COS 9-10 - Logging effects in tropical forests remain largely unknown due to a persistent methodological problem

Monday, August 6, 2012: 4:40 PM
D137, Oregon Convention Center
Benjamin Ramage, Biology Department, Randolph-Macon College, Ashland, VA, Douglas Sheil, Institute of Tropical Forest Conservation, Kabale, Uganda, Hannah Salim, Environmental Science, Policy, and Management, University of California, Berkeley, Berkeley, CA and Matthew D. Potts, Department of Environmental Science, Policy, and Management, University of California, Berkeley, Berkeley, CA
Background/Question/Methods

Ecosystems worldwide are experiencing elevated extinction rates, and tropical forest species are especially threatened as a result of rapid and widespread habitat loss.  Primary forests generally support the highest levels of biodiversity, but tropical timber production forests, which account for the majority of the remaining natural forest area protected from conversion, can augment the conservation value of primary forest reserves.  However, studies of logging effects often yield contradictory findings, fostering vigorous debate and frustrating efforts to develop clear conservation strategies.  Here, we assess the extent to which this variation may be due to simple pseudoreplication (i.e., the confounding of treatment effects with pre-existing differences between treated and untreated areas).  First, we conducted a literature review to determine the proportion of recent peer-reviewed tropical logging-effects studies that suffer from this problem (N=78).  Second, we asked: If one part of a primary forest had been logged, and if this treatment had no effect on species composition, what is the probability that a pseudoreplicated comparison of logged and unlogged areas would incorrectly infer logging effects?  We addressed this question with field data for seven taxa (trees, ants, dung beetles, moths, stingless bees, bats, and birds), collected from a primary forest in Peninsular Malaysia.

Results/Conclusions

We found that 54 papers (69%) were definitively guilty of simple pseudoreplication with regard to logging effects, 19 papers (24%) could not be definitively categorized, and only five studies (6%) were definitively free of simple pseudoreplication. Via our analyses of field data, we found that the probability of falsely inferring a treatment effect was > 0.2 for five taxa, > 0.3 for four taxa, > 0.4 for three taxa, and > 0.5 for two taxa.  These results were consistent regardless of whether species composition was analyzed as presence-absence or abundance data.  We also found that rates of false inference can exceed 0.40 when the correlation between physical distance and compositional distance is non-significant, and can exceed 0.25 even when there is an inverse correlation.  Our findings demonstrate that tropical conservation strategies are being informed by a body of literature that is rife with unwarranted inferences, and suggest that tropical forest ecologists need to place greater emphasis on true replication.  Confronting pseudoreplication is essential for accurately assessing biodiversity in logged forests, identifying the relative merits of specific management practices and landscape configurations, and effectively balancing conservation with timber production in tropical forest landscapes.