Critical analysis of arthropod field project

From BiodivBorneo09

Hypothesis: there are more ground-dwelling arthropod species in closed canopy rainforest than in gaps. Leaf litter under the forest canopy is less likely to be affected by environmental disturbance and temperature extremes, thereby providing a more stable habitat.

H0: there are no differences in the diversity of ground dwelling arthropods found in the rainforest compared with those found in gaps.

Is the hypothesis well written (or practical)?

• What is primary rainforest? [is it variable?]
• What are gaps? [are they variable?]
• Was it easy to find these different forest types on the ground? [Sounds OK in theory but in practice it is more difficult]

If there is variability in ‘rainforest’ and ‘gap’ habitats then this variability could affect ground-dwelling arthropods and our results.

Does our sampling design really test our hypothesis?

Did we sample ‘ground-dwelling arthropod diversity’?

• Could other sampling methods add anything to our results?
• Different size pitfalls – some arthropods may be too big to fit in the pitfall
• Pitfalls with ethanol rather than caustic detergent
• Guided pitfall trap
• Sticky traps
• Bait traps (honey, faeces, fruit – buried or on the surface)
• Deeper soil samples
• Winkler sorting
• Tolgren or Berlese funnels
• Did we miss anything, is the arthropod community variable?

What about the rationale for our study (environmental disturbance and temperature extremes), if we find a significant difference in arthropod diversity, can we attribute this to the different disturbance regimes? What other ecological variables could explain/contribute to the results?

• soil type
• age of the gap
• size of the gap
• slope
• aspect
• altitude
• soil PH
• weather (rain)
• season
• fruiting
• how far from the edge should we sample in the rainforest (edge effect)
• diversity of tree species
• time of day (litter sample)
• length of time for sample (consistent?)

Have we controlled for any of these variables? Do we need to control for all of these variables? With three replicates, noise (variation) could be a factor in our results. More samples or more specific sampling (controlling for variables) could reduce this noise.

Assuming the study is well conceived and well designed, and we have controlled for all the other variables, what other human variables could bias the results?

• Bias in selecting the exact sample site within each habitat type.
• Experience – how well were the traps set?
• Different people setting traps – how consistently were the traps set? (Dave and I split the trap setting so that both groups did a forest and gap – we did not have one group setting gap traps and a different group setting rainforest traps).
• How accurate was the measurement of leaf litter depth?
• Order in which we set the traps – different time span.
• Sampling effort (enthusiasm, time expended)

Have we controlled for these variables (bias with different people setting traps)? Could we control for all these variables?

Was there error in gathering our results?

• How long did each trap remain – did we take down in same order they were installed (so that traps remained in the field for the same amount of time)?
• Were we consistent in sorting the samples?
• Pre-sorting leaf litter – time expended, consistency of disturbance, thoroughness of emptying bag
• Sub-sampling consistency
• Emptying pitfall traps
• Identifying arthropods – taxonomic competence
• Enthusiasm – of course everyone gave full attention to the sorting!
• Counting accuracy!
• Recording data – KEEPING TRACK OF THE DATA!

What other tests could we do with our sample design and results?

• Partition by site (since samples were paired)
• Partition by order in which we set up the traps
• Partition by group (Dave vs Rod)
• Partition by sampling method
• Other?

Can we extrapolate our results to some general rule?

• Will our results hold true in all cases, in all forest types, in all places (e.g. Neotropical)
• Will our results hold true at all times (e.g. every season, every year, El Nino, during mass fruiting)?

How could we improve the sampling and interpretation?

• Pseudoreplication [If we took all three samples in the same large gap – would this be acceptable?]
• Sampling design
• Randomisation
• Number of replicates
• Control for variables
• RECORD THE COORDINATES FOR SAMPLE SITES IN THE PLOT – we should know where we took our samples (plot sites on map) so we know what trees, soil type, etc., are present.

What practical application could such a study have?