The current pandemic is causing widespread disruptions to research, as travel, universities, labs and field stations shut-down. As a PhD student your timeline for submission of your thesis probably won’t be moved back because of the disruptions. Funding for scholarships still has the same limits as always. From what I’ve seen, universities are expecting students to adapt their work to the disruption and will not be putting forward blanket extensions to all PhDs.
Understandably this situation creates a lot of anxiety.
One way to adapt your research is to replace field or lab work with desktop based research work. Whether this is appropriate for your PhD is a case-by-case decision and something you should discuss with your supervisor and mentors.
A desktop based project might involve analysis of existing datasets, modelling or a literature review.
It seems like pursuing desktop based work is the advice university graduate schools and many supervisors will be giving their students. So thousands of researchers worldwide are about to embark on desktop based work, whether they have experience in this style of research or not.
As an ecosystem modelling, desktop based work is my bread and butter. So here’s a few things to think about if you are considering switching some of your field or lab research to modelling or data analysis.
Many people underestimate how long it takes to develop models or data-analysis, just as I tend to under-estimate how complex field work will be. A story. Researchers in Australia wanted to watch fruit bats with a $7000 thermal camera. Sounds easy to me, just stick the camera up a tree. Problem is, coconut crabs might find the camera and rip it apart.
There are set backs in modelling too, that can be just as time consuming (though often less exciting). Like, the spatial data you want to process fills up your RAM, sending you on a 3 week journey looking for a suitable cloud-based processing service. Or the R package you’ve been using turns out to have a major bug.
I commonly get asked to review or edit modelling papers. There are lots of common mistakes I see from groups not used to modelling. I could fill a whole blog with tips on that, but here’s a few major ones
Applying an inappropriate modelling to the research question you want to answer. There are lots of models out there you can just download and start using. But models are usually developed for quite specific research questions, it may not be appropriate to use that same model for addressing different questions.
Basic statistical mistakes, like treating psuedo-replicates as true replicates or mis-interpreting p-values.
Absence of sensitivity analysis, in the general sense (if you don’t know what that means, time to do some reading).
There’s a couple of things you can do to avoid these mistakes.
First, get quantitative collaborators on board early and communicate with the regularly. Maybe your supervisor or lab group already has expertise in modelling, that is great, make use of it. Otherwise, you’ll have to put yourself out there and go looking (or ask your supervisor for help).
Collaborators with expertise in the modelling you want to do will help you save time by avoiding common mistakes and help you find the right kind of methods for your research question.
Second, do your reading. Ecological modelling is a very large field with established norms for analysis. So read the literature in your area and find our what flies, what methods people commonly use.
Give yourself time to upskill. Take training appropriate to the type of modelling you want to do. And give yourself time in your PhD timeplan to do the training. It takes time to do modelling, and even more time if you are learning by doing.
In fact, if you are used to working on a well set-up study system, a modelling study may well take longer than doing the field or lab study. There is a lot of back and forwards between developing the model, finding suitable parameter settings, and updating your analysis.
Modelling studies are never really done, because you can always go back and tweak something and create different results. It takes time to get it right, longer than many people may think. (By right I mean results you can explain and make sense in terms of the ecology).
Some good basic skills to have are of course in the R program, especially data wrangling skills and plotting skills (e.g. see my online notes for these things).
Then, beyond that, the skills you will need depend a lot on the type of modelling you will do.
(I’ll look to run some online training in the R program for marine scientists in the next few months. I was signed up for doing this anyway at 2 conferences, but now they are off! Stay tuned here for updates)
You may well be looking for data to analyse too. Good news is there is a lot of freely available data out there these days. But there are a few traps to avoid.
The biggest and best known datasets (e.g. like Landsat or the world database of protected areas) are very well analysed already. So it might be hard to find a novel question just using this data. This is also why it’s important to do your background reading. As they used to say, 30 minutes in the library can save weeks of fieldwork (or in this case modelling).
Combining existing data together in new ways can be a good way to make something novel out of old data. So if you bring together landsat with some local data you have access too, you may well have a novel study.
Remember that if you are using data, even if you’ve downloaded it ‘for free’ it is generally courteous to contact the data providers and let them know what you are doing. Obviously, you don’t need to ask to use landsat data (but do read the data license). But, say you want to use data published with a study, you should write the authors and tell them what you are doing. They may already be working on that analysis, so this simple step is both polite and will save you time.
Otlet, a data sharing service, are offering a ‘match-making’ service to help students and postdocs find data. Thanks Gretta Pecl for sharing this tip.
If you are a field or lab ecologists about to embark on a modelling study, you have some unique strengths to bring to a modelling study.
You probably have a different perspective than most full-time modellers. You may have a better understanding of how the data is generated, and the issues that poses to making ecological interpretations. You may also see ecology operating at a different ‘scale’. For instance, I find as a modeller I tend to see more uniformity and generalities across different places, whereas field scientists see more of the details and interesting differences. Neither of us is right or wrong, we just bring different perspectives.
Bringing your perspective to your modelling will help make it novel.
You might also have knowledge about datasets that other people can’t access. Make the most of these opportunities, because they will add more ‘novelty’ to your analysis. For instance, maybe your supervisor is sitting on some ‘old’ data they never published. Or maybe there is a unique dataset sitting with a local government monitoring authority that you can get with the right paperwork.
The pandemic is a major disruption to all research. But there are some opportunities to pursue a new research direction and upskill. Talk to your supervisor about how to adapt your PhD to keep it on track. I hope you can make something positive out of this situation that is forced upon us, and be successful in your PhD.