Tag Archives: RDA

The Spatial Ecology of the Comanche Harvester Ant

I have successfully presented my dissertation work and am currently finishing up the revisions for the final submission to the University of Texas at Arlington for the PhD degree. I expect the final dissertation to be available from the university library by July 2015.

The title of the dissertation is: The Spatial Ecology of the Comanche Harvester Ant, Pogonomyrmex comanche (Hymenoptera, Formicidae)

Dr. Esther Betran was the chair of my committee (UTA).

Other committee members were:

Dr. Jonathan Campbell (UTA)

Dr. Paul Chippindale (UTA)

Dr. Sophia Passy (UTA)

and Dr. Walter Tschinkel (FSU)

Here is the slide presentation and the notes which are numbered to correspond to the slides. I have included some of the corrections that came out of the discussion with my committee and otherwise have noted where there are other problems which I am addressing in the revision.

The slides:

and the notes:

Ant Presence and Abundance in the Fort Worth Nature Center

I sampled ants using pitfall traps in 17 sites in the Fort Worth Nature Center monthly in June, July, and August 2012.

I used CANOCO to run redundancy analyses (RDA) on ant presence with abiotic and biotic environmental variables and on ant presence and abundance with soil type to look for ant preference for soil. I used forward selection of variables and Monte Carlo significance tests to select the variables for the final RDA models.

RESULTS

1) RDA for ant presence and environmental variables

RDA Summary Table

Axes

1

2

3

4

Total variance

 Eigenvalues                     

0.122

0.062

0.026

0.014

1.000

Species-environment correlations

0.820

0.872

0.672

0.582

Cumulative percentage variance of species data

12.2

18.4

21.0

22.4

Cumulative percentage variance of species-environment relation 

51.5

77.6

88.8

94.5

Sum of all eigenvalues     

1.000

Sum of all canonical eigenvalues     

0.237

Triplot

2) RDA for ant presence and soil type

RDA Summary Table

Axes                                    1      2      3      4 Total variance
Eigenvalues

0.076

0.023

0.011

0.007

1.000

Species-environment correlations 

0.788

0.603

0.424

0.417

Cumulative percentage variance    of species data

7.6

9.9

11.0

11.7

Cumulative percentage variance    of species-environment relation 65.2   84.9   93.8 100.0
Sum of all eigenvalues

1.000

Sum of all canonical eigenvalues

0.117

Triplot

3) RDA for ant abundance and soil type

RDA Summary Table

Axes                                    1      2      3      4 Total variance
Eigenvalues

0.070

0.031

0.016

0.003

1.000

Species-environment correlations 

0.777

0.655

0.456

0.265

Cumulative percentage variance    of species data

7.0

10.1

11.7

12.0

Cumulative percentage variance    of species-environment relation

58.4

84.6

97.9

100.0

Sum of all eigenvalues

1.000

Sum of all canonical eigenvalues

0.120

Triplot

24% of species presence is explained by the environmental variables with percent litter cover and drainage being the significant variables. Sampling sites by date clumped together indicating a lack of seasonality — which seems a bit unusual since late July and August become quite hot and ant activity seems reduced  at this time.

12% of species presence was explained by soil type with the Aquilla soil being the only significant soil. This soil is the only soil type where the Comanche harvester ant (Pogonomyrmex comanche) is found. All other species are more generalist with respect to soil type.

7.4% of species abundance was explained by soil type again with the Aquilla soil being the only significant soil. This result further supports the result with species presence: only the Comanche harvester ant has such narrow soil preference.

CONCLUSIONS

Though the eigenvalues are low this is not unusual for ecological data. The low level of explanatory value of these variables is likely due to the generalist nature of these species (and more temperate species in general) and the below-ground nesting of most ant species.

The Comanche harvester ant (Pogonomyrmex comanche) was the only species to show strict preference for soil type. Exactly what this species’ preference or requirement is remains unresolved.

Redundancy Analysis on Ant Assemblage Data

Here are the preliminary results of the first redundancy analysis (RDA) for all the sites sampled over the summer of 2012. I performed a PCA with only environmental variables and one with only species data to look for an underlying pattern — which was confirmed. Then I ran an RDA with the full set of environmental variables and species data to determine the most significant quantitative environmental variables. Here are the results of the final RDA for this set of data which includes samples from 21 sites sampled monthly over the summer.

The Summary Table:

Axes

1

2

3

4

Total variance

Eigenvalues

0.358

0.129

0.060

0.020

1.000

Species-environment correlations 

0.977

0.956

0.923

0.720

Cumulative percentage variance of species data               

35.8

48.8

54.7

56.8

Cumulative percentage variance of species-environment relation

61.8

84.1

94.4

97.9

Sum of all eigenvalues     

1.000

Sum of all canonical eigenvalues     

0.580

Marginal effects of environmental variables:

Marginal Effects

Variable Var. N Lambda1
LiC 2 0.32
BG 4 0.27
LA 6 0.21
ToC 3 0.17
SP 5 0.07

Conditional effects of environmental variables:

Conditional Effects

Variable

Var. N

Lambda A

P

F

LiC 2 0.32 0.002 25.00
LA 6 0.13 0.002 12.20
BG 4 0.07 0.002 7.90
ToC 3 0.04 0.002 4.68
SP 5 0.02 0.028 1.81
LiC 2 0.32 0.002 25.00

Ordination Plots:

Species and environmental variables:

Species and Sites:

RDA with interaction terms: Third Level of Analysis for Species Presence

Of the environmental variables measured, percent litter cover (LiCov) and drainage (DRN) came up as the most significant variables affecting ant species presence. Here is the summary results for the RDA including an interaction term for these variables.

Summary Table

**** Summary ****

Axes

1

2

3

4

Total variance

Eigenvalues 0.158 0.120 0.100 0.084 1.000
Species-environment correlations 0.936 0.971 0.978 0.921
Cumulative percentage variance of species data             15.8 27.8  37.8 46.2
Cumulative percentage variance of species-environment relation 22.1 38.9 52.9 64.7
Sum of all eigenvalues      1.000
Sum of all canonical eigenvalues      0.715

 

 

 

Biplots

Species Presence and Environmental Variables

 

Species Presence and Sites

Sites and Environmental Variables

 

 

 

RDA: Second Level of Analysis for Ant Assemblage

After determining with DCCA that RDA was the correct ordination follow-up to the PCA, these are the preliminary results with diagrams. The RDA ordination evaluates the ant assemblages in terms of environmental variables. The RDA indicated that the most significant factors were soil drainage and percent litter cover, so I only included these on the diagrams.

Full RDA:

**** Summary of RDA for Summer Species Presence****

Axes     1 2 3 4 Total variance
Eigenvalues

0.158

0.119

0.099

0.082

1.000

Species-environment correlations 

0.937

0.969

0.979

.9020

Cumulative percentage variance of species data               

15.8

27.7

37.6

45.8

Cumulative percentage variance of species-environment relation

23.6

41.4

56.2

68.5

Sum of all eigenvalues     

1.000

Sum of all canonical eigenvalues     

0.669

Biplot of Species and Environmental Variables:

 

Biplot of Species and Sites:

 

Biplot of Sites and Environmental Variables:

 

 

Beginning RDA for Ant Assemblages

I am continuing the analysis of the ant assemblage data in the Fort Worth Nature Center with an RDA (redundancy analysis) on species occurrence (or presence/absence data). I conducted a DCCA (detrended canonical correspondence analysis) for this data.  All the DCCA’s had short segments (less than 4) and thus, indicated that RDA was the appropriate follow-up to the PCA. All these tests were done using CANOCO.

The PCA for the environmental variables (a mix of soil and vegetation characteristics) indicated that the variables chosen were pretty good at discriminating the sites. The hypothesis is that the ant species use these same characteristics or some combination of them in their choice of habitat. Thus, the ants should also be indicative of the different habitats and therefore, possibly useful  as indicator species.

The PCA on the ant species had rather low values which I originally considered abysmal. However, low values are not unexpected for occurrence data. At any rate, it is the RDA in which both the environmental variables and species occurrences are considered which is the crucial point. The PCA on species occurrence did not include the environmental data.

I conducted RDA on the species occurrences for each month (June, July, and August) and for the summer pooled. The RDA also included forward selection of variables and a Monte Carlo test for significance. Although it is interesting to see a comparison of the significant factors by month, for species occurrence it is really the pooled data (summer) that is of importance.  The by month data is related to seasonality or activity  of the species and not just occurrence. (This is one of the difficulties of ant data — colonies can hang out in the ground or their nest for days, weeks, or months and since they are social, counting individual ants is not the same as counting individuals of other kinds of organisms.).

So, the results of the RDA:

There were 10 environmental variables used in the RDA: 5 for soil and 5 for vegetation. Soil variables included: drainage, soil penetration, latitude, slope, and depth to the restrictive layer. Vegetation variables included: ecological site, percent bare ground, percent litter cover, percent plant cover, and percent total cover. Of these variables, drainage and percent litter cover were consistently significant in the full RDA and when these sets of variables were used as co-variates. Drainage and percent litter cover accounted for 24% of the variation in the species occurrence data for the summer. This percentage varied somewhat from month to month.

The species-environment correlations for each RDA varied from about 70% – 99% with most in the 80s – 90s range. This indicates a strong relationship between the axes of the ordination and the species.

Still more of the RDA to think through and more follow-up is needed to look at the contributions of the variables to each axis. I also need to figure out how to clean-up the biplots and triplots.

Here are the RDA summaries and current biplots

Summer, full RDA:

**** Summary of RDA for Summer Species Presence****

Axes     1 2 3 4 Total variance
Eigenvalues

0.158

0.119

0.099

0.082

1.000

Species-environment correlations 

0.937

0.969

0.979

.9020

Cumulative percentage variance of species data               

15.8

27.7

37.6

45.8

Cumulative percentage variance of species-environment relation

23.6

41.4

56.2

68.5

Sum of all eigenvalues     

1.000

Sum of all canonical eigenvalues     

0.669

 

 

 

Summer, Soil:

**** Summary of Summer RDA with Partials ****

Axes   

1

2

3

4

Total variance

Eigenvalues

0.103

0.089

0.042

0.033

        1.000

Species-environment correlations

0.949

0.978

0.747

0.904

Cumulative percentage variance of species data

15.0

28.0

34.0

38.9

Cumulative percentage variance of species-environment relation

35.3

65.6

79.9

91.2

Sum of all eigenvalues     

0.686

Sum of all canonical  eigenvalues     

0.292

 

 

 

 

Summer, Vegetation

**** Summary of August RDA with Partials ****

Axes   

1

2

3

4

Total variance

Eigenvalues

0.123

0.070

0.055

0.037

        1.000

Species-environment correlations

0.937

0.923

0.927

0.898

Cumulative percentage variance of species data

18.1

28.4

36.5

42.0

Cumulative percentage variance of species-environment relation

43.2

67.7

86.9

100.0

Sum of all eigenvalues     

0.678

Sum of all canonical  eigenvalues     

0.285