Work done under the broad umbrella term of systematic conservation planning (SCP) has a lot to deliver. In the face of accelerating biodiversity crisis the original SCP concept has been applied and augmented with computational tools that hold a promise of more cost-efficient – and indeed more efficient – conservation while keeping the process transparent and understandable for manager and policy-makers alike. But anyone who as ever worked with real-life conservation problem knows all too well that nagging feeling of “what if we got it wrong”? Working with heterogeneous sets of input data, empirical and/or modeled, seems to be the norm and the massive uncertainties involved are rarely quantified let alone reported. This needs to change. Or at least that is what Langford et. al (2011) think in a very recent opinion paper “Raising the bar for systematic conservation planning” published in Trends in Ecology and Evolution and since the topic is very important for many members of our journal club we spent this Friday’s session discussing the paper – with pulla as always.

Undeniably there are benefits in the various methods associated with SCP. However, the authors point out that not knowing the amount of error and uncertainty involved can lead to severe problems, both from scientific as well as real-life application point of view. Because many SCP methods lack good error models for input data, predictive models and generality for outputs, and include unknown amounts of error, decision-makers are left with little chance of reliably estimating the distribution of errors or their effects on predicted outcomes. Most SCP work applies to a specific case studies and does not easily – if at all – generalize to other and potentially more complex circumstances. From a practitioner’s point of view it would be important to know whether the decision and outcome would change if the amount of error would be reduced (e.g. by gathering more data). This is, of course, impossible when the amount, distributions and models of error remain unknown. Fixing the issue would call for something short of a paradigm shift on how we do conservation planning, but one has to start from somewhere. Rightly so, the authors conclude that:

If the problem is so complex that reliable bounds on the performance of a proposed SCP method in a new, unseen situation are impossible, then it is also impossible to claim a priori that the proposed method will give reliable performance in new unseen situations.

So how could one go about in addressing these issues? According to the authors, no new mathematical methods are required but rather “[…] they require a shift in the culture and orientation of SCP research.” On a more general level the authors believe that great improvements could be achieved simply by “moving away from the descriptive, anecdotal, case-study culture to one that is predictive and evidence-based, striving to support generalizable results”. Authors also suggest a list of more detailed improvements, such as explicitly predicting probability, structure and amount of SCP output error and testing for generalization of predictions, using active (machine) learning methods to explore different parts of the problem space, and measuring the results on data with known amounts of error.

All of us agreed that the paper was a refreshing read addressing an important issue. Since many of us are working with quantitative conservation planning methods (Zonation and the like) and/or species distribution modelling techniques, the robustness of our results is of high interest. Our discussion was mostly revolving around two issues: how would making the SCP methods more complex affect their applicability in the eyes of managers and policy-makers, and how feasible it is in the first place to strive for better understanding of underlying error models and distributions.

Existing SCP methods and software implementing them are already complex beasts that require intricate knowledge both on the input data as well as on how to setup the analysis and operate the software. Getting software like Zonation or Marxan up and running requires more know-how than most managers are able to contribute. Now the problem is that complicating the situation with even more layers of complexity as the Langford et al. are suggesting could potentially deter managers from using the available tools. Instead of lowering the threshold of using quantitative tools we would actually be raising it. Although these is a real and practical concern, it does not in any way void any of the arguments of the paper. The development of more robust, correct and generalizible tools should be in every manager’s and decision-maker’s best interest. If depicting and modelling the world around us is complex then we should rather try to tackle the complexity than regarding the task hopeless.

The other issue we had is a more profound one: can we really extract the information of underlying error structure from the data and circumstances we are working with? Can we ever hope to be able to understand the error and uncertainty to a degree that Langford et al. think is necessary? Many conservation problems are unique case-studies confined into their respective socio-economical and ecological domain, and thus generalizability may not even be a worthwhile objective. On the other hand it is possible that many SCP problems do indeed share similar patterns that can be studied and quantified for the information to be used in other similar contexts. With accumulating scientific information we might eventually even be able to use the information for predictive purposes as the authors are suggesting.

So are Langford et al. raising the bar for systematic conservation planning too high? In one step, yes, but the authors are not suggesting doing it in one step. Incremental steps will eventually get the bar high enough, but first we have to change our mindsets on how we do SCP. Overall we concluded that the paper was a good read giving us a lot to think about and we are interested in seeing what kind of response this opinion paper will generate in the conservation community. Even more interesting is to see whether the changes proposed by the authors will take foothold in SCP procedures around the globe.