Time for another Monday Maggie*, where I write my thoughts about something I read or encountered in the last week. 

Last week (okay, really Sunday), I read "A hidden Markov framework for joint identification of animal activity modes and movement phases" (Cisneros-Araujo et al 2026) which is, I will admit, a very niche paper for a blog post. In a nutshell: A lot of movement ecology revolves around GPS data, and we can use the distance and angle between consecutive GPS points (or "fixes") to infer what animals are doing. For example, if an animal isn't moving much between GPS points, and it's noon, and we know it's a nocturnal species, perhaps the animal is asleep. In this paper, Cisneros-Araujo and colleagues detail a new way of modeling movements at two scales simultaneously -- think, delineating whether a whale is resting and it's in a migration phase, vs. a resting whale at the breeding grounds. 

I found the paper really useful for my work, as I try to wrangle the movements of elephant which might be migrating from place to place vs elephant that are just moving around their home ranges. And yet, I also got to thinking about model selection, how ecologists choose which model to use. We can use very simple models (e.g. assuming an animal is always resting at night) or extremely complicated models like the one detailed in this paper. When thinking of the ultimate story we're going to craft in our publications, it can be helpful to think about (1) the audience, (2) the ultimate ecological question, and (3) our own ability to apply methods, before we even start coding. There is a temptation to dump our data into the latest hot model, trying new algorithms or dipping into AI. And there can be good reason for doing so, as long as we remember 1, 2, and 3 and really ask: Is it worth my time and energy to learn and apply a fancy new method?

I recently submitted a manuscript to a remote sensing journal (not naming names) and got a desk rejection that basically read "our journal highlights innovative methods, and your paper doesn't cut it" (I'm paraphrasing but the tone was the same). Oof. I mean, they were right -- my methods aren't innovative from a remote sensing perspective. At first I was really upset, but after some reflection (a 16 hour flight to Johannesburg gives one ample time to ponder) I reaffirmed to myself that this was the whole point. 

The whole point of my paper and methods was to showcase a simple method that researchers worldwide, but especially in southern Africa, can use without needing expert help or advanced knowledge of AI and machine learning. All of the remote sensing data used was publicly available (Sentinel-2, Open Street Maps, etc), all code was written on free software (R and Google Earth Engine), and the method was a small twist on an old approach (binary thresholding). I wanted to write a paper that did a job well, but that was accessible to researchers without my level of computational expertise.

So what am I saying? Sometimes, it's okay to use a tried and true method that's a little boring. Actually, more than sometimes -- most times. Computationally intensive models or AI or machine learning all have their place, and help us to untangle really complicated dynamics and interactions. They're amazing leaps forward in human knowledge, and I applaud authors like Cisneros-Araujo and coauthors for their great work in creating new methods. Heck, I've been on papers like that before (Clark et al 2020)!

I sometimes encounter students or other early career researchers that lament they aren't as good at coding as me, or they aren't using complex methods. That's OKAY. As long as you aren't committing statistical faux pas (e.g., using a linear model for count data, or applying Tukey's HSD when a Kruskal-Wallis test is more appropriate), I think it's much more important that you understand the methods your using and what they mean, than that you use the most fancy, complicated model you can find. You can quote me on that one. :) 

Citations

1. Cisneros-Araujo, P., Gastón, A., Cubero, D. et al. "A hidden Markov framework for joint identification of animal activity modes and movement phases". Landsc Ecol 41, 72 (2026). https://doi.org/10.1007/s10980-026-02304-3
2. Swift, M.E.*, A. Songhurst, G. McCulloch, P. Beytell, R. Naidoo. "Mapping small-scale ephemeral surface water to inform transfrontier conservation planning in southern Africa." Preprint DOI: 10.64898/2026.04.03.715600 
3. J.S. Clark, C.L. Scher, & M. Swift, The emergent interactions that govern biodiversity change, Proc. Natl. Acad. Sci. U.S.A. 117 (29) 17074-17083, (2020). https://doi.org/10.1073/pnas.2003852117 

* It's not Monday, but I bet you future readers didn't even notice. Hah!

I have a fraught relationship with authorship. This stems from several heated conversations when I was a PhD student that I won't get into, but which resulted in high anxiety whenever I contemplate potential author lists or acknowledgements. I've found it helpful to be open and up front about co-authorship early on in the research process, and to lean on frameworks like CRediT to define author roles. Having guide rails has been immensely helpful keeping the panic in check.

All of this would fall apart, of course, if my co-authors weren't human.

The article I've chosen to start this series is Lesley Evans Ogden's Nature Career Feature of Nyikina Warrwa woman and conservationist Anne Poelina, "Why I made a river my co-author" (30 March 2026). In this piece, Ogden speaks with Poelina about her research and her decision to add an unexpected coauthor to her papers: The Martuwarra Fitzroy River, a seasonally-flooding river system in northwestern Australia threatened by fracking, agricultural runoff, and climate change. Martuwarra, RiverOfLife, as it has been registered with ORCID, is a "living Ancestor Being, whose creation stories underpin Kimberley Aboriginal people’s lawful, physical, emotional, and spiritual well-being, and thousands of years of customary practices for Yi-Martuwarra people, the people of the river" (Martuwarra, RiverOfLife et al 2023).
​