Introduction

As a result of biodiversity increasing, it’s essential we continue to investigate the impact this has on surrounding animal species. Biodiversity can be classified as the number of the plant species present, also known as density. Arthropods are the most common animal phylum, thus are highly abundant and easy to examine. We chose to look at ants in particular because they are highly diverse and abundant, well studied and easy to collect and identify different species. Previous studies suggest that urbanisation and plant density affect the abundance of ants and the number of ants morphospecies (Buczkowski & Douglas). We looked at two different areas of density (high and low) to determine if it has an effect on ant morphospecies. It will allow us to improve our understanding of ant ecology and the effect that environment has on ants.

Aim: To determine if biodiversity has an effect on ant morphospecies.

Null hypothesis: The mean of ants present in low diversity area is equal to the mean of ants present in the high diversity area

H0:μ1=μ2

Alternate hypothesis: The mean of ants present in low-diversity area is not equal to the mean of ants present in the high-diversity area

H1:μ1≠μ2

Method

The experiment was conducted over 8 different locations in bushland in the Sydney area, each with 3 replicates. The low-density areas contained 1-2 different plant species and the high-density areas contained 4+ plant species. One 1m X 1m quadrant was randomly placed in a low-density area and one in a high-density area of bushland. Three ant traps (composed of filter paper and honey) were randomly placed within each quadrant. After 30minutes each ant trap was carefully picked up and photos were taken of both sides. The different species of ants were identified based on colour, shape and size.

Analysis

Exploratory Analysis

Treatment = plant density (LOW or HIGH) Replicate = location

library(readxl)
library(readxl)
ant <- read_excel("~/Downloads/stat data .xlsx")

str(ant)
## Classes 'tbl_df', 'tbl' and 'data.frame':    24 obs. of  3 variables:
##  $ Location    : num  1 1 1 2 2 2 3 3 3 4 ...
##  $ low density : num  2 2 1 1 3 2 2 3 2 2 ...
##  $ high density: num  4 4 3 5 4 3 4 3 6 6 ...
ld <- as.factor(ant$`low density`)
hd <- as.factor(ant$`high density`)

location <- ant$Location

library(readxl)
ants <- read_excel("stat data .xlsx", sheet = "Sheet2")

str(ants)
## Classes 'tbl_df', 'tbl' and 'data.frame':    48 obs. of  3 variables:
##  $ Location : num  1 1 1 1 1 1 2 2 2 2 ...
##  $ Abundance: num  2 4 2 4 1 3 1 5 3 4 ...
##  $ Density  : chr  "LOW" "HIGH" "LOW" "HIGH" ...
density <- as.factor(ants$Density)



species <- ants$Abundance
location1 <- ants$Location

cor(ant)
##                 Location low density high density
## Location      1.00000000 -0.08775403   -0.1251252
## low density  -0.08775403  1.00000000    0.2964665
## high density -0.12512521  0.29646654    1.0000000

Hypothesis Testing

A significance level of 0.05 has been selected for hypothesis testing.

T-Test

t.test(species ~ density)
## 
##  Welch Two Sample t-test
## 
## data:  species by density
## t = 7.5327, df = 37.173, p-value = 5.435e-09
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  1.644873 2.855127
## sample estimates:
## mean in group HIGH  mean in group LOW 
##              4.125              1.875

The two-sample T-Test shows a p-value of 5.435e-09, rejecting the null hypothesis and thus accepting the alternative hypothesis.

Boxplot

boxplot(species ~ density, col = c("lightgreen", "lightblue"))

The boxplot shows more a higher abundance of species in areas with high plant density compared to areas of lower plant density, with some overlap of whiskers.

ANOVA

Low Density as a Function of Location

ld.aov <- aov(location ~ ld, data = ant)
summary(ld.aov)
##             Df Sum Sq Mean Sq F value Pr(>F)
## ld           2   1.74   0.872   0.147  0.864
## Residuals   21 124.26   5.917
# check assumptions
plot(ld.aov)

P value not significant.

High density as a Function of Location

model1 <- aov(location ~ hd)
summary(model1)
##             Df Sum Sq Mean Sq F value Pr(>F)
## hd           4  36.36   9.090   1.927  0.147
## Residuals   19  89.64   4.718
plot(model1)

P value is insignificant.

Two-Way ANOVA

ant.aov <- aov(species ~ density, data = ants)
summary(ant.aov)
##             Df Sum Sq Mean Sq F value   Pr(>F)    
## density      1  60.75   60.75   56.74 1.46e-09 ***
## Residuals   46  49.25    1.07                     
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
plot(ant.aov)

library(emmeans)

comp <- emmeans(ant.aov, "density")
plot(comp)

pairs(comp)
##  contrast   estimate    SE df t.ratio p.value
##  HIGH - LOW     2.25 0.299 46 7.533   <.0001

This shows that the number of species is significantly different in areas of low and high plant density, with a very low p-value of < .0001.

loc.aov <- aov(species ~ density + location1, data = ants)
summary(loc.aov) # location as a block
##             Df Sum Sq Mean Sq F value  Pr(>F)    
## density      1  60.75   60.75  56.159 1.9e-09 ***
## location1    1   0.57    0.57   0.528   0.471    
## Residuals   45  48.68    1.08                    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
plot(loc.aov) # check assumptions

ANOVA shows that plant density is significant, whilst location is not significant.

library(lsmeans)
## The 'lsmeans' package is now basically a front end for 'emmeans'.
## Users are encouraged to switch the rest of the way.
## See help('transition') for more information, including how to
## convert old 'lsmeans' objects and scripts to work with 'emmeans'.
lsmeans(loc.aov,
        pairwise ~ density,
        adjust = "tukey")
## $lsmeans
##  density lsmean    SE df lower.CL upper.CL
##  HIGH      4.12 0.212 45     3.63     4.62
##  LOW       1.88 0.212 45     1.38     2.37
## 
## Confidence level used: 0.95 
## Conf-level adjustment: sidak method for 2 estimates 
## 
## $contrasts
##  contrast   estimate  SE df t.ratio p.value
##  HIGH - LOW     2.25 0.3 45 7.494   <.0001

A very significant p-value produced by lsmeans between high and low plant density supports the alternate hypothesis.

The distribution of the high density treatment is normal, with a mean of 4.13, and the confidence intervals (95%) 3.63 and 4.62. Whereas the distribution of low density treatment is skewed to the left, the mean is 1.88, and the confidence intervals (95%) 1.38 and 2.37. The location did not have a statistical effect on the abundance of ant species (p>0.05). The correlation of all variables are weak. A t-test confirmed that there is a significant difference between the high density treatment and low density treatment (p<0.001). The assumptions of the tests were confirmed with graphical post-hoc testing; residuals verses fitted, normal Q-Q, scale-location plot, and residuals verses leverage.

Conclusion

The statistical analysis reinforces the results of previous studies, that have found that ant abundance declines in association with increasing disturbance (Buczkowski & Douglas). Using plant density as a metric for disturbance we highlighted the role of plant density in habitat utilisation of many ant species. Disturbance had a statistically significant effect on the mean abundance of ant species (p<0.001). The location was statistically insignificant (p>0.05), confirming the difference in mean was related to the density of plants and not variation across the locations.

References

Buczkowski, G & Douglas, R.S. 2012. The effect of urbanization on ant abundance and diversity: a temporal examination of factors affecting biodiversity. Urban Ant Diversity, 7(8): 1-9