Deer Exclosures on a Road Edge

One confound is the road-edge placement. All exclosures were within 25 m of a gravel road, while open plots were 20 m farther into the forest. Road edges typically have more light, less leaf litter, and different soil moisture. If oak saplings grow better with more light, the exclosure plots would have higher sapling densities even without any deer effect, exaggerating the apparent benefit of fencing. A second confound is the escaped spring burn. Fire can kill small saplings, and burn severity often depends on fuel load; open forest with more leaf litter may have burned hotter than road-edge plots. If so, the open plots lost more saplings to fire, again inflating the exclosure advantage. The lack of randomization means deer exclusion is entangled with both edge habitat and fire history. A third issue is that the plots are paired only by distance, not by habitat. The reported 42 versus 9 saplings is therefore an association, not a clean estimate of deer limitation. The direction of bias for both main confounds is toward overstating the deer effect.

Randomly select many sites across the preserve, stratified by habitat and distance from road, and at each site randomly assign one of two adjacent 5 m × 5 m plots to exclosure or open control. Pairing should be local, with the same soil, slope, and canopy, and should include a sham fence around open plots if fence artifacts are a concern. Measure pretreatment sapling density, light using canopy photos, leaf litter mass, soil moisture, and deer browse on tagged stems. The key measurement for separating deer from light is direct browse pressure: count browsed twigs or use unbrowsed controls, and analyze sapling survival and growth with light and litter as covariates. Randomize the burn or exclude recently burned areas; do not apply an escaped fire. Run at least three years with repeated measures and enough sites for statistical power. If possible, add a light-reduction treatment to test whether light alone produces the pattern. This design makes deer exclusion independent of edge and fire effects.

The evidence does not justify fencing all 60 ha. The current comparison confounds deer with road edge and fire, and it has only eight pairs. Fencing 10 ha plus monitoring is the better decision: it treats fencing as an adaptive experiment, with pretreatment data, randomized sites, browse measurements, and a comparison of fenced and unfenced areas over time. This approach can estimate whether deer removal actually increases oak regeneration and at what cost. Scaling to 60 ha assumes the effect is uniform, but deer behavior, seed sources, light, and fire history vary. A large fence could also have unintended effects on other herbivores or people. Tell the land trust that the 42-versus-9 difference is suggestive but not conclusive, and that a 10-ha trial with clear success criteria is both more defensible and more informative before committing to whole-preserve fencing.