Like fossil-hunting scientists, ecologists regularly search for things (individuals of species) that are rare and difficult to find, because surveying the big wide world for biodiversity is a challenge that we have faced since the dawn of our discipline. In fact, much of the mathematical development of ecology stems from this probabilistic challenge — for example, species distribution models are an increasingly important component of both observational and predictive ecology.
But the palaeo types generally don’t rely on mathematical models to ‘predict’ where fossils might be hiding just under the surface. Even I’ve done what most do when trying to find a fossil — go to a place where fossils have already been found and start fossicking. I’ve done this now with very experienced sedimentary geologists in the Flinders Rangers looking for 550 million year-old Ediacaran fossils, and most recently searching for Jurassic fossils (mainly ammonites) on the southern coast of England (Devon’s Jurassic Coast). My prized ammonite find is shown in the photo to the left.
If you’ve read anything on this blog before, you’ll probably know that I’m getting increasingly excited about palaeo-ecology, with particular emphasis on Australia’s late-Pleistocene and early Holocene mass-extinction of megafauna. So with a beautiful, brand-new, shiny, and quality-rated megafauna dataset1, we cheekily decided to take fossil hunting to the next level by throwing mathematics at the problem.
Just published2 in PloS One, I’m happy to announce our newest paper entitled Where to dig for fossils: combining climate-envelope, taphonomy and discovery models.
Of course, we couldn’t just treat fossil predictions like ecological ones — there are a few more steps involved because we are dealing with long-dead specimens. Our approach therefore involved three steps:
- Fossils were once living things with physiological-tolerance thresholds to their environments and habitat preferences. In those respects, we can treat them as if they were living today if (and that’s a big ‘if’) we have some idea of their physical environment during the times they were alive. In the case of now-extinct species having lived within the last 120,000 years, we luckily have hindcasted global circulation models that at least provide us with predicted temperature and precipitation values for periods within that past. Provided we have reliable dates for enough fossil specimens, we can construct series of species distribution models for particular time-slices to predict the time-dependent spatial suitabilities of those species when they were alive.
- For a fossil to form and be preserved, the right geological conditions must exist — a fossil cannot form within a granite batholith, for example. In fact, Australian megafauna fossils are found exclusively within sedimentary rocks and regoliths, and in caves and lake deposits. If we have an idea of the distribution of the suitable geological formations in which fossils can form and be preserved, then we can use the spatial distribution of fossils to create a sort of ‘geological’ suitability map.
- The final phase of the process relates to predicting the suitability of the present-day environment for fossil discovery. There are really two parts to this question: (1) often fossils are found when erosion occurs and exposes them at the surface — for this then we need some kind of ‘erosion’ suitability map, which we can derive from high-resolution rainfall maps. Soils and rocks that are covered in dense vegetation are also less likely to give up their fossils easily, so looking for bare soil or areas with sparse vegetation can also assist. (2) We also have to account for the bias imposed by the distribution of living human fossil-hunters for already discovered fossils. For this we simply correct for the distribution of roads and cities — suitable places closer to roads and cities are probably more like to give up their fossils than extremely remote locations.
Once we have a probability (suitability) map for each of these layers, we can combine them to create a ‘suitability for fossil discovery’ map. Of course, we are really referring to a broad-scale spatial brush because we are limited by the resolution of the input data for each layer (for example, the paleao-climate data are only available at a resolution of 1 ° latitude). Our predictions are therefore most useful at continental scale and could assist in finding entirely new fossil sites. We are certainly not suggesting that you could use our model to find a new fossil in a particular lake or river bed near to a known site — for that, the traditional skills of the palaeontologist are still definitely required.
As a way to show that our model is useful, however, I can point to a serendipitous (but unrelated) discovery of new Genyornis3 fossils (actually, fossil egg shells) by Giff Miller in the far west of Australia that happened as we were having our paper reviewed. As it turns out, our modelled predicted a high suitability for this species3 based on the distribution and dates of previous fossils. The fact that Giff found them there is a nice little ‘validation’ of our model’s potential.
I really hope we can do more of these types of fossil-suitability models for other assemblages in other parts of the world. The instances might be restricted mainly by the quality of the fossil dates and the ability to hindcast climates, but we’re getting better at developing such datasets every year, so I’m certain we can find new applications.
1To be published shortly in Scientific Data.
3A recent paper challenges the identity of the species that produced these shells, so it might not actually be Genyornis that we modelled. Regardless of the species however, if the known shells are all from the same species, our modelled predictions are still valid.