Discover more from The Ecological Mind: ecology, research, creativity
How to do 'good' ecological research: 13 points to consider
Thirteen general points to consider when you are about to start a new experiment or study
(Updated 6.2.2023 and 19.08.2023; point 12 was input received by Justine Karst, and point 13 was suggested by anaraa_sh on twitter)
Our ambition as a lab is to do the best possible research in ecology. This presents an ongoing challenge, struggling to find the best ideas and questions, or refining existing questions in novel ways, and then to match these questions with the most promising experimental or study design.
Over the years (now over 15 years, actually) of reading papers in our lab’s journal club, we found that there are a number of traits that we tend to find appealing in others’ work, and I have assembled a list of aspects which we strive for also in our own studies.
So here are my main points to think about when starting a new ecological study
Consider the fundamental trade-off in empirical work: mechanistic resolution vs. ecological realism; you can’t optimize both at the same time, so you will need to decide what’s important to you. And it will take multiple studies to cover the entire spectrum.
Maximize ecological context (in our case this typically means using soil, plants, and soil organisms from the same ecosystem when setting up a greenhouse experiment, for example). Match this context to your question, for example use an agricultural soil, not a forest soil, when performing an experiment with a focus on agroecosystems.
Ask general questions of interest to many ecologists, like about biodiversity, global change, or ecosystem process rates. This way more people will tend to be interested.
Seek firm grounding in applicable ecological theory; no study exists in a vacuum.
Familiarize yourself with a group of organisms or an ecological system so that you can ask better questions in your study, and so that you can judge if your system is free of bias.
Pick a study system that is suitable for the question you want to answer; this may mean deviating from your ‘favorite’ system (this often remains a noble goal). One could also refer to this as the ‘primacy of the question’.
Maximize external validity of the study, that is the degree to which your results can be generalized. This is of course always limited by logistics and cost.
Think carefully through the sampling and experimental design before you start; for example to avoid pseudo-replication at any level (replicating at the right level is sometimes more difficult that expected). Think about statistical analysis before you begin with your study, not only when you have to analyze the data.
Pay attention to data management and archiving (does anybody want to look at my impressive collection of 3 1/2 inch floppy disks from my PhD?). This can help make your data useful also to others.
Consider setting up experiments that can also last for longer periods of time to study potential long-term effects; these tend to accrue value with time, but be careful, since such long-term studies can also come with a lot of commitment.
When faced with experimental (or sampling) design choices, favor the choice that promises to deliver recommendations for management or other applied uses of your data - all else being equal.
Consider alternative explanations or mechanisms. Include different possible explanations for the mechanism or process of interest in the study design. This helps to examine the problem from multiple perspectives and avoid confirming one’s own assumptions.
View your experiment as an opportunity to collaborate with others. What could you get done if you connected with colleagues on this particular study? This can be included in the study design; in addition to the ‘must-haves’ of your list of response variables, you can plan for additional measurements, and that also means securing samples in the appropriate ways (e.g. freezing).
What would you add to the list?
The ten points you proposed are very interesting and important - definitely a great resource! Here's my two cents:
I particularly like point 10, as it is not often done but so many patterns change with time and cost/reward can be high.
I believe the first point is of particular relevance, and it would be hard but important to determine the scope/boundary of your study and how many studies would actually be needed to cover the entire spectrum. Performing a literature review beforehand is essential in order to gain valuable insights and plan the study, but also by knowing the methodologies used. Here, an approach based on systematic quantitative literature review (https://doi.org/10.1080/07294360.2013.841651) that could help for example. There are various tools available to do a detailed literature search. It could be beneficial to have a separate point dedicated to this!
It's the current trend by the use of AI could probably be advantageous for the first points. Actually, something chatGPT suggested that my be relevant: "Gain permission to access necessary sites" and "consider the ethical nature of the research".
With respect to the statistical analyses, it would be useful to run analyses on previous data obtained from literature or even toy examples in order to evaluate if the experimental design is appropriate and well adapted to the question. Additionally, it may be useful to conduct a power analysis to know if any signal could actually be detected and therefore if the hypothesis would be tested appropriately. Finally, when all is well prepared, on might consider preregistering the study (https://www.cos.io/initiatives/prereg), it is likely that in the future, the study with such considerations, will become the norm, for good reasons.
Thanks!
1. Application of modeling approach to make precise predictions beforehand. Not always is it necessary to assess expensive and valuable data, but playing through the scenario would help a lot.
2. Handling every Ressource as if they are absolutely valuable, and lost irreversibly after use. They actually are and keeping this in mind would change the way we deal with resources of any kind.