Analyzing Ecological Data

Visit the link below to download the full version of this book:

https://medipdf.com/product/analyzing-ecological-data/

Click Download Now

To Asterix, Juultje and Poek, for paying more attention to my laptop

To Norma and Juan Carlos, and to Antonio (d9 Aieta) who showed me
that it was worthwhile crossing the great waters...

To Moira, for accepting all the hours shared with my computer that
I should have been sharing with her
Preface

'Which test should I apply?' During the many years of working with ecologists,
biologists and other environmental scientists, this is probably the question that the
authors of this book hear the most often. The answer is always the same and along
the lines of 'What are your underlying questions?', 'What do you want to show?'.
The answers to these questions provide the starting point for a detailed discussion
on the ecological background and purpose of the study. This then gives the basis
for deciding on the most appropriate analytical approach. Therefore, a better start-
ing point for an ecologist is to avoid the phrase 'test' and think in terms of 'analy-
sis'. A test refers to something simple and unified that gives a clear answer in the
form of a p-value: something rarely appropriate for ecological data. In practice,
one has to apply a data exploration, check assumptions, validate the models, per-
haps apply a series of methods, and most importantly, interpret the results in terms
of the underlying ecology and the ecological questions being investigated.
Ecology is a quantitative science trying to answer difficult questions about the
complex world we live in. Most ecologists are aware of these complexities, but
few are fully equipped with the statistical sophistication and understanding to deal
with them.
Even data gathered from apparently simple ecological research can require a
level of statistical awareness rarely taught at the undergraduate or even the post-
graduate level. There is little enough time to teach the essentials of ecology, let
alone finding the time to teach 'advanced' statistics. Hopefully, for post graduates
moving into academia there will be some advanced statistical support available,
but many ecologist end up working in government, a voluntary organisation or
consultancy where statistical support is minimal.
Although, the authors of this book believe that a quantitative approach is at the
core of being a good ecologist, they also appreciate how challenging many ecolo-
gists find statistics. This book is therefore aimed at three levels of reader.
At one level it is aimed at making ecologists aware of how important it is to de-
sign scientifically robust ecological experiments or monitoring programmes, and
the importance of selecting the best analytical technique. For these readers we
hope the book, in particular the case studies, will encourage them to develop their
personal statistical skills, or convince them they need statistical support.
On the next level it is aimed at the statistically literate ecologist, who may not
be fully aware of the techniques we discuss, or when to use them. Hopefully, we
have explained things well enough for these readers to feel confident enough to
use some of the techniques we describe. Often these techniques are presented in a
viii Preface

fairly impenetrable manner, even for the statistically aware ecologist, and we have
tried to make our presentation as 'ecologist friendly' as possible.
Finally, we hope the book will be of value to statisticians, whether they have a
background in ecology or statistics. Ecological data can be particularly challeng-
ing to analyse, and we hope that providing an insight into our approach, together
with the detailed case studies, will be of value to statistician readers, regardless of
their background and expertise.
Overall, however, we hope this book will contribute in some small way to im-
proving the collection and analysis of ecological data and improve the quality of
environmental decision making.
After reading this book, you should be able to apply the following process:
'These are my questions', "This is my statistical approach', 'Here is proof that I
did it all correct (model validation)', 'This is what the data show' and 'Here is the
ecological interpretation'.

Acknowledgement
A large part of the material in this book has been used by the first two authors
as course material for MSc and PhD students, post-docs, scientists, both as aca-
demic and non-academic courses. We are greatly indepted to all 1200-1500
course participants who helped improve the material between 2000 and 2005 by
asking questions and commenting on the material.
We would also like to thank a series of persons who commented on parts of this
book: Ian Jolliffe, Anatoly Saveliev, Barry O'Neill, Neil Campbell, Graham
Pierce, Ian Tuck, Alex Douglas, Pam Sikkink, Toby Marthews, Adrian Bowman,
and six anonymous reviewers and the copy-editor. Their criticisms, comments,
help and suggestions have greatly improved this book.
The first author would like to thank Rob Fryer and FRS Marine Laboratory for
providing the flexibility to start the foundation of this book.
We would also like to thank the people and organizations who donated data for
the theory chapters. The acknowledgement for the unpublished squid data (do-
nated by Graham Pierce, University of Aberdeen) used in Chapters 4 and 7 is as
follows. Data collection was financed by the European Commission under the fol-
lowing projects: FAR MA 1.146, AIR1-CT92-0573, FAIR CT 1520, Study Pro-
ject 96/081, Study project 97/107, Study Project 99/063, and Q5CA-2002-00962.
We would like to thank Roy Mendelssohn (NOAA/NMFS) for giving us a copy of
the data used in Mendelssohn and Schwing (2002). The raw data are summa-
ries calculated from the COADS dataset. The COADS references are Slutz et al.
(1985) and Woodruff et al. (1987). We thank Jaap van der Meer (NIOZ) for allow-
ing us to use the Balgzand data, The Bahamas National Trust and Greenforce An-
dros Island Marine Study for providing the Bahamas fisheries dataset, Chris El-
phick (University of Connecticut) for the sparrow data, and Hrafhkell Eiriksson
(Marine Research Institute, Reykjavik) for the Icelandic Nephrops time series. The
public domain SRTM data used in Chapter 19 were taken from the U.S. Geologi-
cal Survey, EROS Data Center, Sioux Falls, SD. We thank Steve Hare (University
of Washington) for allowing us to use the 100 biological and physical time series
 Preface ix

from the North Pacific Ocean in Chapter 17. A small part of Chapter 13 is based
on Zuur (1999, unpublished PhD thesis), which was partly financed by the EU
project DYNAMO (FAIR-CT95-0710).
A big 'thank you' is also due to the large number of folks who wrote R
(www.r-project.org) and its many libraries. We made a lot of use of the lattice, re-
gression, GLM, GAM (mgcv) and mixed modelling libraries (nlme). This thank
you is probably also on behalf of the readers of this book as everything we did can
be done in R.
Finally, we would like to thank John Kimmel for giving us the opportunity to
write this book, and his support during the entire process. On to the next book.

Alain F. Zuur
Elena N. Ieno
Graham M. Smith

February 2007
Contents

Contributors xix

1 Introduction 1
1.1 Part 1: Applied statistical theory 1
1.2 Part 2: The case studies 3
1.3 Data, software and flowcharts 6

2 Data management and software 7

2.1 Introduction 7
2.2 Data management 8
2.3 Data preparation 9
2.4 Statistical software 13

3 Advice for teachers 17

3.1 Introduction 17

4 Exploration 23
4.1 The first steps 24
4.2 Outliers, transformations and standardisations 38
4.3 A final thought on data exploration 47

5 Linear regression 49
5.1 Bivariate linear regression 49
5.2 Multiple linear regression 61
5.3 Partial linear regression 73

6 Generalised linear modelling 79

6.1 Poisson regression 79
6.2 Logistic regression 88

7 Additive and generalised additive modelling 97

7.1 Introduction 97
7.2 The additive model 101
7.3 Example of an additive model 102
7.4 Estimate the smoother and amount of smoothing 104
7.5 Additive models with multiple explanatory variables 108
xii Contents

7.6 Choosing the amount of smoothing 112

7.7 Model selection and validation 115
7.8 Generalised additive modelling 120
7.9 Where to go from here 124

8 Introduction to mixed modelling 125

8.1 Introduction 125
8.2 The random intercept and slope model 128
8.3 Model selection and validation 130
8.4 A bit of theory 135
8.5 Another mixed modelling example 137
8.6 Additive mixed modelling 140

9 Univariate tree models 143

9.1 Introduction 143
9.2 Pruning the tree 149
9.3 Classification trees 152
9.4 A detailed example: Ditch data 152

10 Measures of association 163

10.1 Introduction 163
10.2 Association between sites: Q analysis 164
10.3 Association among species: R analysis 171
10.4 Q and R analysis: Concluding remarks 176
10.5 Hypothesis testing with measures of association 179

11 Ordination — First encounter 189

11.1 Bray-Curtis ordination 189

12 Principal component analysis and redundancy analysis 193

12.1 The underlying principle ofPCA 193
12.2 PCA: Two easy explanations 194
12.3 PCA: Two technical explanations 196
12.4 Example of PC A 197
12.5 The biplot 200
12.6 General remarks 205
12.7 Chord and Hellinger transformations 206
12.8 Explanatory variables 208
12.9 Redundancy analysis 210
12.10 Partial RDA and variance partitioning 219
12.11 PCA regression to deal with collinearity 221

13 Correspondence analysis and canonical correspondence analysis 225

13.1 Gaussian regression and extensions 225
13.2 Three rationales for correspondence analysis 231
13.3 From RGR to CCA 238
 Contents xiii

13.4 Understanding the CCA triplot 240

13.5 When to use PCA, CA, RDA or CCA 242
13.6 Problems with CA and CCA 243

14 Introduction to discriminant analysis 245

14.1 Introduction 245
14.2 Assumptions 248
14.3 Example 250
14.4 The mathematics 254
14.5 The numerical output for the sparrow data 255

15 Principal coordinate analysis and non-metric multidimensional scaling 259

15.1 Principal coordinate analysis 259
15.2 Non-metric multidimensional scaling 261

16 Time series analysis — Introduction 265

16.1 Using what we have already seen before 265
16.2 Auto-regressive integrated moving average models with exogenous
variables 281

17 Common trends and sudden changes 289

17.1 Repeated LOESS smoothing 289
17.2 Identifying the seasonal component 293
17.3 Common trends: MAFA 299
17.4 Common trends: Dynamic factor analysis 303
17.5 Sudden changes: Chronological clustering 315

18 Analysis and modelling of lattice data 321

18.1 Lattice data 321
18.2 Numerical representation of the lattice structure 323
18.3 Spatial correlation 327
18.4 Modelling lattice data 331
18.5 More exotic models 334
18.6 Summary 338

19 Spatially continuous data analysis and modelling 341

19.1 Spatially continuous data 341
19.2 Geostatistical functions and assumptions 342
19.3 Exploratory variography analysis 346
19.4 Geostatistical modelling: Kriging 358
19.5 A full spatial analysis of the bird radar data 363

20 Univariate methods to analyse abundance of decapod larvae 373

20.1 Introduction 373
20.2 The data 374
20 3 Data exnlnratinn ^nn
xiv Contents

20.4 Linear regression results 379

20.5 Additive modelling results 381
20.6 How many samples to take? 383
20.7 Discussion 385

21 Analysing presence and absence data for flatfish distribution in the Tagus
estuary, Portugal 389
21.1 Introduction 389
21.2 Data and materials 390
21.3 Data exploration 392
21.4 Classification trees 395
21.5 Generalised additive modelling 397
21.6 Generalised linear modelling 398
21.7 Discussion 401

22 Crop pollination by honeybees in Argentina using additive mixed

modelling 403
22.1 Introduction 403
22.2 Experimental setup 404
22.3 Abstracting the information 404
22.4 First steps of the analyses: Data exploration 407
22.5 Additive mixed modelling 408
22.6 Discussion and conclusions 414

23 Investigating the effects of rice farming on aquatic birds with mixed

modelling 417
23.1 Introduction 417
23.2 The data 419
23.3 Getting familiar with the data: Exploration 420
23.4 Building a mixed model 424
23.5 The optimal model in terms of random components 427
23.6 Validating the optimal linear mixed model 430
23.7 More numerical output for the optimal model 431
23.8 Discussion 433

24 Classification trees and radar detection of birds for North Sea wind
farms 435
24.1 Introduction 435
24.2 From radars to data 436
24.3 Classification trees 438
24.4 A tree for the birds 440
24.5 A tree for birds, clutter and more clutter 445
24.6 Discussion and conclusions 447

25 Fish stock identification through neural network analysis of parasite

fauna 449
 Contents xv

25.1 Introduction 449

25.2 Horse mackerel in the northeast Atlantic 450
25.3 Neural networks 452
25.4 Collection of data 455
25.5 Data exploration 456
25.6 Neural network results 457
25.7 Discussion 460

26 Monitoring for change: Using generalised least squares, non-metric

multidimensional scaling, and the Mantel test on western Montana
grasslands 463
26.1 Introduction.. 463
26.2 The data 464
26.3 Data exploration 467
26.4 Linear regression results 472
26.5 Generalised least squares results 476
26.6 Multivariate analysis results 479
26.7 Discussion 483

27 Univariate and multivariate analysis applied on a Dutch sandy beach

community 485
27.1 Introduction 485
27.2 The variables 486
27.3 Analysing the data using univariate methods 487
27.4 Analysing the data using multivariate methods 494
27.5 Discussion and conclusions 499

28 Multivariate analyses of South-American zoobenthic species — spoilt for

choice 503
28.1 Introduction and the underlying questions 503
28.2 Study site and sample collection 504
28.3 Data exploration 506
28.4 The Mantel test approach 509
28.5 The transformation plus RDA approach 512
28.6 Discussion and conclusions 512

29 Principal component analysis applied to harbour porpoise fatty acid

data 515
29.1 Introduction 515
29.2 The data 515
29.3 Principal component analysis 517
29.4 Data exploration 518
29.5 Principal component analysis results 518
29.6 Simpler alternatives to PCA 524
29.7 Discussion 526
xvi Contents

30 Multivariate analyses of morphometric turtle data — size and shape ....529

30.1 Introduction 529
30.2 The turtle data 530
30.3 Data exploration 531
30.4 Overview of classic approaches related to PCA 534
30.5 Applying PCA to the original turtle data 536
30.6 Classic morphometric data analysis approaches 537
30.7 A geometric morphometric approach 542

31 Redundancy analysis and additive modelling applied on savanna tree

data 547
31.1 Introduction 547
31.2 Study area 548
31.3 Methods 548
31.4 Results 551
31.5 Discussion 559

32 Canonical correspondence analysis of lowland pasture vegetation in the

humid tropics of Mexico 561
32.1 Introduction 561
32.2 The study area 562
32.3 The data ....563
32.4 Data exploration 565
32.5 Canonical correspondence analysis results 568
32.6 African star grass 571
32.7 Discussion and conclusion 573

33 Estimating common trends in Portuguese fisheries landings 575

33.1 Introduction 575
33.2 The time series data 576
33.3 MAFA and DFA 579
33.4 MAFA results 580
33.5 DFA results 582
33.6 Discussion 587

34 Common trends in demersal communities on the Newfoundland-Labrador

Shelf 589
34.1 Introduction 589
34.2 Data 590
34.3 Time series analysis 591
34.4 Discussion 598

35 Sea level change and salt marshes in the Wadden Sea: A time series
analysis 601
35.1 Interaction between hydrodynamical and biological factors 601
35.2 The data 603
 Contents xvii

35.3 Data exploration 605

35.4 Additive mixed modelling 607
35.5 Additive mixed modelling results 610
35.6 Discussion 613

36 Time series analysis of Hawaiian waterbirds 615

36.1 Introduction 615
36.2 Endangered Hawaiian waterbirds 616
36.3 Data exploration 617
36.4 Three ways to estimate trends 619
36.5 Additive mixed modelling 626
36.6 Sudden breakpoints 630
36.7 Discussion 631

37 Spatial modelling of forest community features in the Volzhsko-Kamsky

reserve 633
37.1 Introduction 633
37.2 Study area 635
37.3 Data exploration 636
37.4 Models of boreality without spatial auto-correlation 638
37.5 Models of boreality with spatial auto-correlation 640
37.6 Conclusion 646

References 649

Index 667
 Contributors

BASTIDA, R.
Departamento de Ciencias Marinas
Universidad Nacional de Mar del Plata
Consejo Nacional de Investigaciones Cientificas y Tecnicas
Casilla de Correo 43
(7600) Mar del Plata
Argentina

BASUALDO, M.
Fac. Cs. Veterinarias
UNCPBA, Campus Universitario
-7000-Tandil
Argentina

BUDGEY, R.
Central Science Laboratory
Sand Hutton
York,Y041 1LZ
United Kingdom

CABRAL, H.
Universidade de Lisboa
Faculdade de Ciencias, Instituto de Oceanografia,
Campo Grande
1749-016 Lisboa
Portugal
CAMPBELL, N.
School of Biological Sciences
University of Aberdeen
Zoology Building
Tillydrone Avenue
Aberdeen, AB24 2TZ
United Kingdom
Current address:
FRS Marine Laboratory
375 Victoria Road.
Aberdeen, AB11 9DB
United Kingdom

CHIZHIKOVA, N.A.
Faculty of Ecology
Kazan State University
18, Kremlevskaja Street
Kazan, 420008
Russia

CLAUDE, J.
Universite de Montpellier 2
ISE-M, UMR 5554 CNRS, cc 64
2, place Eugene Bataillon
34095 Montpellier cedex 5
France

DEVINE, J.A.
Memorial University
Department of Biology
4 Clark Place
St. John's NL,A1C5S7
Canada
Current address:
Middle Depth Fisheries & Acoustics
National Institute of Water and Atmospheric Research Ltd
Private Bag 14-901 Kilbirnie
Wellington 6241
New Zealand

DIJKEMA, K.S.
Wageningen IMARES, Department Texel
Institute for Marine Resources & Ecosystem Studies
P.O. Box 167
1790 AD Den Burg, Texel
The Netherlands