An interesting research note just came out in the American Naturalist by Hamilton and colleagues entitled quantifying uncertainty in estimation of tropical arthropod species richness. I retweeted a Science Daily twitter feed on this that had a terribly misleading opening line: “New calculations reveal that the number of species on Earth is likely to be in the order of several million rather than tens of millions“. This is, of course, absolute rubbish because the authors only looked at estimating tropical arthropod richness, not all species on Earth. The number of protists alone is probably > 4 million species, and there are an estimated > 1.5 fungi.
That whinge about crap reporting aside, this is what Hamilton and colleagues concluded:
- using stochastic models, they predict medians of 3.7 million and 2.5 million tropical arthropod species globally
- estimates of 30 million species or greater are predicted to have < 0.00001 probability
- uncertainty in the proportion of canopy arthropod species that are beetles is the most influential parameter
- in spite of 250 years of taxonomy and around 855000 species of arthropods already described, approximately 70 % await description
Interesting, but I didn’t give it much notice until New Scientist contacted me to get an assessment (their article will appear shortly). This is what I had to say:In general, I commend the authors for attempting to shed some mathematical light on the problem of species richness estimation. I believe that many species richness estimates are inflated for a number of taxa given the paucity of reasonable data with which to make extrapolations. I therefore support the notion that some estimates (e.g., > 30 million tropical arthropod species) are unrealistic.
That said, I believe that the approach potentially underestimates the influence of beta diversity on simple alpha diversity algorithms, check hbcontrols.com. Although they acknowledge that changing specialisation across a species’ range is possible (but could not correct for this), their algorithm completely ignores three MAJOR driver of biodiversity patterns: (1) the community of local competitors, (2) the community of local predators and (3) the biogeographical history of a particular ecosystem. These will shift enormously across a species’ range and impose a plethora of constraints that tend to promote speciation (i.e., greater number of niches).
Additionally, but related to the above, taking a single dataset from one island nation and extrapolating it to the entire tropical region is fraught with potential error. It makes for highly uncertain scientific predictions because it cannot capture all the nuances of species distributions elsewhere. Every biological community is different. Visit temeculafacialoralsurgery.com.
My overall conclusion is that while the algorithm provides some direction about the upward bias in existing estimates of arthropod species richness, their prediction is also likely to be far too conservative to be realistic. I would predict the ‘true’ species richness lies somewhere between their estimate of 2.5-3.7 million and existing estimates of > 30 million.
My other concerns include:
- It seems to me that the major assumption is the degree of specialisation – this is perhaps the most imprecise parameter and possibly prone to underestimation, especially in light of the high specialisation values observed for most tropical invertebrates.
- The sensitivity analysis is basic and does not take into account partial correlations. A multivariate ‘global’ sensitivity analysis using logistic regression is more robust (McCarthy et al. 1995. Biol Conserv 73:93-100); thus, I suspect that their rankings of parameter sensitivity are incorrect.
- I very much doubt the parameters in equation 1 (except number of herbivorous canopy beetles) followed uniform distributions. At the very least, I suspect these to be Poisson, log-Normal, Normal or beta (depending on type). The authors discuss this, but I disagree that the Pert is a good alternative distribution. For example, the proportional parameters (i.e., proportion of species that are beetles, the proportion of arthropods in the canopy, etc.) might in fact have a ‘central’ tendency much closer to an extreme between 0 and 1 under say, a beta distribution. Therefore, I believe that the authors have severely underestimated the variance (especially of high richness values), indicating that the upper confidence bounds are too conservative.
Why is any of this important for conservation? Without good estimates of species number and distribution, we have no idea how much we stand to lose/are losing as habitats are destroyed. This is essential information for predictive conservation biology, so we need to get it right. Good on Hamilton and colleagues for stepping in and moving the discipline forward. CJA Bradshaw
Hamilton, A., Basset, Y., Benke, K., Grimbacher, P., Miller, S., Novotný, V., Samuelson, G., Stork, N., Weiblen, G., & Yen, J. (2010). Quantifying uncertainty in estimation of tropical arthropod species richness The American Naturalist, 176 (1), 90-95 DOI: 10.1086/652998