Making high-resolution biodiversity maps from low-res maps

By | October 23, 2013

This post advertises our new Ecological Applications paper which is in press.

Imagine that there would be a tool that could make hi-res images out of low-res ones, just like this:

Such tool would be really useful for creating maps of things for which we only have a very crude (low-res or coarse-grain) spatial information, like this:

In geographical ecology most maps are low-res. Atlases of species distributions are compiled at grains of roughly 10 x 10 km (country-level atlases) or 50 x 50 km (EU-wide atlases), or they have the form of scale-free blob maps that become reliable only after they are fitted to a grid no finer than 100 x 100 km. Then there are the compiled point record data (e.g. GBIF). Ten years ago the founding fathers of "niche modelling" established the dubious paradigm that point records are fine-grain data. This has led to a diarrhoea of studies using the twisted combo of presence/pseudo-absences modelling, probability thresholding and pattern-recognition techniques. They have ignored the fact that point records can be scaled-up to areal units that correctly represent uncertainty about the position of presences and absences of a species. But then there is a problem: the resulting maps are too coarse.

So can we refine the coarse-grain maps?

Two years ago I started to work on it. I knew about the paper by McInerny & Purves (2011) who came up with a statistically correct way to infere species' niches from unobserved (latent) fine-grain variation in coarse-grain environmental data. I felt that there has to be a way to apply the approach the other way round: One should be able to make the unobserved (latent) fine-grain species distribution conditional on the observed fine-grain environment and the observed coarse-grain presences and absences, so that we can get this (the maps show distribution of American three-toed woodpecker):


With the help of Jonathan Belmaker, Adam Wilson and Walter Jetz I found the solution. We have published it here. We also have a study in review which we did over larger span of scales, and which adds spatial autocorrelation, informative priors and prediction uncertainty into the picture.

But then I thought, wait, if it is possible to do this with individual species distributions, then maybe... maybe we can refine biodiversity per se. We more or less know how number of species scales with area. Maybe we can use the species-area relationship (SAR) to map the (unobserved) fine-grain patterns of species richness conditional on the (observed) fine-grain environmental conditions, and through SAR also conditional on the observed coarse-grain richness.

And I did it. This is how it looks like for the map of species richness of South African birds (from the paper in press in Ecological Applications. ):


It works. Of course the approach now needs to be refined, critically examined, tested and improved by broader community. It could be implemented in a more user-friendly software tool (now you need to know OpenBUGS, JAGS, or likelihood optimization). Yet the most exciting prospect is that there are other phenomena that in theory could be "downscaled", especially outside of ecology. And maybe someone with a sense for business (not me) can commercialize the method.

I am opened for suggestions and collaboration.



Overview of alternative approaches to downscaling:

  • McPherson, J. M., W. Jetz, and D. J. Rogers. 2006. Using coarse-grained occurrence data to predict species distributions at finer spatial resolutions: possibilities and limitations. Ecological Modelling 192: 499–522.

Example uses of the alternative approaches:

  • Araújo, M. B., W. Thuiller, P. H. Williams, and I. Reginster. 2005. Downscaling European species atlas distributions to a finer resolution: implications for conservation planning. Global Ecology and Biogeography 14:17–30.
  • Bombi, P., and M. D’Amen. 2012. Scaling down distribution maps from atlas data: a test of different approaches with virtual species. Journal of Biogeography.
  • Niamir, A., A. K. Skidmore, A. G. Toxopeus, A. R. Muñoz, and R. Real. 2011. Finessing atlas data for species distribution models. Diversity and Distributions 17:1173–1185.


  • Azaele, S., S. J. Cornell, and W. E. Kunin. 2012. Downscaling species occupancy from coarse spatial scales. Ecological Applications 22:1004–1014
  • Hurlbert, A. H., and W. Jetz. 2007. Species richness, hotspots, and the scale dependence of range maps in ecology and conservation. PNAS 104:13384 –13389.
  • Kunin, W. E. 1998. Extrapolating species abundance across spatial scales. Science 281:1513 –1515.
  • McInerny, G. & Purves, D. 2011. Fine-scale environmental variation in species distribution modelling: regression dilution, latent variables and neighbourly advice. Methods in Ecology and Evolution 2: 248-257.

Self citations:

  • Keil, P., J. Belmaker, A. M. Wilson, P. Unitt, and W. Jetz. 2013. Downscaling of species distribution models: a hierarchical approach. Methods in Ecology and Evolution 4:82–94.
  • Storch, D., P. Keil, and W. Jetz. 2012. Universal species-area and endemics-area relationships at continental scales. Nature 488:78–81.
  • Keil, P. & Jetz, W. (in press) Downscaling the environmental associations and spatial patterns of species richness. Ecological Applications.
(Visited 1,878 times, 2 visits today)

Leave a Reply

Your email address will not be published. Required fields are marked *