The Extent of Contemporary Loss and the Effects of Local in Spatial Networks

A dissertation submitted to the Graduate School of the University of Cincinnati in partial fulfillment of the requirements for the degree of

Doctor of Philosophy

in the Department of Biological Sciences of the College of Arts and Sciences by

Megan Lamkin M.S. Purdue University June 1998

Committee Chair: Stephen F. Matter, Ph.D.

Abstract

Since the development of conservation science nearly four decades ago, leading conservation biologists have warned that human activities are increasingly setting the stage for a loss of life so grand that the mark on the fossil record will register as a mass extinction on par with the previous “big five” mass , including that which wiped out the dinosaurs 65 million years ago. The idea that a “sixth mass extinction” was in progress motivated me to explore the extent of recent extinction and the underpinning of the widely iterated statement that current rates of extinction are 100-1,000 times greater than the background rate. In Chapter 2, I show that the estimated difference between contemporary and background extinction does not align with the number of documented extinctions from which the estimates are extrapolated. For example, the estimate that current extinction rates are 100-1,000 times higher than background corresponds with an estimated loss of 1-10 named eukaryotic species every two days. In contrast, fewer than 1,000 extinctions have been documented over the last 500 years. Given this discrepancy, it may prove politically imprudent to use extraordinarily high rates of contemporary extinction to justify conservation efforts. Conservation efforts are sufficiently justified based on the proportion of habitat that has been destroyed or degraded in recent decades and the proportion of species threatened with extinction.

In addition to examining the current extinction crisis, I evaluated potential mechanisms of extinction. Although mechanisms of population-level extinction and species-level extinction are well-resolved, little is known regarding the effects of extinction in spatial population networks.

A fascinating question that I was surprised had not been thoroughly investigated concerned potential effects of population-level extinction on surrounding of the same species:

How does the extinction of one population affect the risk of additional extinction within a

ii

network and the persistence of the network as a whole? Chapters 3, 4, and 5 document findings

from my experimental approach to evaluating the effects of population extinction and

recolonization on the abundance, synchrony, and stability of surrounding populations. In various

tests of extinction and recolonization in experimental protozoan networks, I found that

population-level extinction does not necessarily increase the risk of additional extinction beyond

the simple loss of a population. Although the loss of an important source of immigrants may

reduce the abundance of surrounding populations, for various reasons, extinction is unlikely to

increase the risk of network-wide extinction through synchronized dynamics. However,

exponential growth of a population during its recovery from extinction can generate a degree of

dispersal that bolsters surrounding populations beyond what would be expected in the absence of

extinction. Because the bolstering is temporary, these results by no means suggest extinction-

recolonization dynamics as a potential conservation strategy. Rather, they emphasize the

importance of habitat restoration and population recovery for the sustainability of population

networks and the persistence of species at risk of extinction. The results of this dissertation illustrate several mechanisms promoting resiliency in spatial population networks which may help them persist despite continuing anthropogenic onslaught.

iii

This Page Intentionally Left Blank

iv

Acknowledgments

My “Cerebral Sensei”, Dr. Stephen F. Matter, guided my academic and professional development over the last five and a half years. Just as a sensei of martial arts trains his/her students in all aspects of the combative arts, Dr. Matter trained me in all aspects of the art of

science. I am forever grateful of his mentorship, particularly for his patience while I fumbled. Dr.

Edna Kaneshiro was a strong source of academic and moral support over the years. She was

always generous with her space, equipment, expertise, and culinary delights. I will always appreciate her simultaneous toughness and warmth. Dr. Arnold Miller inspired my interest in paleontology and encouraged me to pick back up after a few failures through his steadfast belief in my ability to “do anything”. Research advisory committee members Dr. Ken Petren, Dr. Eric

Maurer, and Dr. Thomas Crist challenged me to improve my presentation skills and expand my breadth of knowledge. Mom, Dad, Simone, Diane, Aunt Pat, Eric, Lucia, Rick and Paulette: you kept me grounded along the way. Chris, I couldn’t have done it without your help and support with Simone. Family and friendship are everything. Love and gratitude.

v

Table of Contents

Abstract……………………………………………………………………. …………….. ii

Acknowledgments……………………………………………………………………….... v

Chapter 1………………………………………………………………………………….. 2 Introduction

Chapter 2…………………………………………………………………………………..10 On the challenge of comparing contemporary and deep-time biological extinction rates

Chapter 3…………………………………………………………………………………. 26 An experimental test of local population extinction in spatial population networks

Chapter 4………………………………………………………………………………...... 48 Stochastic extinction of a large population stabilizes but does not synchronize dynamics of small populations in experimental networks

Chapter 5…………………………………………………...... 77 Recolonization of a large population temporarily counters effects of long-term habitat degradation in a spatial population network

Chapter 6……………………………………………………………….…………………107 Conclusions

Supplemental Information (SI) I…….…………………………………………………..115 Supplemental Tables

Supplemental Information (SI) II…………………..…………………………………....119 Supplemental Figures

vi

Chapter 1 Introduction

As the global human population continues to grow in abundance, so too does the human capacity to influence the abundance and distribution of millions of other species on Earth (MEA

2005, Freedman 2014). Anthropogenic impacts on species’ fates (excluding impacts on fellow hominids) date back to more than 50,000 ago when colonization of Pacific islands resulted in the extinction of various large mammals and flightless birds (Diamond et al. 1989, Braje and

Erlandson 2013). Despite a global population of less than one million, hunting and trapping facilitated humankind’s first marks of anthropogenic extinction on the fossil record. Through advances in agriculture, engineering, and medicine over the recent 10,000-15,000-year Holocene epoch, the global population increased from approximately one million to more than 7.3 billion individuals (US Census 2016). During that time, humans have influenced the abundance and distribution of species worldwide (Diamond et al. 1989, Braje and Erlandson 2013). Human mutualists (pets, crops, and livestock) and synanthropes (weeds, rodents) have increased in abundance and distribution, often to the detriment of other species (Simberloff 1990), whereas many other species that have failed to adapt to anthropogenic change have decreased in abundance, often to extinction (Diamond et al. 1989, Freedman 2014, IUCN 2015). Considering the number of species that humans promote is small compared to the number of species humans negatively impact (Diamond 1997), it is not surprising that large proportions of the earth’s major taxonomic groups are threatened with extinction (IUCN 2015).

Despite the long history of anthropogenic extinction, humans have only recently recognized their impact on the natural world. For example, extinction was not accepted until the early 1800s (Sepkoski 2016). Prior to Cuvier’s incontrovertible demonstration that fossilized remains of mammoth and mastodon were those of extinct species and not elephants (Cuvier

2

1796), all fossils were presumed to belong to extant species that, if not locally present, had relocated (Sepkoski 2016). Despite widespread acceptance within Cuvier’s lifetime that extinction was a natural phenomenon, it was not accepted that extinction could be caused by humans. Rather, leading scientists including Charles Darwin believed that extinction merely offset origination and so was an inevitable component of natural law. That is, extinction was required to maintain constancy in the number of species on earth (Sepkoski 2016). In America, pioneering efforts by Theodore Roosevelt and John Muir in the late 1800s-early 1900s advanced the protection of natural resources, but did not prevent the rapid extinction of species such as the passenger pigeon (Ectopistes migratorius). The lack of scientific understanding of ecological principles made it easy to dismiss fears by ornithologists that humans were hunting the bird to extinction (Schulz et al. 2014). Advances in ecology during the 1950s and 1960s (Hutchinson

1959, MacArthur and Wilson 1963), showed the tremendous impact human actions can have on species’ fates (Wilson 1988, Barnosky et al. 2011, Sepkoski 2016). Despite recent widespread recognition among scientists and world leaders of the importance of for the economic, social, and medical well-being of earth’s burgeoning human population (MEA 2005), the conservation of the world’s ecosystem remains hindered by the politics of economic development (Butchart et al. 2010, Cazzolla Gatti 2016). As a result of the inability of nations to achieve goals pertaining to habitat protection, the anthropogenic depletion of biodiversity on

Earth goes on (Butchart et al. 2010, Dirzo et al. 2014).

Since the development of conservation science nearly four decades ago (Soulé and

Wilcox 1980), leading conservation biologists have warned that human activities are increasingly setting the stage for a loss of life so grand that the mark on the fossil record will register as a mass extinction on par with the previous “big five” mass extinctions, including the end-

3

Cretaceous event that wiped out the dinosaurs 65 million years ago (Wilson 1988, Diamond

1989, Leaky and Lawton 1995, Barnosky et al. 2011, Kolbert 2014). The idea that a “sixth mass extinction” was in progress motivated me to explore the extent of recent extinction and the underpinning of the widely iterated statement that current rates of extinction are 100-1,000 times greater than the background rate of extinction gleaned from the fossil record. In Chapter 2, “On the challenge of comparing contemporary and deep-time biological-extinction rates”, I examine the appropriateness of comparing extinction rates based on 100 or 500-year time intervals

(contemporary rates) to rates based on million-year time intervals that indicate the “background rate of extinction”, or the rate of extinction expected between mass extinction intervals (Pimm et al. 1995). Because the comparison of contemporary to deep-time extinction requires the extrapolation between short and long time intervals, the magnitude difference may exaggerate contemporary extinction (Barnosky et al. 2011). The recent proposal by Pimm et al. (2014) and

De Vos et al. (2015) to reduce the background rate for comparative purposes would increase the estimated difference between contemporary and background extinction 10-fold. A major objective in Chapter 2 is to clarify the meaning of a 10-fold increase in the difference between background and contemporary extinction.

In addition to understanding the extent of the current extinction crisis, I was also interested in understanding mechanisms of extinction. Species-level extinction is the consequence of a series of population-level events, where the extinction of the last population marks the extinction of a species. Although mechanisms of population-level extinction and species-level extinction are well-resolved, little is known regarding the mechanisms of extinction of spatial population networks. Most populations in nature are naturally or anthropogenically fragmented, and populations that occur within dispersal distance of one another comprise a

4

population network. Therefore, within a network, the growth and abundance of local populations

is affected not only by local processes (birth and death) but also by immigration and emigration

(Levins 1969, Hanski 1991). A fascinating question that I was surprised had not been thoroughly

investigated concerned potential effects of population-level extinction on surrounding

populations: How does the extinction of one population affect the risk of additional extinction

within a network and the persistence of the network as a whole?

In Chapter 3, I introduce the protozoan model system I used to experimentally test

hypotheses pertaining to the effects of extinction and recolonization in population networks.

Also in Chapter 3, I document a test of extinction in networks comprised of similarly sized

populations (homogeneous networks). Here I tested two hypotheses: (1) that extinction would

reduce the abundance of surrounding populations through the loss of immigrants (Thomas et al.

1996, Matter and Roland 2010), and (2) that populations that were reduced would have increased

synchrony in their dynamics (i.e., growth would be more positively correlated; Matter and

Roland 2009). Because decreased abundance increases extinction risk (Richter-Dyn and Goel

1972) and synchrony increases the risk of correlated extinction (Heino et al. 1997), a finding of decreased local abundance and/or increased synchrony in the dynamics of remaining populations would indicate that local population extinction increases the extinction risk for population networks beyond the simple loss of a population (Matter and Roland 2009).

Chapters 4 and 5 build on Chapter 3, and test the effects of and recolonization in increasingly perilous circumstances. In Chapter 4, the population subjected to extinction was either four times larger or eight times larger than the surrounding populations.

Here, I hypothesized that the extinction of increasingly larger populations would have greater impacts on surrounding populations and network persistence than were observed in networks

5 with populations all of the same size (Chapter 2). In these networks we found the dynamics of the small populations were destabilized by immigration from the large population. Therefore, in addition to testing the hypothesis that extinction of the large population would reduce the abundance of surrounding populations and synchronize dynamics, we also tested the hypothesis that extinction would stabilize dynamics. In Chapter 5, I “upped the ante” and created a condition of persistent, network-wide habitat degradation combined with the extinction and recolonization of a large population and tested how the combination of these factors affected surrounding populations and the network as a whole. General conclusions and a synthesis of in the results are presented in the final chapter.

References

Barnosky AD, Matzke N, Tomiya S, Wogan GO, Swartz B, Quental TB, Marshall C, McGuire

JL, Lindsey EL, Maguire KC, Mersey B (2011) Has the Earth/'s sixth mass extinction

already arrived? Nature 471:51-7 doi:10.1038/nature09678.

Braje TJ, Erlandson JM (2013) Human acceleration of animal and plant extinctions: A Late

Pleistocene, Holocene, and continuum. Anthropocene. 4:14-23

doi:10.1016/j.ancene.2013.08.003.

Butchart SH, Walpole M, Collen B, Van Strien A, Scharlemann JP, Almond RE, Baillie JE,

Bomhard B, Brown C, Bruno J, Carpenter KE (2010) Global biodiversity: indicators of

recent declines. Science. 328:1164-8 http://dx.doi.org/10.1126/science.1187512.

Cazzolla Gatti R (2016) Trends in human development and .

International J Environmental Studies. 73:268-76 doi:10.1080/00207233.2016.1148447.

6

Cuvier, G (1796) Memoir on the Species of Elephants, both Living and Fossil. English

Translation by Rudwick MJS (1998) in Georges Cuvier, Fossil Bones, and Geological

Catastrophes: New Translations and Interpretations of the Primary Texts, Univ. of

Chicago Press.

DeVos JM, Joppa LN, Gittleman JL, Stephens PR, Pimm SL (2014) Estimating the normal

background rate of species extinction. 29: 452-462.

Diamond JM, Ashmole NP, Purves PE (1989) The present, past and future of human-caused

extinctions [and discussion]. Phil Trans Royal Soc London B: Biol Sciences. 325:469-77

http://www.jstor.org/stable/2396936.

Diamond JM (1997) Guns, Germs and Steel. W.W. Norton and Company.

Dirzo R, Young HS, Galetti M, Ceballos G, Isaac NJ, Collen B (2014) Defaunation in the

Anthropocene. Science. 345:401-6 doi:10.1126/science.1251817.

Freedman B (2014) Population growth and global change. pp. 571-577 in Global Environmental

Change. Springer Netherlands. doi:10.1007/978-94-007-5784-4.

Hanski I (1991) Single-species dynamics: concepts, models, and observations.

Biol J Linn Soc 42:17-38.

Heino M, Kaitala V, Ranta E, Lindström, J (1997) Synchronous dynamics and rates of extinction

in spatially structured populations. Synchronous dynamics and rates of extinctions in

spatially structured populations. Proc Roy Soc Lond B 264:481-486. doi:

10.1098/rspb.1997.0069.

Hutchinson GE (1959) The concept of pattern in ecology. Proceedings of the Academy of

Natural Sciences of Philadelphia. 105:1-2 http://www.jstor.org/stable/4064466.

IUCN (2015) The IUCN Red List of . Version 2015-4.

7

Kolbert, E (2014) The Sixth Extinction: An Unnatural History. Bloomsbury Publishing

Leakey R and Lewin R (1995) The Sixth Extinction: Patterns of life and the future of

humankind. Doubleday Press.

Levins R (1969) Some demographic and genetic consequences of environmental heterogeneity

for biological control. Bull Entomol Soc Am 15:237-240

doi:http://dx.doi.org/10.1093/besa/15.3.237.

MacArthur RH, Wilson EO (1963) An equilibrium theory of insular zoogeography. Evolution.

1:373-87 http://www.jstor.org/stable/2407089.

Matter SF, Roland J (2009) Local extinction synchronizes population dynamics in spatial

networks. Proc Roy Soc Lond B 277: 729-737 doi: 10.1098/rspb.2009.1520.

Matter SF, Roland J (2010) Effects of experimental extinction for the spatial population

dynamics of the butterfly Parnassius smintheus. Oikos 119:1961-1969

doi: 10.1111/j.1600-0706.2010.18666.x.

MEA (2005) Ecosystems and Human Well-being: Synthesis in the Millenium Ecosystem

Assessment, Island Press.

Pimm S, Russell GH, Gittleman, JL Brooks TM (1995) The future of biodiversity. Science 269:

347-350.

Pimm SL, Jenkins CN, Abell R, Brooks TM (2014) The biodiversity of species and their rates of

extinction, distribution, and protection. Science 344.6187:1246752.

Richter-Dyn N, Goel NS (1972) On the extinction of a colonizing species. Theor Population

Biology 3:406-433. doi:10.1016/0040-5809(72)90014-7.

8

Schulz JH, Otis DL, Temple SA (2014) 100th anniversary of the passenger pigeon extinction:

Lessons for a complex and uncertain future. Wildlife Society Bulletin. 38:445-50

doi:10.1002/wsb.453.

Sepkoski D (2016) Extinction, diversity, and endangerment. In Endangerment, Biodiversity and

Culture. Eds F. Vidal, N. Dias pp 62-86.

Simberloff D (1990) Community effects of biological introductions and their implications for

restoration. Pages 128-136 in Ecological restoration of New Zealand islands. Towns DR,

Atkinson IAE, and Daugherty CH eds. Department of Conservation, Wellington.

Soule ME, Wilcox BA (1980) Conservation biology. An evolutionary-ecological perspective.

Sinauer Associates, Inc.

Thomas CD, Singer MC, Boughton DA (1996) Catastrophic extinction of population sources in a

butterfly metapopulation. Am Nat 1:957-75.

US Census Bureau (2016) US and World Population Clock http://www.census.gov/popclock.

Wilson EO (1988) The current state of biological diversity. Pages 3-18 in Biodiversity. Wilson

EO and Peter FM eds. National Academy Press.

9

Chapter 2 On the challenge of comparing contemporary and deep-time biological-extinction rates

Abstract To assess whether Earth is currently experiencing a human-induced “sixth” mass extinction, scientists over the past twenty years have compared modern rates of extinction to the widely accepted average global background rate of 1 Extinction (E)/Million Species-Years

(MSY). Application of the comparative method has led to the widely iterated estimate that contemporary global extinction rates are 100-1,000 times higher than the background rate.

Recent analyses indicate that the average background rate is closer to 0.1 E/MSY, suggesting that the difference between contemporary and background extinction is actually about ten times greater than previously thought. Here, we review the historical development and mathematical underpinning of these estimates and show that, regardless of the baseline measure, there have been fewer documented extinctions in the recent 100-500 years than the comparative measure implies. Although anthropogenic activities have reduced the abundance and distribution of countless species and have caused more species extinctions than would be expected in the absence of humans, we conclude that the most appropriate interpretation of existing data is that the global rate of contemporary extinction is closer to 100 times greater than the (revised) background rate of extinction, rather than 1000 times greater.

10

In coming to grips with the magnitude of contemporary global biological extinction, it

has been widely suggested that Earth has entered the early stages of a mass extinction rivaling

the “big five” of the geological past, during which more than half of all species went extinct over

the course of thousands or millions of years (Bambach et al. 2004). A commonly cited indicator

that a modern mass extinction is underway is the estimate that contemporary rates of global

extinction are 100-1,000 times greater than the average global background rate of extinction

gleaned from the past (Pimm et al. 1995, MEA 2005, Wagler 2007, Kolbert 2015). For these

comparisons, the background rate derives from estimated lifespans of species in the fossil record

during non-mass extinction intervals (Raup 1991). Mean lifespans range from ≈0.2-16 million

years depending on taxonomic group, which yields an average background longevity generally

expressed as ≈1-10 million years (summarized in May et al. 1995 and May 2002). For the last

two decades, the lower value, 1 million years, has been the accepted benchmark lifespan from which background extinction is inferred and against which contemporary rates of extinction are evaluated. This shorter lifespan has been favored because it translates into a higher rate of background extinction than does a longer lifespan, and therefore yields a more conservative estimate of the difference between background and contemporary extinction (May et al. 1995,

Pimm et al. 1995, 2006; Burkhead 2012). Furthermore, for practical reasons related to preservation and the state of species-level in the fossil record, direct assessments of

longevity based on first and last appearances over the sweep of the Phanerozoic eon have

generally been conducted at or above the genus level and extended for species (Raup 1991, May

et al. 1995). The million-year estimate, therefore, is more likely than a 10 million-year estimate

to reflect the lifespan of a species, the desired taxonomic level for comparing background and

contemporary extinction.

11

In separate, but related, articles, Pimm et al. (2014) and deVos et al. (2014) summarized

several lines of evidence supporting a longer lifespan (10 million years) as a more appropriate benchmark for inferring historical background rates. Both papers cited observations that genus- level lifespans for major groups (e.g., Cenozoic mammals) tend to be longer (Alroy 1996), not shorter, than a million years, with major marine groups typically having genus lifespans longer than 10 million years (e.g., cetaceans, marine carnivores, and brachiopods; Harnik et al. 2012).

In addition, they relayed Russell et al.’s (1998) suggestion that species- and genus-level background extinction should not differ appreciably because most extinctions documented in the fossil record are based on monotypic or species-poor genera, and, furthermore, that species of the same genus tend to have similar vulnerabilities to extinction. Finally, de Vos et al. (2014) presented results from molecular phylogeny analyses that showed low rates of extinction (most approaching zero extinctions per lineage per million years) and diversification (<0.2 species per species per million years) within five major taxonomic groups as additional evidence that the longer (10 million-year) rather than the shorter (million-year) lifespan is the more appropriate benchmark from which to infer background extinction. Extinction, after all, cannot have been greater than diversification, or else species richness in the fossil record would not increase over time (Rosenzweig 1995). Given the recent controversy pertaining to whether molecular phylogenies confer reasonably accurate estimates of extinction and diversification (Rabosky

2010, Beaulieu and O'Meara 2015, Rabosky 2015), it is relevant to note that de Vos et al. (2014) omitted dubious phylogenies and included a broad range of scenarios in their analysis to show that their results were not dependent on a narrow range of assumptions.

Acceptance of the longer lifespan implies that the background rate of extinction used for the aforementioned comparative purposes is 10 times lower than the previously accepted

12 benchmark, which, in turn, implies that the difference between contemporary and background extinction is ten times greater than previously thought (deVos et al. 2014). Nevertheless, the authors did not revise the benchmark range of contemporary extinction from 100-1,000 times the background rate (Pimm et al. 1995) to 1000-10,000 times background. Instead, they and various secondary authors (Alford 2014, Gutierrez 2014, Orenstein 2014) maintained that the finding suggested that contemporary extinction was 1,000 times the background rate with future rates expected to be as high as 10,000 times the background rate (deVos et al. 2014). Nevertheless, since the upper bound of the previous estimate was already 1,000 times the background rate, it is not surprising that authors in the popular press have interpreted the 10-fold increase to mean that the rate of global extinction is now thought to be 1,000-10,000 times the background rate

(Jivanda 2014, Spotts 2014). As the scientific community and general public become increasingly accustomed to the much higher value, it is worthwhile to consider the historical development and mathematical underpinning of this family of estimates. In doing so, we illustrate why decreasing the background rate of extinction does not significantly increase the contemporary rate beyond the previous upper bound of 1,000 times the background rate. In addition, we show that, regardless of the baseline measure, there have been fewer documented extinctions in the last 100 to 500 years than the comparative measure implies (Smith et al. 1993,

IUCN 2015). We fully accept that anthropogenic activities have reduced the abundance and distribution of countless species and have caused more species extinctions than would be expected in the absence of humans. However, the most appropriate interpretation of the data is that the global rate of contemporary extinction is closer to 100 times greater than the (revised) background rate of extinction, rather than 1000 times greater.

13

Calculating a background rate of extinction from the fossil record

Inferring background extinction from taxonomic longevity in the fossil record follows the

straightforward logic that if the average lifespan of a species in the fossil record is one million

years, then, in a pool of one extant species, there would be, on average, one extinction every one million years. If the size of the species pool were enlarged, say, to one million species, each of independent origin, then a rate of extinction on par with background would be 1 extinction/year

(Raup 1991). Pimm and colleagues used this logic to compare deep time and contemporary rates of extinction in terms of “species years”, equivalent to the product of the number of species and the number of years (Nott et al. 1995, Pimm et al. 1995). The background rate in terms of this novel unit, given an average species lifespan of one million years, is 1 extinction/million species years (1E/MSY). The format is convenient because it expresses rate of extinction in species and years simultaneously, so that neither the number of species nor the number of years must be constant to compare rates.

The application of “species years” to compare contemporary and deep-time extinction

The magnitude of the difference between contemporary and deep-time extinction had been considered prior to the publications by Pimm and colleagues, but the estimate was either inferred as a function of habitat loss (e.g., Wilson 1988) or was communicated in terms less clearly conveyed than E/MSY. For example, Reid (1992) estimated that 60 bird and mammal extinctions between the years 1900-1950 “greatly exceeds” the background rate of “one extinction per 100-1,000 years” for those taxonomic groups, but with the caveat that broad extrapolation of these numbers to claim an extinction crisis may not be appropriate. Rather, Reid suggested that it would be more productive to estimate extinction as a function of habitat loss

14

(i.e., species-area relationships). Early estimates of species loss derived from species-area relationships helped bring attention to the accelerating rate of tropical deforestation in the 1970s-

1980s, but the uncertainty associated with the indirect measure of species loss was widely recognized as a limitation to conferring more accurate estimates (Burgman et al. 1988, Reid

1992). Predictions of species extinctions based on species-area relationships, such as the loss of

15-20% of all species between the years 1980-2000 (Lovejoy 1979), eventually proved excessive

(Table 1).

The application of “species years” to compare contemporary and deep-time extinction

(Nott et al. 1995, Pimm et al. 1995) was an appealing methodological advance relative to previous methods used to estimate species loss because (1) the results were based on documented extinctions versus predicted extinctions and (2) the new unit allowed for direct comparison between contemporary and deep-time extinction.. Results derived from the application of the method were used to convey that recent extinction rates were 100-1,000 times higher than the background rate of 1 E/MSY (Pimm et al. 1995). This finding bolstered the notion that the degree of species loss driven by anthropogenic activities was, as previously suggested (e.g.,

Myers 1979, Wilson 1988), causing the Earth’s sixth mass extinction (e.g., Leaky and Lawton

1995, Zimmer 1996: Discover Magazine’s Top 100 Science Stories of 1995). The following example clarifies the method (1995).

In a pool of 297 North American freshwater mollusk species, 21 extinctions were

recorded during a 100-year time span (Williams et al. 1992). Therefore,

21 Extinctions Extinctions = ; = 707 E/MSY 297 Species 100 Years 1,000,000 Species Years 𝑥𝑥 𝑥𝑥 𝑋𝑋 15

Although this result indicated that the contemporary rate of extinction for these freshwater mollusks, indeed, fell within the range of 100-1,000 times greater than the background rate, measures for the remaining taxonomic groups considered by Pimm et al. (1995) were all less than 200 E/MSY. Their conclusion that contemporary extinction could range up to as high as 1,000 times greater than the background rate of 1 E/MSY was derived by averaging their estimates with those previously derived from species-area relationships (e.g., Myers 1979,

Lovejoy 1980, Wilson 1988), the latter of which conferred much higher rates of 1,000 E/MSY or more (Fig. 2 in Pimm et al. 1995). Subsequent comparisons between contemporary and deep- time extinction based on extrapolating the number of documented extinctions over 100- or 500- year time spans, however, found extinction rates within most taxonomic groups to be less than

100 E/MSY (Table 2). Higher rates are commonly encountered when the estimate is extrapolated from a small species pool covering a limited area (e.g., the freshwater mussel example above; island species). Lower rates are commonly encountered when the estimate is extrapolated from species that occupy a broad range (e.g., continental species) or from a longer time interval (e.g., the recent 500 versus 100 years; Table 2).

Despite the extrapolation method providing a more direct estimate of global extinction rates than species-area relationships, the upper bound of the decades-old estimate that extinction rates are 100-1000 times above background nevertheless was based on species-area relationships

(Pimm et al. 1995). Therefore it would have been logical for a researcher interested in communicating extinction rates based on documented extinction only to focus on the lower bound. By doing so it would follow logically that, compared to a background rate of 1 E/MSY, contemporary extinction is 100 times the background rate, but compared to a background rate of

16

0.1 E/MSY, contemporary extinction is 1,000 times the background rate. Those who suggest that

the rate of contemporary extinction is now 1,000-10,000 the background rate (Jivanda 2014,

Spotts 2014) are likely unaware that the higher bound of the estimate was based on expected

rather than documented extinction.

The extrapolation problem

Broad acceptance that contemporary extinction rates are 1,000 times higher than the

background rate (i.e., 100 E/MSY) relies on extrapolation of both species and time, and it

follows that the relative weights of the size of the species pool and the span of time influence the

overarching signficance of a 100 E/MSY. For example, a contemporary rate of extinction equal

to 100 E/MSY could be derived from 100 extinctions out of 200 species in 5,000 years or 10,000 extinctions out of 1 million species in 100 years:

100 Extinctions 10,000 Extinctions = (200 Species)(5,000 Years) (1,000,000 Species)(100 Years)

The latter might convey a stronger impression than the former of a mass extinction in progress.

Since approximately 1.9 million (mostly multicellular) contemporary species have been

named (Chapman 2009, as used in Pimm et al. 2014), it follows that, at a rate of 100 E/MSY we

should have lost 19,000 named species over the last century (i.e., 190 species per year or one

species every two days):

19,000 Extinctions 100 Extinctions = (1,900,000 Species)(100 Years) 1,000,000 Species Years

∗ 17

However, according to the IUCN (2015) there have been fewer than 1,000 documented extinctions since 1500 (834 extinct + 69 ), resulting in a global extinction rate that approximates 1 E/MSY:

1,000 Extinctions 1.05 Extinctions = (1,900,000 Species)(500 Years) 1,000,000 Species Years

∗

Considering that most of these extinctions occurred in the last 100 years, and that an additional

950 species are suspected of being extinct because they have not been detected for decades, a more reasonable estimate of global extinction would be about 10 E/MSY:

1,950 Extinctions 10.26 Extinctions = (1,900,000 Species)(100 Years) 1,000,000 Species Years

∗

Adding the approximately 20,000 species presently threatened with extinction (IUCN 2015) in the next 100 years to the 1,950 species already lost, , the estimate of global extinction still falls below 100 E/MSY:

21,950 Extinctions 57.76 Extinctions = (1,900,000 200 Years) 1,000,000 Species Years

∗ ∗

The point is not to suggest that current rates of extinction are not high, but rather to demonstrate the difficulty of extrapolating between short and long time spans and/or small and large species pools. We are not the first to recognize difficulties comparing contemporary and background

18 extinction. Barnosky et al. (2011) used the highly resolved mammalian record for the most recent

1,000 years to estimate E/MSY from documented extinctions spanning 1 to 1,000 year time bins and found that shorter intervals yielded higher estimates of mammalian extinction than longer intervals. Similarly, we show that when considering the entirety of documented species on earth, there have been fewer documented extinctions in the last 100 to 500 years than would be expected from the estimate that the average contemporary global extinction rate is 1,000 times the background rate. Although many species have become extinct without recognition, this estimate is based on documented extinction, and therefore should reasonably align with those numbers.

It has long been understood that the nature of contemporary, anthropogenic extinction is qualitatively different than what we observe for mass extinctions in the fossil record. Whereas the “big five” are known to have affected taxa with broad, often global geographic distributions, anthropogenic activities in recent time have resulted primarily in the extinction of endemic species with narrow distribution (Jablonski 2001). In the same vein, despite incontrovertible consensus that biodiversity is in a state of deepening crisis caused by human activities, it is risky to proclaim that contemporary global extinction rates are 1,000 times greater than background when the data suggest otherwise. The future of biodiversity rests on the ability of financial, political, social, environmental, and academic leaders to collectively acknowledge this crisis and then devise sweeping societal shifts that prioritize healthy ecosystems (MEA 2005). In today’s politically volatile atmosphere, the development of such a consensus is proving problematic, and it might be prudent to avoid presenting estimates that, perhaps ironically, may one day provide grist for those who seeking to undermine broad conservation efforts.

19

Acknowledgments

We thank Stephen F. Matter, Natasha Brown, David C. Sepkoski and three anonymous reviewers

for constructive comments and guidance. Dedicated in memory of David M. Raup.

References Alford J. 2014. Extinction rates ten times higher than previously thought. IFL Science (3

September, www.iflscience.com/plants-and-animals/current-extinction-rate-10-times-

worse-previously-thought).

Alroy J. 1996. Constant extinction, constrained diversification, and uncoordinated stasis in North

American mammals. Palaeogeography Paleaeoclimatology Palaeoecology 127: 285-311.

Bambach RK, Knoll AH, Wang SC. 2004. Origination, extinction, and mass depletions of

marine diversity. Paleobiology. 30:522-42.

Barnosky AD, et al. 2011. Has the earth’s sixth mass extinction already arrived? Nature 471: 51-

57.

Beaulieu J, O'Meara BC. 2015. Extinction can be estimated from moderately sized phylogenies.

Evolution 69: 1036–1043.

Burgman MA, Akcakaya HR, Loew SS. 1988. The use of extinction models for species

conservation. Biological Conservation 43:9-25.

Burkhead, NM. 2012. Extinction rates in North American freshwater fishes, 1900-2010.

BioScience 62: 798-808.

Butchart SH, Walpole M, Collen B, Van Strien A, Scharlemann JP, Almond RE, Baillie JE,

Bomhard B, Brown C, Bruno J, Carpenter KE (2010) Global biodiversity: indicators of

recent declines. Science. 328:1164-8.

20

Ceballos G, Ehrlich PR, Barnosky AD, Garcia A, Pringle RM, Palmer TM. 2015. Accelerated

modern human-induced species losses: Entering the sixth mass extinction. Science

Advances. 1.5: e1400253.

Chapman AD. 2009. Number of living species in Australia and the World, 2nd ed. Biodiversity

Information Services.

DeVos JM, Joppa LN, Gittleman JL, Stephens PR, Pimm SL. 2014. Estimating the normal

background rate of species extinction. Conservation Biology 29: 452-462.

Gutierrez, D. 2014. Humans have increased animal extinction rates by 1,000 times. Natural

News.

Haag WR, Williams JD. 2014. Biodiversity on the brink: an assessment of conservation

strategies for North American freshwater mussels. Hydrobiologia. 735: 45-60.

Harnik PG, et al. 2012. Extinctions in ancient and modern seas Trends in Ecology and Evolution

27: 608-617.

IUCN (2015) The IUCN Red List of Threatened Species. Version 2015-4. Jablonski D. 2001.

Lessons from the past: evolutionary impacts of mass extinctions. Proceedings of the

National Academy of Sciences. 98: 5393-8.

Jivanda, T. 2014. World ‘on verge of next mass extinction’: Humans have caused extinction

rates to increase by up to 10,000 times. Independent UK.

Kolbert, E. 2014. The Sixth Extinction: An Unnatural History. Bloomsbury Publishing.

Leakey R and Lewin R. 1995. The Sixth Extinction: Patterns of life and the future of humankind.

Doubleday Press.

Loehle C and Eschenbach W. 2012. Historical bird and terrestrial mammal extinction rates and

causes. Diversity and Distributions 18: 84–91.

21

Lovejoy TE. 1980. A projection of species extinctions. In Council on Environmental Quality.

The Global 2000 Report to the President, Vol 2, pp 328-331, Washington DC.

May RM, Lawton JH, Stork NE. 1995. Assessing extinction rates. Pages 1-24 in Extinction

Rates. Lawton JH and May RM, eds. Oxford University Press Inc.

May R. 2002. The future of biological diversity in a crowded world. Current Science 82: 1325-

1331.

McCallum M. 2007. Amphibian decline or extinction: Current declines dwarf background

extinction rate. Journal of Herpetology 41: 483-491.

Millenium Ecosystem Assessment (MEA). 2005. Ecosystems and Human Well-being: Synthesis.

Island Press.

Nott MP, Rogers E, Pimm S. 1995. Modern extinctions in the kilo death range. Current Biology

5: 14-17.

Orenstein, D. 2014. Extinction during human era worse than thought. News from Brown (3

September, www.news.brown.edu/articles/2014/09/extinctions).

Pimm S, Russell GH, Gittleman, JL Brooks TM. 1995. The future of biodiversity. Science 269:

347-350.

Pimm SL, Jenkins CN, Abell R, Brooks TM. 2014. The biodiversity of species and their rates of

extinction, distribution, and protection. Science 344.6187:1246752.

Rabosky DL. 2010. Extinction rates should not be estimated from molecular phylogenies.

Evolution 64: 1816–1824.

Rabosky DL. 2015. Challenges in the estimation of extinction from molecular phylogenies: A

response to Beaulieu and O'Meara. Evolution 70: 218-228.

Raup DM. 1991. A kill curve for Phanerozoic marine species. Paleobiology 17: 37-48.

22

Regan HM, Lupia R, Drinnan AN, Burgman MA. 2001. The currency and tempo of extinction.

The American Naturalist 157: 1-10.

Reid VW. 1992. How many species will there be? Pages 55-71 In Tropical Deforestation and

Species Extinction. Whitmore TC and Sayer JA, eds. Chapman and Hall.

Rosenzweig ML. 1995. Species diversity in space and time. Cambridge University Press.

Russell GJ, Brooks TM, McKinney MM, Anderson CG. 1998. Present and future taxonomic

selectivity in bird and mammal extinctions. Conservation Biology 12: 1365-1376.

Smith FD, May RM. Estimating extinction rates. Nature 364:494-6.

Spotts, P. 2014. Extinction rates higher than previously thought, but not all doom and gloom.

Christian Science Monitor.

Wagler R. 2007. The Anthropocene mass extinction: an emerging curriculum for science

teachers. The American Biology Teacher 73: 78-83.

Williams JD, Warren MI, Cummings KS, Harris JL, Neves RJ. 1992. of

freshwater mussels of the United States and Canada. Fisheries 18: 6-22.

Wilson EO. 1988. The current state of biological diversity. Pages 3-18 in Biodiversity. Wilson

EO and Peter FM eds. National Academy Press.

Zimmer, C. 1996. 1995: State of the Earth. Discover Magazine, January edition. Disney

Magazine Publishing, Inc.

23

Table 1. Early estimates of species extinction rates as a consequence of .

Estimate Reference Extinction of 1 million species by 2000 Myers 1979 Extinction of 1/5 of all species by 2000 Lovejoy 1980 Extinction of 50% of all species by 2000 Ehrlich and Ehrlich 1981 Extinction of 25-30% of all species by 2000 Myers 1983 Extinction of 20-25% of species by 2010 Norton 1986 Extinction of 27,000 species a day in the tropics Wilson 1988

24

Table 2. Estimates of extinction rates for various taxonomic groups in terms of Extinctions/Million Species-Years (E/MSY). Contemporary extinctions from 1500 through publication year unless otherwise noted. Background rate of comparison may be slightly higher or lower than 1 E/MSY. Taxonomic Group Estimate (E/MSY) Reference Vertebrates 30 Ceballos et al. 2015 Mammals 39 Ceballos et al. 2015 72 (from 1900-2014: 243) Pimm et al. 2014 82-702 (island species) Loehle and Esenbach 2012 0.89-7.4 (continental species) 36-78 Regan et al. 2001 Birds 30 Ceballos et al. 2015 49 (from 1900-2014: 132) Pimm et al. 2014 98-844 (island species) Loehle and Esenbach 2012 0.69-5.9 (continental species) 26 (from 1850-2006: ≈100) Pimm 2006 Amphibians 45 Ceballos et al. 2015 66 (from 1900-2014: 132) Pimm et al. 2014 12 McCollum 2007 Reptiles 16 Ceballos et al. 2015 Freshwater Fish of North 305 (from 1900-2010) Burkhead et al. 2012 America Freshwater Gastropods of 954 (from 1900-2010) Johnson et al. 2013 North America Angiosperms of Australia 3.6-7.1 Regan et al. 2001

25

Chapter 3 An Experimental Test of Local Population Extinction in Spatial Population Networks

Abstract

There have been few studies of the effects of population-level extinction within spatial

population networks. We developed a five-population linear network in which the single-celled caudatum showed distant-dependent dispersal. We used replicates of this network to test the hypothesis that sudden, sustained extinction would reduce the abundances of remaining populations. In addition we tested the hypothesis that the loss of immigrants to remaining populations would synchronize population growth. We monitored eight networks

(four treatment and four control) for three consecutive 10-day periods: pre-extinction, extinction, and reestablishment. During the extinction period, we imposed population-level extinction in treatment networks only. We found reduced network abundance and increased synchrony between the two populations closest to the extinction in treatment networks only. The proximity of the populations with increased synchrony to the extinction suggests a distant-dependent effect as expected from the sudden loss of immigrants. However, these populations had no corresponding reduction in abundance. Rather, the two populations most distant from the extinction accounted for the reduction in network abundance. Our results show that population- level extinction has the potential to decrease persistence of spatial population networks by causing reduced abundance and increased synchrony. However, the sudden loss of dispersers as a mechanism of synchrony was not supported.

Key words

Dispersal, extinction, metapopulation, microcosm, Paramecium caudatum, spatial population dynamics, synchrony

26

Introduction

General factors affecting population-level extinction risk are well understood and relate to population size, growth rate, variability in growth, and susceptibility to environmental stochasticity (Richter-Dyn and Goel 1972; Lande 1993; Griffen and Drake 2008). Conversely,

we know relatively little about the effects of a local population extinction within spatial

population networks. From a classical metapopulation perspective, an extinction simply opens a

habitat patch for recolonization and reduces the number of populations producing colonists

(Levins 1969; Hanski 1991). However, in spatial networks where migration affects population

size and dynamics, population-level extinction can affect both the size and dynamics of

remaining populations (Thomas et al. 1996).

Within a spatial population network, the extinction of a local population is expected to

reduce the total number of immigrants to surrounding populations. If immigrants contribute to

local population abundance, the abundance of surrounding populations will decrease following

extinction. If local abundances are reduced simultaneously by the sudden removal of immigrants,

synchronized growth could ensue. Reduced abundances of local populations increases their risk

of extinction (Richter-Dyn and Goel 1972; Lande 1993). Increased synchrony among local

populations decreases network persistence (Heino et al. 1997; Earn et al. 2000; Pearson et al.

2014). Thus, the extinction of a local population may increase the risk of extinction in spatial

population networks beyond the simple loss of a population.

Despite these potential effects of local extinction in a network context, there have been

few studies of local extinction in living networks (versus theoretical networks). Testing long- term effects of extinction in natural networks is a complicated endeavor. In a long-term, large scale field experiment, Matter and Roland (2009, 2010) showed that extinction of two

27 contiguous, relatively large local populations of Rocky Mountain Apollo butterfly (Parnassius smintheus) decreased immigration, reduced abundance, and synchronized populations within dispersal distance of the extinction. They hypothesized that the underlying mechanism of synchrony was the synchronized loss of immigrants to extant populations and showed theoretical support for this hypothesis. However, because the experiment could not be replicated at the network level, they could neither rule out the possibility of an idiosyncratic result nor dismiss alternative explanations. For example, conspecific attraction could have caused the butterflies to avoid the large, vacant patch, thereby increasing dispersal and synchrony among extant surrounding populations. Alternatively, the removal of a relatively large population could have created a spatial structure more conducive to synchronized fluctuations because the remaining populations were relatively small and therefore similar in their capacity to produce and receive dispersers (Matter 2001).

Given the limitations of field studies, we used a replicated microcosm approach to examine the effects of local population extinction within spatial population networks.

Specifically we tested two hypotheses. First, we tested the hypothesis that sudden, long-term local extinction decreases the abundance of populations within immigration distance of the extinction. Second, we tested the hypothesis that the reduction in abundance would synchronize population growth.

Methods

Focal Organism

Paramecium caudatum is a globally cosmopolitan, single-celled bacteriovorous ciliate of freshwater habitats (Wichterman 1986). Individuals can reproduce asexually by binary in

28

as few as 8 hours when resources are abundant (personal observation in our system) and are

inclined to sexually reproduce (conjugate) when resources are scarce (Mikami and Hiwatashi

2007). We reared cultures in a wheatgrass-based nutrient broth infused with a laboratory strain of

Klebsiella pneumoniae (Sonneborn 1972), a non-motile species that, among bacterial species,

excretes relatively high amounts of folate (El-Sheekh et al. 2013), a chemical attractant

paramecia use to find food (Preston and Van Houten 1986). We initiated all experimental

population networks from a single 3-L culture at stationary phase (≈400 paramecia/mL),

indicated by zero growth for 2 consecutive 24-hour periods, frequent conjugation, and infrequent binary fission (Mikami and Hiwatashi 2007).

Network Configuration

Four aspects of network configuration were important for testing our hypothesis: (1) that the physical dimensions supported distant-dependent dispersal; (2) that dispersal direction was unbiased; (3) that populations fluctuated somewhat independently for the duration of the experiment; and (4) that the number of populations allowed us to detect the spatial impact of extinction treatments relative to dispersal distance. Based on results of preliminary experiments

intended to discern the appropriate configuration according to the above criteria, we adopted a five-population linear array with extinction imposed on the center population (Fig. 1).

Each network consisted of five 50-mL Pyrex filter flasks and a 5 mm diameter “migration

corridor” made of plastic t-type connectors attached to Tygon® tubing (Saint-Gobain S.A., La

Défense, France). Flasks were filled just below the side-arm with 65 mL of well-mixed stock

culture at stationary phase, and corridors were filled with 15 mL of sterile water. The networks

were tilted forward to rest on corridors throughout the experiment to keep air bubbles from

29

developing in the corridor. Following network assembly, an additional 5 mL of culture was

added to each flask so that local population volume (70 mL) surpassed the flask’s side-arm, allowing paramecia to enter and exit the migration corridor. Flasks were loosely capped with aluminum foil so that oxygen was not limiting.

Preliminary experiments showed that dispersal of paramecia in this network configuration is distant-dependent and directionally unbiased (Fig. 2). Approximately 20% of a population at a moderate density (≈400 paramecia/mL) emigrated from its initial flask in a 24-h period, with ≈10% dispersing to a new flask in that period and the other 10% remaining in the dispersal corridor (Fig. 2A). Of the 10% dispersing to flasks within a 24-h period, the majority

(≈5-6%) dispersed into the adjacent flask, and very few (<0.01%) dispersed to the most distant flask (Fig. 2A). Because dispersal is distant-dependent with few emigrants traveling beyond a two-flask distance in a 24-hour period, extinction imposed on center populations will have the greatest effect on the two adjacent populations, and a lesser effect on the two distal populations

(Fig. 2B).

Experimental Design

We used four treatment networks and four control networks for the 30-day experiment.

The day after network assembly, we began a daily feeding regime for each population by removing 5 mL of mixed culture volume and replenishing with 5 mL fresh, bacterized medium.

We elected this replenishment volume because preliminary experiments showed that it maintained a stable carrying capacity.

Preliminary experiments also showed that growth fluctuated with greater amplitude during the first 10 days after network assembly compared to the following 20 days. Therefore,

30

following network assembly we informally spot-checked densities daily for two weeks, but

waited two weeks to start the experiment. Density estimates were conducted daily on all

populations for the next 31 consecutive days in order to calculate growth over thirty 24-hour

periods. We extrapolated density (paramecia/mL) and abundance from the mean number of

paramecia captured in three 0.25 mL samples. We used an automated pipette dispenser

(Drummond Scientific Co., Broomall, PA) to sample each population. We vigorously mixed the

population with the pipette prior to withdrawing each sample. We distributed each 0.25 ml

sample into several small drops over a 100 mm Pyrex plate and tallied the total number of

paramecia under 7.5X magnification using a Leica stereoscope (Weitzlar, Germany). Mixing

prior to sampling did not cause the flask contents to mix with corridor contents based on tests

conducted using a tracer dye.

Our experiment consisted of three 10-day periods: pre-extinction, extinction, and reestablishment. After density estimates and small volume removals on the last day of the pre-

extinction period, center flasks were detached from networks with extinction and control

networks. Center populations from treatment networks only were then microwaved to near

boiling (95 °C) and quickly cooled in a cold water bath, effectively killing paramecia and . When the liquid reached room temperature (23 °C), the flasks of all networks were

reconnected to their former position within their respective networks and replenished with 5

mL of fresh bacterized medium. Individuals detected within the center flask during the

extinction period were killed using the same heating procedure. Following the 10-day

extinction period, paramecia were allowed to reestablish the center population in treatment

networks. There was no extinction imposed on any population in control networks. Except

during data collection all networks were maintained at 23 ◦C under 16:8 h light:dark cycles in

31 a temperature controlled culture room. Data were collected under normal laboratory conditions.

Statistical Analysis

Population growth was calculated as Rt = ln , where N = population abundance Nt+1 � Nt � (Royama 1992). We used Pearson’s product moment correlation (r) of population growth as our measure of synchrony, and calculated network synchrony as the mean of the six pairwise correlations among the four populations surrounding the center population (Bjørnstad et al.

1999). Correlations were z-transformed to meet distributional requirements (Zar 1999). We used paired t-tests to evaluate changes in abundance and synchrony between intervals within treatment and control networks. We compared differences in synchrony and abundance between treatment and control networks using independent t-tests. We used a one-tailed test when our prediction was directionally specific (e.g., that abundance would be lower and synchrony higher during extinction than during pre-extinction); otherwise we used 2-tailed tests.

All data collected for statistical evaluation is available at the UCScholar Data Repository

(scholar.uc.edu, doi:10.7945/C2J01Q).

Results

Abundance

Abundance in networks with extinction was slightly lower during the extinction period than it was during pre-extinction (t = 2.37, df = 3, P < 0.05). Abundance was higher in control networks during extinction compared to the pre-extinction period (t= - 5.28, df = 3, P = 0.01).

Correspondingly, network abundance was lower in networks experiencing local population

32 extinction than it was in control networks (t = 4.21, df = 6, P < 0.01). Examining networks experiencing extinction at the population level showed that the distal populations had reduced abundance during the 10-day extinction period (t = 3.63, df = 3, P = 0.02; Fig. 3A), but populations closest to extinction had the same mean abundance they had during the pre- extinction period (t = 0.04; df = 3; P = 0.48; Fig. 3B).

During the reestablishment period, abundance of treatment networks increased relative to the extinction period (t = -7.43, df = 3, P < 0.01), and was somewhat higher than the abundance in control networks (t = -2.30, df = 6, P = 0.06). In control networks, abundance did not change between extinction and reestablishment intervals (t = -0.25, df = 3, P = 0.83).

Network Synchrony

Populations tended to fluctuate with some degree of synchrony for the duration of the experiment, (Figs. 4-5). Contrary to our expectation, synchrony in treatment networks was not higher during extinction compared to the pre-extinction period (t = 0.25, df = 3, P = 0.41; Fig.

5A). Synchrony was variable among networks, such that there was no network-wide change in synchrony through time (Fig. 5A) or any difference in network synchrony between treatment and control networks during pre-extinction (t=0.40, df = 6, P = 0.70), extinction (t = 0.09, df = 6, P =

0.93), or reestablishment (t = -0.30 , df = 6, P = 0.77) intervals (Figs. 4, 5A).

Pairwise Synchrony

The two populations closest to the extinct population had higher synchrony during the extinction period (r = 0.71 ± 0.02, N=4 pairs) compared to the pre-extinction period (r = 0.42 ±

0.15, N=4 pairs; Fig. 5B). Examination at a finer temporal scale revealed a trend for higher

33 synchrony the first five days (r = 0.78 ± 0.03) compared to the second five days (r = 0.67 ± 0.05) of extinction (t = 2.0, df = 3, P=0.07). Owing in part to the decay, the increase in synchrony was not significant over the entire 10-day scale of comparison (t = -2.32, df = 3, P = 0.05), but was significant the first five days (t = -5.49, df = 3, P < 0.01). Also during extinction, the populations closest to the extinction were more synchronized in treatment than control networks (t = 2.25, df

= 6, P = 0.03; Fig. 5B). Except for the population pair closest to the extinct population, no other pair in treatment networks had increased synchrony associated with extinction (Fig. 6). In control networks, there were no statistical changes in synchrony for any pair between pre-extinction and extinction intervals (Fig. 6).

During reestablishment, the pair closest to extinction was less synchronized than it was during extinction (t = 6.58, df = 3, P < 0.01; Fig. 5B). No other population pair in treatment networks had a change in synchrony between extinction and reestablishment periods (Fig. 6).

During the reestablishment period in control networks, no statistical changes in synchrony occurred for any population pair relative to the previous 10-day (extinction) period (Fig. 6).

Discussion

Here we document the first replicated, experimental test of the effects of a local population extinction in spatial population networks. Our finding of reduced abundance within networks during extinction demonstrates that even in networks comprised of similarly sized populations, local extinction can measurably reduce abundances of remaining populations. Our finding of increased synchrony between the population pair closest to extinction supports the hypothesis that the extinction of a local population can increase synchrony in population networks (Matter and Roland 2009). Support for the sudden loss of dispersers as the direct

34 mechanism of synchrony, however, is lacking. The proximity of the populations with increased synchrony to the extinction suggests a distant-dependent effect as would be expected from the sudden loss of immigrants; however there was not a corresponding reduction in abundance for these populations (Fig. 4A).

We cannot observe migration directly in our system, and so cannot explain with certainty why the pair closest to extinction had no measurable decrease in abundance during extinction.

The result is especially surprising given that during the 10-day extinction period, an average of

7.7 ± 0.78% (N=4 networks) of the previous day’s network abundance dispersed into the vacant flask. Despite daily extirpation of these individuals, treatment networks lost only 4.4 ± 0.02%

(N=4) of network abundance over the 10-day interval. Therefore, it is possible that abundance reductions occurred, but due to the high reproductive capacity of single-celled organisms, the reductions could not be detected at 24-hour sampling intervals.

Although the above speculation may explain the lack of detectable abundance reductions for the populations closest to extinction, it fails to explain why the populations with measureable abundance reductions did not become synchronized. Therefore, taken together, our results call into question whether the sudden loss of dispersers was the direct mechanism of synchrony in this experiment. There are two alternate mechanisms by which local extinction could synchronize remaining populations in a network context: (1) increased dispersal due to avoidance of the vacant patch (Matter and Roland 2009); and (2) increased homogeneity in the capacity of populations to produce and receive dispersers (Matter 2001). Neither mechanism can account for the increased synchrony observed in our experiment. First, the number of individuals detected in the center flask during extinction dismisses the possibility of patch avoidance.

Second, we conducted the experiment in networks comprised of populations that were

35 homogeneous in their ability to produce and receive dispersers; thereby eliminating the potential for extinction to cause increased homogeneity in dispersal dynamics by our experimental design.

So why did the populations closest to extinction become synchronized but maintain consistent abundance, whereas populations farthest from extinction had reduced abundance but did not become synchronized? We propose that in addition to decreasing the number of immigrants to extant populations, the extinction opened an “ecological trap” within the network insofar as the unoccupied patch was attractive but fatal to migrants (Dwernychuk and Boag

1972). The populations closest to the extinction likely provided a constant flow of immigrants into a patch unable to produce emigrants. Considering this loss, it is feasible that abundance reductions in center-adjacent populations were strong enough to synchronize growth in the hours following extinction, but because reproduce rapidly, the populations had rebounded by the time of the next sampling interval. Meanwhile, distal populations continued to exchange dispersers with center-adjacent populations. Here it is relevant to note that emigration from distal populations is unidirectional with most individuals dispersing into the populations adjacent to extinction, and few dispersing farther (Figs. 1, 2). Emigration from populations adjacent to extinction, however, is bidirectional with ≈half dispersing toward the distal flask and the other

≈half dispersing toward the center (Figs. 1, 2). Therefore, during extinction the disparity between emigration and immigration in distal flasks was exaggerated by the loss of immigrants from the center population. Although the disparity between emigration and immigration in distal populations caused reduced abundance over time, the loss was not strong enough to synchronize growth.

Sudden, long-term population-level extinction could occur by anthropogenic or natural means. In the event that habitat formerly supporting a thriving population was rendered

36

unsuitable (e.g., by exploitation or contamination), individuals migrating into the patch could

succumb to patch conditions, as they did during our experimental extinctions and during those of

Matter and Roland (2009, 2010). Sudden extinction with unencumbered rapid recovery,

however, would likely produce different results from those we observed. For example, Matter

(2013) demonstrated theoretically that for species with high intrinsic rates of increase, sudden

local extinction with rapid recovery was unlikely to synchronize network dynamics. Species with

low intrinsic rates of increase, however, were slow to recover the vacated space and as a

consequence remained synchronized in the long-term (Matter 2013). The rapid recovery from extinction during the first two days of the reestablishment period in our experiment (Fig. 3) and coincident de-synchronization of the nearest population pair (Fig. 5B) aligns with Matter’s

(2013) theoretical result.

Just as recovery could occur slowly or rapidly, so too could extinction. A population subjected to gradual depletion in advance of extinction would likely cause different effects than those we observed with rapid extinction. Had we simulated gradual extinction of the center flask habitat, say by warming or increasing acidity, it is reasonable to speculate that the flood of emigrants from the deteriorating habitat could have increased immigration to extant populations, perhaps synchronizing local fluctuations.

We have demonstrated that local extinction can have regional effects within a spatial

population network. Our findings align with previous findings of decreased abundance (Thomas

et al. 1996; Matter and Roland 2010) and increased synchrony (Matter and Roland 2009) during

sudden, sustained population-level extinction. Similar to the configuration of our experimental networks, Matter and Roland’s (2009, 2010) Parnassius network approximated a linear arrangement with extinction imposed on the centrally positioned population. The experiments

37 differed, however, in that the butterfly population (actually two contiguous populations) Matter and Roland (2009, 2010) subjected to extinction was several times larger than extant populations whereas the populations in our networks were equal in terms of habitat volume and carrying capacity. Relevant to that distinction in design is the primary distinction in our results: in the

Matter and Roland (2009) experiment, four remaining population pairs became highly synchronized during extinction treatments, whereas only the most proximate population pair became synchronized during our extinction treatments. While it does stand to reason that sudden extinction of a larger population would have stronger effects than sudden extinction of a smaller population, our results indicate that even within a network of similarly sized populations, sudden local extinction can have measurable consequences on remaining populations. If the population we subjected to extinction had been larger in size or had a greater dispersal rate than extant populations, local extinction may have synchronized the entire network.

Our results show that local population extinction can affect remaining populations in such a way that increases the risk of additional extinction and compromises network persistence.

Therefore, from a management perspective, it may be important to emphasize population-level protection or minimize the duration of population-level extinction in order to maximize network persistence. Given that so few studies have examined the effects of extinction in spatial population networks, more work is needed to understand the conditions and extent to which local extinction is risky to population networks.

Acknowledgments

We thank Elliott King, Ian Thrush and Amanda Eisenhart for assisting with daily counts and

Thomas Doak (Indiana University) for providing the cultures. In addition we wish to express

38

gratitude to the University of Cincinnati Department of Biological Sciences Weiman, Wendel,

Benedict fund for financial support.

References Bjørnstad O, Ims RA, Xavier L (1999) Spatial population dynamics: analysing patterns and

processes of population synchrony. Trends Ecol Evol 14:427–432. doi:10.1016/S0169-

5347(99)01677-8

Dwernychuk LW, Boag DA (1972) Ducks nesting in association with gulls-An ecological trap?

Can J of Zool 50:559-563. doi:10.1139/z72-076

Earn DJ, Levin SA, Rohani P (2000) Coherence and conservation. Science 290: 1360-1364.

doi:10.1126/science.290.5495.1360

El-Sheekh MM, Abd-Elsalam IS, Shabana S, Zaki A (2013) Production of vitamin B12 and folic

acid from agricultural wastes using new bacterial isolates. Afr J Microbiol Res 7:966-973

Griffen BP, Drake JM (2008) A review of extinction in experimental populations. J Animal

Ecology 77:1274-1287. doi:10.1111/j.1365-2656.2008.01426.x

Hanski I (1991) Single-species metapopulation dynamics: concepts, models, and observations.

Biol J Linn Soc 42:17-38

Heino M, Kaitala V, Ranta E, Lindström, J (1997) Synchronous dynamics and rates of extinction

in spatially structured populations. Synchronous dynamics and rates of extinctions in

spatially structured populations. Proc Royal Soc Lond B 264:481-486. doi:

10.1098/rspb.1997.0069

Lande R (1993) Risks of population extinction from demographic and environmental

stochasticity and random catastrophes. Am Nat 142:911-927 doi:10.1093/besa/15.3.237

39

Levins R (1969) Some demographic and genetic consequences of environmental heterogeneity

for biological control. Bull Entomological Soc Am 15:237-240

doi:10.1093/besa/15.3.237

Matter SF (2001) Synchrony, extinction, and dynamics of spatially segregated heterogeneous

populations. Ecol Modelling 141: 217-226 doi:10.1016/S0304-3800(01)00275-7

Matter SF (2013) Increases and decreases in network synchrony associated with local population

extinction and recolonization. Ecol Modelling 226:68-72

doi:10.1016/j.ecolmodel.2013.06.021

Matter SF, Roland J (2009) Local extinction synchronizes population dynamics in spatial

networks. Proc Royal Soc Lond B 277: 729-737. doi: 10.1098/rspb.2009.1520

Matter SF, Roland J (2010) Effects of experimental extinction for the spatial population

dynamics of the butterfly Parnassius smintheus. Oikos 119:1961-1969.

doi: 10.1111/j.1600-0706.2010.18666.x

Mikami K, Hiwatashi K (1975) Macronuclear regeneration and cell division in Paramecium

caudatum. J Protozool 22: 536-540. doi: 10.1111/j.1550-7408.1975.tb05227.x

Pearson RG, Stanton JC, Shoemaker KT, Aiello-Lammens ME, Ersts PJ, Horning N, Fordham

DA, Raxworthy CJ, Ryu HY, McNees J,Akçakaya HR (2014) Life history and spatial

traits predict extinction risk due to climate change. Nat Clim Change 4:217-221.

doi:10.1038/nclimate2113

Preston RR, Van Houten JL (1986) Chemoreception in Paramecium tetraurelia: acetate and

folate-induced membrane hyperpolarization. J Comp Phys 160: 525-535

Richter-Dyn N, Goel NS (1972) On the extinction of a colonizing species. Theor Popul Biol

3:406-433 doi:10.1016/0040-5809(72)90014-7

40

Royama T (1992) Analytical Population Dynamics, 1st ed. Chapman and Hall, London.

doi:10.1007/978-94-011-2916-9

Sonneborn TM (1970) Methods in Paramecium Research In: Prescott DM (ed) Methods in Cell

Phys Vol.4. Academic Press Inc, New York, pp. 241-399

Wichterman R (1986) The Biology of Paramecium, 2nd edn. Plenum Press, New York doi:

10.1007/978-1-4757-0372-6

Zar JH (1999) Biostatistical Analysis, 4th edn. Prentice Hall, Upper Saddle River NJ

41

Figure Captions

Figure 1. Diagram of the protozoan population network used to test the hypothesis that sudden,

long-term local extinction increases synchrony of extant populations in population network. In

this experiment populations in center flasks were subjected to extinction treatments.

Figure 2. Proportion (± SE; N=3) of a local population that dispersed from a distal (A) or center

(B) flask into other network locations in a 24-h period. Results indicate that dispersal in this

system is distant-dependent (A) and directionally unbiased (B).

Figure 3. Daily density estimates (Paramecia/mL) for local populations in treatment (left panel)

and control (right panel) networks for the 31-day experiment (30 growth periods). Mean

networks synchrony (r) provided for each time period. Transparent grey box over each panel

indicates the extinction period. Fluctuations are shown for center (grey lines), center-adjacent

(red lines), and distal (black lines) populations. Circle diagrams above panels denote positions of

populations by color.

Figure 4. Mean densities (Paramecia/mL ± SE; N=4) of center-adjacent (A) and distal (B) populations for the three consecutive 10-day periods in treatment and control networks. An asterisk (*) above a bar indicates mean density is significantly different from the previous period according to a paired t-test (P<0.05). Grey bars represent the pre-extinction and reestablishment periods; red bars represent the extinction period. There were no extinction treatments in control networks.

Figure 5. Mean network synchrony (A: r, ±SE; N=4) and pairwise synchrony for the populations closest to the extinction (B: r, ±SE; N=4) during consecutive 10-day periods. Grey bars indicate

42 pre-extinction and reestablishment periods; red bars show the extinction period. Center populations excluded from analysis. There were no extinction treatments in control networks.

Circle diagrams denote positions of populations included in analysis.

Figure 6. Mean synchrony (r, ±SE; N=4) for each population pair during consecutive 10-day periods. Grey bars indicate pre-extinction and reestablishment periods; red bars represent the extinction period. There were no extinction treatments in control networks. Circle diagram denotes positions of populations included in analysis.

43

Figures

Figure 1.

1.0 A B

0.8

0.6

Proportion 0.4

0.2

0.0 Corridor 0 1 2 3 4 2 Left 1 Left 0 1 Right 2 Right

Dispersal Distance Dispersal Distance Figure 2.

44

r= 0.48 r= 0.45 r= 0.24 r= 0.14 r= 0.25 r= 0.13

300

200

100

0 r= 0.07 r= 0.23 r= 0.16 r= 0.46 r= 0.29 r= 0.50

300

200

100

0 r= 0.49 r= 0.29 r= -0.04 r= 0.37 r= 0.64 r= 0.23

300 Density(Paramecia/mL) 200

100

0 r= 0.50 r= -0.01 r= 0.25 r= 0.46 r= 0.29 r= 0.24

300

200

100

0 0 5 10 15 20 25 30 0 5 10 15 20 25 30 Day Day Networks with Extinction Control Networks

Figure 3.

45

400 A 2 4 B 1 5 Pre-extinction Extinction Pre-extinction Extinction 300 Reestablishment Reestablishment * * * *

200

100 Density(Paramecia/mL)

0 Treatment Control Treatment Control

Figure 4.

1.0 A 1 2 4 5 B 2 4 0.8 Pre-extinction Extinction

) Reestablishment r 0.6

0.4 Synchrony ( 0.2

0.0

Treatment Control Treatment Control

Figure 5.

46

1.0 1 2 4 5 Pre-extinction Extinction Reestablishment 1 2 4 5 ) r 0.8

0.6

0.4

0.2 PairwiseSynchrony (

0.0

1&2 1&4 1&5 2&4 2&5 4&5 1&2 1&4 1&5 2&4 2&5 4&5 Networks with Extinction Control Networks

Figure 6.

47

Chapter 4 Stochastic extinction of a large population stabilizes but does not synchronize dynamics of small populations in experimental networks

Abstract

Many species exist in spatial population networks, where dispersal influences the dynamics of local populations and the network. Within heterogeneous networks, large populations produce relatively many dispersers and therefore have a greater capacity than small populations to influence population dynamics. Because large populations wield large influence within spatial population networks, extinction of a large population may profoundly change the dynamics and persistence of remaining populations. We used experimental protozoan networks to test the hypothesis that a sudden, sustained extinction of a large population would reduce abundance and increase the synchrony and stability in the dynamics of remaining smaller populations. In one treatment, the density of the large population was similar to that of small populations; in a second treatment, it was lower. In neither did extinction reduce abundance or synchronize growth, but for both treatments extinction stabilized the dynamics of remaining populations. Regardless of the large population’s density, the focal organism’s preference for small patches over the large one culminated in weak interactions between large and small populations. Therefore, the immediate loss of dispersers did not reduce abundance or synchronize dynamics. However, the continuous loss of dispersers from the sustained extinction stabilized local dynamics. These results emphasize the importance of understanding population interactions in order to predict the effects of extinction within spatial population networks.

Key words

Spatial population networks, dispersal, extinction, microcosm, Paramecium, synchrony, stability

48

Introduction

Many species exist in spatial population networks, where dispersal influences the

dynamics of local populations and the dynamics of the network (Hanski 1991, Liebhold et al.

2004). In natural settings, networks are comprised of populations that vary in their mean

abundance due to factors such as spatial heterogeneity in the size and quality of habitat patches

(Harrison 1991, Fronhofer et al. 2012). A landscape fragmented into a large and small patches

that are qualitatively similar support “mainland-island” or “patchy” population networks, the former of which have true spatial structure and latter of which are difficult to distinguish from a single population due to a high degree of mixing among subpopulations (Harrison 1991,

Fronhofer et al. 2012). In mainland-island networks, large “mainland” populations are important for network persistence because, compared to small “island” populations, they produce relatively many dispersers (Gyllenberg and Hanski 1997) and are relatively resistant to disturbance-driven

and demographic extinction (Richter-Dyn and Goel 1972). Therefore, although dispersers from

any population can colonize new habitat, reestablish extinct populations, and/or bolster the

abundances of surrounding populations, large populations have a greater, more persistent

capacity to do so than small populations (Thomas et al. 1996, Öckinger and Smith 2007).

As a persistent source of relatively many dispersers, large populations not only bolster the

average abundance of surrounding populations; they also influence the stability and synchrony

(positive correlation) of their dynamics (Matter 2001, Matter and Roland 2009, 2010; DeRoissart

et al. 2015, Wang et al. 2015). For example, a large number of immigrants from a large

population can destabilize the abundance of a small population by forcing its abundance beyond

the carrying capacity of the local patch (DeRoissart et al. 2015, Wang et al. 2015). After reaching

peak abundance, the population would decrease to (or below) an abundance the habitat could

49

support. By generating higher amplitude fluctuations in the small population than would occur

by local processes, immigrants from the large population increase the small population’s

vulnerability to disturbance-driven or demographic extinction (Lande 1993). On a regional scale,

spatial uniformity in the distribution of dispersers from a large population into surrounding

populations would cause synchrony (positively correlated growth) in their dynamics and

therefore increase the risk of a synchronized crash (Higgins 2009). For example, the equal

proportioning of immigrants into surrounding populations would cause the populations to grow

similarly. Conversely, spatial non-uniformity in the distribution of dispersers from the large

population would cause asynchrony (uncorrelated growth) in local dynamics and decrease the

risk of a synchronized crash (Higgins 2009). Because the number of immigrants an outlying

population receives is affected by factors such as distance from the source and the dispersers’

perception of habitat quality, the perfectly uniform distribution of dispersers from a large

population is improbable. Therefore, although spatially correlated immigration could generate statistical synchrony in surrounding populations (where the mean correlation of pairwise growth, r, is greater than zero), some degree of spatial non-uniformity in the distribution of dispersers is expected to dampen the effect of dispersal on synchrony.

Given the potential for large populations to wield large influence within heterogeneous networks, the extinction of a large population has the potential to profoundly change the dynamics and persistence of the network. First, the immediate loss of dispersers would reduce immigration into surrounding populations (Thomas et al. 2006, Matter and Roland 2010). A loss of immigrants that was strong enough to force some or all of the surrounding populations synchronously below carrying capacity would synchronize growth as long as the populations were similarly density-dependent (Moran 1953, Matter and Roland 2009). Following a

50 synchronized reduction, populations would continue to fluctuate synchronously short- or long- term depending on the relative frequency and strength of subsequent local (desynchronizing) and regional (synchronizing) environmental fluctuations (Moran 1953). Furthermore, populations that were slow to recover their abundances in the absence of the large population would be reduced long-term (Thomas et al. 1996). In the event that the habitat supporting the large population was rendered unsuitable for reestablishment, populations could be further reduced through increased migration mortality associated with longer dispersal distance (Matter 2006) and/or immigration into a patch that could not sustain growth, as in an ecological trap

(Dwernychuk and Boag 1972, Robertson and Hutto 2006). Finally, in the absence of destabilizing immigrants, the dynamics of remaining populations would stabilize (Ruxton et al.

1997), an effect that would buffer populations from the increased risk of extinction associated with decreased abundance and increased synchrony (Lande 1993).

To date there has been no test of the effects of extinction of a large population in controlled, replicated living networks. Improved understanding of how extinction impacts dynamics within spatial population networks is important for improved understanding of spatial population dynamics and persistence Here, we use experimental protozoan networks to test the hypotheses that extinction of a large population would reduce local abundance, increase synchrony and increase stability in the dynamics of surrounding populations that were similarly small.

Methods

Focal Organism

51

Paramecium caudatum is a free-swimming single-celled ciliate common in freshwater habitats (Wichterman 1986). Our cultures were obtained from Indiana University and maintained in wheatgrass-based broth infused with a laboratory strain of the bacteria species Klebsiella pneumoniae (Sonneborn 1972). We initiated all experimental networks from a single stock culture near stationary phase (≈200 paramecia/mL). We maintained a constant environment by replenishing 5 mL per 70 mL of population volume with fresh, bacterized medium each day

(Lamkin et al. under review).

Experimental Design

Each network consisted of 5 filter flasks aligned linearly and connected to a 5 mm diameter “migration corridor” made of plastic tubing (SI Fig. 1). As a preliminary step we examined dispersal from smaller and larger populations to ensure that more dispersers were produced by increasingly large populations (Fig. 1). Comparing dispersal over a 24-hour period from a 70 mL, 140 mL (2X), or 560 mL (8X) population, we found that increasing the size of a dispersing population increased the number of immigrants into neighboring flasks that were vacant prior to the dispersal event (Fig. 1A-B). In addition, we found that although dispersal was

distant-dependent in homogeneous networks (i.e., five 70 mL populations), the effect of distance dampened as the size of the dispersing population increased relative to neighboring populations

(Fig. 1). From a relatively large population, 24-hour dispersal was global, but not necessarily

spatially uniform (Fig. 1). Lastly, these experiments showed that after the 24-hour dispersal

period, the 560 mL (8XL) population had reduced density compared to the density of a smaller

start population (Fig. 1B).

52

As a second preliminary step, we monitored networks in which the center flask contained a culture volume that was either four (4XL networks) or eight times larger (8XL networks) than the culture volume of non-center flasks. Although the first preliminary experiment showed the large population did not disperse individuals uniformly over a 24-hour period, it was important for our test of extinction on synchrony that networks did not become highly synchronized by dispersal alone. Results from this second preliminary experiment showed growth in 4XL and

8XL networks remained uncorrelated over a 10-day observation period (SI Fig. 2). In addition, results showed that in 8XL networks but not 4XL networks, the large population maintained a lower density than small populations (SI Fig. 3). Because larger populations in nature often occur at lower density than smaller populations (Bowman et al. 2002), we thought it would be valuable to test the effects of extinction for both 4XL and 8XL networks.

To assemble networks for the experiments, we filled flasks to just below the sidearm from a single stock culture at stationary phase. We filled the corridors with 15 mL sterile water and connected each sidearm of the corridor to the sidearms of the five flasks. After the corridor was connected to the flasks, additional culture was added to each population so that the volume surpassed the sidearm and paramecia could enter the corridor. For example, each small population occupied a 70 mL “habitat patch” in a 50 mL flask. The 4XL populations occupied a

280 mL habitat patch in a 250 mL flask, and the 8XL populations occupied a 560 mL habitat in a

500 mL flask. Smaller flasks sat on blocks so that the side arms and corridor were level and flask size did not affect dispersal distance. Finally, all flasks were loosely capped with aluminum foil so that oxygen was not limiting.

We began the experiment two weeks after network assembly to allow populations to stabilize in the network configuration (Lamkin et al. under review). The experiment consisted

53

of three consecutive 10-day periods: pre-extinction, extinction, and reestablishment. During

the extinction period we caused a 10-day extinction of the large population in half of the 4XL

networks and half of the 8XL networks. Networks without extinction served as controls. To

create an extinction, the large population was separated from the network, microwaved to near

boiling (≈95 °C), and quickly cooled in a cold water bath. When the liquid reached room

temperature (≈23 °C), all flasks were reconnected to their respective networks and replenished

with 5 mL of fresh bacterized medium per 70 mL culture volume. Individuals were free to

migrate into the center flask habitat during the extinction period, but were extirpated daily.

Therefore, the extinction and subsequent removals functioned as an ecological trap

(Dwernychuk and Boag 1972, Robertson and Hutto 2006). Following the 10-day extinction

period, large populations were allowed to reestablish. Networks were maintained in normal

laboratory conditions throughout the experiment (≈23 °C, 10:14 hours, light:dark).

Abundance

Daily abundance estimates were extrapolated from the mean density of paramecia

captured in three 0.25 mL samples withdrawn from a thoroughly mixed population. Each sample

was distributed into multiple drops over a Petri dish, and individuals were tallied under 7x

magnification. We accepted the mean density of the three samples when the variance to mean

ratio was less than 20%, i.e., when the coefficient of variation, = 100, was less than 20, s 𝐶𝐶𝐶𝐶 �𝜇𝜇� ∗ where s is the standard deviation and μ is the mean density. If the coefficient of variation was greater than 20%, we resampled the population and used the combined mean of the six samples as our estimate. The CV of our mean estimate was uncommonly greater than 20%, representing only 9% of estimates. Note that in our figures we present density rather than abundance for ease

54 of comparison between large and small populations, but unless stated otherwise comparisons were based on abundance.

Synchrony

We calculated growth (Rt) for each population according to Rt = ln , where N = Nt+1 � Nt � population abundance (Royama 1992). We used Pearson’s product moment correlation (r) of population growth between pairs of populations as our measure of pairwise synchrony, and calculated network synchrony as the mean of the six pairwise correlations (Bjørnstad et al.

1999). Prior to statistical analysis, all correlation coefficients were z-transformed to meet distributional assumptions (Zar 1999).

Stability

Since stability is the inverse of variability, we used the coefficient of variation (CV) in population abundance over time as our measure of instability, where, as stated previously (see

Abundance). Because CV measures instability, lower values indicate more stable dynamics

(Abbott 2011, Wang et al. 2015). We calculated the stability of small populations using the daily abundance estimates of all small populations within a network over a given timespan, i.e., 40 abundance measures for four populations over a 10-day period. We calculated network stability using the summed abundance of small populations over time. Prior to statistical analysis, coefficients were arcsine square-root transformed to meet distributional assumptions (Zar 1999).

Statistical Analysis

55

Within treatment and control networks, we tested for differences in mean abundance, stability, and synchrony between consecutive 10-day periods using paired t-tests. In addition, we evaluated short-term changes that may have been masked by the longer observation period. Since measures of abundance, synchrony and stability during each of the consecutive time periods would be affected by antecedent conditions, we did not consider differences between networks with extinction and control networks within each time period. For example, increased synchrony associated with extinction would not necessarily increase the mean measure above the mean measure of synchrony in control networks. However, if synchrony increased in networks with extinction and control networks we could not attribute the change to the treatment of extinction.

We considered whether the relative largeness of the population in 4XL versus 8XL networks had an effect on abundance, synchrony and stability using Welch’s unequal variance t- tests for which modified degrees of freedom were rounded down to the nearest integer (Ruxton

2006). For this calculation of degrees of freedom, as variances converge (i.e., become more similar), the degrees of freedom increase to approximate those of an equal variance t-test. As variances diverge, the degrees of freedom decrease to approximate those of a paired t-test.

Therefore, Welch’s test is appropriate for comparisons of groups with unequal as well as equal variance, and is preferred over the test of equal variance when n is small as in this experiment

(Ruxton 2006).

Because data collection was labor intensive, the experiment was conducted in two blocks. The first block was conducted from October to December 2013 and the second block was conducted from January to March 2014. Each block used eight networks: four 4XL networks and four 8XL networks. Within blocks, two 4XL and two 8XL networks had extinction and the others served as controls. We used one-way ANOVAs to test whether

56

abundance, synchrony, and stability through time varied between blocks. Since we found no

block effect, we pooled the data from both blocks for analyses (SI Tables 1-12). All data

collected for statistical evaluation are available at the UCScholar Data Repository

(scholar.uc.edu, doi:10.7945/C2J01Q).

Results

Abundance

Whether the population subjected to extinction was four or eight times larger than

surrounding populations by volume (4XL and 8XL networks, respectively), its extinction had no

measurable effect on the abundance of remaining populations (Figs. 2, 3, 4). The mean

abundance of small populations during the extinction of the large population was no different

than during the pre-extinction period in 4XL (t = -0.77, df = 3, P = 0.25) or 8XL (t = 2.35, df = 3,

P = 0.26; Fig. 4) networks. Furthermore, there was no change in the abundance of small

populations between the last day of the pre-extinction period and the first day of the extinction period in either 4XL (t = -0.72, df = 3, P = 0.52) or 8XL (t = -2.62, df = 3, P = 0.08) networks with extinction, indicating that there was no immediate loss of dispersers to small populations followed by fast recovery. In addition, there were no spatial (pairwise) effects of extinction on abundance (results not shown). Finally, there was no difference in the abundance of small populations between 4XL and 8XL networks during the pre-extinction (t = -0.07, df = 5, P =

0.95) or extinction periods (t = -1.24, df = 3, P = 0.30; Fig. 4). Therefore, despite the relatively high dispersal potential of the 8XL population, its effect on small populations during the pre- extinction period was not stronger than the effect of the 4XL population.

57

During the reestablishment period, the large population regained its influence over the

abundance of surrounding populations in 4XL and 8XL networks. In both network types,

immigration bolstered the abundance of small populations relative to the previous two periods

(e.g., vs. extinction in 4XL: t = -3.11, df = 3, P < 0.05 and 8XL: t = -7.43, df = 3, P < 0.01; Figs.

2-4). The increase was greater in 8XL networks than it was in 4XL networks (t = -5.14, df = 3, P

= 0.01; Fig. 4).

In 4XL and 8XL control networks there were no changes in abundance between pre-

extinction and extinction (4XL: t = -0.61, df = 3, P = 0.59; 8XL: t = -2.08, df = 3, P = 0.13) or

between extinction and reestablishment (4XL: t = 0.08, df = 3, P = 0.94; 8XL: t = 0.54, df = 3, P

= 0.62).

Synchrony

The effect of extinction on synchrony was not consistent between 4XL and 8XL

networks, but in neither case did the extinction of a large population increase synchrony in the

dynamics of surrounding populations (Figs 2, 3, 5, 6). In contrast to our hypothesis, mean

network synchrony in 4XL networks was lower during the extinction period than it was during

the pre-extinction period (t = 3.69, df = 3, P = 0.03), and reestablishment of the large population had no effect on synchrony (t = -0.37, df = 3, P = 0.73; Fig. 5A). Furthermore, there were no spatial (pairwise) effects of extinction on synchrony (Fig. 6A). Similarly, just as there were no short-term effects of extinction on abundance, there were no short-term effects of extinction on

synchrony that were masked by the longer observation period (results not shown).

In 8XL networks, dynamics were generally asynchronous (i.e., r ≤ 0.0; Figs. 3, 5B).

Unlike the 4XL networks, there was no change in mean network synchrony between the pre-

58

extinction and extinction periods (t = -0.29, df = 3, P = 0.79). As in 4XL networks,

reestablishment of the large population had no effect on synchrony (t = 1.13, df = 3, P = 0.34;

Fig. 5B). There were no pairwise changes between consecutive 10-day intervals in 8XL

networks (Fig. 6B) or short-term patterns masked by the longer observation period (results not

presented).

In 4XL and 8XL control networks there were no changes in synchrony between pre-

extinction and extinction (4XL: t = 0.44, df = 3, P = 0.69; 8XL: t = 0.81, df = 3, P = 0.48) or

between extinction and reestablishment (4XL: t = 0.93, df = 3, P = 0.42; 8XL: t = 0.83, df = 3, P

= 0.47, Figs. 2, 3, 5)

Stability

As predicted, the dynamics of small populations were more stable when the large population was extinct than when it was extant (Figs. 2, 3, 7). In both 4XL and 8XL networks, the dynamics of small populations were more stable during the extinction period than during either pre-extinction (4XL: t = 9.56, df = 3, P < 0.01; 8XL: t = 7.95, df = 3, P < 0.01; Fig. 7) or

reestablishment (4XL: t = -7.36, df = 3, P < 0.01; 8XL: t = -3.38, df = 3, P = 0.04). The network- wide abundance of small populations was relatively stable throughout the experiment in 4XL and

8XL networks with or without extinction (Fig. 7C-D), which often occurs when populations have asynchronous dynamics.

In control networks, small populations in 4XL networks were more stable during the extinction period than the pre-extinction period (t = 5.47, df = 3, P = 0.01; Fig. 7A), but there was no change in the stability of small populations in 8XL control networks (t = 0.60, df = 3, P =

0.59). During the reestablishment period, stability did not change in 4XL (t = 1.26, df = 3, P =

59

0.30; Fig. 7A) or 8XL (t = -0.49, df = 3, P = 0.66, Fig. 7B) control networks. It is unclear why

populations stabilized in 4XL control networks during the extinction period.

Discussion

The results of our experiment show that the extinction of a large population and the

maintenance of this habitat as an ecological trap stabilize the dynamics of remaining populations in a spatial population network (Ruxton et al. 1997, DeRoissart et al. 2015, Wang et al. 2015).

Contrary to theory (Matter and Roland 2009) and observation (Thomas et al. 1996, Matter and

Roland 2009, 2010), we found that extinction of a large population did not reduce the abundance of surrounding populations (Thomas et al. 1996, Matter and Roland 2010) or synchronize their dynamics (Matter and Roland 2009). The inability of the extinction of a large population, but not

a similar sized population (Lamkin et al., In Review) to reduce abundance or synchronize

surrounding populations in this system is likely the result of the dispersal behavior of our focal

organism from large populations relative to expectations in theoretical systems (Matter and

Roland 2009) and in natural networks where extinction has been shown to affect abundance and

dynamics (e.g., Thomas et al. 1996, Matter and Roland 2009, 2010, Lamkin et al., In Review). In

the discussion that follows, we present evidence that large populations of paramecium produce

pulses in dispersal which likely contain groups of migrants, rather than individuals migrating

between populations. Aside from these pulses, large populations may not have had high enough

levels of emigration to significantly alter the dynamics of surrounding populations, despite the

preliminary results showing large numbers of emigrants from large populations to vacant flasks.

Evidence for pulsed and aggregated emigration from large populations following network

assembly follows from density (not abundance) estimates. Culture densities for all populations

60

began at 200/mL. Preliminary counts taken 3 days following network assembly showed a

decrease in mean density from initial density to 26.7 ± 10.5/mL (SD) for the large populations in

the 8XL networks and an increase in mean density in smaller surrounding populations to 401.0 ±

26.5/mL. Somewhat less dramatically in the 4XL networks the large population decreased to

173.3 ± 31.1/mL while the small populations increased to 250.5 ± 47.1/mL. While these changes

in density could be due to higher mortality or lower natality in the large population and vice

versa in the small populations, higher equilibrium densities commonly occurring in isolated

flasks up to at least 4L (Lamkin, personal observation) strongly suggest that these changes in

density were due to migration.

The second pulse of emigration from the large populations occurred during the

reestablishment period. The effects of this event paralleled those of the first, but followed a burst

of exponential growth (Figs. 2 and 3). Over a three day period after peak abundance, there was a

decrease of 11 ± 0.22 and 21 ± 0.50%, in the large populations in 4XL and 8XL networks,

respectively and a corresponding increase in the mean abundance of small populations (e.g.,

between days 23-26 and 27-30, Figs. 2-3). Again, a great deal of this decline after growth

appears to have resulted from emigration from the large population into the network.

These emigration events are consistent with the positively thigmotropic behavior of paramecium; they use cilia to detect and maintain proximity to conspecifics and habitat structure, both of which protect from (Wichterman 1986). Because of their thigmotropism, migration was both en masse and biased toward small flask habitats where paramecia would have more encounters with “structure” than in the larger flask. Based on this, we might expect corridors to be optimal habitat, but paramecia also use cilia to detect the chemical environment.

For example, they follow increasing concentrations of oxygen and folate (produced by bacteria),

61 and decreasing concentrations of carbon dioxide (Van Houten and Preston 1987). Therefore, the folate-rich flask habitats should serve as chemical attractants versus the physically preferable corridor.

It is unclear what triggers these large scale dispersal episodes from larger populations, but they appear to be based on a density threshold and being connected to smaller populations.

Importantly for the dynamics of the networks, prior to the extinction period the large populations did not re-attain their original density (i.e., as if they were in isolation) and thus a density where mass emigration would occur. While there likely was always some movement out of large populations, they were not providing a consistent source of migrants as expected theoretically and seen in most other spatial systems. These pulses in emigration and attraction to smaller volumes also may explain why preliminary experiments showed high dispersal from large populations to smaller, vacant flasks (Fig. 1).

Interestingly, although expected from theory (Moran 1953, Matter 2013), the episodes of pulsed, mass emigration did not consistently synchronize population growth. Immigration to small populations arising from the exodus following network assembly may have generated some synchrony; during the pre-extinction period as evidenced by half of the 4XL and three of the eight 8XL networks being synchronous based on their mean pairwise correlation in growth

(Figs. 2-3). However, the mass emigration following reestablishment had no effect on synchrony.

The lack of a consistent effect on synchrony relates again to the thigmotropic behavior of paramecium. Rather than dispersing as individuals, paramecia likely disperse in groups (Ogata et al. 2008), particularly from the larger populations. Occasionally these groups are equally distributed among populations and produce synchrony, but in general this type of dispersal is not

62 uniformly distributed among populations, particularly from 8XL populations, and thus is de- synchronizing (Higgins 2009).

When observations do not meet theoretical predictions, either the model needs to be reconsidered or the observations are not in alignment with model assumptions. The results of the current study contrasts with theory, but rather than calling the theory into question, they illustrate novel ways in which biological systems can differ from theoretical assumptions. Prior to conducting these experiments there was little evidence that paramecia showed mass or clumped dispersal, neither of which is generally considered in theoretical population dynamics. It is unclear how prevalent these types of dispersal are in nature, but they deserve more consideration.

Following decades of effort aimed at understanding risk factors associated with population-level extinction (Dyn-Richer and Goel 1972, Griffen and Drake 2008, Peterson et al.

2014) and the effect of population-level extinction in a community context (MacArthur 1955;

Cardinale et al. 2012), it is appropriate to direct attention toward understanding the effects of population-level extinction in the context of spatial population networks. Results from this experiment support the long-held understanding that large “mainland” populations are important to the abundance and dynamics of surrounding “island” populations, but also remind that, like all ecological phenomena, interactions between large and small populations can be fleeting. When dispersal between a large and small populations is strong but intermittent, extinction of a large population is unlikely to increase the risk of additional extinction though synchrony or reduced abundance in contrast to systems where dispersal is consistently strong (Thomas et al. 1996,

Matter and Roland 2009, 2010).

Acknowledgments

63

We would like to thank N. Brown, H-Y Chen, and B. Merritt at the University of Cincinnati for

thoughtful comments on this manuscript. In addition we would like to thank T. Doak at Indiana

University for providing the paramecia and culture methods.

References

Abbott KC (2011) A dispersal-induced paradox: synchrony and stability in stochastic

. Ecol Letters 14:1158-1169. doi:10.1111/j.1461-0248.2011.01670.x

Bjørnstad O, Ims RA, Xavier L (1999) Spatial population dynamics: analysing patterns and

processes of population synchrony. Trends Ecol Evol 14:427–432. doi:10.1016/S0169-

5347(99)01677-8

Bowman J, Cappuccino N, Fahrig L (2002) Patch size and population density: the effect of

immigration behavior. Cons Ecology 6:9

Brown JH, Kodric-Brown A (1977) Turnover rates in insular biogeography: effect of

immigration on extinction. Ecology 58:445-9 doi:10.2307/1935620

Cardinale BJ, Duffy JE, Gonzalez A, Hooper DU, Perrings C, Venail P, Narwani A, Mace GM,

Tilman D, Wardle DA, Kinzig AP (2012) Biodiversity loss and its impact on humanity.

Nature 486:59-67 doi:10.1038/nature11148

De Roissart A, Wang S, Bonte D (2015) Spatial and spatiotemporal variation in metapopulation

structure affects population dynamics in a passively dispersing arthropod. J Animal

Ecology 84:1565-74. doi:10.1111/1365-2656.12400

Dwernychuk LW, Boag DA (1972) Ducks nesting in association with gulls-An ecological trap?

Can J Zool 50:559-563. doi:10.1139/z72-076

64

Fronhofer EA, Kubisch A, Hilker FM, Hovestadt T, Poethke HJ (2012) Why are

metapopulations so rare? Ecology 93:1967-78. doi:10.1890/11-1814.1

Griffen BP, Drake JM (2008) A review of extinction in experimental populations. J Animal

Ecology 77:1274-1287. doi:10.1111/j.1365-2656.2008.01426.x

Gyllenberg M, Hanski I (1997) Habitat deterioration, habitat destruction, and metapopulation

persistence in a heterogeneous landscape. Theor Population Biol 52:198-215.

doi:10.1006/tpbi.1997.1333

Hanski I (1991) Single-species metapopulation dynamics: concepts, models, and observations.

Biol J Linn Soc 42:17-38

Harrison S (1991) Local extinction in a metapopulation context: an empirical evaluation. Biol J

Linnean Soc 42:73-88. doi: 10.1111/j.1095-8312.1991.tb00552.x

Higgins K (2009) Metapopulation extinction risk: dispersal’s duplicity. Theor Pop Biol 76:146-

155. doi:10.1016/j.tpb.2009.05.006

Lamkin MK, Kaneshiro ES, Matter SF (Chapter 3 of this Dissertation) An experimental test of

local population extinction in spatial population networks.

Lande R (1993) Risks of population extinction from demographic and environmental

stochasticity and random catastrophes. Am Nat 142:911-927.

doi:http://dx.doi.org/10.1093/besa/15.3.237

Liebhold A, Koenig WD, Bjørnstad ON (2004) Spatial synchrony in population dynamics.

Annual Rev Ecol, Evol, and Systematics. 35:467-490. doi:10.1146/annure

MacArthur R (1955) Fluctuations of animal populations and a measure of community stability.

Ecology 36:533–536. doi: 10.2307/1929601

65

Matter SF (2001) Synchrony, extinction, and dynamics of spatially segregated heterogeneous

populations. Ecol Modelling 141: 217-226. doi:10.1016/S0304-3800(01)00275-7

Matter SF (2006) Changes in landscape structure decrease mortality during migration. Oecologia

150:8-16. doi:10.1007/s00442-006-0491-3

Matter SF (2013) Increases and decreases in network synchrony associated with local population

extinction and recolonization. Ecological Modelling 266:68-72.

doi:10.1016/j.ecolmodel.2013.06.021

Matter SF, Roland J (2009) Local extinction synchronizes population dynamics in spatial

networks. Proc Roy Soc Lond B 277: 729-737. doi: 10.1098/rspb.2009.1520

Matter SF, Roland J (2010) Effects of experimental extinction for the spatial population

dynamics of the butterfly Parnassius smintheus. Oikos 119:1961-1969.

doi: 10.1111/j.1600-0706.2010.18666.x

Moran PA (1953) The statistical analysis of the Canadian lynx cycle. Aust J of Zool 1:291-298.

Öckinger E, Smith HG (2007) Semi‐natural grasslands as population sources for pollinating

insects in agricultural landscapes. J Applied Ecol. 44:50-9. doi:10.1111/j.1365-

2664.2006.01250.x

Ogata M, Hondou T, Hayakawa Y, Hayaski Y, Sugawara K (2008) Adaptation-induced

collective dynamics of a single-cell protozoan. Physical Review E 77:011917.

doi:10.1103/PhysRevE.77.011917

Pearson RG, Stanton JC, Shoemaker KT, Aiello-Lammens ME, Ersts PJ, Horning N, Fordham

DA, Raxworthy CJ, Ryu HY, McNees J,Akçakaya HR (2014) Life history and spatial

traits predict extinction risk due to climate change. Nat Clim Change 4:217-221.

doi:10.1038/nclimate2113

66

Richter-Dyn N, Goel NS (1972) On the extinction of a colonizing species. Theor Popul Biol

3:406-433. doi:10.1016/0040-5809(72)90014-7

Robertson BA, Hutto RL (2006) A framework for understanding ecological traps and an

evaluation of existing evidence. Ecology 87:1075-1085. doi: jstor.org/stable/20069046

Ruxton GD (2006) The unequal variance t-test is an underused alternative to Student's t-test and

the Mann–Whitney U test. Behavioral Ecology. 17:688-90. doi: 10.1093/beheco/ark016

Ruxton GD, Gonzalez-Andujar JL, Perry JN (1997) Mortality during dispersal and the stability

of a metapopulation. J Theor Biol 186:389-96. doi:10.1006/jtbi.1997.0402

Van Houten J, Preston RR (1987) Chemoreception: Paramecium as a receptor cell. In Mol Mech

of Neuronal Responsiveness, 1st ed. Springer, New York.

Richter-Dyn N, Goel NS (1972) On the extinction of a colonizing species. Theor Popul Biol

3:406-433. doi:10.1016/0040-5809(72)90014-7

Royama T (1992) Analytical Population Dynamics, 1st ed. Chapman and Hall, London.

doi:10.1007/978-94-011-2916-9

Sonneborn TM (1970) Methods in Paramecium Res. In: Prescott DM (ed) Methods in Cell Phys

Vol.4. Academic Press Inc, New York, pp. 241-399

Thomas CD, Singer MC, Boughton DA (1996) Catastrophic extinction of population sources in a

butterfly metapopulation. Am Nat 1:957-75.

Wichterman R (1986) The Biology of Paramecium, 2nd edn. Plenum Press, New York. doi:

10.1007/978-1-4757-0372-6

Zar JH (1999) Biostatistical Analysis, 4th ed. Prentice Hall, Upper Saddle River NJ

67

Figure Captions

Figure 1. Density of paramecia in a five-flask linear array (N=3 networks) following a 24-hour dispersal from a single “start” population connected to four “receiving” flasks (x-axis: 1, 2, 3, 4, where “1” was closest to the start population and “4” was most distant). The density of the start population in A and B was 644 paramecia/mL and 620 paramecia/mL, respectively. Black bars in A and B show distant-dependent dispersal from a 70 mL start population (X), with most immigrants arriving in the neighboring flask (x-axis: “1”). The effect of distance dampened when the start population was twice the volume (2X) of receiving flasks (A, grey bars) and disappeared when the start population was eight times the volume (8X) of receiving flasks (B, grey bars).

Straight density (vs mean density) shown to explicitly show the variability in the results among networks. Note the relatively low density in the 8X population following the 24 hour dispersal period.

Figure 2. Daily density estimates (paramecia/mL) for local populations in 4XL networks with extinction (left, N=4) and control networks (right, N=4) for the 31-day experiment (30 growth periods). Mean network synchrony for the six pairwise measures (r≤0 ± SE, *r>0.0 ± SE;

**r>0.5 ± SE) shown for each 10-day period for each network. Transparent grey box in the center of each panel indicates the extinction period (note: no extinction in control networks).

Data are shown for center (grey lines), center-adjacent (red lines), and distal (black lines) populations. Circle diagrams above panels denote positions of populations by color.

Figure 3. Daily density estimates (paramecia/mL) for local populations in 8XL networks with extinction (left, N=4) and control networks (right, N=4) for the 31-day experiment (30 growth

68 periods). Mean network synchrony for the six pairwise measures (r≤0 ± SE, *r>0.0 ± SE) shown for each 10-day period for each network. Transparent grey box in the center of each panel indicates the extinction period (note: no extinction in control networks). Data are shown for center (grey lines), center-adjacent (red lines), and distal (black lines) populations. Circle diagrams above panels denote positions of populations by color.

Figure 4. Mean density (paramecia/mL) of small populations in 4XL (A) and 8XL (B) networks during each of the three 10-day periods in networks with extinction and control networks

(Density ± SE; N=4). A statistical difference from the previous period indicated by “*” or “**” for P<0.05 and P<0.01, respectively.

Figure 5. Mean network synchrony (r, ±SE; N=4) in 4X (A) and 8XL (B) networks with extinction and control networks during consecutive 10-day periods (small populations only). A statistical difference from the previous period indicated by “*” (0.01≤ P<0.05).

Figure 6. Mean synchrony (r, ±SE; N=4) for each pair of small populations in 4X (A) and 8XL

(B) networks with extinction (left) and control networks (right) during consecutive 10-day periods. Circle diagram denotes positions of populations included in analysis. There were no statistical changes in pairwise synchrony during the experiment.

Figure 7. Mean variability in the abundance of local populations (A-B) and of the network (C-D) expressed as a coefficient of variation (CV ± SE; N=4) where greater variability indicates less stable abundance fluctuations. A and C show variability in 4XL networks; B and D show

69 variability in 8XL networks (N=4 ± SE). Center populations excluded from analysis. A statistical difference from the previous period indicated by “*” (0.01

70

Figures

800 A B X 1 2 3 4 X 1 2 3 4 600

2X 1 2 3 4 8X 1 2 3 4 400

200 Density (#Paramecia/mL)

0 X ¦ 2X 1 2 3 4 X ¦ 8X 1 2 3 4

Figure 1.

71

4XL 4XL

0.24 ± 0.27 -0.08 ± 0.13 -0.18 ± 0.08 0.08 ± 0.15 0.13 ± 0.14 0.26 ± 0.08* 1000

800

600

400

200

0 -0.01 ± 0.22 -0.18 ± 0.22 0.25 ± 0.14 0.03 ± 0.27 -0.06 ± 0.21 -0.03 ± 0.25 1000

800

600

400

200

0 0.56 ± 0.12* 0.32 ± 0.16 0.01 ± 0.22 0.43 ± 0.10* 0.19 ± 0.10 -0.05 ± 0.22 1000

800 Density(# paramecia/mL)

600

400

200

0 0.73 ± 0.05** 0.29 ± 0.14* 0.51 ± 0.10* 0.27 ± 0.13* 0.40 ± 0.06* 0.09 ± 0.19 1000

800

600

400

200

0 0 5 10 15 20 25 30 0 5 10 15 20 25 30 Day Day Networks with Extinction Control Networks

Figure 2.

72

8XL 8XL

-0.07 ± 0.14 0.25 ± 0.13 0.08 ± 0.11 0.31 ± 0.15* -0.22 ± 0.14 0.00 ± 0.12 1000

800

600

400

200

0 -0.16 ± 0.16 0.11 ± 0.17 -0.01 ± 0.14 -0.10 ± 0.21 -0.01 ± 0.25 -0.06 ± 0.16 1000

800

600

400

200

0 0.18 ± 0.13 0.11 ± 0.14 -0.05 ± 0.19 0.36 ± 0.15* 0.28 ± 0.08* -0.06 ± 0.14 1000

800 Density(# paramecia/mL) 600

400

200

0 0.38 ± 0.17* 0.06 ± 0.18 0.21 ± 0.11 0.11 ± 0.07 0.19 ± 0.13 -0.16 ± 0.07 1000

800

600

400

200

0 0 5 10 15 20 25 30 0 5 10 15 20 25 30 Day Day

Networks with Extinction Control Networks

Figure 3.

73

600 A Pre-extinction B Pre-extinction * Extinction ** Extinction 500 Reestablishment Reestablishment

400

300 Density 200

100

0 With Extinction Control With Extinction Control

4XL Networks 8XL Networks

Figure 4.

1.0 A Pre-extinction B Pre-extinction 0.8 Extinction Extinction

) Reestablishment Reestablishment

0.6

0.4 * Synchrony ( r 0.2

0.0

With Extinction Control With Extinction Control

4XL Networks 8XL Networks

Figure 5.

74

1.0 A A B 4XL D E Pre-extinction A B 4XL D E Extinction Reestablishment

0.5

0.0

-0.5

1.0 Pre-extinction

Synchrony (r) A B D E B A B 8XL D E Extinction 8XL Reestablishment

0.5

0.0

-0.5

A&B A&D A&E B&D B&E D&E A&B A&D A&E B&D B&E D&E

Networks with Extinction Control Networks

Figure 6.

75

0.8 Pre-extinction A B Pre-extinction Extinction Extinction Reestablishment 0.6 Reestablishment ** * CV

0.4 Population 0.2

0.0 Pre-extinction Pre-extinction C Extinction D Extinction 0.6 Reestablishment Reestablishment CV

0.4 Network 0.2

0.0 Treatment Control Treatment Control

4XL Networks 8XL Networks

Figure 7.

76

Chapter 5 Recolonization of a Large Population Temporarily Counters Effects of Long-term Habitat Degradation in a Spatial Population Network

Abstract

Anthropogenic habitat degradation is an on-going, global phenomenon that contributes to the loss of biodiversity and ecosystem function in terrestrial, marine and freshwater systems. For a species under persistent degradation on a regional scale, the extinction of a large population would have significant

effects on the species’ regional trajectory to extinction. The renewal of the population would have

significant effects as well, but the degree to which the renewal of a large population may counter

the effects of extinction and habitat degradation remain unexplored. Given the commonality of

both abrupt and gradual change occurring within habitats the world over, it is a valuable exercise

to consider how the two may interact. We used a microcosm approach to test the null hypothesis that

extinction and recolonization of a large population in a declining network would have no net effect on

regional persistence. We compared changes in abundance and dynamics through time in networks with

and without extinction/recolonization. We found the renewal process balanced the losses incurred by

extinction and recovered most of the loss that would have occurred by habitat degradation alone.

However, by the end of the experiment, networks with extinction were declining similarly to

networks with no extinction so that there was no ultimately no difference in the projected

persistence of the two networks. Therefore, for a fast growing, dispersive species, the

recolonization of a large population can temporarily bolster regional abundance, but ultimately

the effects of habitat degradation will off-set gains produced by the recolonization, growth and

dispersal.

77

Key words

Spatial population networks, habitat degradation, disturbance, extinction, microcosm,

Paramecium, synchrony

Introduction

Anthropogenic habitat degradation is an on-going, global phenomenon that contributes to

the loss of biodiversity and ecosystem function in terrestrial, marine and freshwater systems

(MEA 2005, Gaston and Fuller 2008, Butchart et al. 2010; Macusi et al. 2016). Distinct from

abrupt changes associated with habitat destruction, losses incurred by habitat degradation are

gradual and difficult to detect in the short term (Doak 1995, Mortelleti et al. 2010). For example,

a slight but persistent directional change in the pH, temperature, nutrient concentration, and/or

toxic load can slowly reduce the abundance of sensitive species (Biesmeijer et al. 2006, Fabricius et al. 2011, Li et al. 2011, Macusi et al. 2016) and alter community composition to favor more

tolerant species. Under persistent degradation on a regional scale, the persistence of a sensitive

species will depend not only on its life history but also on its abundance and spatial distribution

(Bender et al. 1998, Vuilleumier et al. 2007, Robert 2009, Franzén and Nilsson 2010).

In nature, large populations often support networks of smaller populations (Harrison

1991, Fronhofer et al. 2012). However, within these “mainland-island” networks, large

“mainland” populations and surrounding “island” populations are important to regional abundance (Verboom et al. 2001). Large populations have a numbers advantage that buffers them from extinction by environmental or demographic disturbance (Richter-Dyn and Goel

1972, Lande 1993), but small populations ensure, to a degree, that extreme localized disturbance aimed at the large population will not extirpate the species on a regional scale (Brown and

78

Kodric-Brown 1977, Hanski 1994, Rösch 2015). In a declining network where large populations

are particularly important for network persistence, the extinction of a large population is

expected to increase the network trajectory to extinction (Doak 1995). Although recolonization

of the population would counter the effects of extinction to a degree, whether or not extinction

and recolonization would produce a net effect on regional persistence remains unexplored. Given

the commonality of both abrupt and gradual change occurring within habitats the world over, it is

a valuable exercise to consider how the two may interact.

In a declining network, the extinction of a large population is expected to decrease network persistence through decreased abundance and increased rate of decline. First, the loss of the population would decrease regional abundance and therefore decrease resilience to regional

disturbance (Folke et al. 2004). In addition, the sudden loss of immigrants from the large

population into surrounding populations may decrease their abundance (Thomas et al. 1996,

Matter and Roland 2010, but see Lamkin Chapter 4). If the vacant patch attracted immigrants

that it could not support, as in an ecological trap (Robertson and Hutto 2006), network

abundance would decrease further and so too would persistence. Contrastingly, the trap may

dampen oscillations of small populations, thus promoting network persistence (Lande 1993,

Lamkin et al. Chapter 4). The sudden loss of immigrants into surrounding populations could

synchronize growth (i.e., more positively correlated growth; Matter and Roland 2009), which

would increase the risk of correlated extinction (Heino et al. 1997, Earn et al. 2000). Overall, the

long-term extinction of a large population is expected to increase the rate of decline in the

network and vulnerability to extinction.

Just as the extinction of a large population is predicted to have strong effects on dynamics

and persistence of a network in decline, so too is its recovery. Recolonization of the large

79 population would increase regional abundance and therefore increase the species’ resilience to regional disturbance (Folke et al. 2004). Increased immigration into surrounding populations would bolster their abundance therefore decrease their risk of extinction (Thomas et al. 1996,

Matter and Roland 2010). Small patches that were rendered vacant during extinction of the large population would be recovered (Hanski 1994). Distinct from the synchronizing effect of the loss of immigrants associated with extinction (Matter and Roland 2009), the return of immigrants through recolonization could increase, decrease or have no effect on synchrony (Matter 2001,

Higgins 2009, Matter 2013). For example, a spatially uniform (i.e., proportionate) distribution of dispersers from the large population into surrounding populations would increase synchrony, but a spatially non-uniform (i.e., disproportionate) distribution of dispersers from the large population into surrounding populations would decrease synchrony (Matter 2001, Higgins 2009,

Lamkin et al. Chapter 4). Since various factors in natural settings (e.g., distance, clumped dispersal behavior, and perception of habitat quality) generate non-uniformity in the distribution of dispersers, dispersal from a large population is not expected to be persistently synchronizing.

For example, populations most proximate to the large population may be positioned to absorb disproportionately more dispersers than populations located farther away. Alternatively, species with clumped dispersal behavior and no distance-dependent limitations may immigrate disproportionately into surrounding populations by chance. The degree that recolonization of the large population decreased synchrony in the dynamics of surrounding populations would be the degree to which it reduced their risk of correlated extinction. Overall, the effects produced by the renewal of the population would counter those of extinction.

Given the propensity for extinction and recolonization of a large population to affect regional persistence in opposing ways, it is relevant to consider whether the combined effects

80

would have a net effect on regional persistence. The renewal may counter the effects of

extinction so that there was no artifact of extinction on the network trajectory to extinction.

Alternatively, the effects of renewal may be insufficient to counter those of extinction, or they

may be sufficient to counter those of extinction and prolong regional persistence beyond that of

an equivalent network with no extinction. To date there has been no test of extinction and

recolonization of a large population in declining networks. We used experimental protozoan

population networks to determine how the extinction and recolonization of large populations in declining networks may affect regional persistence.

Methods

Focal Organism

Paramecium caudatum is a free-swimming single-celled ciliate common in freshwater habitats (Wichterman 1986). Our cultures were obtained from Indiana University and maintained in wheatgrass-based broth infused with a laboratory strain of the bacteria species Klebsiella pneumoniae (Sonneborn 1972). We initiated all experimental networks from a single stock culture at stationary phase indicated by zero growth for 2 consecutive days (≈400 paramecia/mL).

Experimental Design

We constructed eight networks (N=4 with extinction; N=4 control), each of which consisted of 5 filter flasks aligned linearly and connected to a 5 mm diameter “migration corridor” made of plastic tubing (SI Fig. 1). The center position held a 500 mL flask and each of the four non-center positions a 50 mL filter flask. To assemble networks, we filled flasks to just

81

below the sidearm from a single stock culture at stationary phase. After filling the corridor with

15 mL sterile distilled water, we connected the sidearms of the corridor to the sidearms of the

five flasks. Small flasks sat on blocks so that the side arms and corridor were level and flask size

did not affect dispersal distance. We then added additional volume to each population so that

culture volume surpassed the entry to the corridor. Specifically, a 70 mL population occupied a

50 mL flask and a 560 mL population occupied a 500 mL flask. Flasks were loosely capped with

aluminum foil so that oxygen was not limiting. The experiment began two weeks after network

assembly to allow for the populations to stabilize in the network configuration.

We maintained a stable carrying capacity in the network until the start of the experiment

by replacing a small volume of each population (5 mL/70 mL) with bacterized medium once per

day that was proportionately equivalent in small in large populations (5 mL/70 mL = 40 mL/560

mL). Based on previous results (Lamkin et al. Chapter 4) and paramecium’s preference for the physical dimension of the small flasks, the large population would not maintain dispersal to smaller populations unless we bolstered the large population’s growth by provisioning it with a more concentrated medium than that used to provision small populations. Specifically, we replenished the large population with undiluted bacterized medium and replenished the small

populations with a 3:1 dilution of the same medium (i.e., 75% bacterized medium: 25% buffered

water). Through differential provisioning that favored the production and absorption of

dispersers in the large population, we expected the large population to remain an important

source of immigrants to surrounding populations throughout the experiment.

Network-wide habitat degradation began on the first day of the experiment (day 0/30)

and continued throughout the three consecutive 10-day periods of the experiment: pre-extinction, extinction, and recolonization. In order to gradually degrade habitats, we simply diluted the

82

replenishments given to populations during the two-week stabilization period by 25%. By

diluting the replenishment allocated to both large and small populations by 25%, we essentially

deprived each population of 1.25 mL of bacterized medium/70 mL per day (5 mL * 0.25)

without compromising culture volume. Therefore, each local habitat was subjected to a uniform

rate, R, of degradation equal to 1.79% per day (1.25 mL/70 mL). As in economics where an asset

depreciates over time at a constant rate, using this method caused habitat quality to depreciate at

t a constant rate. Therefore, we can estimate degradation over time as: Qt = Q(1-(R/100) , where

Q is habitat quality, R is the rate of decline and t is time (number of days). Therefore, assuming habitat quality, Q, was 100% on day 1, habitat quality would be reduced by 16.5% at the end of the 10-day pre-extinction period and by 41.8% on the last day of the experiment. Therefore, degradation may not generate detectable reductions in regional abundance between days but is expected to generate detectable reductions within each of the 10-day periods without compromising regional persistence in the 30-day span of the experiment.

After the 10-day pre-extinction period we caused a 10-day extinction of the large population in four of the eight networks. Networks without extinction served as controls. To create an extinction, the large population was separated from the network, microwaved to near boiling (≈95 °C), and quickly cooled in a cold water bath. When the liquid reached room temperature (≈23 °C), all flasks were reconnected to their respective networks Individuals were free to migrate into the center flask habitat during the extinction period, but were extirpated daily. Therefore, the vacant patch functioned as an ecological trap (Dwernychuk and Boag 1972, Robertson and Hutto 2006). Following the 10-day extinction period, large populations were allowed to be recolonized. Networks were maintained in normal laboratory conditions throughout the experiment (≈23 °C, 10:14 hours, light:dark).

83

Abundance

Each day we estimated the density of paramecia within each population as the mean number of individuals captured in three 0.25 mL samples. Each sample was withdrawn independently and distributed into multiple drops over a Petri plate so that individuals could be easily tallied under 7.5x magnification. We accepted the mean density of the three samples when the sample variance was less than 20%, measured as a coefficient of variation, CV, given by

= 100, where s was the standard deviation of the sample densities and μ was the mean 𝑠𝑠 𝐶𝐶𝐶𝐶 �𝜇𝜇� ∗ density. When CV ≥ 20%, we resampled the population and used the combined mean of the six samples as our estimate. We extrapolated abundance from measures of density in order to calculate growth, rates of decline, and synchrony (below) as well as compare local and regional abundance through time or between extinction and control networks. We compared the mean proportion of network or population abundance lost or gained through time. Because proportions could be greater than one, we arc-sin square-root transformed them prior to analysis to meet distributional assumptions (Zar 1999). To show changes in abundance for large and small populations in the same figure, we presented abundance as density (#/mL).

Rate of Decline and Time to Extinction

We calculated the daily per capita growth rate, R, for each population according to

R = ln , where N = population abundance and t=time (Royama 1992). We used the 𝑁𝑁𝑡𝑡+1 𝑁𝑁𝑁𝑁 average� per capita� daily rate over a specified time period to describe the mean rate of decline

(or increase). Because we were interested in whether extinction-recolonization dynamics affected regional persistence and were not interested in time to extinction per se, we used the

84

same asset depreciation formula we used to estimate habitat degradation to project persistence

in networks with extinction and control networks. Specifically, we rearranged the

t depreciation formula, Nt = N(1-R) to solve for the number of days it would take for regional

𝑁𝑁 abundance, Nt, to be less than one. Therefore, persistence was estimated as t= ( 𝑡𝑡 ) , where ln� 𝑁𝑁 � ln 1+𝑅𝑅 Nt=0.25, R=the mean rate of decline, and N=the final regional abundance upon which the rate

of decline was imposed.

Synchrony

We used Pearson’s product moment correlation, r, of population growth between pairs of populations as our measure of pairwise synchrony, and calculated network synchrony as the mean of the six pairwise correlations (Bjørnstad et al. 1999). Prior to statistical analysis, all correlation coefficients were z-transformed to meet distributional assumptions (Zar 1999).

Statistical Analysis

Within treatment and control networks, we tested for differences in mean abundance, the proportional changes in abundance, growth (i.e., rate of decline or increase), and synchrony using paired t-tests. Generally we were interested in differences between consecutive 10-day periods (i.e., between pre-extinction, extinction and recolonization), but we also evaluated short- term changes that may have been masked by the longer observation period. We tested for differences between treatment and control networks using Welch’s unequal variance t-test for

which degrees of freedom (df) were rounded down to the nearest integer (Ruxton 2006).

All data collected for statistical evaluation are available at the UCScholar Data

Repository (scholar.uc.edu, doi:10.7945/C2J01Q).

85

Results

Persistent habitat degradation generated a long-term decline in networks with extinction

and in control networks (Fig. 1). Extinction and recolonization of the large population increased

network abundance above what would be expected from habitat degradation alone (Figs. 2-4),

but had no effect on synchrony (Figs. 5-6). By the end of the experiment there was no statistical

difference between the projected persistence in networks with extinction and control networks

(Fig. 3). Below, we report the effects of extinction and recolonization separately.

The effects of extinction

The effects of extinction on network-wide abundance were strong and immediate. In the

first three days of the period, networks with extinction lost 68.1 ± 2.0% of their final pre-

extinction abundance, after which no further losses occurred; Figs. 2-3). Conversely, control networks remained in gradual decline throughout the extinction period. By the end of the period control networks decreased 37.5 ± 2.2% below their final pre-extinction abundance, which was less than the loss of reginal abundance in networks with extinction (t=9.89, df=6, P<0.001; Fig.

2-3). Although the dramatic loss of regional abundance in networks with extinction was due mostly to the loss of the large population itself, surrounding populations incurred significant reductions as well. For example, within the first 24 hours of extinction, populations surrounding the vacant patch lost 15.2 ± 2.6% of their previous day’s abundance (Fig. 4A). By comparison, small populations in control networks lost 5.6 ± 1.4% of their abundance over the same 24 hour period (Fig. 4B). Therefore, the immediate loss of dispersers from the large population caused

86 greater reductions in surrounding populations than were generated by habitat degradation alone

(reduction in treatment vs control networks: t=3.18, df=6, P=0.02).

Over the 10-day extinction period, the mean per capita rate of decline in small populations was similar in extinction and control networks (t=0.48, df=3, P=0.67), but the pattern differed depending on whether the decline was driven by the loss of immigrants and degradation or habitat degradation alone (Figs. 3-4). In networks with extinction, small populations declined at a daily rate of 26.9 ± 0.83% for the first three days, after which their abundance stabilized, i.e., mean growth was statistically zero for the remainder of the period (t=-

0.76, df=3, P=0.50; Figs. 3, 4A). By the end of the 3-day reduction, small populations in networks with extinction lost 55.5 ± 1.5% of their final pre-extinction measure. During the same

3-day span, small populations in control networks were not so strongly reduced (18.6 ± 1.2%, t=8.23, df=5, P<0.001; Fig. 4B). However, due to the persistent habitat degradation, small populations in control networks continued to decline so that, by the end of the period, the abundance of small populations in control networks had converged to the abundance of small populations in networks with extinction (t=-2.54, df=3, P=0.08; Fig. 4A-B). Therefore, despite differences in short-term rates of decline in networks with extinction and control networks, by the end of the ten-day period, small populations in both networks had been similarly reduced

(t=1.76, df=6, P=0.13; Fig. 3, 4A-B).

Because the regional rate of decline changed dramatically in networks with extinction during the extinction period, so too did the projected persistence of the network. For example, considering mean growth rates over the 10-day span of the period, the projected persistence in networks with extinction (101 ± 6.0 days) was lower than the projected persistence in control networks (273 ± 23 days: t=-7.21, df=3, P<0.01). However, considering the last five days of the

87

extinction period only, when the small populations in networks with extinction had stabilized

(Fig. 3, 4A), there was no difference in the projected persistence of networks with extinction and

control networks (t=-1.37, df=3, P=0.27). Therefore, despite the loss of the large population in

networks with extinction, the gradual declines associated with habitat degradation had

comparable effects on regional persistence.

Surprisingly, the sudden, network-wide loss of immigrants associated with extinction of the large population was not synchronizing (Figs. 1, 5A-B). There was no difference in network synchrony between pre-extinction and extinction at the 10-day scale of comparison (t=0.80,

df=3, P=0.48) nor was there a short-term (i.e., 5-day) synchronization that was masked by the

longer observation period (t=-0.75, df=3, P=0.51; Fig. 5A-B). Note also during the first three

days of the extinction period when small populations were in rapid decline, the local per capita

rate of decline was not correlated (r=-0.08 ± 0.16). Although there was a trend for higher

synchrony in networks with extinction than in control networks during the extinction period

(t=2.43, df=5, P=0.06), no network in either the extinction or control group was statistically

positively correlated (i.e., r>0.0) during the extinction period (Fig. 1, 5). Therefore, rather than a

trend of greater synchrony in networks with extinction, it would more accurately be described as

lesser asynchrony. Since pairwise synchrony varied within and among networks during the

extinction period (Fig. 6), we considered whether pairwise reductions during the first 24 hours of

the extinction period correlated with pairwise synchrony during the first five days of the

extinction period. We found no relationship between the two variables in networks with

extinction (t=0.70, df=3, P=0.53) or control networks (t=-1.0, df=3, P=0.38). Therefore, the

synchronized loss of immigrants had no effect on how synchronously populations increased and

decreased through time.

88

The effects of recolonization

Recolonization of the large population bolstered network abundance above that of control

nteworks, but by the end of the recolonization period, the effects of habitat degradation had

begun to dominate dynamics. Despite on-going habitat degradation during the recolonization period, the large population recovered its pre-extinction abundance within three days (t=-0.75,

df=3, P=0.51; Fig. 4A) and bolstered the abundances of surrounding populations above what

they had been at the end of the extinction period (t=-3.25, df=3, P<0.05). By the end of the 10-

day period, networks with extinction recovered 93.3 ± 2.1% of the regional abundance lost

during the extinction period (i.e., nearly all of the 68.1 ± 2.0% of regional abundance lost

between days 10 and 20). Accounting for both losses and gains, networks with extinction lost a

net 12.4 ± 4.4% of their pre-extinction regional abundance by the end of the experiment (i.e., day

10 vs day 30). Conversely, control networks decreased by16.5 ± 9.0% during the recolonization

period (i.e., day 30 vs day 20), accumulating a net loss of 48.2 ± 4.5% of their pre-extinction

regional abundance by the end of the experiment (i.e., day 10 vs day 30). Therefore, in networks

with extinction, the effect of recolonization not only countered loss from extinction, but also

countered loss due to habitat degradation so that by the end of the experiment, networks with

extinction were 166 ± 15.7% larger than control networks (t=5.60, df=6, P=0.001). However,

during the last two days of the experiment, networks with extinction were declining at a rate that

was almost as strong as the rate of decline in control networks (t=2.58, df=6, P=0.04; Fig. 3).

Based on final abundance measures and regional decline during these final two periods, the

projected persistence was statistically similar in networks with extinction (116 ± 16.4 days) and

control networks (76.3 ± 7.65 days; t=2.23, df=4, P=0.09). Therefore, despite net gains produced

89

by extinction and recolonization of the large population, habitat degradation limited the long-

term effect of those gains.

There was a trend for higher synchrony during recolonization than during extinction in networks with extinction (t=-2.67, df=3, P=0.08) and control networks (t=-2.88, df=3, P=0.06;

Fig. 5A-B). In both network types negatively correlated growth (r<0.0) during the last five days of extinction was followed by positively by increasingly correlated growth. Just as we found no

correlation between the loss of immigrants and pairwise synchrony during the extinction period,

we found no correlation between the gain of immigrants and pairwise synchrony during the

recolonization period in networks with extinction (t=0.74, df=3, P=0.51) or control networks

(t=0.27, df=3, P=0.80). Therefore, neither the loss nor gain of immigrants was a mechanism of

synchrony in this system.

Discussion

Compared to habitat loss and fragmentation, the effects of habitat degradation have

received considerably less attention (Mortellitti et al. 2010). Here, we took a microcosm

approach to explore how a network in decline due to habitat degradation responds to local

population extinction and recolonization. Somewhat surprisingly we found that exponential

growth and dispersal following recolonization of large populations bolstered network abundance

and persistence relative to networks without extinction. It is important to note that recolonization

did not completely offset the effects of habitat degradation. Rather, the exponential growth

allowed the networks to attain an abundance that exceeded the abundance of networks under

constant decline, but did not recover the network to pre-extinction levels. Thus, by no means are

90 extinction-recolonization dynamics a strategy to offset habitat degradation, though they may temporarily forestall its effects.

Importantly for network persistence, extinction and recolonization of large populations in declining networks did not synchronize their dynamics. Despite the widely demonstrated effects of correlated environmental conditions promoting synchrony (Moran 1953, Benton et al. 2001,

Liebhold et al. 2004), neither the degradation of regional habitat nor the synchronized loss of immigrants generated synchrony in local dynamics. In this system where the focal organism is thought to disperse in groups rather than individuals (Ogata et al. 2008, Lamkin et al. Chapter 4), it is not surprising that dispersers from the large population were not a consistent source of synchrony in the dynamics of surrounding populations during the pre-extinction or recolonization periods (Higgins 2009; Fig. 1). However, it was surprising that contrary to previous theoretical and empirical findings, the synchronized reduction of small populations caused by extinction had no effect on synchrony. For example, Matter and Roland (2009) showed theoretically that the increased synchrony they observed in relatively small butterfly populations during experimental extinction of a large population could have been induced by the synchronized loss of immigrants. Similarly, we observed previously (Lamkin et al. Chapter 3) that extinction of a small population in experimental protozoan networks synchronized the dynamics of its nearest neighbors which were similarly small. The difference between the previous experiments and this one is that following the loss of immigrants in those experiments, surrounding populations had the capacity to reproduce. The stabilization rather than growth of small populations after they were reduced indicates reproduction within small populations was stunted. That small populations supported a larger abundance when the large population was intact than when it was extinct indicates habitat degradation had stronger effects on the

91 reproductive capacity of small populations than the absolute carrying capacity of the flask. A second indication of a discrepancy in the reproductive capacity of the large versus small populations is that we frequently observed conjugation in samples withdrawn from small populations but not large populations (Lamkin personal observation). Paramecia are known to exchange genetic material by conjugation when resources are scarce and divide by binary fission only when resources are abundant (Mikami and Hiwatashi 1975). Since populations could not reproduce in response to the synchronized forcing, dynamics did not become synchronized.

Rather, any fluctuation in the abundance of small populations during the latter part of the extinction period occurred in response to emigration and immigration (or lack of precision in abundance estimates). Since dispersal is not synchronizing in this system, synchrony could only be induced by a synchronized reduction that produced synchronized growth.

It is interesting that habitat degradation stunted reproduction in the small populations but appeared to have no effect on the large population. This discrepancy in the reproductive capacity of the large versus small populations is explained by the relatively low density at which the large population was maintained. We learned in a previous experiment that a population in a large flask maintains a higher abundance when it is isolated than when it is connected to smaller populations (Lamkin et al. Chapter 4). When connected to a network of smaller populations, a large population is reduced from a higher stable abundance to a lower stable abundance through a massive dispersal event (Lamkin et al. Chapter 4). For reasons related to the behavior of paramecium, the large population maintains an abundance that is less than the food-based carrying capacity of the habitat. Because of its reduced abundance relative to the absolute carrying capacity of the habitat, habitat degradation over the 30-day span of the experiment reduced the surplus of food in the large flask but the baseline needed to maintain the population

92

at its stable abundance was not compromised (Fig. 4). As the surplus was depleted by on-going

habitat degradation, so too was the number of dispersers the large population could produce.

Therefore, habitat degradation within the large flask did not compromise the abundance of the

large population but reduced immigration into surrounding populations. Because small

populations lacked sufficient resources to reproduce, they declined in-step with the gradual loss

of immigrants. Therefore, despite equivalent degradation treatments in large and small

populations, the relatively low population density in the large flask allowed the population to

maintain its reproductive capacity, at least within the time span of this experiment.

Habitat degradation and stochastic events will continue to threaten the persistence

population networks as well as species (MEA 2005, Gaston and Fuller 2008, Butchart et al.

2010; Macusi et al. 2016). Reassuringly, the results of our study suggest that there are not interactive effects of habitat degradation and local population extinction and recolonization on the persistence of spatial population networks, at least for species that have a high rate of reproduction (insects, annual plants, amphibians, small mammals). For slow growing species, the

population may fail to recover before the habitat becomes too degraded to support growth.

Regardless of an organism’s growth rate, our results indicate the opening or restoration of a large

habitat patch could prolong the regional persistence of a species in decline (Doak 1995,

Verboom 2001, Folke et al. 2004).

Acknowledgments

We would like to thank Maqeda Campbell, Muriel LeMaitre, and Ashley Joa for assisting with

data collection.

93

References

Bjørnstad O, Ims RA, Xavier L (1999) Spatial population dynamics: analysing patterns and

processes of population synchrony. Trends Ecol Evol 14:427–432. doi:10.1016/S0169-

5347(99)01677-8

Bender DJ, Contreras TA, Fahrig L (1998) Habitat loss and population decline: a meta-analysis

of the patch size effect. Ecology 72:517-533 doi:10.1890/0012-9658

Benton TG, Lapsley CT, Beckerman AP (2001) Population synchrony and environmental

variation: an experimental demonstration. Ecol Letters 4:236-243 doi: 10.1046/j.1461-

0248.2001.00225.x

Benton TG, Solan M, Travis JM, Sait SM. Microcosm experiments can inform global ecological

problems (2007) Trends Ecol & Evol 22:516-21 doi:10.1016/j.tree.2007.08.003

Biesmeijer JC, Roberts SP, Reemer M, Ohlemüller R, Edwards M, Peeters T, Schaffers AP,

Potts SG, Kleukers R, Thomas CD, Settele J (2006) Parallel declines in pollinators and

insect-pollinated plants in Britain and the Netherlands. Science. 313:351-4.

doi:10.1126/science.1127863

Brown JH, Kodric-Brown A (1977) Turnover rates in insular biogeography: effect of

immigration on extinction. Ecology 58:445-9 doi:10.2307/1935620

Bowers MA, Matter SF (1997) Landscape ecology of mammals: relationships between density

and patch size. J Mammology. dio:10.2307/1383044 78:999-1013

Butchart SH, Walpole M, Collen B, Van Strien A, Scharlemann JP, Almond RE, Baillie JE,

Bomhard B, Brown C, Bruno J, Carpenter KE (2010) Global biodiversity: indicators of

recent declines. Science 328:1164-8 http://dx.doi.org/10.1126/science.1187512

94

Conner E, Courtney AC, Yoder JM (2000) Individuals-area relationships: the relationships

between animal population density and area. Ecology 81:734-748 doi: 10.2307/177373

Doak DF (1995) Source‐sink models and the problem of habitat degradation: general models and

applications to the Yellowstone grizzly. Cons Biol 9:1370-9 doi: 10.1046/j.1523-

1739.1995.09061370.x

Dwernychuk LW, Boag DA (1972) Ducks nesting in association with gulls-An ecological trap?

Can J of Zool 50:559-563 doi:10.1139/z72-076

Earn DJ, Levin SA, Rohani P (2000) Coherence and conservation. Science 290: 1360-1364

doi:10.1126/science.290.5495.1360

Fronhofer EA, Kubisch A, Hilker FM, Hovestadt T, Poethke HJ (2012) Why are

metapopulations so rare? Ecology 93:1967-78 doi:10.1890/11-1814.1

Griffen BP, Drake JM (2008) A review of extinction in experimental populations. J Anim Ecol

77:1274-1287 doi:10.1111/j.1365-2656.2008.01426.x

Fabricius KE, Langdon C, Uthicke S, Humphrey C, Noonan S, De’ath G, Okazaki R,

Muehllehner N, Glas MS, Lough JM (2011) Losers and winners in coral reefs

acclimatized to elevated carbon dioxide concentrations. Nat Clim Change 1:165-9 doi:

10.1038/nclimate1122

Folke C, Carpenter S, Walker B, Scheffer M, Elmqvist T, Gunderson L, Holling CS (2004)

Regime shifts, resilience, and biodiversity in ecosystem management. Annual Rev of

Ecol, Evol, and Systematics 31:557-581 http://www.jstor.org/stable/30034127

Franzén M, Nilsson SG (2010) Both population size and patch quality affect local extinctions

and colonizations. Proc. of the Royal Soc. of London B: Biol. Sci. 277:79-85

doi:10.1098/rspb.2009.1584

95

Fronhofer EA, Kubisch A, Hilker FM, Hovestadt T, Poethke HJ (2012) Why are

metapopulations so rare? Ecology 93:1967-78 doi:10.1890/11-1814.1

Gaston KJ, Fuller RA (2008) Commonness, population depletion and conservation biology.

Trends Ecol & Evol 23:14-9 doi:10.1016/j.tree.2007.11.001

Hanski I (1994) Spatial scale, patchiness and population dynamics on land. Phil Trans Royal Soc

B 343:19-25 doi:10.1098/rstb.1994.0003

Harrison S (1991) Local extinction in a metapopulation context: an empirical evaluation. Biol J

Linnean Soc 42:73-88 doi: 10.1111/j.1095-8312.1991.tb00552.x

Heino M, Kaitala V, Ranta E, Lindström, J (1997) Synchronous dynamics and rates of extinction

in spatially structured populations. Synchronous dynamics and rates of extinctions in

spatially structured populations. Proc Royal Soc Lond B 264:481-486 doi:

10.1098/rspb.1997.0069

Higgins K (2009) Metapopulation extinction risk: dispersal’s duplicity. Theor Pop Biol 76:146-

155 doi:10.1016/j.tpb.2009.05.006

Lamkin MK, Kaneshiro ES, Matter SF (Chapter 3 of this Dissertation) An experimental test of

local population extinction in spatial population networks.

Lamkin MK, Kaneshiro ES, Matter SF (Chapter 4 of this Dissertation) Stochastic extinction of a

large population stabilizes but does not synchronize dynamics of small populations in

experimental networks.

Lande R (1993) Risks of population extinction from demographic and environmental

stochasticity and random catastrophes. Am Nat 142:911-927

doi:http://dx.doi.org/10.1093/besa/15.3.237

96

Li S, Verburg PH, Lv S, Wu J, Li X (2012) Spatial analysis of the driving factors of grassland

degradation under conditions of climate change and intensive use in Inner Mongolia,

China. Regional Env Change 12:461-74 doi: 07/s10113-011-0264-3

Liebhold A, Koenig WD, Bjørnstad ON (2004) Spatial synchrony in population dynamics.

Annual Rev of Ecol, Evol, Systematics 35:467-490. doi:10.1146/annure

Macusi ED, Cuenca GC, Andam MB, Cardona LC (2016) Nutrient Enrichment, Sedimentation,

Heavy Metals and Plastic Pollution in the Marine Environment and its Implications on

Philippine Marine Biodiversity: A Review. IAMURE: Intl J Ecol and Cons 15(1)

Matter SF (2001) Synchrony, extinction, and dynamics of spatially segregated heterogeneous

populations. Ecol Modelling 141: 217-226. doi:10.1016/S0304-3800(01)00275-7

Matter SF (2013) Increases and decreases in network synchrony associated with local population

extinction and recolonization. Ecol Modelling 266:68-72.

doi:10.1016/j.ecolmodel.2013.06.021

Matter SF, Roland J (2009) Local extinction synchronizes population dynamics in spatial

networks. Proc Roy Soc Lond B 277: 729-737 doi: 10.1098/rspb.2009.1520

Matter SF, Roland J (2010) Effects of experimental extinction for the spatial population

dynamics of the butterfly Parnassius smintheus. Oikos 119:1961-1969

doi:10.1111/j.1600-0706.2010.18666.x

Mikami K, Hiwatashi K (1975) Macronuclear regeneration and cell division in Paramecium

caudatum. J Protozool 22: 536-540. doi: 10.1111/j.1550-7408.1975.tb05227.x

Millenium Ecosystem Assessment (MEA). 2005. Ecosystems and Human Well-being: Synthesis.

Island Press.

Moran PA (1953) The statistical analysis of the Canadian lynx cycle. Aust J Zool 1:291-298.

97

Mortelliti A, Amori G, Boitani L (2010) The role of habitat quality in fragmented landscapes: a

conceptual overview and prospectus for future research. Oecologia 163:535-47.

doi:10.1007/s00442-010-1623-3

Ogata M, Hondou T, Hayakawa Y, Hayaski Y, Sugawara K (2008) Adaptation-induced

collective dynamics of a single-cell protozoan. Phys Rev E 77:011917

doi:10.1103/PhysRevE.77.011917

Richter-Dyn N, Goel NS (1972) On the extinction of a colonizing species. Theor Popul Biol

3:406-433 doi:10.1016/0040-5809(72)90014-7

Robert A (2009) The effects of spatially correlated perturbations and habitat configuration on

metapopulation persistence. Oikos 118:1590-600 doi: 10.1111/j.1600-0706.2009.17818.x

Robertson BA, Hutto RL (2006) A framework for understanding ecological traps and an

evaluation of existing evidence. Ecol 87:1075-1085. doi: jstor.org/stable/20069046

Rösch V, Tscharntke T, Scherber C, Batáry P (2015) Biodiversity conservation across taxa and

landscapes requires many small as well as single large habitat fragments. Oecologia

179:209-22. doi:10.1007/s00442-015-3315-5

Royama T (1992) Analytical Population Dynamics, 1st edn. Chapman and Hall, London

doi:10.1007/978-94-011-2916-9

Ruxton GD (2006) The unequal variance t-test is an underused alternative to Student's t-test and

the Mann–Whitney U test. Behav Ecol 17:688-90 doi: 10.1093/beheco/ark016

Sonneborn TM (1970) Methods in Paramecium Res. In: Prescott DM (ed) Methods in Cell Phys

Vol.4. Academic Press Inc, New York, pp. 241-399

Thomas CD, Singer MC, Boughton DA (1996) Catastrophic extinction of population sources in a

butterfly metapopulation. Am Nat 1:957-75. www.jstor.org/stable/2463557

98

Verboom J, Foppen R, Chardon P, Opdam P, Luttikhuizen P (2001) Introducing the key patch

approach for habitat networks with persistent populations: an example for marshland

birds. Biol Cons 100:89-101 doi:10.1016/S0006-3207(00)00210-X

Vuilleumier S, Wilcox C, Cairns BJ, Possingham HP (2007) How patch configuration affects the

impact of disturbances on metapopulation persistence. Theor Population Biol 72:77-85

doi:10.1016/j.tpb.2006.11.001

Wichterman R (1986) The Biology of Paramecium, 2nd edn. Plenum Press, New York. doi:

10.1007/978-1-4757-0372-6

Zar JH (1999) Biostatistical Analysis, 4th edn. Prentice Hall, Upper Saddle River NJ

99

Figure Captions

Figure 1. Daily density estimates (paramecia/mL) for local populations in networks with

extinction (left) and control networks (right) for the 31-day experiment (30 growth periods).

Mean networks synchrony for the six pairwise measures (r≤0 ± SE) provided for each of the

three 10-day periods (pre-extinction, extinction, and recolonization). Transparent grey box in the center of each panel indicates the extinction period (again, no extinction in control networks).

Fluctuations are shown for the large center population (grey lines), surrounding small populations (red and black lines). Circle diagrams above panels denote positions of populations by color.

Figure 2. Daily network abundance (ln(N)±SE, N=4) in networks with extinction (filled circles) and control networks (open circles). Grey panel between days 11-20 denotes the extinction

period (note: no extinction in control networks).

Figure 3. Network growth rate (R±SE, N=4) averaged over five-day intervals in networks with

extinction (black bars) and control networks (grey bars). Circled area notes decreased growth the

last two days of the experiment. P<0.05*, P<0.01**, P<0.001***.

Figure 4. Daily population density (# paramecia/mL) for the large center population (open

circles) and the small outer populations (closed circles) in networks with extinction (A) and

control networks (B). Grey panel between days 11-20 denotes the extinction period. Although

the large population was less dense than small populations during the pre-extinction period, it was much larger than small populations in terms of abundance (560 mL vs 70 mL).

100

Figure 5. Mean network synchrony (r, ±SE; N=4) in the growth of small populations during pre- extinction (black bars), extinction (grey bars) and recolonization (dark grey bars) of the large population in networks with extinction (left ) and control networks (right). Figure 5A shows mean synchrony at the 10-day scale of observation, and 5B shows synchrony at the 5-day scale of observation (i.e., the two consecutive 5-day intervals within each 10-day period). There were no statistical differences between consecutive intervals or between networks with extinction and control networks.

Figure 6. Mean synchrony (r, ±SE; N=4) for each pair of small populations during pre- extinction (black bars), extinction (grey bars) and recolonization (dark grey bars) of the large population in networks with extinction (left) and control networks (right). Circle diagrams above bars denote positions of each pairwise measure. There were no statistical changes in pairwise synchrony during the experiment.

101

Figures

-0.16 ± 0.18 -0.17 ± 0.14 0.19 ± 0.11 -0.15 ± 0.08 -0.11 ± 0.13 0.30 ± 0.18 300

200

100

0 0.30 ± 0.17 0.19 ± 0.13 0.34 ± 0.13 0.19 ± 0.10 -0.16 ± 0.18 0.08 ± 0.08 300

200

100

0 0.14 ± 0.16 0.11 ± 0.15 0.16 ± 0.12 -0.24 ± 0.18 -0.04 ± 0.09 -0.03 ± 0.23 300

200

100

0 0.05 ± 0.14 -0.05 ± 0.18 0.39 ± 0.04 -0.06 ± 0.28 -0.21 ± 0.20 0.19 ± 0.12 300

200

100

0 0 5 10 15 20 25 30 0 5 10 15 20 25 30 Day Day Networks with Extinction Control Networks

Figure 1

102

Networks with Extinction Control Networks 12.5 ) N

12.0

11.5 Network Abundance ln(

11.0

0 5 10 15 20 25 30 Day

Figure 2

Networks with Extinction Control Networks 0.2 *** * *** * *

0.1

0.0 Network Growth -0.1

-0.2

1-5 6-10 11-15 16-20 21-25 26-30 28-30 Days

Figure 3

103

700 Networks with Extinction Small Outer Populations 600 Large Center Population 500 400 300

200 100

0

700 Control Networks (No Extinction) Density (#/mL) Small Outer Populations 600 Large Center Population 500 400 300 200 100 0 0 5 10 15 20 25 30

Day

Figure 4

104

0.8 A Pre-extinction ) r Extinction 0.6 Reestablishment

0.4

0.2

10-Day Synchrony ( 0.0

-0.2

0.8 B Pre-extinction

) Extinction r 0.6 Reestablishment

0.4

0.2

5-Day Synchrony ( 0.0

-0.2

With Extinction Control Networks

Figure 5

105

1.0 1 2 4 5 Pre-Extinction 1 2 4 5 Extinction Reestablishment 0.5

0.0 Synchrony(r)

-0.5

1&2 1&4 1&5 2&4 2&5 4&5 1&2 1&4 1&5 2&4 2&5 4&5

Networks with Extinction Control Networks

Figure 6

106

Chapter 6 Conclusions

Anthropogenic habitat destruction and degradation since 1950 has transformed the natural world at a scale and pace unmatched by any previous period in the Earth’s history (MEA

2005). There is no doubt that previous and on-going anthropogenic extinction of species will distinguish the Holocene epoch from the preceding and subsequent epochs (Braje and Erlandson

2013). Whether or not anthropogenic extinction will generate a mass extinction on par with the previous “big 5”, however, depends on the human capacity to conserve thousands of species that remain at risk of extinction (Barnosky et al. 2011, Kolbert 2014, IUCN 2015). As conservation biologists have known for decades, the best way to conserve biodiversity is through habitat protection and restoration (Myers 1979, Wilson 1988), but international agreements to decrease the loss of biodiversity have proven difficult to uphold (Butchart et al. 2010, Cazzolla Gatti

2016).

Despite unprecedented habitat destruction and degradation over the last several decades, the number of documented extinctions is surprisingly small (≈1,000 including species that are extinct in the wild; IUCN 2015). The estimate that the current rate of extinction is 100-1,000 times higher than the background rate implies that between 1 and 10 named eukaryotic species are going extinct every two days. Lowering the background rate 10-fold as proposed by Pimm et al. (2014) and De Vos et al. (2015) would increase the contemporary rate of extinction to 1,000-

10,000 times higher than the background rate. At this rate, all named eukaryotic species would be extinct in 100 years. Since these estimates do not align with the data on which they are based, lowering the background rate for comparative purposes does not more accurately inform policy makers. Rather, increasing the difference between contemporary and background extinction by lowering the baseline rate ramps rhetoric in a way that could ultimately stymie conservation

107

efforts. Certainly, the more conservative background rate should be used for comparative

purposes when comparisons between contemporary and background extinction are desired.

However, since the number of documented extinctions is low, it may be better to avoid using

documented extinction as evidence that extinction rates are extraordinarily high. Rather, the

status of anthropogenic degradation of biodiversity on Earth may be more appropriately

expressed, as it often is, in terms of the proportion of degraded habitat and the proportion of

species that are threatened with extinction (MEA 2005, Butchart et al. 2010, IUCN 2015).

Because habitats are often destroyed piecemeal, the resulting fragmentation supports

multiple populations that occur within dispersal distance of one another (Hanski 1991). It is well

understood that the time lag between habitat destruction and species extinction, or “”, is broadly attributed to metapopulation dynamics where dispersal among local populations within regional space maintains regional persistence longer than would be expected if the populations existed in isolation (Diamond et al. 1972, Tillman et al. 1994, Hanski 1998). Given the importance of dispersal to the regional persistence of a species, it is reasonable that the loss of a population and its dispersers could influence the dynamics of surrounding populations in a way that could increase the risk of additional extinction, e.g., via reduced immigration and increased synchrony (Thomas et al. 1996, Matter and Roland 2009, 2010). However, in various tests of extinction in experimental protozoan networks, I found that population-level extinction does not necessarily increase the risk of additional extinction through reduction in abundance or synchronization of surrounding populations (Chapters 3, 4). Although extinction reduced and synchronized populations within in homogeneous networks, the reduction was not substantial to increase local or network-wide extinction risk. Furthermore, the synchronization of local dynamics was not network-wide (Chapter 3). Since there was both synchrony and asynchrony

108

within the network during the period of extinction, rescue potential was not compromised (Ben-

Zion et al. 2012). However, the localized increase in synchrony associated with extinction in

homogenous networks led me to predict that (1) extinction of a relatively large population would

reduce the abundance of all remaining populations and (2) through the synchronized reduction,

all populations in the network would have increased synchrony in their dynamics.

Surprisingly, I found no effect of extinction of a large population on the abundance or

synchrony of surrounding populations (Chapter 4). Since populations were not synchronously

reduced, they were not synchronized as if by a correlated forcing (Matter and Roland 2009).

However, the ongoing loss of surplus individuals into the vacant patch was stabilizing (Ruxton et

al. 1997, Wang et al. 2015). Therefore, extinction of the large population decreased the risk of

additional extinction by stabilizing local dynamics (Lande 1993). Since the expectation of

reduced abundance and increased synchrony in surrounding populations relied on the large

population being a persistent source of immigrants into surrounding populations, the results of

this experiment challenged the assumption of constant dispersal used in many theoretical

systems. Ecologists have recognized a tendency for relatively lower emigration rates, but not

number of emigrants, from large versus small populations (e.g., due to a low edge-to-area ratio:

Kareiva 1985, Matter 1996, Englund and Hambäck 2004), but the emigration rate indicated from preliminary migration experiments did not support a tendency for low migration from the large population in our system (Chapter 4, Fig. 1). Although variability in the dispersal potential of populations through time is rarely considered (but see Bowler and Benton 2005), our finding of no effect of extinction on the abundance of surrounding populations is feasible in natural systems where large, low-density populations have low dispersal rates. For fast-growing species, the effect of high dispersal rates following recovery from disturbance is feasible in natural systems

109

as well (Bowler and Benton 2005). Therefore, for species with a high reproductive rate, the

benefit of increased abundance following exponential growth during recovery from extinction

can out-weigh the cost of extinction, at least temporarily.

That recolonization may outweigh the cost of extinction led me to test the hypothesis that

extinction and recolonization within declining networks could possibly be beneficial in the short-

term (Chapter 5). In response to lessons learned in Chapter 4, I maintained the dispersal potential

of the large population by provisioning it with a more concentrated medium than was used to

provision surrounding populations that were small. In effect, extinction of the large population

resulted in immediate, dramatic abundance reductions in surrounding populations (Chapter 5,

Figs. 2-3). Surrounding populations did not synchronously increase in response to the reductions,

however, because habitat degradation reduced the reproductive capacity of the small populations.

Therefore, as can occur in natural systems (especially for long-lived species), habitat degradation

quickly reduced reproduction but not survival (Williams et al. 1993, Anouk Simard et al. 2008).

The large population maintained its reproductive capacity despite habitat degradation because it

maintained a reduced density compared to what the bacteria in the habitat could support (Chapter

5, Fig. 2). This result indicates that large, low-density populations that commonly occur within spatial population networks (Bowers and Matter 1997, Fronhofer et al. 2012) may have a delayed response to habitat degradation. Although the true effect of habitat degradation on small populations could be masked by immigration, persistent reductions in small populations could

indicate regional habitat degradation despite stable abundance in large, low-density populations.

Collectively, these results do not rule out the possibility of local population extinction as a mechanism of synchrony. Rather, I have shown where the capacity of extinction to increase synchrony in population networks is limited. For example, extinction of a population will not

110

decrease network persistence through the synchronization of local dynamics when the population

going extinct is not a consistent and sustained source of immigrants to surrounding populations

(Chapters 3, 4) or when remaining populations are unable to increase in numbers in response to

the loss of immigrants (Chapter 5). Furthermore, assuming some degree of non-uniformity in the

distribution of dispersers, a high rate of global dispersal following a burst of exponential growth

during recovery from extinction can bolster surrounding populations without synchronizing

them. Because this boost in abundance is temporary, these results by no means suggest

extinction-recolonization dynamics as a potential conservation strategy. Rather, they emphasize

the importance of habitat restoration and population recovery for the sustainability of population

networks and the persistence of the multitudes of species at risk of extinction (Tilman et al. 1994,

Hanski and Ovaskainen 2002, Ewers and Didham 2006).

References

Anouk Simard M, Côté SD, Weladji RB, Huot J (2008) Feedback effects of chronic browsing on

life‐history traits of a large herbivore. J Animal Ecology. 77:678-86 doi:10.1111/j.1365-

2656.2008.01374.x.

Barnosky AD, Matzke N, Tomiya S, Wogan GO, Swartz B, Quental TB, Marshall C, McGuire

JL, Lindsey EL, Maguire KC, Mersey B (2011) Has the Earth/'s sixth mass extinction

already arrived? Nature 471:51-7 doi:10.1038/nature09678.

Ben-Zion Y, Fried Y, Shnerb NM (2012) Migration, coherence and persistence in a fragmented

landscape. Theoretical Ecology 5:481-93 doi:10.1007/s12080-011-0140-2.

Bowers MA, Matter SF (1997) Landscape ecology of mammals: relationships between density

and patch size. J Mammalogy 78:999-1013 doi:10.2307/1383044.

111

Bowler DE, Benton TG (2005) Causes and consequences of animal dispersal strategies: relating

individual behaviour to spatial dynamics. Biol Reviews 80:205-25

doi:10.1017/S1464793104006645.

Braje TJ, Erlandson JM (2013) Human acceleration of animal and plant extinctions: A Late

Pleistocene, Holocene, and Anthropocene continuum. Anthropocene 4:14-23

doi:10.1016/j.ancene.2013.08.003.

Butchart SH, Walpole M, Collen B, Van Strien A, Scharlemann JP, Almond RE, Baillie JE,

Bomhard B, Brown C, Bruno J, Carpenter KE (2010) Global biodiversity: indicators of

recent declines. Science 328:1164-8 doi:10.1126/science.1187512.

Cazzolla Gatti R (2016) Trends in human development and environmental protection. Intl J Env

Studies 73:268-76 doi:10.1080/00207233.2016.1148447.

DeVos JM, Joppa LN, Gittleman JL, Stephens PR, Pimm SL. 2014. Estimating the normal

background rate of species extinction. Conservation Biology 29: 452-462

doi:10.1111/cobi.12380.

Diamond JM (1972) Biogeographic kinetics: estimation of relaxation times for avifaunas of

southwest Pacific islands. Proc Nat Acad Sci 69:3199-203.

Englund G, Hambäck PA (2004) Scale dependence of emigration rates. Ecology 85:320-7.

doi:10.1890/03-3039.

Ewers RM, Didham RK (2006) Confounding factors in the detection of species responses to

. Biol Reviews 81:117-42. doi:10.1017/S1464793105006949.

Fronhofer EA, Kubisch A, Hilker FM, Hovestadt T, Poethke HJ (2012) Why are

metapopulations so rare? Ecology 93:1967-78 doi:10.1890/11-1814.1

112

Hanski I (1991) Single-species metapopulation dynamics: concepts, models, and observations.

Biol J Linn Soc 42:17-38.

Hanski (1998) Metapopulation dynamics. Nature 396:41-9 doi:10.1038/23876.

Hanski I, Ovaskainen O (2002) Extinction debt at extinction threshold. Conservation Biology

16:666-73 doi:1 0.1046/j.1523-1739.2002.00342.x.

Heino M, Kaitala V, Ranta E, Lindström, J (1997) Synchronous dynamics and rates of extinction

in spatially structured populations. Synchronous dynamics and rates of extinctions in

spatially structured populations. Proc Roy Soc Lond B 264:481-486.

doi:10.1098/rspb.1997.0069.

IUCN (2015) The IUCN Red List of Threatened Species. Version 2015-4.

Kareiva P (1985) Finding and losing host plants by Phyllotreta: patch size and surrounding

habitat. Ecology. 1:1809-16 http://www.jstor.org/stable/2937376.

Kolbert, E (2014) The Sixth Extinction: An Unnatural History. Bloomsbury Publishing

Matter SF (1996) Interpatch movement of the red milkweed beetle, Tetraopes tetraophthalmus:

individual responses to patch size and isolation. Oecologia 105:447-453.

Matter SF, Roland J (2009) Local extinction synchronizes population dynamics in spatial

networks. Proc Roy Soc Lond B 277: 729-737 doi:10.1098/rspb.2009.1520.

Matter SF, Roland J (2010) Effects of experimental extinction for the spatial population

dynamics of the butterfly Parnassius smintheus. Oikos 119:1961-1969

doi:10.1111/j.1600-0706.2010.18666.x.

Millenium Ecosystem Assessment (MEA) (2005) Ecosystems and Human Well-being:

Synthesis. Island Press.

113

Pimm SL, Jenkins CN, Abell R, Brooks TM. (2014) The biodiversity of species and their rates of

extinction, distribution, and protection. Science 344:987-997

doi:10.1126/science.1246752.

Richter-Dyn N, Goel NS (1972) On the extinction of a colonizing species. Theor Popul Biol

3:406-433. doi:10.1016/0040-5809(72)90014-7.

Ruxton GD, Gonzalez-Andujar JL, Perry JN (1997) Mortality during dispersal and the stability

of a metapopulation. J Theor Biol 186:389-96. doi:10.1006/jtbi.1997.0402.

Tilman D, May RM, Lehman CL, Nowak MA (1994) Habitat destruction and the extinction debt.

Nature 371:65-66 doi:10.1038/371065a0.

Wang S, Haegeman B, Loreau M (2015) Dispersal and metapopulation stability. PeerJ.

1;3:e1295 doi:10.7717/peerj.1295.

Williams TD, Cooch EG, Jefferies RL, Cooke F (1993) Environmental degradation, food

limitation and reproductive output: juvenile survival in lesser snow geese. J Animal Ecol

1:766-77 http://www.jstor.org/stable/5395.

Wilson EO (1988) The current state of biological diversity. Pages 3-18 in Biodiversity. Wilson

EO and Peter FM eds. National Academy Press.

114

Supplemental Information (SI) I: Tables

Table 1. ANOVA shows no effect of block on the difference in the density of small (outer) populations between the pre-extinction and extinction periods in 4XL networks. ANOVA Source of SS df MS F P-value Variation Between 265.1 1 265.14 3.684 0.103 Blocks Within 431.8 6 71.97 Blocks

Total 696.9 7 337.11

Table 2. ANOVA shows no effect of block on the difference in the density of small (outer) populations between the extinction and reestablishment periods in 4XL networks. ANOVA Source of SS df MS F P-value Variation Between 69 1 69.03 0.367 0.567 Blocks Within 1128 6 188.08 Blocks

Total 1197 7 257.11

Table 3. ANOVA shows no effect of block on the difference in the synchrony of small (outer) populations between the pre-extinction and extinction periods in 4XL networks. ANOVA Source of SS df MS F P-value Variation Between 0.0076 1 0.0076 0.19 0.678 Blocks Within 0.2340 6 0.0400 Blocks

Total 0.2416 7 0.4076

115

Table 4. ANOVA shows no effect of block on the difference in the synchrony of small (outer) populations between the extinction and reestablishment periods in 4XL networks. ANOVA Source of SS df MS F P-value Variation Between 0.1628 1 0.1628 2.791 0.146 Blocks Within 0.3500 6 0.0583 Blocks

Total 0.5128 7 0.2211

Table 5. ANOVA shows no effect of block on the difference in the stability of small (outer) populations between the pre-extinction and extinction periods in 4XL networks. ANOVA Source of SS df MS F P-value Variation Between 17.31 1 17.31 0.431 0.536 Blocks Within 241.04 6 40.17 Blocks

Total 258.35 7 57.48

Table 6. ANOVA shows no effect of block on the difference in the stability of small (outer) populations between the extinction and reestablishment periods in 4XL networks. ANOVA Source of SS df MS F P-value Variation Between 163.1 1 163.05 3.238 0.122 Blocks Within 302.1 6 50.35 Blocks

Total 465.2 7 213.4

116

Table 7. ANOVA shows no effect of block on the difference in the density of small (outer) populations between the pre-extinction and extinction periods in 8XL networks. ANOVA Source of SS df MS F P-value Variation Between 82.4 1 82.39 0.812 0.402 Blocks Within 608.6 6 101.43 Blocks

Total 691.0 7 183.82

Table 8. ANOVA shows no effect of block on the difference in the density of small (outer) populations between the extinction and reestablishment periods in 8XL networks. ANOVA Source of SS df MS F P-value Variation Between 7.9 1 7.9 0.016 0.903 Blocks Within 2961 6 493.5 Blocks

Total 2969 7 501.4

Table 9. ANOVA shows no effect of block on the difference in the synchrony of small (outer) populations between the pre-extinction and extinction periods in 8XL networks. ANOVA Source of SS df MS F P-value Variation Between 0.0380 1 0.0380 0.417 0.542 Blocks Within 0.5466 6 0.0911 Blocks

Total 0.5846 7 0.1291

117

Table 10. ANOVA shows no effect of block on the difference in the synchrony of small (outer) populations between the extinction and reestablishment periods in 8XL networks. ANOVA Source of SS df MS F P-value Variation Between 0.0190 1 0.0109 0.226 0.651 Blocks Within 0.2890 6 0.0482 Blocks

Total 0.3080 7 0.0591

Table 11. ANOVA shows no effect of block on the difference in the stability of small (outer) populations between the pre-extinction and extinction periods in 8XL networks. ANOVA Source of SS df MS F P-value Variation Between 0.0104 1 0.1037 2.01 0.206 Blocks Within 0.0310 6 0.0052 Blocks

Total 0.0414 7 0.1089

Table 12. ANOVA shows no effect of block on the difference in the stability of small (outer) populations between the extinction and reestablishment periods in 8XL networks. ANOVA Source of SS df MS F P-value Variation Between 0.0139 1 0.0139 4.491 0.0784 Blocks Within 0.0185 6 0.0031 Blocks

Total 0.0324 7 0.0170

118

Supplemental Information (SI) II: Figures

SI Fig. 1. Picture of 4XL and 8XL networks. The large population in 4XL and 8XL networks occupy a 250 mL flask and 500 mL flask, respectively.

0.6

0.5

0.4 ) r 0.3

0.2 Sy nchrony (

0.1

0.0

-0.1 1XL 4XL 8XL

SI Fig. 2. Mean synchrony of four outer populations (i.e., mean of six pairwise measures) over a 10-day period when the center population was the same size as outer populations (1X), four times larger than outer populations (4X), and eight times larger than outer populations (8X).

119

600 Center Population Outer Populations 500

400

300 Density

200

100

0 1XL 4XL 8XL

SI Fig. 3. Mean population density (#paramecia/mL ± SE, N=4 networks) of center (black bars) and outer (grey bars) populations over a 10-day period when the center population is the same volume as outer populations (1X), four times larger than outer populations (4X), and eight times larger than outer populations (8X). When the center population is eight times larger than outer populations, the mean density of the large population is reduced.

120