framework in island biology, assumes species richness to result from a dynamic equilibrium between immigration (plus speciation) and extinction. Immigration and speciation rates are higher and extinction rates lower on larger islands, while immigration rates are lower and extinction and speciation rates higher on more isolated islands1,16–18. The equilibrium theory assumes island geography not to change, and limitations of this static view have widely been recognized13,19. More recent models emphasize the dynamic nature of islands and relate biogeographical rates and biodiversity to island ontogeny-related changes in area and topographical
b Hypothesis 2
c Hypothesis 3
Variable importance
0 –20,000 years
Natives
Endemics
Species number
Sea level (m)
No. of endemics
ΔLGM ΔLGM
P
Species number
P
No. of natives
Natives
Endemics
–120 Now 0
50
100
Time (kyr before present)
150 ΔArea (LGM – P)
Figure 1 | Hypothesized effects of post-LGM sea-level changes on island biodiversity. a, We hypothesize that present (P) plant diversity has been affected by LGM island characteristics (sea-level curve in a after ref. 3). Changes in island characteristics since the LGM (ΔLGM) should therefore be important predictors of present diversity patterns. These effects should
ΔIsolation (P – LGM)
be stronger for endemic than for native non-endemic species richness. kyr, thousand years. b, c, Hypothesized effects of changes in island area (Δarea) (b) and stepping-stone isolation (Δisolation; considering also currently submerged seamounts and changes in mainland coastlines) (c) from LGM to present (P) on island plant diversity.
1
Biodiversity, Macroecology & Conservation Biogeography Group, University of Göttingen, 37077 Göttingen, Germany. 2Systemic Conservation Biology, University of Göttingen, 37073 Göttingen, Germany. 3Section for Ecoinformatics and Biodiversity, Department of Bioscience, Aarhus University, 8000 Aarhus, Denmark. 4Department of Biogeography, BayCEER, University of Bayreuth, 95440 Bayreuth, Germany. 5Synthesis Centre of the German Centre for Integrative Biodiversity Research (iDiv) Halle-Jena-Leipzig, 04103 Leipzig, Germany. 7 A p r i l 2 0 1 6 | VO L 5 3 2 | NAT U R E | 9 9
© 2016 Macmillan Publishers Limited. All rights reserved
RESEARCH Letter a
67° W
66° W
65° W
19° N Anegada
Puerto Rico 18° N
b
160° W
159° W
158° W
157° W
156° W
155° W 22° N
Ni‘ihau
Kaua‘i Maui
O‘ahu
21° N
Big Island 53° E
54° E
55° E
56° E
c
57° E 20° N
4° S
Mahé
5° S
19° N
6° S Lanzarote 29° N
d Fuerteventura
28° N
18° W
17° W
16° W
15° W
14° W
Figure 2 | Examples of spatial arrangements of present landmasses (blue) and modelled landmasses during the LGM (orange). LGM coastlines were estimated based on global bathymetry data31 assuming a sea-level decrease of 122 m at 20,000 years before present4 (see Methods). Present coastlines were taken from the GADM database of Global Administrative Areas (http://www.gadm.org/version1). a, Puerto Rico and British Virgin Islands. b, Hawaiian Islands. c, Seychelles. d, Canary Islands. Note the different changes in area, isolation and connectivity of islands. Exemplary islands mentioned in the text are named.
heterogeneity over long geological time scales2,6. Also shallow-time changes (that is, changes over short geological time scales) in island area, isolation, elevation and climate caused by Late Quaternary oscillations, have recently been proposed to have influenced immigration, speciation and extinction5, but whether this has left an imprint in global island biodiversity remains untested. Here, we evaluate present and shallow-time past drivers of species richness and endemism of angiosperms (flowering plants) on 184 islands worldwide. We hypothesize post-LGM changes in physical (area, isolation, elevation) and bioclimatic (temperature, precipitation) island characteristics to be strong predictors of present angiosperm diversity (hypothesis 1; Fig. 1a). Especially endemic species richness, which relates to evolutionary dynamics, should reflect past physical and bioclimatic conditions9. Considering that islands were larger during the LGM, islands with greater area change (Δarea) should harbour both more native and endemic species than predicted by their present size (hypothesis 2; Fig. 1b). Islands with greater changes in isolation (Δisolation) should harbour more native, but fewer endemic species (hypothesis 3; Fig. 1c), owing to higher LGM connectivity of
landmasses increasing inter-island dispersal as well as colonization from and to the mainland. We estimated LGM coastlines of islands and continents assuming a eustatic sea level drop of 122 m (ref. 4). The 184 studied islands experienced strong post-LGM changes in most physical and bioclimatic conditions owing to warming climate and resulting sea level rise (Fig. 2 and Extended Data Fig. 1). Island area decreased by 5,035 km2 and 71.1% on average (ranging from −4 to −61,377 km2; −6.3% to −99.997%). Islands that are currently small, but merged with other islands during the LGM, experienced particularly strong proportional reductions in area (for example, −99.8% for Anegada, British Virgin Islands; −99.6% for Mahé, Seychelles; Fig. 2a, c). These post-LGM changes have been much faster than average rates of geologic processes of island formation, subsidence, and erosion (Extended Data Fig. 2). Island isolation measured as distance to the nearest landmass >100,000 km2 accounting for present stepping-stone islands12 and currently submerged seamounts increased on average by 107 km (+3 to +513 km; Extended Data Fig. 1). Annual temperatures increased on average by 2 °C and annual precipitation by 163 mm changing in both directions (−2,836 to +2,684 mm). Changes in elevation were high if a focal island was connected to a high-elevation island (+327 m on average). Regression analyses showed that present island biodiversity is strongly related to post-LGM changes in island characteristics (Fig. 3), supporting hypothesis 1. Endemism measured as the number and proportion of endemic species, in particular, was strongly affected by LGM-related variables (sum of scaled Akaike’s information criterionweighted standardized regression coefficients >0.5; Fig. 3a) and could be well explained by a combination of post-LGM changes and present island characteristics (adjusted deviance explained (D2adj) all > 0.8; Extended Data Table 1). Effects of area change via sea-level change on endemism were strong, whereas direct effects of changes in tempera ture and precipitation were weak. Species richness, in contrast, was mostly related to present environmental characteristics, and post-LGM changes showed only minor effects (Fig. 3a). The strong effects of present environments on species richness might be explained by diversity dynamics closely following climate-induced changes20, suggesting that overall species richness on islands is close to equilibrium with present conditions. Island area was the variable that changed most since the LGM, and the correlation between LGM and present area was weak (r = 0.32, P native non-endemics; Fig. 3a, b). Island endemics usually result from in situ differentiation rather than
1 0 0 | NAT U R E | VO L 5 3 2 | 7 A p r i l 2 0 1 6
© 2016 Macmillan Publishers Limited. All rights reserved
Letter RESEARCH a
b
ΔLGM
Present
4
+
SIE
+
+ −
+
−
+
−
+
+
0.4
2 1
*** ***
c
−
2 3 4 ΔArea (log10[km2])
4
− 0.6
*** ***
1
+
0.2
3
0
+
+
0.0
+
+ +
PIE
+
+
+
pSIE
pPIE
+
log10[species number + 1]
N
−
log10[species number + 1]
+
S
0.8
1.0
Relative importance
3 2 1
of
**
n.s.
** 0.5
s
n
ie
tit
en
n
re
tu
tio
ta
ra
n
tio
va
le
o.
N
ΔE pi
ci
re
ΔP
pe
n
re
tio
n
tio
la
em
ΔT
a
re
so
ΔI
ΔA
io
at
ev
El
tu
ra
ita
ip
ec
Pr
n
io
pe
m
at
ol
ea
Te
Is
Ar
0
n.s.
1.0 1.5 2.0 2.5 ΔIsolation (log10[km2])
Figure 3 | Predictors of angiosperm diversity on 184 islands worldwide. a, Relative importance of present island characteristics (blue) and postLGM changes (ΔLGM; orange) as predictors of angiosperm diversity, that is, number of native species (S), native non-endemics (N), singleisland endemics (SIE), species endemic to past island units (PIE), and proportions of single-island endemics (pSIE) and species endemic to past islands (pPIE). ΔLGM variables quantify the difference in island characteristics between LGM and today. No. of entities, number of present
island entities that were part of single LGM islands; +/−, direction of important effects. b, c, Effects of post-LGM changes in island area (Δarea) (b) and stepping-stone isolation (Δisolation) (c) controlling for all covariables in generalized linear models. Relationships and 95% confidence intervals (dashed) are shown for S (black), N (blue), SIE (light green) and PIE (green). **P
