It is an interesting and tricky question. I think we should first be able to define what do we mean by "natural" habitat (and landscape), especially considering the extensively human-altered Palaearctic region.

Yes you could create a myriad different slew of definitions appropriate for different criteria and where one puts an arbitrary break in the multidimensional gradient of natural to human landuses is going to be contentious. But, OK, lets call not heavily modified areas where human activities are minimal, i.e. largely functional natural ecosystems. Such as definition would not well consider semi-natural habitats, but these are part of the anthropogenic world so perhaps should be considered so. Where one draws the line is hard to say, is a twice logged forest no-longer a natural habitat compared to a once-logged or pristine forest? Is an unimproved pasture equivalent to a forest glade? One could say protected areas instead of natural habitats to sneak around the problem but then that would exclude the large proportion of non-protected natural habitat out there. Given the global distribution of biodiversity a general definition of natural habitat should probably reflect tropical systems?

Perhaps I should just take the inverse of how much biodiversity is found in human dominated landscapes? But then ones runs into the issue of gardening and issues of shifting baseline. Take someone from India or Kenya to a nature reserve in the UK and they will often look confused as to why they are standing in a farmland and ask you what happened to nature reserve? clearly you've different baselines as to what is a natural habitat and land-use and you would choose a different arbitrary break in the continuum of habitat degradation before saying stop! that's not a natural habitat anymore.

Anyhow I found a paper:

http://onlinelibrary.wiley.com/doi/10.1046/j.1523-1739.2002.00530.x/pdf

BROOKS, T. M., MITTERMEIER, R. A., MITTERMEIER, C. G., DA FONSECA, G. A., RYLANDS, A. B., KONSTANT, W. R., FLICK, P., PILGRIM, J., OLDFIELD, S. & MAGIN, G. 2002. Habitat loss and extinction in the hotspots of biodiversity. Conservation biology, 16, 909-923.

This suggests >50% of threatened terrestrial plant and mammal diversity is endemic to biodiversity hotspots and implies this is in "natural habitat" in biodiversity hotspot. 

Of course a huge amount of global biodiversity is concentrated in a relative small fraction of the planet, mainly referred to "natural habitats". Conversely, some semi-natural habitats (e.g., European dry grasslands) can host impressive absolute values of species diversity and a large proportion of endemic species at a smaller scale...but I think this falls outside the issue.

 'Biodiversity Hotspots' are a tool for prioritising conservation, and their study (e.g. Myers, 1988, Myers et al., 2000, Myers, 2003) highlights "exceptional concentrations of endemic species"  found in a tiny proportion of global terrestrial area. It may be worthwhile to limit your study to hotspots? But, as Rocco mentioned, defining 'natural' - specifically non-anthropogenic - habitats, is problematic. I take non-anthropogenic to mean, in this context, uninhabited by people (Peter did mention areas that are not influenced by humans, including semi-natural areas). Hotspots by their very nature are threatened by humankind. And anthropogenic effects are almost completely global (via climate change, hydrological cycle, pollution etc etc). It's a tough one! But good luck ... I'll post the above references, but got to go now ...

Here they are:

MYERS, N. (1988) Threatened biotas: `hotspots' in tropical forests. Environmentalist, 8, 187-208 (1988)

MYERS, N., MITTERMEIER, R.A., MITTERMEIER, C.G., DE FONSECA, G.A.B. and KENT, J. (2000) Biodiversity hotspots for conservation priorities. Nature, 403, 853-858

MYERS, N. (2003) Biodiversity Hotspots Revisited, BioScience, 53 (10), 916-917

The flipside, of course, is to find species that have adapted wholly to persist solely in anthropogenic environments.

a few quick points - re definitions.  there is quite a literature on novel, recombinant (including one of mine in a book on urban envts), hybrid and ragamuffin communities. the conclusion is that there is hardly an ecosystem in the world that hasn't been affected by humans and introduced spp. the whole thing is recombinant to varying degrees.

second I agree with one of yr correspondents that there is a difference between biodiversity and species richness. I have argued for a while that biodiversity is a global concept whereas spp richness is local. thus we could replace the 2500 native plant spp in nz (most of which are endemic) with the 30 000 exotic spp and have a perfectly functioning ecosystem. many farmers and lay people would say we've increased our biodiversity, but hang on! all that has happened is that global diversity has been reduced by around 2000 spp. in other words, adding widespread exotic spp to a biota ipso facto decreases (global) biodiversity. in this sense biodiversity conservation applies almost exclusively to indigenous spp.

another pet gripe is equating the envt with biodiversity. ecology and biology have almost been totally usurped by envtal engineers. and the debate about envtal quality and sustainability is increasingly about air and water quality and green buildings or green miles. it is often assumed that a good envt, by definition, means biodiversity. one example tells us why not - clean up the Thames and salmon return; clean up a nz river and salmon (and trout) do even better - and eat up the native fish!!

finally, I have also argued that urban (and rural) envts can be the saviour of especially lowland, endangered herbaceous plant spp - especially in an oceanic envt like nz where exotic competitors are a major hazard to biodiversity. The importance of urban ecology (see the Erfurt Declaration etc) is because we can create and maintain stressed envts there, that generally create a competitive advantage for the indigenous spp

enough !  colin 

Thanks Julian, Rocco and Colin.

I am really just looking for some quick and dirty figures about the proportion of terrestrial biodiversity that needs protected areas or relatively undisturbed natural habitats to persist. I am not doing a study on the diversity with protected areas or hotspots per se, I am just writing some introductory material regarding a thesis chapter on trade-offs between land-uses in the buffer zones of protected areas and was hoping to say something about the % of biodiversity that is difficult to integrate into strongly human modified habitats (i.e., protected areas are a still a fundamental tenant of conservation, but happily complemented by urban or agro biodiversity).

I think focusing on endemics in hotspots might be enough for that purpose, I wonder if there has been an analysis of the number of threatened species that can persist agricultural systems vs. those that cannot?

Pete