On Friday the 28th of August we discussed a paper by Cawardine et al. (2010) which explores the sensitivity of conservation priorities to uncertain costs. The analysis is fairly straightforward: Use multiple conservation scenarios where the cost of a site is varied (from x0.001 to x4), prioritize across the landscape using targets (15% of original extent of occurrence of biological features) and MARXAN software, and look how the priority (or selection frequency) of a site changes as costs are changed.
This paper was very good selection to our journal club as it created a LOT of discussion :). Unfortunately to the authors, most of the discussion was not positive (sorry!). While discussing this paper, quite a few concerns came up, and I’ll try to summarize the most critical ones here:
- Overall, we were slightly surprised by the quality of review process and proof reading of Conservation Biology regarding this article. There seem to be typos here and there, two lines in Figure 3 seemed to have swapped places (lines d and e, based on the description in the text), and the description of material and methods leaves the reader quite a bit puzzled whether it would be possible to repeat this analysis with the same data and simply reading the paper. These types of errors should show up in review and editorial process.
- The results seem to us somewhat trivial: If fixed targets are used and there are no budget limitations, then the software will always select those sites that are required to meet the target, regardless how much they cost. The true sensitivity of sites to changing costs would have emerged if the total budget was fixed – i.e. when irreplaceable site simply becomes too expensive to buy within a restricted budget (like budgets usually are in real life).
- The high priority sites (that are also most robust) cover only a tiny part of the total landscape. How much of the targets do these sites cover alone? This would give some estimate on how much attention is truly needed to the other sites that do not stand out as highest priority and are somewhat sensitive to costs. We are not at all convinced that uncertain cost data should always be used simply because the small top (or, consequently, bottom) fraction is robust to cost variation (when using target based planning and apparently non-limited total budget). To our opinion, in this particular case the true sensitivity of conservation planning to uncertain cost data is measured with the priority level of e.g. 70-90%, which covers fairly large area and is mostly highly or moderately sensitive to costs.
- The analysis totally ignores the final outcomes of different planning scenarios: When assigning different costs to sites, what is it that eventually ends up being protected (beyond the fixed targets) and what is the final prize? It would have been very interesting to see a comparison of final costs and the levels of biodiversity features protected between each planning scenario.
- Some of the data conversions seem problematic to us, and due to fairly simplified description of methodologies, they bring out several concerns such as: (1) How is the cost-efficiency of a cell really calculated? As far as we understood, the price of the 10×10 km cell was determined by multiplying the average land cost of that cell with the proportion of the cell that is still covered by native vegetation (+ transaction costs). Thus, if only a very small parcel (e.g. 1%) of a cell was still covered by native vegetation, the relative cost of protecting that cell would be very low. But how is the biodiversity value of a cell like this determined? As authors do not provide any details about their biological data we assume that the biodiversity values of cell are not filtered in the same way as costs, but instead are given as presences and absences (although this is not clear either, the data could also be abundances, probabilities occurrences etc.). This would badly compromise the cost-efficiency calculations of a cell (e.g. if two sites have the same species composition, but the site A has only 1% of native vegetation left whereas site B has 80%, then in most cases site A would be selected as more cost-efficient option regardless of the fact that site B would be biologically clearly valuable). (2) Also the simple threshold to define protected and unprotected cells is problematic. Using a cut off level of 50% means that many of the already existing PAs (e.g. in cells that have a coverage of 49%) are being re-selected for protection (or, in worst case, if the cost-efficiency of a cell is determined as speculated above, already existing PAs are not selected because they hold large areas of native vegetation). On the other hand, areas right next to protected areas, which could have lot of potential in terms of high biodiversity value and reserve network aggregation, are now left out of the analysis because they are assumed to be already protected.
Final conclusion: The article tries to address a very important and interesting topic, but fails to convince us that cost data should always be used – even if it is uncertain. On the contrary, we found the results of this paper to indicate just the opposite.
Link to the paper:
Cawardine et el. (2010) Conservation planning when costs are uncertain. – Cons. Biol. DOI: 10.1111/j.1523-1739.2010.01535.x (Early Online)