spatial synchrony of population fluctuations: causes and consequences
DESCRIPTION
Spatial synchrony of population fluctuations: causes and consequences. Jeremy Fox University of Calgary Website: homepages.ucalgary.ca/~jefox/Home.htm Blog: dynamicecology.wordpress.com. Collaborator: David Vasseur, Yale University. - PowerPoint PPT PresentationTRANSCRIPT
Spatial synchrony of population fluctuations: causes and consequences
Jeremy FoxUniversity of Calgary
Website: homepages.ucalgary.ca/~jefox/Home.htm
Blog: dynamicecology.wordpress.com
With thanks to: Tara Janes, Jessica Scharein, Joyce MacNeil,Stephen Hausch, Jodie Roberts, Geoff Legault
Collaborator: David Vasseur, Yale University
"An odd kind of sympathy": Huygens' clocks
Synchrony
Spatial synchrony in population ecology
Lynx Gypsy moth
0
10
1994 1995 1996 1997 1998 1999 2000
Year
Lem
min
g a
bu
nd
ance
ind
ex Collared lemming
Measles
Blasius et al. 1999, Johnson et al. 2006, Rohani et al. 1999, Paradis et al. 2000, Krebs et al. 2002
Wren
Causes of spatial synchrony
•Dispersal
•Spatially-synchronous environmental fluctuations (Moran effect)
•Interspecific interactions
Stochastic predator-prey model
N1
P1
N2
P2
Dispersal
Patch 1 Patch 2
iPijiPi
iii
iNiji
iiiNii
i
PtPPdPtmhN
PeaN
dt
dP
NtNNdhN
PaNNtmNN
dt
dN
ii
ii
1
Envi. flucts. Envi. flucts.
Growth Mortality Predation DispersalDemogr.stochas.
Model predictions for prey synchrony
•Dispersal is synchronizing
•Moran effect is synchronizing
•Predation increases the synchronizing effect of dispersal
Vasseur & Fox 2009 Nature
Sync. envi., + dispersal
Time (arbitrary units)
Mod
el p
rey
dens
ity
Sync. envi., no dispersal
Patch 1 Patch 2
Predator-prey oscillations are synchronized (‘phase locked’) by dispersal
• No predatorsno cycleslittle effect of dispersal
Summary of model predictions
•Dispersal is synchronizing
•Moran effect is synchronizing
•Predators that generate oscillations greatly increase the synchronizing effect of dispersal
-Statistical signature of phase locking
Protist microcosm experiment
•Prey: Tetrahymena pyriformis
•Predator: Euplotes patella
•Microcosms: 80 ml, semi-continuous cultures
•Small samples taken on weekdays
•Dispersal of 10% of individuals, 3x/week
•Daily temperature fluctuations (independent or perfectly synchronous)
•6 replicate bottle pairs/ttmt. combination
•Experimental units: pairs of bottles
•2x2x2 factorial design crossing pres./abs. of dispersal, Moran effect, predator
Conducting dispersal events
102
103
104
0 9 18 27 36 45 54 63
2030
Tem
p. (°C)T
et./
ml
(log
scal
e)
Day
Illustrative population dynamics
Day
0
600
1200
0 9 18 27 36 45 54 630
10
20
30
Tem
p. (°C)
Eupl./m
lTet
./m
l
Experimental results vs. model predictions
Vasseur & Fox 2009 Nature
Phase-locked oscillations
0 630
700
Day
Tet
rahy
men
a/m
l Patch 1Patch 2
102
103
104
0 9 18 27 36 45 54 63
2030
Tem
p. (°C)T
et./
ml
(log
scal
e)
Day
Prey densities did not track temperature fluctuations
Summary so far
Synchrony
Dispersal Moran effect
Species interactions
Population dynamics(cyclic vs.
not)
Lynx
Phase drift at low dispersal rates: data
Day
Pre
y de
nsity
(m
l-1)
Fox et al. in press Plos One
Phase drift at low dispersal rates: model
Fox et al. in press Plos One
Scaling up
Synchrony usually decays with distance
Syn
chro
ny
Distance between populations
Ranta et al. 1995
•Links between pattern of decay and underlying mechs.?
Questions
• Why does synchrony decay with distance?– Decay of environmental synchrony– Limited dispersal distance
• Phase locking across long distances?
Methods
1 2 3 4 5 6•Exptl. units:
•2 x 2 factorial design (y/n Moran effect, y/n dispersal)
•Stepping-stone dispersal
•Moran effect with spatially-decaying synchrony
•Predators + prey
Illustrative prey dynamicsLo
g(T
etra
./m
l + 1
)
Time
+M +D +M -D -M +D -M -D
Moran Disp. n n y n n y y y
Prey synchrony
0
0.9
1.8
1 2 3 4 5Spatial lag
Mea
n pr
ey s
ynch
rony
±S
E
•High mean sync. (init. conds.)•Higher sync. at even lags (init. conds.)
+ dispersal
- dispersal
•Dispersal increases sync.•Same effect at all lags (phase locking)•Moran eff. increases short-distance sync.•Spat. decay of sync. in +Moran ttmts.•No Moran x disp. interaction
Fox et al. 2011 Ecol. Lett.
Take-home points
• Dispersal generates long-distance phase locking• Distance-decay of synchrony due to Moran effect
– Same likely true in many natural systems– Short-distance dispersal either phase-locks cycles, or
produces little synchrony at all
Summary:Spatial predator-prey cycles work like this:
Consequences of synchrony for metapopulation persistence:
the spatial “hydra effect”
The “hydra effect”
The usual story: intermediate dispersal rates maximize metapopulation persistence
Met
apop
ulati
on p
ersi
sten
ce ti
me
Dispersal rateZero/low Intermediate High
Indep. patches(async.)
Coloniz.-extinction(async.)
“One big patch”(sync.)
Big patch persistent
Big patch extinction-prone
Yaari et al. 2012
Intermediate dispersal rates maximize metapopulation persistence
Huffaker 1958
Intermediate dispersal maximizes metapopulation persistence
Holyoak and Lawler 1996:
A puzzle: How are asynchronous colonization-extinction dynamics possible?
An answer: A spatial hydra effect
Local extinctions are desynchronizing
• Anything that reduces synchrony promotes recolonization, and thus persistence
• Empirical examples of colonization-extinction dynamics involve extinction-prone subpopulations
• Empirical examples of synchrony at low dispersal rates involve persistent subpopulations
An illustration of the spatial hydra effect
• Nicholson-Bailey host-parasitoid model with demogr. stochas. (Yaari et al. 2012)
• 4 patches
• Global density-independent dispersal of both spp. after births & deaths
• At end of timestep: random subpop. destruction
Subpopulation dynamics under low dispersal, no subpop. destruction
Subpopulation dynamics under intermediate dispersal, no subpop. destruction
0 50 100 150
05
00
10
00
15
00
Index
n.h
[, 1
]
Timestep
Hos
t su
bpop
ulat
ion
abun
danc
e
0 10 20 30 40
01
00
02
00
03
00
04
00
0
Index
n.h
[, 1
]
Subpopulation dynamics under high dispersal, no subpop. destruction
Timestep
Hos
t su
bpop
ulat
ion
abun
danc
e
0 10 20 30 40 50 60
01
00
20
03
00
40
05
00
60
0
Index
n.h
[, 1
]
Subpopulation dynamics under high dispersalwith random subpopulation destruction
Timestep
Hos
t su
bpop
ulat
ion
abun
danc
e
0
90
0.0001 0.001 0.01 0.1 1
Dispersal rate (log scale)
Met
apop
ulat
ion
pers
iste
nce
time
(mea
n)
Subpopulationdestruction rate
00.0250.50.0750.1
A spatial hydra effect
Hydra effect summary
• Hydras are real
• Effect can vary in strength, be swamped by other effects-Matter & Roland 2010 Proc Roy Soc B
• Biological details only matter via effects on colonization and extinction rates
-local extinctions affect coloniz. rate via effect on synchrony
Really exists.
Future directions• Interplay of determinism and stochasticity• Embedding of Euplotes-Tet. cycle in larger food webs• Environmental heterogeneity• Larger spatial arrays?• Hydra effect under different forms of envi. stochasticity• Comparisons with nature
-changes in synchrony as cycles collapse?
0
800
0 1 0 1
Dispersal rate
Mea
n m
etap
op.
pers
ist.
tim
e
Stochastic Ricker Stochastic logistic map
0
0.025
0.05
0.075
0.1
Destruct. rate
Weak spatial hydra effect
0 2 4 6 8 10 12
0.0
0.2
0.4
0.6
0.8
1.0
Dispersal rate (% per event)
Pre
y sy
nch
ron
y
Even low dispersal rates can rapidly synchronize cycling populations
Fox et al. unpublished
Prey synchrony vs. dispersal rate
Fox et al. in press Plos One
Data analysis
1 2 3 4 5 6
r(1,2)
1. Calculate prey synchrony (cross-correl. of log abundance) for every pair of jars in an array
-predator abundances too noisy to analyze
r(3,6)
2. Calculate mean sync. at every spatial lag within an array-vector of 5 cross-correl. coeffs.
3. z-transform to normalize
4. MANOVA for treatment effects, follow-up ANOVAs
5. Spatial decay: regress z-transformed cross-correlation on spatial lag, ANOVA on slopes
Day
Log(
Eup
lote
s/m
l +1)
Illustrative predator dynamics
0
2
4
0 25 50Day
Mea
n E
uplo
tes/
ml
0
1000
2000
3000
4000
5000
0 1 2 3 4 5 6
Day
Tetrahymena
/ml
No dispersal
Direct demonstration of dispersal-generated phase locking
-2000
0
2000
4000
0 1 2 3 4 5 6
Day
De
ns
ity
dif
fere
nc
e
No dispersal
0
1000
2000
3000
4000
5000
0 1 2 3 4 5 6
Day
Tetrahymena
/ml + dispersal
-2000
0
2000
4000
0 1 2 3 4 5 6
Day
De
ns
ity
dif
fere
nc
e
+ dispersal
Desync. Sync. Little desync.
“Leading”patches
“Trailing”patches
Phase drift at the cycle nadir in the absence of dispersal
Illustrative examples of prey synchrony
0 630
1000
Day
Tet
./m
lIndep. envi., no disp.
No
pred
ator
s
0
1000
0 63Day
Tet
./m
l
+ p
reda
tors
0 630
1400
Day
Tet
./m
l
Sync. envi., + disp.
0 630
700
Day
Tet
./m
l
-pred. +pred.
Mod
el p
rey
sync
hron
y
Indep. envi. Sync. envi.
-pred. +pred.
No disp.Pred. disp.Prey disp.Both disp.
Dispersal × predator interaction not due to prey tracking synchronized predators
DispersalNo disp.
Predator synchrony
Vasseur & Fox 2009 Nature
Robust qualitative match between model and data
Monte Carlo simulns.Exptl. data
Vasseur & Fox 2009 Nature
Monte Carlo simulations: ANOVA main effects
Monte Carlo simulations: ANOVA interaction terms
-pred. +pred.
Mod
el p
rey
sync
hron
y
Indep. envi. Sync. envi.
-pred. +pred.rand. mort.
rand. mort.
No dispersalDispersal
Dispersal × predator interaction not due to increased prey variability in presence of predators
-1.2
0
0.6
0 25 50
Day
Det
rend
ed lo
g pr
ey d
ensi
ty 0
2 (=0)
Estimating prey cycle phase
Dispersal entrains the phases of predator-prey cycles(and the Moran effect doesn’t)
0
/2
3/2
0
/2
3/2
0
/2
3/2
0
/2
3/2
+M +D
-M +D
+M -D
-M -D
Fox et al. 2011 Ecol. Lett.
Var
ianc
e in
pha
se0.05
0.20
- dispersal + dispersal
- Moran eff.+ Moran eff.
Predators generated prey oscillations
No
rma
lize
d sp
ect
ral p
ow
er
Frequency (1/d)
Predator-prey cycle
Illustrative prey power spectrum