As the saying goes “don’t put all your eggs in one basket”. Instead —to continue along with the egg analogy—, by putting your eggs in multiple baskets you are diversifying your risk: drop one basket, and eggs in the other baskets will still remain intact.

So does this idea of ‘risk diversification’ have relevance for conservation planning? In their 2012 PNAS paper Amy Ando and Mindy Mallory argue that yes it does, using a case study involving the prioritisation of land for conservation under climate change uncertainty to demonstrate how and why. Their paper, “Optimal portfolio design to reduce climate-related conservation uncertainty in the Prairie Pothole Region” focuses on adopting a well-established technique from finance – Modern Portfolio Theory (MPT), to develop a diversified land portfolio for conservation in the US Prairie Pothole Region that hedges against future uncertainty due to climate change impacts.

As Ando explains “In one climate scenario, you might have one tract of land that yields the best results, in another climate scenario, the other tract of land might be better. So if you have a little bit of both, you’ve hedged your risk.”[1] Using this risk diversification approach, it is often possible to substantially reduce the overall uncertainty in your future outcomes by sacrificing just a small bit of your expected conservation benefits (i.e. by choosing good, but slightly less optimal sites to those that might be selected under the more traditional conservation prioritisation approaches).

As the paper shows, simple, intuitive diversification strategies such as choosing equal amounts of different land types, will typically fail to provide the full benefits of diversification. The advantage of MPT is that it capitalises on information about the way future ecological outcomes co-vary across the landscape in order to most efficiently spread the risk. In the case of the Prairie Pothole Region, the authors found that MPT allowed for much greater gains per dollar spent than would a simple diversification approach for a desired level of risk.

Now moving on to our journal club’s discussion of the paper. Our group didn’t find this paper as easy to read as other similarly technical papers. It was a bit jargon heavy, and could have done with greater explanation of, and captions for, the figures. A disappointing aspect of the paper was that there was not much focus on the potential uncertainty involved in attempting to parameterise the covariance matrices used to specify how risks co-vary. The article’s methods required the assigning of probabilities to the likelihood of various climate change scenarios, which we noted was something that the scientific community is very reluctant to do. This received very little discussion in the paper, as did the follow-on estimation of the covariance parameters.

The authors were presumably focused on demonstrating the MPT-based approach, but particularly as they have pitched it as a tool for dealing with climate change related uncertainty (and not just general uncertainty about conservation action effectiveness, for example), the issue of whether a method based on probabilities can be truly applicable is important. The paper’s own results clearly demonstrate the impact that the probability estimates will have on which portfolio solution is best, so that it appears that getting them wrong might lead MPT to do more harm than good. A greater understanding of the sensitivity of the results to errors in the estimated probabilities would likely be necessary before considering it a workable approach in practice.

A second topic in our discussion referenced points made in the response letter by Dunkel and Weber (2012), about the fact that the simple MPT techniques demonstrated in the paper have largely been superseded by more advanced risk spreading techniques, which —along with the accurate estimation of climate change scenario probabilities and inter-dependencies across a region noted above—, would also be necessary for reliable practical implementation. In the article, the approach used equates risk with variance, and by doing so assumes that we are equally concerned with the consequences of upside and downside fluctuations.

For conservation applications, this will rarely hold: we are, for example, usually far more concerned about downside fluctuations in threatened population numbers, habitat, qualify, etc., than upside fluctuations.[2] Dunkel and Weber suggest that improved shortfall risk measures (SRMs) that have been developed in the last decade should be used in place of variance as a risk measure in future MPT applications for conservation to handle this, and other related issues.

Finally, JC-ers also noted that the paper was lacking in sophistication in other areas of the manuscript. The cost calculations and value calculations, for example, were probably not as thorough or inline with current best practice as they could have been. Again, clearly the authors were focusing on the demonstration of the MPT method, but this did appear to detract from people’s perceptions of the realism of the prioritisation and analysis. Given that it falls somewhat outside what many of the methods currently in place for setting conservation priorities are, it will be interesting to see if, and how, this method gets picked up by the spatial conservation planning community.

Further reading

Ando, Amy W., and Mindy L. Mallory (2012). “Reply to Dunkel and Weber: Probability distributions and shortfall risk measures in conservation portfolio analysis.”Proceedings of the National Academy of Sciences of the USA 109.35: E2305.

Dunkel, Jörn, and Stefan Weber (2012). “Improving risk assessment for biodiversity conservation.” Proceedings of the National Academy of Sciences 109.35: E2304-E2304.

Hoekstra, Jonathan (2012). “Improving biodiversity conservation through modern portfolio theory.” Proceedings of the National Academy of Sciences 109.17 : 6360-6361.

For an interesting perspective on the implications of portfolio theory for species valuation and management, see Figge, Frank (2004). “Bio-folio: applying portfolio theory to biodiversity.” Biodiversity & Conservation 13.4: 827-849.