Practical 3 – Part B – Research and analyses: Feedback

This is a feedback document for the research and analyses Part B practical. Please use the information provided here to help you with your write-up (Part C).

Sections 1-2

These sections dealt with the literature review and referencing.

The South African Journal of Botany referencing standard follows:

Bond, W.J., Midgley, G.F., Woodward, F.I., 2003. What controls South African vegetation - climate or fire? South African Journal of Botany 69, 79-91.

Note the following:

  • Journal names are written in full (i.e. not abbreviated) and, more importantly, are NOT in italics (styles for other journals often want the journal name to be in italics, this is not the case for SAJB)
  • You must provide both the issue number and page numbers, separated by a comma.
  • There is no comma between the journal name and issue number.
  • No first names for authors, only their initials separated by “.”
  • Comma separates the authors from the year of publication.
  • Learn this referencing style - there will be an exam question on reference formatting…

Section 3 – Life History Trait Comparisons

In the small study site where we conducted Practical 2 and Practical 3, 61 species were recorded. Of these, 54 could be clearly assigned to either Thicket or Fynbos. The others were discarded as they were alien invasive species, or invading indigenous grasses due to disturbance (probably the grazing animals on the reserve).

Obligate seeders consist of plants which are either killed out-right by fire and recruit thereafter from banks of seed buried in soil or encapsulated in woody fruits in the above-ground canopy (i.e. serotiny). Resprouters employ a very different strategy whereby rootstocks, thick trunks and branches survive fire and replace entire shoot canopies by sprouting from heat-resistant buds. Below is a summary of the obligate seeders and sprouters in our sample…

Table 1. Summary of the percentage (within each biome) of obligate seeders and sprouters in the Dune Fynbos and Thicket at the Practical 2 & 3 study site.

Resprouter.Reseeder

Fynbos

Thicket

Seeder

76

0

Sprouter

16

100

Unknown

8

0

Table 2. Summary of dispersal agents in the Dune Fynbos and Thicket at the Practical 2 & 3 study site

Dispersal.Mechanism

Fynbos

Thicket

Ant

32

0

Bird

4

86

Bird/tortoise

4

0

Mammal

4

0

Passive

24

3

Rain

8

0

Wind

24

10

Table 3. Summary of dispersal mechanisms versus seeding or sprouting in the Dune Fynbos and Thicket at the Practical 2 & 3 study site

Dispersal.Mechanism

Seeder

Sprouter

Unknown

Ant

31.6

0.0

100.0

Bird

5.3

75.8

0.0

Bird/tortoise

5.3

0.0

0.0

Mammal

5.3

0.0

0.0

Passive

21.1

9.1

0.0

Rain

5.3

3.0

0.0

Wind

26.3

12.1

0.0

Leaf traits

Table 4. Summary of percentage cover of the different Leaf Consistency Indices across Thicket and Fynbos sites (analysed separately)

LeafConsistencyIndex

Fynbos

Thicket

Fleshy (Semi-succulent)

0.1

10.8

Orthophyll

50.6

25.8

Sclerophyll

47.7

61.7

Succulent

1.5

1.7

Table 5. Summary of percentage cover of the different Leaf Size Indices across Thicket and Fynbos sites (analysed separately).

LeafSize

Fynbos

Thicket

Nanophyll

49.4

1.6

Leptophyll

34.7

1.4

Microphyll

15.9

97.0

HINT: The tables above would make great bar graphs.

4a) Moisture analyses

The crux of this question was the fact that the soil (and other) weights included the crucible (or bag/paper) weights too. This biases your results (the crucible has no moisture, so lowers your moisture values). So, you need to deduct the container’s mass from the wet weight and the dry weight before calculating the percentage moisture. This usually applies to other calculations too.

Thus, \[ Soil Moisture (\%) = \frac{(WetMass-CrucibleMass)-(Dry Mass - Crucible Mass)}{(Wet Mass - Crucible Mass)}*100\]

To calculate the percentage change (the symbol for change is \(\Delta\)),it is always:

\[ \Delta (\%) = \frac{(original_{value} - second_{value})}{original_{value}}*100 \]

**Figure 1.** *The difference in soil moisture content if the crucible mass is not removed. Failing to remove the mass of the container generates a significant bias in the moisture estimations. For these calculations, the wet mass is 10 g, the dry mass is 9 g, and thus the soil moisture is 10%.*

Figure 1. The difference in soil moisture content if the crucible mass is not removed. Failing to remove the mass of the container generates a significant bias in the moisture estimations. For these calculations, the wet mass is 10 g, the dry mass is 9 g, and thus the soil moisture is 10%.

4b) Light infiltration

Why were light levels measured above and below the canopy? If light levels remained constant all the time, then we would not need to measure light levels above the canopy. However, light levels fluctuate quite substantially through the day (think of mid-day sun versus mid-morning). Also, light levels may fluctuate at much shorter time scales — e.g. changes in cloud cover. Thus, light levels measured below the canopy are influenced by light levels above the canopy. So, what we are primarily interested in is the degree of light infiltration — i.e. the degree of light that makes its way through the canopy. So, to reiterate, light was measured above the canopy to control for general fluctuations in light levels.

**Figure 2.** *The distribution of light intensity readings above and below the canopy (lumping Fynbos and Thicket sites)*

Figure 2. The distribution of light intensity readings above and below the canopy (lumping Fynbos and Thicket sites)

Why were there three replicates of light intensity? For two reasons. The first I was not expecting you to know: 1) the light meter is quite sensitive to the angle that it is held to the sky — have a look (in the dataset) at the high variation in the above-canopy readings within a plot. If these were taken in quick succession, they really should be very similar. But they aren’t because the slight shifts in angle of the light meter. However, the second reason should be more obvious: 2) the light penetration under the canopy is highly variable (e.g. dappled shade with some sunny spots and some shady spots). We took three readings to average out these effects to get an indicative below-canopy reading.

How should you calculate light infiltration? It is the proportion of light that starts above the canopy and makes its way below the canopy to the ground. Thus, \[ Light Infiltration = \frac{Avg(LightIntensity_{BelowCanopy})}{Avg(LightIntensity_{AboveCanopy})} \]

4c) Sample size variance at the species level in the flammability dataset.

There is a problem when you are likely to have a high degree of variance in your data (i.e. low precision). To get an accurate value for a parameter, one therefore needs to have a number of replicates and we assume that the average of those replicates are close to the true value of that given parameter. Thus, species with only one or two values per trait cannot provide reliable averages for those traits and must be discarded. Table 6. The number of replicates per species for the flammability traits (i.e. flammability index, branch moisture, size of material)

Species

Replicates

Coleonema pulchellum

5

Euclea racemosa

4

Helichrysum teretifolium

3

Metalasia muricata

4

Osteospermum moniliferum

4

Phylica ericoides

4

Sideroxylon inerme

4

Tarchonanthus littoralis

3

4d) How to calculate the flammability index?

The flammability index is a composite of three different parameters, specifically the maximum temperature (MT), the burn rate (BR) and burnt biomass (BB) category. These were relativised, i.e. their values were shifted to be between 0 and 1.

Thus the way that the flammability index (FI) was calculated was:

\[ FI = \frac{MT_{sample}}{max(MT_{all values})}+\frac{BR_{sample}}{max(BR_{all values})}+\frac{BB_{sample}}{max(BB_{all values})} \]

Section 5. Figures…

**Figure 5a.** *The mean and standard deviation (whiskers) of the percentage light infiltration between Fynbos and Thicket plots (n=15 per vegetation type)*

Figure 5a. The mean and standard deviation (whiskers) of the percentage light infiltration between Fynbos and Thicket plots (n=15 per vegetation type)

**Figure 5b.** *The mean and standard deviation (whiskers) of soil temperature at 5 cm depth between Fynbos and Thicket.*

Figure 5b. The mean and standard deviation (whiskers) of soil temperature at 5 cm depth between Fynbos and Thicket.

**Figure 5b.** *The temperature difference between ambient temperature and soil temperature at 5 cm depth*

Figure 5b. The temperature difference between ambient temperature and soil temperature at 5 cm depth

**Figure 5c.** *Soil moisture (%) between Fynbos and Thicket plots.*

Figure 5c. Soil moisture (%) between Fynbos and Thicket plots.

Shew, why do we get a standard deviation bar going into negative soil moisture in Fynbos? That makes no sense. Remember, a similar thing happened in Prac 1. Standard deviations are used to indicate the variation around the mean, and it assumes a normal distribution (i.e. most values are located around a central mean). In this case, there a few odd readings — one in particular — see the figure below.

## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.

If we remove that outlier (~50% soil moisture is really really high for the conditions we were sampling in), then the graph above looks like this…
**Figure 5cUpdate.** *Soil moisture (%) between Fynbos and Thicket plots.*

Figure 5cUpdate. Soil moisture (%) between Fynbos and Thicket plots.

So, whenever you get really large sd values – there’s like an explanation. Go and look for possible errors in the data!

**Figure 5d.** *Soil Organic Carbon (%) between Fynbos and Thicket plots.*

Figure 5d. Soil Organic Carbon (%) between Fynbos and Thicket plots.

**Figure pH (A new figure).** *Mean and standard deviation (whiskers) of soil pH between Fynbos and Thicket plots. Note that it is a convention **not** to use barplots for pH values*.

Figure pH (A new figure). Mean and standard deviation (whiskers) of soil pH between Fynbos and Thicket plots. Note that it is a convention not to use barplots for pH values.

**Figure 5e.** *Species-level flammability between Fynbos and Thicket species.*

Figure 5e. Species-level flammability between Fynbos and Thicket species.

If we run a Student’s t-test (in this case non-paired), this will test whether there are signficant differences between the flammability indices of the two vegetation types.

## 
##  Welch Two Sample t-test
## 
## data:  FIu by VegType
## t = 2.2456, df = 4.3489, p-value = 0.08269
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  -0.0857531  0.9509364
## sample estimates:
##  mean in group Fynbos mean in group Thicket 
##              1.701883              1.269292

You will notice that the p-value is low (~0.08), but not significant. This is largely due to the small sample size and that Tarchonanthus appears to be highly flammable. So, one sample that bucks the trend means there are not enough degrees of freedom (in this case, the number of species per biome) to be able to obtain a value <0.05 (the golden threshold where we can say something is statistically significant.). Oh, well - we were close. But more species would have clarified the matter.

**Figure 5f.** *Species-level mean flammability (n=3) and plant moisture content (%) between Fynbos and Thicket species.*

Figure 5f. Species-level mean flammability (n=3) and plant moisture content (%) between Fynbos and Thicket species.

The r values above can be transformed into an \(r^2\) value (\(r^2=(-0.79)^2=0.62\)). The \(r^2\) value of a linear regression can be interpreted as the percentage that the dependent variable (in this case the Plant Moisture) explains the independent variable (in this case Flammability index). In ecology, \(r^2>0.4\) are quite rare.

**Figure 5g** Species-level flammability versus percentage fine plant part between Fynbos and Thicket species.

Figure 5g Species-level flammability versus percentage fine plant part between Fynbos and Thicket species.

What do \(r^2\) values mean?

The \(r^2\) value tells you how strongly correlated to variables may be, in the cases above this is the correlation between plant moisture (%) and flammability index, or fine plant material (%) and the flammability index. Another way to think of it is what percentage of your y value (in this case the flammability index) is explained by your x value (plant moisture or fine fuel). We have values values ~0.62 (or ~62%). This might seem quite low — only 62% of the flammability index is explained by plant moisture — but such \(r^2\) values are actually quite high for an ecological study. Anything from 30-60% is considered biologically meaningful. Values greater than 70% are usually considered with suspicion, i.e. the researcher must have done something wrong because such such high \(r^2\) values are very very rare in ecology. Many students obtained an \(r^2\) values of ~90% in their Part B write-ups.

What do standard deviation bars on the bargraphs mean?

One statistical fact that is often overlooked is that mean values are meaningless without some indication of the variation in the data. For example, if the mean height for all the students in the class was 170 cm this does not tell us whether there are sort or tall people in the class. We could also report the range (i.e. minimum and maximum values) of height values. Thus, if we compared two classes with the same mean value, the range values tells us (somewhat) how well this mean value is representative of the class. Thus, minimum and maximum values of 165 cm and 175 cm, respectively, tells us that the mean value is very representative of the students in the class. However, values of 105 cm and 220 cm, respectively, indicate that the mean value is certainly not representative of all students.

So, where does standard deviation come in? A definition of standard deviation is “a quantity expressing by how much the members of a group differ from the mean value for the group”. Another definition is “a measure that is used to quantify the amount of variation or dispersion of a set of data values”. It is less affected by outliers than range (minimum and maximum values). Thus, reporting mean values with standard deviations — either in your figures or tables — should be your standard practice when you have more than six replicates (standard deviations vary and in themselves become meaningless when replicates sizes are low).

One use of the standard deviation is to eyeball whether values are significantly different from one another. If the standard deviation bars do not overlap, then there is a high likelihood that there is a significant difference between the recorded values.

For example…

**Additional Figure A.** *The mean and standard deviation between Fynbos and Thicket for some variable. The standard deviation bars do not overlap, and this strongly suggests that the values between the two vegetation types are **significantly** different from one another.*

Additional Figure A. The mean and standard deviation between Fynbos and Thicket for some variable. The standard deviation bars do not overlap, and this strongly suggests that the values between the two vegetation types are significantly different from one another.

**Additional Figure B.** *The mean and standard deviation between Fynbos and Thicket for some variable. The standard deviation bars, in this case, **do overlap**, and this suggests that although the mean values for the parameter differ between the vegetation type, this difference is **not significant**.*

Additional Figure B. The mean and standard deviation between Fynbos and Thicket for some variable. The standard deviation bars, in this case, do overlap, and this suggests that although the mean values for the parameter differ between the vegetation type, this difference is not significant.

This is probably a good time to remind you about the definition of “significant”: “sufficiently great or important to be worthy of attention; noteworthy.” Thus, only when you have a statistically significant result can you say that there are major differences in the values between populations of data (in this case values recorded for some parameter from the Fynbos and Thicket).

Thus, for your write-up return to your figures, make sure you have correctly calculated the percentage moisture etc. (Remember about the issue with the ultra-fine, fine and coarse plant parts – see above). Include standard deviation bars in your bar plots, and then use these to write-up your scientific manuscript (Part C of Practical 3). To make things easier, split the results and your discussion up into:

  • Environmental data, and
  • Plant trait data.

The overarching question for this practical is: why do we get two completely physiognomically (structurally) different vegetation types (i.e. Fynbos and Thicket) in these dune systems?

There are two potential hypotheses that we will explore in this practical to answer this question:

  • Thicket and fynbos vegetation occur in different microenvironments.
  • Thicket and fynbos communities have different functional traits (e.g. leaf morphology, plant moisture, flammability, dispersal and recruiting mechanisms). It is these traits, and the repercussions thereof, that allow for the coexistence of these communities.

Remember to include relevant topics in your introduction and cite relevant literature. In your results, you must write-out your general findings (e.g. “No significant difference was observed between Fynbos and Thicket variable (Figure 1)”). Your discussion is not a rehash of your results, rather you discuss your findings in relation to other studies or extrapolate your results into possible causes; e.g. you are likely to find that Fynbos and Thicket do differ in their physiognomic traits – so what? How does this allow these two vegetation types to coexist?

Something else for you to consider — are all of our soil measurements (i.e. moisture content, organic content, pH) good representations of the environment. They were taken at a snapshot in time. Maybe their values may shift over time and we’re missing an important driver simply due to when we’ve sampled. Something always to think about in tightly-linked correlative studies.

An additional point, the questions in section 4 (sample size, above and below canopy readings, etc.) are a very minor component of your write-up. These were meant to get you thinking about your data.

Due date for your scientific write-up: Monday 7th May via TurnItIn…