This presentation represents a talk given by Michael … · This presentation represents a talk given by Michael Anderson ... Our slide show has been biased towards the large open

1

This module summarizes a study that utilized structural equation

modeling, motivated by the desire to evaluate a integrated set of

hypotheses that had long been debated. The study in question is

Anderson, T.M., Hopcraft, J.G.C., Mayemba, E., Eby, S., Ritchie, M.E.,

Grace, J.B., and Olff, H. 2010. Landscape-scale analyses suggest both

nutrient and antipredator advantages to Serengeti herbivore hotspots.

Ecology, 91:1519-1529.

http://onlinelibrary.wiley.com/doi/10.1890/09-0739.1/full

Notes: IP-056512; Support provided by the USGS Climate & Land

Use R&D and Ecosystems Programs. I would like to acknowledge

formal review of this material by Jesse Miller and Phil Hahn,

University of Wisconsin, as well as collaboration with Michael

Anderson, senior author of the paper. The use of trade names is for

descriptive purposes only and does not imply endorsement by the U.S.

Government.

Last revised 17.02.27.

Source: https://www.usgs.gov/centers/wetland-and-aquatic-research-

center/science/quantitative-analysis-using-structural-equation

2 2

This presentation represents a talk given by Michael Anderson

on the work in 2009 prior to publication of the work in 2010.

Modest narration has been added to give some flow to the

presentation. For additional details, consult the primary

publication at

https://www.researchgate.net/publication/44631643_Landscape-

level_analyses_suggest_both_nutrient_and_antipredator_advant

ages_to_Serengeti_herbivore_hotspots

Here is a screen shot of the official publication.

3

4 4

The Serengeti ecosystem is one of the world’s charismatic

places. It is home to the mass migrations that are featured on so

many nature shows.

5

Because of the portails in documentaries and also in movies, we

think of the Serengeti in terms of the massive herds of millions of

animals.

6

As the animals move through the system, we can see a variety

of landscapes. It is often the wildebeest that leads the way,

willing to eat nearly any type of forage.

7

Often the grass is indeed past its prime and of lesser quality.

Other herbivores are mixed in the large multi-species herds.

8

The Serengeti ecosystem is contained within both Tanzania and

Kenya.

9 9

The migratory path of the animals spans a large gradient in

conditions.

From:

High rainfall

High standing plant biomass

Granitic soils (sandy clay loams)

Low soil nutrients and SOC

Low quality forage

Grazing: infrequent

To:

Low rainfall

Low standing plant biomass

Volcanic soils (silty clay loams)

High soil nutrients and SOC

High quality forage

Grazing: frequent and severe

10

While our attention is drawn to the animals in their migration,

there are large swaths of the system that are less visited.

11

Some of the unused areas are heavily vegetated, though the

vegetation is somewhat different from the more used locations.

12

Viewing locations within the landscape, we observe a variety of

conditions.

13

14

15

16

Our slide show has been biased towards the large open areas,

but there are also plenty of woodlands as well. These woodlands

are particularly important as well-used spots for predators, as

well as certain herbivores (especially those with substantial

predator defenses like water buffalo and elephant.

17

18

Considering all this, we see that understanding the consumer

dynamics of the system requires more than an understanding of

the great migration.

19 19

-In reality there exists considerable heterogeneity in Serengeti

-Vast areas of the ecosystem are devoid of ungulate herbivores

-Habitats include plains, savannas, woodlands, and gallery

forests

-Most people would be surprised to learn just how many trees

there are in this savanna

-The story told here is about one type of heterogeneity that was

noticed early on in the ecological investigation of the Serengeti,

but that has more or less existed under the radar because of the

overwhelming focus on the migration. This heterogeneity in the

spatial distribution of permanent resident habitats of many of

Serengeti’s large herbivores

20

While the great herds are migrating, and thus concentrated in the

larger landscape, smaller groups of animals appear to be doing

their own thing.

21

These non-migratory animals include members of all the major

herbivores.

22

While predators certainly get their fill when the migrating herds

pass by, during the bulk of the year, the non-migrating animals

provide the consistent food base for resident lions, leopards,

cheetahs, hyenas, and other predators. Herbivores engaged in

this part of the system cycle are exposed to much greater

predation risk and forage much more selectively and cautiously.

23

Most often, herbivores are clearly visible, seeming to prefer open

areas where predators can be easily observed.

24

25

26

This system within the larger system is comprised of numerous

“hotspots” of permanent resident grazers.

ZB = zebra

WB = wildebeest

TP = topi

HB = hartebeest

GG = Grant’s gazelle

TG = Thompson’s gazelle

WH = warthogs

27

There are a number of known characteristcs of the hotspot

areas.

28

Numerous theories exist, but can be classified into those related

to feeding and avoiding predation.

29 29

As stated earlier, this all happens within the context of a very

large and continuous macro-gradient of conditions.

High rainfall

High standing plant biomass

Granitic soils (sandy clay loams)

Low soil nutrients and SOC

Low quality forage

Grazing: infrequent

Low rainfall

Low standing plant biomass

Volcanic soils (silty clay loams)

High soil nutrients and SOC

High quality forage

Grazing: frequent and severe

30

In this study, mapping was used to help establish sample

locations.

31

For each site surveyed, measures were made to represent the

key elements of various competing theories.

32

Looking closely at the map shows the locations that were

identified, as well as where sample transects were selected for

data collection.

33

Data from other studies support some key assumptions about

predation risk and landscape characteristics.

34

Based on hotspot-landscape relations, it is possible to predict

their probabily of occurrence based solely on climate and scape

conditions.

35

For SEM studies, we begin with a general representation of

the hypotheses being evaluated. Care is taken to describe

the potential intepretations of the relationships that might

be found relative to the network of hypothesized links.

A: Direct influence of predation risk created by structural

variation in the environment

B & C: indirect influences of landscape features on hotspots

mediated through their effects on resource quality and quantity -

i.e. catena variation (Bell 1970)

D & E: Main paths representing forage quality and quantity, both

hypothesized to be positive

F: Note that plant biomass represents several potentially

contrasting processes

(1) herbivores benefit from biomass once the negative effect of

biomass on forage quality has been statistically controlled

(positive path)

(2) vegetation biomass is related to increased predator kill

success (negative path)

(3) Only indirect effects of biomass (path insignificant)

36

The study by Anderson et al. serves as an excellent example

regarding the documentation of their logic and procedures. Here I

point to the underlying principles that they followed, which are

expounded upon in the tutorials on this site. This slide is taken from

the tutorial entitled, “Overview of th e Modeling Process”.

Briefly, this slide is meant to convey that SEM is a process designed

to lead to scientifically interpretable models. It relies on sequential

learning and a general multi-step process to build confident

knowledge. In SEM, we first translate our ideas into models, test

those models, modify our models if need be, and then use that

knowledge to inform where we start with the next study.

37

The results in this case show that many processes are operating

at once. At a general level, it appears that animals are more

influenced by predator avoidance than by forage, though

certainly both are important.

38

Conclusions emphasize that hotspots require certain landscape

features favorable for predator detection. Soils and positions

within the larger gradients of soils and rainfall also play a role.

A major point made in the paper is that SEM provides a way past

the dichotomy of previous debates which have been fueled in

part by the reductionist nature of traditional statistical models.

39 39

Support for this work included numerous groups and individuals.