+ All Categories
Home > Documents > Student: Paul Welle Collaborators: Ines Azevedo Mitchell Small Sarah Cooley Scott Doney

Student: Paul Welle Collaborators: Ines Azevedo Mitchell Small Sarah Cooley Scott Doney

Date post: 23-Feb-2016
Category:
Upload: alamea
View: 32 times
Download: 0 times
Share this document with a friend
Description:
The Impact of Climate Stressors on Coral Bleaching and Mortality : A Case Study of the 2005 Caribbean Summer. Student: Paul Welle Collaborators: Ines Azevedo Mitchell Small Sarah Cooley Scott Doney. Background. Eakin et al. (2010) Caribbean summer 2005 - PowerPoint PPT Presentation
33
Student: Paul Welle Collaborators : Ines Azevedo Mitchell Small Sarah Cooley Scott Doney THE IMPACT OF CLIMATE STRESSORS ON CORAL BLEACHING AND MORTALITY : A CASE STUDY OF THE 2005 CARIBBEAN SUMMER 1
Transcript
Page 1: Student:  Paul  Welle Collaborators: Ines  Azevedo Mitchell Small Sarah Cooley Scott  Doney

1

Student: Paul Welle

Collaborators:Ines AzevedoMitchell SmallSarah CooleyScott Doney

THE IMPACT OF CLIMATE STRESSORS

ON CORAL BLEACHING AND MORTALITY:

A CASE STUDY OF THE 2005 CARIBBEAN SUMMER

Page 2: Student:  Paul  Welle Collaborators: Ines  Azevedo Mitchell Small Sarah Cooley Scott  Doney

2

Eakin et al. (2010) Caribbean summer 2005 Bleaching, Mortality (dependent variables) Temperature (independent variable)

BACKGROUND

Reproduced from Eakin (2010)

Page 3: Student:  Paul  Welle Collaborators: Ines  Azevedo Mitchell Small Sarah Cooley Scott  Doney

3

TEMPERATURES 2005

Retrieved from http://coralreefwatch.noaa.gov/satellite/dhw.php

Page 4: Student:  Paul  Welle Collaborators: Ines  Azevedo Mitchell Small Sarah Cooley Scott  Doney

4

THE DATAn=2945

Page 5: Student:  Paul  Welle Collaborators: Ines  Azevedo Mitchell Small Sarah Cooley Scott  Doney

5

(1) Limited by the functional form of OLS

We expand to a non-linear model.

(2) Uncontrolled spatial correlation

We add in fixed effects.

(3) Limited number of explanatory variables

We extend the dataset to include photosynthetically active radiation (PAR) and pH. We also recalculate

DHW.

ANALYSIS CAN BE IMPROVED

Page 6: Student:  Paul  Welle Collaborators: Ines  Azevedo Mitchell Small Sarah Cooley Scott  Doney

6

DHW correlates well with bleaching and mortality, although there are indications that the 12-week interval should be lengthened.

PAR Anomaly correlates well with bleaching and mortality (PAR does not), but seems to be of less importance than DHW.

In predicting mortality, it is best to use the maximum value of a stressor, while predicting bleaching the recent (observed) temperatures are more important.

Depth is very protective against PAR. For deep corals (13.5m, or 80 th percentile), PAR plays almost no part in predicting bleaching. For shallow corals (5m, or 20 th percentile), PAR is roughly as important a stressor as DHW.

WHAT WE LEARNED

Page 7: Student:  Paul  Welle Collaborators: Ines  Azevedo Mitchell Small Sarah Cooley Scott  Doney

7

METHOD (OLS VS FRACTIONAL LOGIT)

100100

𝑦𝑖 = 11+ 𝑒−(𝛽0+σ 𝛽𝑗𝑥𝑖𝑗+ σ 𝛽𝑘𝑑𝑖𝑘 + 𝜖𝑖)𝑘𝑗

Page 8: Student:  Paul  Welle Collaborators: Ines  Azevedo Mitchell Small Sarah Cooley Scott  Doney

8

METHOD (MANIPULATION OF CONTINUOUS DATA)

PAR, DHW, pH…

time

observed

maximum

Page 9: Student:  Paul  Welle Collaborators: Ines  Azevedo Mitchell Small Sarah Cooley Scott  Doney

9

METHOD(VARIABLES)

Four stressor formulations Temperature - Degree Heating Weeks (DHW) – 12 week Photosynthetically Active Radiation – PAR 12 week average Photosynthetically Active Radiation – PAR Anomaly Simulated pH – Monthly average

Each formulation has 2 forms “Maximum” – Hypothesized to be important for mortality “Observed” – Hypothesized to be important for bleaching

Bleaching-and-

Mortality =MaxDHW, ObsDHW, MaxPAR, MaxPAR Anomaly, ObsPAR,

ObsPAR Anomaly, MaxPH, ObsPHf( )

Page 10: Student:  Paul  Welle Collaborators: Ines  Azevedo Mitchell Small Sarah Cooley Scott  Doney

10

RESULTS

General Model:

Mortality Model

Bleaching Model

ݕ ൌ�� ͳͳ � �௭

ݖ ൌߚ�� ଵߚ ܦܪ ݔܯ ଶߚ ܣ ݔݔݔݔݔݔݔݔݔݔݔݔݔܯ ݔ ݕ � ܣ ଷߚ ݐݐܦ ߚ � Kݖ ൌߚ�� ଵߚ ܦܪ� ଶߚ ܣ� ଷߚ ݐݐݐݐݐݐݐݐݐݐݐݐݐݐܦ ସߚ ݐݐݐݐݐݐݐݐݐݐݐݐݐݐܦ ൈ� ܦܪ� ହߚ ݐݐݐݐݐݐݐݐݐݐݐݐݐݐܦ ൈ� ܣ� ߚ � K

ܦܪ� ൌ����ఈభή�ௗ�௧ ή� ܦܪ ݔܯ ͳൌ���ఈభή�ௗ�௧ ή� ܦܪ�� ݏ ܣ� ൌ����ఈమή�ௗ�௧ ή� ܣ ݔݔݔݔݔݔݔݔݔݔݔݔݔܯ ݔ ݕ� ܣ ͳൌ�� �ఈమήݐ�� � ή� ܣ�� ݏ ݕ� ܣ

Page 11: Student:  Paul  Welle Collaborators: Ines  Azevedo Mitchell Small Sarah Cooley Scott  Doney

11

RESULTS - MORTALITY

Robust standard errors in parentheses*** p<0.01, ** p<0.05, * p<0.1

Page 12: Student:  Paul  Welle Collaborators: Ines  Azevedo Mitchell Small Sarah Cooley Scott  Doney

12

RESULTS - MORTALITY

Page 13: Student:  Paul  Welle Collaborators: Ines  Azevedo Mitchell Small Sarah Cooley Scott  Doney

13

RESULTS - BLEACHING

Robust standard errors in parentheses*** p<0.01, ** p<0.05, * p<0.1

Page 14: Student:  Paul  Welle Collaborators: Ines  Azevedo Mitchell Small Sarah Cooley Scott  Doney

14

RESULTS - BLEACHING

Page 15: Student:  Paul  Welle Collaborators: Ines  Azevedo Mitchell Small Sarah Cooley Scott  Doney

15

RESULTS BLEACHING

Depth = 5 m

Depth = 13.5 m

Page 16: Student:  Paul  Welle Collaborators: Ines  Azevedo Mitchell Small Sarah Cooley Scott  Doney

16

DHW correlates well with bleaching and mortality, although there are indications that the 12-week interval should be lengthened.

PAR Anomaly correlates well with bleaching and mortality (PAR does not), but seems to be of less importance than DHW.

In predicting mortality, it is best to use the maximum value of a stressor, while predicting bleaching the recent (observed) temperatures are more important.

Depth is very protective against PAR. For deep corals (13.5m, or 80 th percentile), PAR plays almost no part in predicting bleaching. For shallow corals (5m, or 20 th percentile), PAR is roughly as important a stressor as DHW.

WHAT WE LEARNED

Page 17: Student:  Paul  Welle Collaborators: Ines  Azevedo Mitchell Small Sarah Cooley Scott  Doney

17

Eak in , C . M. , Morgan, J . a , Heron, S . F. , Smith , T. B . , L iu , G. , A lvarez-Fi l ip , L . , … Bouchon, C. (2010) . Car ibbean cora ls in c r is is : record thermal s tress , b leach ing, and morta l i ty in 2005. PloS one , 5 (11) , e13969.

Hoegh- Guldberg, O. , Mumby, P. J . , Hooten, a J . , S teneck , R. S . , Greenfield , P. , Gomez, E . , … Hatz io los , M. E . (2007). Cora l reefs under rapid c l imate change and ocean ac id ificat ion . Sc ience (New York , N.Y. ) , 318 (5857) , 1737–42.

McWil l iams, J . , Côté, I . , & Gi l l , J . (2005). Accelerat ing impacts o f temperature-induced cora l b leach ing in the Car ibbean. Eco logy , 86 (8) , 2055–2060.

Wilk inson, C . "Cora l b leach ing and morta l i ty–The 1998 event 4 years la ter and b leach ing to 2002."  Status o f cora l reefs o f the wor ld   (2002) : 33-44.

Wilk inson, C l ive R. , and Dav id Souter , eds .  Status o f Car ibbean cora l reefs a f ter b leach ing and hurr icanes in 2005 . G loba l Cora l Reef Monitor ing Network, 2008.

Yee, S . H. , Santavy, D. L . , & Barron , M. G. (2008). Compar ing envi ronmenta l influences on cora l b leach ing across and wi th in species us ing c lustered b inomia l regress ion . Eco log ica l Model l ing , 218 (1 -2) , 162–174.

Yee, S . H. , & Barron, M. G. (2010) . Pred ic t ing cora l b leaching in response to env ironmenta l s t ressors us ing 8 years o f g loba l -sca le data . Envi ronmenta l moni tor ing and assessment , 161 (1-4) , 423–38.

REFERENCES

Page 18: Student:  Paul  Welle Collaborators: Ines  Azevedo Mitchell Small Sarah Cooley Scott  Doney

18

This work would not be possible without support by

SUPPORT

Page 19: Student:  Paul  Welle Collaborators: Ines  Azevedo Mitchell Small Sarah Cooley Scott  Doney

19

DATA

>30%<30%& >0%

0%

Page 20: Student:  Paul  Welle Collaborators: Ines  Azevedo Mitchell Small Sarah Cooley Scott  Doney

20

BLEACHING

Page 21: Student:  Paul  Welle Collaborators: Ines  Azevedo Mitchell Small Sarah Cooley Scott  Doney

21

MORTALITY

Page 22: Student:  Paul  Welle Collaborators: Ines  Azevedo Mitchell Small Sarah Cooley Scott  Doney

22

DESCRIPTIVE STATISTICS

Variable [units] Min Median Mean Max Dependent Variables:

Bleaching [%] 0.0 26.3 33.2 100 Mortality [%] 0.0 0.0 2.0 68.3

Candidate Explanatory Variables: Depth [m] 0.9 9.2 10.0 42.7 Maximum DHW [°C] 0.0 3.7 4.6 17.2 Observed DHW [°C] 0.0 0.5 2.8 16.6 Maximum PAR [Einsteins/m2] 35.6 47.7 47.9 53.7 Maximum PAR Anomaly [Einsteins/m2]

1.6 16.0 17.9 56.0

Observed PAR [Einsteins/m2] 26.3 39.8 40.4 52.4 Observed PAR Anomaly [Einsteins/m2]

0.0 0.75 4.67 36.5

Simulated Maximum pH [-] 8.07 8.10 8.10 8.20 Simulated Observed pH [-] 8.03 8.07 8.07 8.13 Base PAR [Einsteins /m2] 42.3 51.1 50.4 54.2

Page 23: Student:  Paul  Welle Collaborators: Ines  Azevedo Mitchell Small Sarah Cooley Scott  Doney

23

DEGREE HEATING WEEKS

Typical Hottest Month

Page 24: Student:  Paul  Welle Collaborators: Ines  Azevedo Mitchell Small Sarah Cooley Scott  Doney

24

DATA

Page 25: Student:  Paul  Welle Collaborators: Ines  Azevedo Mitchell Small Sarah Cooley Scott  Doney

25

DATA

>30%<30%& >0%

0%

Page 26: Student:  Paul  Welle Collaborators: Ines  Azevedo Mitchell Small Sarah Cooley Scott  Doney

26

CORRELATIONS

Max DHW

Obs DHW

Max PAR

Obs PAR

Max PAR Anomaly

Obs PAR Anomaly Max pH Obs pH Base

PAR Depth

Max DHW ---- 0.67*** -0.02 -0.09*** 0.19*** 0.11*** 0.00 0.41*** -0.39*** 0.15***

Obs DHW *** ---- -0.11*** -0.39*** -0.06*** -0.33*** -0.05** 0.51*** -0.26*** 0.15***

Max PAR *** ---- 0.33*** 0.21*** 0.21*** 0.15*** -0.15*** 0.59*** -0.01

Obs PAR *** *** *** ---- 0.03* 0.70*** -0.12*** -0.71*** 0.27*** 0.06*** Max PAR Anomaly *** *** *** * ---- 0.36*** 0.22*** 0.02 -0.46*** -0.10***

Obs PAR Anomaly *** *** *** *** *** ---- -0.06*** -0.51*** -0.09*** 0.00

Max pH ** *** *** *** *** ---- 0.23*** 0.11*** -0.11***

Obs pH *** *** *** *** *** *** ---- -0.18*** 0.06***

Base PAR *** *** *** *** *** *** *** *** ---- 0.05***

Depth *** *** *** *** *** *** *** ----

Page 27: Student:  Paul  Welle Collaborators: Ines  Azevedo Mitchell Small Sarah Cooley Scott  Doney

27

Questions:Which stressor form fits best- maximum, observed, or weighted average?Bleaching – Weighted AverageMortality – MaximumDoes PAR or PAR anomaly fit the data better?Bleaching – PAR AnomalyMortality – PAR AnomalyDoes measuring independent maximums of temperature and radiation suffice, or must one account for simultaneously high peaks?Bleaching – IndependentMortality – IndependentIs there evidence for a depth-stressor interaction?Bleaching – YesMortality - No

SUMMARY

Page 28: Student:  Paul  Welle Collaborators: Ines  Azevedo Mitchell Small Sarah Cooley Scott  Doney

28

MORTALITY MODELS

Model 4

Model 5

General Model:

Model 1

Model 2

Model 3

ݕ ൌ��ͳ

ͳ ��௭

ݖ ൌߚ�� ଵߚ ܦܪ ݔܯ ଶߚ ܣ ݔݔݔݔݔݔݔݔݔݔݔݔݔܯ ݔ ݕ � ܣ ଷߚ ݐݐܦ ߚ � K

ݖ ൌߚ�� ଵߚ ܦܪ ݔܯ ଶߚ ܣ ݔݔݔݔݔݔݔݔݔݔݔݔݔܯ ݔ ݕ � ܣ ଷߚ ݐݐݐݐݐݐݐݐݐݐݐݐݐݐܦ ସߚ ݐݐݐݐݐݐݐݐݐݐݐݐݐݐܦ ൈ� ܦܪ ݔܯ ହߚ ݐݐݐݐݐݐݐݐݐݐݐݐݐݐܦ ൈܣ� ݔݔݔݔݔݔݔݔݔݔݔݔݔܯ ݔ ݕ� ܣ ߚ � K

ݖ ൌߚ�� ଵߚ ܦܪ ݔܯ ଶߚ ܣ ݔݔݔݔݔݔݔݔݔݔݔݔݔܯ ݔ ݕ � ܣ ଷߚ ݐݐݐݐݐݐݐݐݐݐݐݐݐݐܦ ସߚ ܦܪ ݔܯ ൈ� ܣ� ݕ � ܣ ݓܨ ହߚ ܣ ݔݔݔݔݔݔݔݔݔݔݔݔݔܯ ݔ ݕ � ܣ ൈ� ܦܪ � ݓ ߚ � K

ݖ ൌߚ�� ଵߚ ܦܪ ݔܯ ଶߚ ܣ ݔݔݔݔݔݔݔݔݔݔݔݔݔܯ ݔ ଷߚ ݐݐݐݐݐݐݐݐݐݐݐݐݐݐܦ ߚ � K

ݖ ൌߚ�� ଵߚ ܦܪ ݔܯ ଶߚ ܣ ݔݔݔݔݔݔݔݔݔݔݔݔݔܯ ݔ ଷߚ ݐݐݐݐݐݐݐݐݐݐݐݐݐݐܦ ସߚ ݏ ܤ ܣ� ߚ � K

Page 29: Student:  Paul  Welle Collaborators: Ines  Azevedo Mitchell Small Sarah Cooley Scott  Doney

29

MORTALITY MODELS

(1) (2) (3) (4) (5) VARIABLES Model Model Model Model Model

Maximum DHW 0.164*** 0.166*** 0.188*** 0.176*** 0.187***

(0.0279) (0.0295) (0.0286) (0.0628) (0.0289) Maximum PAR 0.0452 0.156***

(0.0352) (0.0576) Depth -0.0535*** -0.0501*** -0.0479*** -0.173*** -0.0477***

(0.0112) (0.0121) (0.0118) (0.0670) (0.0117) Base PAR -0.167***

(0.0641) Maximum PAR 0.0283*** -0.0276* 0.0261***

Anomaly (0.00851) (0.0149) (0.00939)

Depth x 0.000333 Maximum DHW (0.00486)

Depth x 0.00551**

*

Maximum PAR (0.00125)

Maximum DHW 0.00109 x Maximum PAR Anomaly Follow

(0.00215)

Maximum PAR

Anomaly x Max DHW Follow

-0.00449 (0.00595)

Constant -6.648*** -3.823** -5.496*** -4.068*** -5.515***

(1.547) (1.845) (0.430) (0.784) (0.468)

Log-Likelihood -62.4294 -62.0861 -62.0672 -61.4354 -62.0358

AIC 0.163501 0.164758 0.162808 0.165427 0.166576

Observations 1,045 1,045 1,045 1,045 1,045

Page 30: Student:  Paul  Welle Collaborators: Ines  Azevedo Mitchell Small Sarah Cooley Scott  Doney

30

BLEACHING MODELS

Model 5

Model 6

General Model:

Model 1

Model 2

Model 3

Model 4

ݕ ൌ��ͳ

ͳ ��௭

ݖ ൌߚ�� ଵߚ ݏܦܪ�� ଶߚ ܣ�� ݏ ݕ� ܣ ଷߚ ݐݐܦ ߚ � K

ݖ ൌߚ�� ଵߚ ܦܪ�� ݏ ଶߚ ܣ ݔݔݔݔݔݔݔݔݔݔݔݔݔܯ ݔ ݕ� ܣ ଷߚ ݐݐݐݐݐݐݐݐݐݐݐݐݐݐܦ ߚ � K

ݖ ൌߚ�� ଵߚ ݏܦܪ�� ଶߚ ܣ�� ݏ ଷߚ ݐݐݐݐݐݐݐݐݐݐݐݐݐݐܦ ߚ � K

ݖ ൌߚ�� ଵߚ ݏܦܪ�� ଶߚ ܣ�� ݏ ଷߚ ݐݐݐݐݐݐݐݐݐݐݐݐݐݐܦ ସߚ ݏ ܤ ܣ� ߚ � K

ݖ ൌߚ�� ଵߚ ݏܦܪ�� ଶߚ ܣ ݔݔݔݔݔݔݔݔݔݔݔݔݔܯ ݔ ݕ� ܣ ଷߚ ݐݐݐݐݐݐݐݐݐݐݐݐݐݐܦ ସߚ ݐݐݐݐݐݐݐݐݐݐݐݐݐݐܦ ൈ� ݏܦܪ�� � ହߚ ݐݐݐݐݐݐݐݐݐݐݐݐݐݐܦ ൈܣ� ݔݔݔݔݔݔݔݔݔݔݔݔݔܯ ݔ ݕ� ܣ ߚ � K

ݖ ൌߚ�� ଵߚ ܦܪ� ଶߚ ܣ� ଷߚ ݐݐݐݐݐݐݐݐݐݐݐݐݐݐܦ ସߚ ݐݐݐݐݐݐݐݐݐݐݐݐݐݐܦ ൈ� � �ܦܪ ହߚ ݐݐݐݐݐݐݐݐݐݐݐݐݐݐܦ ൈ� ܣ� ߚ � K

ܦܪ� ൌ����ఈή�ௗ�௧ ή� ܦܪ ݔܯ ͳൌ���ఈή�ௗ�௧ ή� ݏܦܪ�� ܣ� ൌ����ఈή�ௗ�௧ ή� ܣ ݔݔݔݔݔݔݔݔݔݔݔݔݔܯ ݔ ݕ� ܣ ͳൌ�� �ఈή�ௗ�௧ ή� ܣ ݔݔݔݔݔݔݔݔݔݔݔݔݔܯ ݔ ݕ � ܣ

Page 31: Student:  Paul  Welle Collaborators: Ines  Azevedo Mitchell Small Sarah Cooley Scott  Doney

31

BLEACHING MODELS (1) (2) (3) (4) (5) (6)

VARIABLES Model Model Model Model Model Model

Observed DHW 0.145*** 0.148*** 0.138*** 0.170*** 0.204*** (0.00982) (0.00974) (0.00982) (0.00924) (0.0155)

Observed PAR -0.0273*** -0.0245*** (0.00601) (0.00601)

Depth 0.00962** 0.0103** 0.00863* 0.0106** 0.0731*** 0.0739*** (0.00464) (0.00461) (0.00463) (0.00463) (0.0113) (0.0115)

Base PAR -0.0364** (0.0179)

Observed PAR Anomaly

-0.0213*** (0.00377)

Maximum PAR

Anomaly 0.00909**

(0.00359) 0.0340*** (0.00593)

Depth x Observed

DHW -0.00333***

(0.000999)

Depth x Observed

PAR -0.00269***

(0.000492)

Weighted Average DHW (α=0.0282)

0.213*** (0.0152)

Weighted Average PAR (α=0.0034)

0.0416*** (0.00662)

Depth x Weighted

Average DHW -0.00322***

(0.000967)

Depth x Weighted Average PAR

-0.00301*** (0.000559)

Constant -0.332 1.329 -1.264*** -1.680*** -2.258*** -2.452***

(0.266) (0.894) (0.107) (0.144) (0.175) (0.173)

Log-Likelihood -1267.13 -1266.48 -1265.58 -1268.99 -1263.85 -1260.05 AIC 0.87683 0.877067 0.875777 0.878091 0.875961 0.87472

Observations 2,945 2,945 2,945 2,945 2,945 2,945

Page 32: Student:  Paul  Welle Collaborators: Ines  Azevedo Mitchell Small Sarah Cooley Scott  Doney

32

Fixed Effects Fractional Logit Model Logit – Used for binary dependent variables Fractional Logit – Repurposed for bounded dependent

variable Fixed Effects – Used to control for homogeneity within

groupsMaximize quasi-likelihood function:

Returns sigmoid in range (0,1)

MODEL

𝑦𝑖 = 11+ 𝑒−(𝛽0+σ 𝛽𝑗𝑥𝑖𝑗+ σ 𝛽𝑘𝑑𝑖𝑘 + 𝜖𝑖)𝑘𝑗

Page 33: Student:  Paul  Welle Collaborators: Ines  Azevedo Mitchell Small Sarah Cooley Scott  Doney

33

VARIABLES

Maximum DHW

Maximum PAR Anomaly

Depth

Constant

Log-LikelihoodAIC

Observations

Coefficients

0.188***(0.0286)

0.0283***(0.0085)

-0.0479***(0.0118)-5.50***(0.430)-62.070.16281,045

Marginal Effects(at means)

0.00152***(0.000197)

0.000229***(0.0000587)-0.000139***(0.0000342)

 

   


Recommended