Let me explain.
As a person who cut his teeth in field ecology (with all the associated dirt, dangers, bites, stings, discomfort, thrills, headaches and disasters), I’ve had my fair share of fun and excitement collecting ecological data. There’s something quaintly Victorian (no, I am not referring to the state next door) about the romantic and obsessive naturalist collecting data to the exclusion of nearly all other aspects of civilised life; the intrepid adventurer in some of us takes over (likely influenced by the likes of David Attenborough) and we convince ourselves that our quest for the lonely datum will heal all of the Earth’s ailments.
Bollocks.
As I’ve matured in ecology and embraced its mathematical complexity and beauty, the recurring dilemma is that there are never enough data to answer the really big questions. We have sampled only a fraction of extant species, we know embarrassingly little about how ecosystems respond to disturbance, and we know next to nothing about the complexities of ecosystem services. And let’s not forget our infancy in understanding the synergies of extinctions in the past and projections into the future. Multiply this uncertainty by several orders of magnitude for ocean ecosystems.
The upshot is that ecologists have been searching for proxies and indicators of biodiversity patterns and processes, with the ultimate aim of being able to predict dynamics (at least at broad spatial scales) from decidedly non-biological features. A case in point is one that I’m most familiar with – the use of ‘surrogates’ in marine ecology – that is, using a species/taxon to predict the distribution of many more species. We have also done some work to predict coral reef fish diversity using little more than the position of the reef (latitude and distance to shore), and we have inferred inherent susceptibility to extinction using nothing more than the shape and isolation of reefs.
I even remember once that Hugh Possingham wished out loud at a conference that he hoped we’d never have to collect a species again if we got our maths right. That might be a little far-fetched, but it gets at my main point.
So the theme of this post introduces a new-ish paper that my productive post-doc, Dr. Camille Mellin, has recently written in Ecological Applications entitled Multi-scale marine biodiversity patterns inferred efficiently from habitat image processing. Here the idea is fairly simple – by taking a photo of an area where animals hang out, you can estimate how many different types there are (diversity).
Using a multi-scale dataset of images taken of the Great Barrier Reef – from the scale of a single transect to satellite images of multiple reefs – we measured the amount of ‘complexity’ in the photo using something called the ‘mean information gain’. After accounting for spatial non-independence, it turns out that we could explain up to 29 % of the variance in fish species richness, 33 % in total fish abundance, and 25 % in fish community structure at multiple scales.
Now, this might not seem like a lot to some, but take it from me, it is a remarkably high predictive ability for most ecological studies. And all from merely taking a photograph? I’m not suggesting (as the title of this post implies) that we need to abandon all ecological sampling studies, but we should be constructing ever-more-efficient ways to estimate biodiversity patterns and processes using such short-cuts. They’re less time-consuming, more cost-effective and potentially cover areas that are difficult or impossible to sample directly.
Oh, and before I leave this one – I’m foreshadowing a great example of data proxies for inferring the effectiveness of tropical nature reserves; this is a manuscript currently in review in a very high-profile journal (lead by Bill Laurance), and I hope to have some good news about it soon.
CJA Bradshaw
-34.917731 138.603034