The need for biological realism in habitat modeling : a reinterpretation of Zharikov et al. (2006). Authors' reply

Zharikov et al. (2006: Landscape Ecology 21:107-120) modeled the nest-site habitat use of marbled murrelets (Brachyramphus marmoratus) in Desolation Sound (DS) and Clayoquot Sound (CS), British Columbia. They compared known nest sites, located with radio-telemetry, with randomly-located points within the same areas. Their conclusions suggest that murrelets tended to nest in disproportionately smaller fragments within the more disturbed DS landscape; streams, steeper slopes, and lower elevations were selected in both landscapes; murrelets nested closer to recent clearcuts than would be expected in the DS landscape; and survivorship of nestlings was greater in areas with recent clearcuts and was positively correlated with recent habitat fragmentation. These conclusions are contrary to current management guidelines in British Columbia, and therefore require close scrutiny. Our detailed examination reveals flaws in their use of data, application of modeling, and most seriously, interpretation of the results. Problems include: conceptual errors in the interpretation of models; inappropriate spatial resolution; confusing use and interpretation of fragmentation and patch size data; overemphasis of statistically significant but biologically trivial results; and ignoring some contradictory studies. We conclude that it would be risky to apply the results from Zharikov et al. in the selection of murrelet nesting habitat for management purposes in British Columbia. Our review identifies issues that may arise in other ecological modeling studies and stresses the need for biological realism in addition to statistical rigour.