Microsoft Research has figured out how to save the maximum number of plant species by protecting the smallest areas possible. 

We all want to save as much biodiversity as possible. It isn't a just a matter of sentiment - diversity ensures that as many genes are available for our future study and use. The big problem is that conservation by protecting areas is costly because it deprives other humans of areas to support their particular way of life. Its a trade-off but there is no reason why we shouldn't try to optimize the choices.

Now Microsoft Researchers, Lucas Joppa and Piero Visconti of Microsoft Research Cambridge, Clinton N. Jenkins of North Carolina State, and Stuart L. Pimm of Duke, have used techniques more commonly found in AI to work out which areas should be protected.

The conclusion produced by the AI methodology is striking - if you protect 17% of the land on Earth you can preserve 67% of endemic plant species. This is a good trade-off. 

The algorithm used was a combination of a simple "greedy algorithm" and a "genetic algorithm. The initial allocation was performed using a greedy heuristic - i.e. find areas with the maximum plant species in the smallest area. Then the genetic algorithm was used to "breed" better solutions. The regions chosen were enlarged by adding neighbouring regions with maximum plant species density.

As Joppa comments:

“This is a fairly straightforward approach, but the results are often non-obvious. This is especially true for ‘endemic species,’ those species found in a region or a set of regions and nowhere else. Thus our algorithm was constantly trying to trade off picking small areas with just a few species—such as islands—where all the species are endemic only to that island, versus picking larger regions—such as the central Amazon—that have very many species, but proportionately not as many species found only within that region.”

The data on over 100,000 species was obtained from data compiled by the Royal Botanic Gardens in London. A map showing endemic species density color-codes to show high priority regions was also produced: