Universal Patterns in Food Webs?

1
Universal Patterns in Food Webs?

• Benjamin Good

2
What is the evidence?

• Paper by Camacho et al. claims to demonstrate
  some traits common to all food web networks
• Uses the niche model to make theoretical
  predictions and then compares predictions to data
  from 7 different food webs

3
The Model

• Niche Parameter chosen bynirandom(0,1),
  i1,,S
• Species eat prey in a range given byrixni
  wherep(x) b(1-x)(b-1) , 0 lt x lt 1,
  andb(S2/2L 1)
• This range centered on cirandom(ri/2, ni)
• Hence, a list of prey or a list of predators can
  be calculated for each species

Sample Web
4
The Predictions

• In another paper, they used this model to derive
  expressions for Pprey(k) and Ppred(k)
• Pprey(k) is the probabilitythat a given species
  hask or more prey, Ppred(k) isthe probability
  that aspecies has k or more predators.

5
Initial Results

• Using data from the St. Martin Island web, with
  an empirically measured value for z, the data fit
  the model quite well.
• No fitting parameters at all!

6
Taking the idea further

• Instead of the distribution of number of prey,
  they use the scaled number of prey k/2z and the
  scaled number of predators m/2z and
  obtain
• Now the same function works for all 7 food webs
  and still without any fitting parameters!

7
and further

• By pooling the data from all 7 webs, they
  construct a master plot again with no
  fit parameters

8
Havent seen a fit this good since
especially when compared to the modelswe have
been dealing with so far.
9
Other predictions

• Amazingly, the same model predictsseveral
  other network characteristicsquite well number
  of links per species, raverage distance,
  dclustering coefficient, C

10
Compared to older studies

• Rejects the scale-free hypothesis that had been
  established earlier
• Also rejects the small world hypothesis for food
  webs
• Hypothesizes a single functional form for various
  food web properties

11
Response to this article

• A later study by Dunne, Williams, and Martinez
  extended the data to include 16 data sets in
  total.

12
Results from Dunne, et al
Three differentfunctional forms(instead of 1)
13
Yet some similarities remain

• Confirmed the earlierresults for a certain
  levelof connectedness
• Also rejected bothscale-free and
  small-worldhypotheses
• Some of the differencesin functional form
  attributedto poor data collectionor extremely
  specializedwebs

14
The End