advances.sciencemag.org/cgi/content/full/4/7/eaar4349/DC1

Supplementary Materials for

Temperature-dependent adaptation allows fish to meet their food across their

species’ range

Anna B. Neuheimer*, Brian R. MacKenzie, Mark R. Payne

*Corresponding author. Email: abneuheimer@gmail.com

Published 25 July 2018, Sci. Adv. 4, eaar4349 (2018)

DOI: 10.1126/sciadv.aar4349

This PDF file includes:

Table S1. Observed and estimated predator timing and prey timing used in this study. Box S1. A description of the TDF model. Box S2. Examples of using the TDF model to estimate predator timing. Fig. S1. Illustrating MMH. Fig. S2. Predator-prey match across a species’ range with source-specific estimates. Fig. S3. Predator-prey match across a species’ range for northern populations. Fig. S4. Predator-prey match across a species’ range using alternate prey species. Fig. S5. Predator-prey match across a species’ range using observed predator-prey stages. Fig. S6. Predator-prey match across a species’ range using timing estimates based on constant SD predictions. Fig. S7. Illustration of method to estimate timing of unobserved stages via the TDF metric. Fig. S8. Temperature-dependent SDs for Atlantic cod. Fig. S9. Intra-annual variation in population-specific temperatures estimated from temperature observations between 5- and 100-m depth. References (63, 64)

SUPPLEMENTARY TABLE

Table S1. Observed and estimated predator timing and prey timing used in this study. Also

given are Northwest Atlantic Fisheries Organization (NAFO) and International Council for the

Exploration of the Sea (ICES) area descriptors by population.

Population Code Area

Predator timing Prey timing

Observed -

Spawning time

(ICES 2005)42

with mean

spawning time

(spawning

time range)

Estimated -

First-feeding

larvae with

mean timing

and estimated

variability due

to variability

in temperature

phenology

Observed -

copepodite stages

C1-C3 with

among study

range in observed

timing. HL = Heath & Lough

(2007)10;

A = Anderson

(1990)21; M = Melle

et al (2014)22

Estimated -

naupliar stage

N3 with mean

timing and

estimated

variability due

to variability in

temperature

phenology.

Georges Bank GEO NAFO

5Zj, 5Zm

61

(305-151)

83

(79-88) 128 M

109

(103-113)

Gulf of Maine GOM NAFO

5Y

76

(305-151)

102

(99-106) 128 M -135HL

109

(104-112) spring-spawning

western Scotian

Shelf WSS1

NAFO

4X

61

(32-151)

86

(76-101) 98M-135HL 93

(73-101)

Northern Gulf

of St. Lawrence NSL

NAFO

3Pn 4RS

135

(91-181)

171

(164-182) 197M

167

(132-181)

South

Newfoundland SNL

NAFO

3PS

121

(60-243)

155

(148-164) 166 HL -174M

137

(114-149)

Grand Banks GB NAFO

3NO

129

(91-181)

154

(145-171) 166 HL -174M

147

(116-157)

Flemish Cap FC NAFO

3M

63

(32-120)

75

(69-87) 105 HL-126A

102

(89-106) southern

Labrador &

eastern

Newfoundland

LAB NAFO 2J,

3KL

113

(60-212)

154

(147-163) 189M -196HL

153

(121-170)

West Greenland

offshore WGO NAFO 1*

105

(60-181)

138

(131-146) 135HL -189M

127

(112-136)

West Greenland

inshore WGI NAFO 1**

90

(32-212)

126

(124-128) 135HL -189M

125

(120-130)

Iceland ICE ICES Va 105

(60-151)

124

(122-126) 135HL-174M

133

(124-138)

Faroe Plateau FP ICES VbI 90

(32-151)

107

(105-109) 135HL-144M

124

(123-126)

Northeast Arctic NEA ICES I &

II

92

(32-151)

113

(109-117) 105HL-128M

97

(92-100)

Western Baltic

Sea WBS

ICES 21

to 24

76

(32-120)

104

(100-107) 98M

68

(61-74)

North Sea NS ICES IV 49 67 105HL-123M 98

& VIId &

IIIa

(1-120) (65-70) (95-99)

Irish Sea IRS ICES

VIIa

90

(30-151)

107

(106-108) 135HL

119

(119-121)

Celtic Sea CEL ICES

VIIe to k

83

(32-120)

94

(93-95) 105HL

95

(93-95)

* restricted to 59-70°N as per ICES (2005)42

** restricted to 62.5-69°N as per ICES (2005)42

SUPPLEMENTARY FIGURES

Box S1. A description of the TDF model. An illustration comparing methods of estimating

development time (after 63). (a) When an individual’s progression through a stage is tracked in

time (Age, e.g. days), an individual advances to its next life-history stage when Age equals the

estimated constant stage duration (SD). However, in a dynamic environment, development rate

(and SD) may change from one time-step to another due to e.g. changing temperature, leading to

unrealistic development timing (e.g. developing too early) if constant stage durations are assumed.

(b) Instead, an individual's progression through a stage may be tracked with the thermal

development fraction (TDF, after 38, 63, 64), a proportional measure of age that allows for

temporally varying development rates (due to e.g. temperature, food) and realistic development

timing.

The Thermal Development Fraction (TDF) Model

Box S2. Examples of using the TDF model to estimate predator timing. Using the TDF metric to estimate larval fish timing from spawning time estimates. Examples are shown for the Irish Sea (IRS, cyan), western Scotian Shelf (WSS1, green) and western Greenland inshore (WGI, purple) populations. 1) Population-specific daily temperature is estimated and temperatures post-spawning are isolated. 2) Stage duration variation in time is determined from daily temperatures e.g. via eqn. 2. 3) The daily thermal development fraction is estimated as the ratio between the time-step (∆t, here 1 day) and the variable stage duration (days). 4) The TDF is tracked (cumulative sum) from spawning. The timing of larval fish is estimated as the day of year when TDF ≥ 1.

Thus, larval fish timing estimates depend both on spawning time as well as the temperature-dependent stage durations such that populations with similar spawning times (e.g. in 1, WGI & IRS, spawning time = day of year 90), can lead to variability in larval fish timing (e.g. in 4, WGI & IRS larval fish timing = day of year 126 and 107 respectively) due to temperature-dependent development. Similar methods were used to estimate prey (naupliar) timing by using a TDF metric to back-calculate larval copepod timings from observations of juveniles (see Materials & Methods).

Fig. S1. Illustrating MMH. The Match-Mismatch Hypothesis (MMH; 1,2) states that population

abundance and production (grey) varies with the overlap (purple) of larval fish (red) and their

prey (blue). (a) Years with high overlap between larvae and their prey results in more energy

transfer between the trophic levels and higher fish biomass later in life. (b) Mismatches in larval-

prey timing results in less energy transfer between trophic levels and lower fish biomass later in

life. In the latter case, a higher proportion of energy is recycled among the lowest trophic levels

and/or leaves the system (e.g. sinking out of the water column).

Fig. S2. Predator-prey match across a species’ range with source-specific estimates.

Estimates of larval copepod (N3 stage) and fish larvae (first-feeders) timing for populations of

Atlantic cod (Gadus morhua) across the north Atlantic. Estimates of N3 stage copepods and

first-feeding fish larvae timing are made via TDF metrics (see Materials and Methods) using

observations of copepodite timing (via 10, purple circle; 21, orange square and 22, teal triangle) and

spawning time42 respectively. Estimates are made using stage durations based on population-

specific mean daily temperature estimates (data points with error-bars ± 1 standard deviation

from mean temperature phenology; data labels refer to populations in Fig. 1). Also given is the

standard major axis (SMA) line-fit (solid line, with 95% confidence intervals around the slope as

dashed lines; P<0.001; R2=0.52; Slope not different from 1: P=0.38; Intercept not different from

0: P=0.52). The 1:1 line (i.e. slope of 1, intercept of 0) is given in the solid grey line.

CEL

FCFC

FPFP

GBGB

GEO

GOMGOM

ICEICE

IRS

LABLAB

NEANEA

NSNS

NSL

SNLSNL

WBS

WGIWGI

WGOWGO

WSS1 WSS1

45

70

95

120

145

170

195

45 70 95 120 145 170 195

Mean prey timing (day of year)

Mea

n p

red

ato

r tim

ing (

day o

f ye

ar)

Heath & Lough 2007

Anderson 1990

Melle et al. 2014

Fig. S3. Predator-prey match across a species’ range for northern populations. Estimates of

larval copepod (N3 stage) and fish larvae (first-feeders) timing for populations of Atlantic cod

(Gadus morhua) for northern populations which feed primarily on Calanus finmarchicus based

on a) stomach contents data10 or b) habitat modelling57. Estimates of N3-stage copepods and

first-feeding fish larvae timing are made via TDF metrics (see Materials and Methods) based on

observations of copepodite timing10,21,22 and spawning time42 respectively. Estimates are made

using stage durations based on population-specific mean daily temperature estimates (data points;

data labels refer to populations in Fig. 1). Also given are the standard major axis (SMA) line-fits

(solid line, with 95% confidence intervals around the slope as dashed lines; a: P=0.021; R2=0.62;

Slope not different from 1: P=0.64; Intercept not different from 0: P=0.71; b: P=0.0019; R2=0.82;

Slope not different from 1: P=0.73; Intercept not different from 0: P=0.14). The 1:1 line (i.e.

slope of 1, intercept of 0) is given in the solid grey line.

GB

FP

NEA

LAB

ICE

SNL

WGI

WGO

a

45

70

95

120

145

170

195

45 70 95 120 145 170 195

Mean prey timing (day of year)

Mea

n p

red

ato

r tim

ing (

day o

f ye

ar)

GB

NEA

LAB

ICE

SNL

NSL

WGI

WGO

b

45

70

95

120

145

170

195

45 70 95 120 145 170 195

Mean prey timing (day of year)

Mea

n p

red

ato

r tim

ing (

day o

f ye

ar)

Fig. S4. Predator-prey match across a species’ range using alternate prey species. Estimates

of larval copepod (N3 stage) and fish larvae (first-feeders) timing for populations of Atlantic cod

(Gadus morhua). Estimates of N3 stage copepods and first-feeding fish larvae timing are made

via TDF metrics (see Materials and Methods) based on observations of copepodite timing10,21,22

and spawning time42 respectively. Estimates are made using stage durations based on population-

specific mean daily temperature estimates (data points with error-bars ± 1 standard deviation

from mean temperature phenology; data labels refer to populations in Fig. 1). Prey timing is

estimated from observed timing of a) Calanus finmarchicus for populations where mean annual

temperatures are <7˚C and Pseudocalanus spp. for populations where mean annual temperatures

are >7˚C57; and b) Calanus finmarchicus for populations where mean annual temperatures are

<7˚C and averaged across species (C. finmarchicus, Pseudocalanus spp.) for populations where

mean annual temperatures are >7˚C57; Also given are standard major axis (SMA) line-fits (solid

line, with 95% confidence intervals around the slope as dashed lines; a: P<0.001; R2=0.67; Slope

not different from 1: P=0.57; Intercept not different from 0: P=0.80; b: P<0.001; R2=0.68; Slope

not different from 1: P=0.37; Intercept not different from 0: P=0.96). The 1:1 line (i.e. slope of 1,

intercept of 0) is given in the solid grey line.

CEL

GOM

GEO

GB

FP

FC

NS

NEA

LAB

IRSWBS

ICE

WSS1

SNL

NSL

WGI

WGO

a

45

70

95

120

145

170

195

45 70 95 120 145 170 195

Mean prey timing (day of year)

Mea

n p

red

ato

r tim

ing (

day o

f ye

ar)

CEL

GOM

GEO

GB

FP

FC

NS

NEA

LAB

IRSWBS

ICE

WSS1

SNL

NSL

WGI

WGO

b

45

70

95

120

145

170

195

45 70 95 120 145 170 195

Mean prey timing (day of year)

Mea

n p

red

ato

r tim

ing (

day o

f ye

ar)

Fig. S5. Predator-prey match across a species’ range using observed predator-prey stages.

Observations of juvenile copepods (C1-C3 stage)10,21,22 and fish spawning time42 for populations

of Atlantic cod (Gadus morhua, data labels refer to populations in Fig. 1). Also given is the

standard major axis (SMA) line-fit (solid line, with 95% confidence intervals around the slope as

dashed lines; P<0.001; R2=0.70; Slope not different from 1: P=0.21; Intercept different from 0:

P<0.001). The 1:1 line (i.e. slope of 1, intercept of 0) is given in the solid grey line.

CEL

GOM

GEO

GB

FP

FC

NS

NEA

LAB

IRS

WBS

ICE

WSS1

SNL

NSL

WGI

WGO

45

70

95

120

145

170

195

45 70 95 120 145 170 195

Observed average C1−C3 timing (day of year)

Ob

se

rved

avera

ge s

paw

nin

g tim

e (

day o

f ye

ar)

Fig. S6. Predator-prey match across a species’ range using timing estimates based on

constant SD predictions. Estimates of N3-stage copepods and first-feeding fish larvae timing

are made based on observations of copepodite timing10,21,22 and spawning time42 respectively.

Timing estimates are made using stage durations based on population-specific overall mean

(time-invariant) temperature estimates (data labels refer to populations in Fig. 1) for comparison

with the TDF method in Fig. 3. Also given is the standard major axis (SMA) line-fits (solid line,

with 95% confidence intervals around the slope as dashed lines; P<0.001; R2=0.72; Slope

different from 1: P=0.022). The 1:1 line (i.e. slope of 1, intercept of 0) is given in the solid grey

line.

CEL GOM

GEO

GB

FP

FC

NS

NEA

LAB

IRS

WBS

ICE

WSS1

SNL

NSL

WGI

WGO

45

70

95

120

145

170

195

45 70 95 120 145 170 195

Prey timing estimated with average temperature (day of year)

Pre

da

tor

tim

ing e

stim

ate

d w

ith a

ve

rage

tem

pera

ture

(day o

f yea

r)

Fig. S7. Illustration of method to estimate timing of unobserved stages via the TDF metric.

Observed stages (light colored distributions; red: fish eggs at spawning, blue: juvenile copepods

or copepodites) are back- or forward-calculated (1) to estimate timing (and temporal match - 2) of

observed stages (dark distributions; red: fish larvae, blue: larval copepods or nauplii).

Fig. S8. Temperature-dependent SDs for Atlantic cod. Stage duration (days) from spawning to

yolk absorption measured at various temperatures (°C) for cod from the Gulf of Maine (GOM,

yellow), northeast Arctic (NEA, purple), southern Newfoundland (SNL, green) and western

Scotian Shelf (WSS1, pink) populations (locations given in Fig. 1). Fitted line is 𝑆𝐷𝑌𝑜𝑙𝑘 =

𝑒𝑎∙𝑇+𝑏 where SDYolk is the stage duration (days) from spawning to yolk absorption estimated at a

temperature T (°C), and a = -0.125°C-1 (standard error = 0.0086°C-1) and b = 3.72 (standard error

= 0.055). Shown is fit with 95% confidence intervals (P<0.001; R2=0.87). Yolk absorption stage

duration is estimated from time to 50% yolk absorption (6, 48; open symbols), time to feeding (46;

+), or time to total yolk absorption (45, 47; closed symbols). Timing reported as “days after hatch”

(45, 48) was converted to “days after fertilization” based on a temperature-dependent egg

incubation relationship (estimated from observations across 6 cod populations) reported in

Geffen et al. (2006)27 as 𝑆𝐷𝐻𝑎𝑡𝑐ℎ = 74 ⋅ (𝑇 + 2)−0.79 where T is incubation temperature.

0 2 4 6 8 10 12

10

20

30

40

50

Temperature °C

Yo

lk a

bsorp

tion

sta

ge d

ura

tion

, d

ays

●●●●

GOM

NEA

SNL

WSS1

Davenport & Lonning 1980

Hunt von Herbing et al. 1996

Jordaan & Kling 2003

Laurence 1978

Pepin et al. 1997

Month

Tem

pe

ratu

re (

°C)

GEO

J F M A M J J A S O N D

−5

05

10

15

20

Month

Tem

pe

ratu

re (

°C)

GOM

J F M A M J J A S O N D

−5

05

10

15

20

Month

Tem

pe

ratu

re (

°C)

WSS1

J F M A M J J A S O N D

−5

05

10

15

20

Month

Tem

pe

ratu

re (

°C)

NSL

J F M A M J J A S O N D

−5

05

10

15

20

Month

Tem

pe

ratu

re (

°C)

SNL

J F M A M J J A S O N D

−5

05

10

15

20

Month

Tem

pe

ratu

re (

°C)

GB

J F M A M J J A S O N D

−5

05

10

15

20

Month

Tem

pe

ratu

re (

°C)

FC

J F M A M J J A S O N D

−5

05

10

15

20

Month

Tem

pe

ratu

re (

°C)

LAB

J F M A M J J A S O N D

−5

05

10

15

20

Month

Tem

pe

ratu

re (

°C)

WGO

J F M A M J J A S O N D

−5

05

10

15

20

Month

Tem

pe

ratu

re (

°C)

WGI

J F M A M J J A S O N D

−5

05

10

15

20

Month

Tem

pe

ratu

re (

°C)

ICE

J F M A M J J A S O N D

−5

05

10

15

20

Month

Tem

pe

ratu

re (

°C)

FP

J F M A M J J A S O N D

−5

05

10

15

20

Fig. S9. Intra-annual variation in population-specific temperatures estimated from

temperature observations between 5- and 100-m depth. Mean daily temperature (solid line, ±

1 standard deviation, dashed lines; 5-100m depth) estimated for each population from

temperature sampled in the relevant population areas (see also Materials and Methods, Table S1,

Fig. 1). Population codes as in Table S1.

Month

Tem

pe

ratu

re (

°C)

NEA

J F M A M J J A S O N D

−5

05

10

15

20

Month

Tem

pe

ratu

re (

°C)

WBS

J F M A M J J A S O N D

−5

05

10

15

20

Month

Tem

pe

ratu

re (

°C)

NS

J F M A M J J A S O N D

−5

05

10

15

20

Month

Tem

pe

ratu

re (

°C)

IRS

J F M A M J J A S O N D

−5

05

10

15

20

Month

Tem

pe

ratu

re (

°C)

CEL

J F M A M J J A S O N D

−5

05

10

15

20