Comparing decomposition rates of leaf litter from different plant functional types (PFTs)

HYPOTHESES:

1. Broadleaf litter decomposes faster than needle litter.

2. Increasing temperature increases decomposition rate.

3. The effect of temperature on decomposition depends on litter type.

EXPERIMENTAL DESIGN:

Incubate six different litter types at two different temperatures.

For the experiments we will use three species of deciduous broadleaved litter

• Aescutus hippocastenum – Horse Chestnut

• Platanus x hispanica – London Plane

• Tilia cordilolia – Hornbeam

And three species of evergreen needleleaved litter:

• Xanthocyparis nootkanensis ‘Glauca’ – Nootka Cypress

• Pinus coulteri – Coulter Pine

• Juniperus chinensis piura – Juniper

All tree species grew on the same soils around Sutton Bonington. Moisture conditions will optimal for decomposition, temperatures will be 15 and 20°C. Incubations will run for 2 months. All litter

were gathered early October 2016 and were air dried immediately after collection. Care was taken to only collect recently fallen litter as assessed by color and litter condition.

PROCEDURE:

3 October 2016

1. Weigh out two sets of 5 g for each litter type onto mesh cloth.

2. Record the exact initial weight to two decimal places.

3. Tie the mesh cloth up and label each one with your group number, litter type and respective temperature (15°C and 20°C).

4. Wet the samples with deionised water and place one set in the 15°C and one in the 20°C incubation tray.

12 December 2016

1. Weigh the dried samples and record the weights to two decimal places.

DATA ANALYSIS:

In Excel calculate % Mass remaining for the different litter types.

In GenStat create a table of means and SEMs and use this to draw a barchart in Excel.

In GenStat carry out a multifactor ANOVA with temperature and functional group as your fixed effects to test your hypotheses.

Write a report including your graph and two-way ANOVA results and make comments about whether your data fitted the assumptions of ANOVA revealed by model checking plots.

Data analysis work-flow:

1. Calculate % mass in the Excel workbook “Decomposition experiment data_2016. Note that the there are two spreadsheets, one for each temperature condition.

Note also that groups with missing data will be left out from further analysis.

2. Copy the % data for each temperature, using “paste special>values”, into the spreadsheet named “For Genstat”, putting the 20oC data below the 15oC data.

• TIP for highlighting data in Excel: click the top left hand corner cell of your table then use SHIFT/CTRL/right arrow (to select all filled cells to the right) and SHIFT/CTRL/down arrow (to

select all filled cells to the bottom).

3. Copy the “For Genstat” data and import it into Genstat through the Clipboard.

• Remember, convert “Functional Group” and “Temp” into Factors

• Nomenclature – they are factors with two levels i.e. 15oC or 20oC are levels of factor temperature.

4. Make a table of means and standard error of the means (SEMs) using the menu options “Statistics>Summary Statistics>Summary Tables”:

• Double click to enter the variate (mass remaining) and factors (functional group and temp).

• Choose “Means” and in “More” choose “Standard of Error of Mean”, like this:

• Click “Run” (if you can’t see the table go to the output window to find it!)

• Copy this table into Excel spreadsheet “For Graphing”

5. In Excel plot a histogram of the means for each functional group split by temperature.

• Add custom s.e.mean (SEM) error bars specific to each column using “Error Bars>More Options>Custom>Specify value” for each factor level.

6. Now use two-way-ANOVA from the “Analysis of Variance” menu to assess the influence of both temperature and functional group on decomposition (mass remaining %). Include the interaction

(this will be on by default), so you can see if decomposition for the different functional groups responds differently to temperature.

• In “Options” ask for “Model checking plots” in the graphic output.

7. Run the analysis and first look at the model checking plots:

• Is the top left plot roughly bell-shape and the bottom left plot a diagonal line, suggesting normally-distributed residuals (residuals are distances between your data and the mean)?

• Is the top right plot evenly distributed above and below the mean line, suggesting equal variance and are the groups similar in variance? If not, please comment upon this in your report and

read the notes below on page 4 about non-normal data, but we’re not going to do anything more about it today.

8. Look at the results of your ANOVA and copy them to Excel or Word to save them.

9. Write a paragraph to summarize the results of your analysis in your report, referring to the trends seen your graph and quoting the F ratios (remember ‘v.r.’, variance ratio, means F

ratio), degrees of freedom and associated p-values (remember ‘F. pr.’means probability associated with the F ratio), for the main effects (that’s each factor on their own i.e. temperature or

functional group) and the interaction.

Notes on normality and equal variance – or lack of it

ANOVA assumes that you have independent samples and that the residuals (distance between the mean and your data points) are normally distributed, so they make a bell-shaped curve if plotted by

frequency of residuals of each size (like the top left model checking plot) and that they have equal variance (distance from the mean), both above and below the mean for each grouping (here that’s

each temp and functional group combination) and between different groups.

In previous years model checking plots showed that the data was not normally distributed and/or did not have equal variances between groups, which does not fit the assumptions of ANOVA. This may

have been due to extreme outliers AND because percentages will mostly lie between 0 and 100, so it’s not fully continuous data.

We tried log10-transforming the 2014 data, to bunch it up, but this did not make it look more normal.

A non-parametric test could be used that doesn’t assume a normal distribution, but instead uses the ranks of the data (1st, 2nd, 3rd etc), but you can only run the equivalent of a one-way ANOVA

(the non-parametric version is called a Kruskal Wallis test) looking at the influence of one factor at a time on your continuous variable. There is no non-parametric equivalent to the two-way

ANOVA, so you can’t look at the interaction with a non-parametric test. The interaction is the extra information that you get from a two-way-ANOVA that you wouldn’t get from running two one-way

ANOVAs.

The non-parametric ANOVA, Kruskal Wallis test, can be found in GenStat in the “Statistical Tests” menu and for our experiment we’d choose “One variate with grouping factor”. The output gives an H

statistic (rather than F), degrees of freedom (this is calculated as the number of factor levels – just two for temp or functional group – minus 1) and an associated p-value. One could state the n

number per group as well for completeness.

If there is a lack of normality in this year’s data, please show the model checking plots and comment upon their shapes in your report, but we do not expect you to do anything further about it in

your analysis, the above notes about what to do are just for completeness sake.

WRITE UP INSTRUCTION

Formatting:

Write your report in the form of a laboratory report. The report should be no more than two pages long, font 12, and minimum of 2 cm margins. I will not read material beyond the two page. You

should write the report for a target reader with subject knowledge but not prior knowledge of the practical (e.g. one of the Environmental science academics).

Structure of the report:

The report should contain the following sections:

Title: This should describe the essence of the study

Introduction: This section should be brief and outline the rationale of the experiment, it should end with stating the hypotheses you tested.

Methods: This section should briefly outline the experimental set up and how the samples were treatment. You also need to make a brief description of your data analysis, specifically you need to

include the equation for calculating % mass remaining, what type of statistical test you used, which factors/fixed effects you included in the analysis and what stats package you used to analyse

the data. The data analysis section should be the last section in the methods.

Results: In this section you need to describe the main trends in your data based on your statistical analysis and your figures. The results section must include both a figure illustrating the

treatment effects with a correct caption. You need to refer to this when you describe you data in the text. You need to report your statistics in the text or figure caption depending on what is

most appropriate. You should include the barchart of the % litter remaining data in this section.

Discussion and conclusion: In this section you should briefly compare and contrast your results to the wider literature and address your hypothesis. Ensure that your conclusions are underpinned by

your results.

References: You need to list all the references you have used. Make sure you reference correctly, I suggest you use the Harvard referencing style.

Additional guidance

• Based on the formative lab report I would like to give some additional advice on how you can improve your reports:

• Latin species names should always be in italics.

• Do not include title in you figures when you write reports.

• Do not include a border around the figures when you make your figures.

• Make sure the legends are clear and informative.

• Write informative figure captions but do not include results in these.

• Always add units/axis title to your axis depending on what’s appropriate. If you do not include units/axis titles you will not get more than a third class mark!

• Use a sensible number of significant digits when you report mean values, these need to reflect the precision or your data collection (e.g. no more decimal places than you measured the

weights to).

• Keep up good reporting and interpretation of the stats. Make sure you understand what significant means, i.e. p-values less than 0.05. If you are unsure how to do this look at the model

answers from the prevision practical class sessions on Moodle. If you do not include the statistical analysis in the report you will not get more than a third class mark!

• Look at the individual feedback on your formative lab report to improve your work! Lots of these are still in my pigeon hole outside B47, Life Science Building, UP.