pollinator meadow or deer blind?

Problem statement

State DOTs across the country are changing how they mow highway rights-of-way — mowing on rotations less frequent than once a year and avoiding grassland-bird nesting and monarch-rearing seasons — to create early-successional pollinator habitat, cut fuel and labor, improve stormwater quality, and slow invasive species. Taller roadside vegetation for longer periods raises an obvious safety question that maintenance staff, the traveling public, and resource agencies keep asking: does it change how many deer are struck by vehicles, either by attracting deer to the roadside or by hiding them from drivers until they step out? According to the NYSDOT problem statement, "the true effect of altered vegetation practices upon DVCs has not been significantly researched," and the one federally funded evaluation — a 2012 before/after comparison at ten sites in Maryland and New York — found no relationship but warned that its samples were small, its treatment incremental, and deer-vehicle collisions so rare and multi-causal that any mowing effect could be masked. Agencies are therefore adopting conservation mowing statewide on the basis of an absence of evidence rather than evidence of absence.

Why this matters

Deer-vehicle collisions cause an average of 148 deaths a year nationally, and in New York State alone about 65,000 collisions with white-tailed deer per year at an average cost of $4,300 each, plus about six deaths a year. If conservation mowing measurably raises collision rates in some settings, states are trading a safety cost for an ecological benefit without knowing the exchange rate — and if it does not, DOTs are absorbing public and maintenance-staff pushback and hedging their pollinator programs for no reason. The 2012 survey embedded in the FHWA study found 21 of 24 responding DOTs had already reduced mowing, most of them within the prior three years, so the practice has outrun its safety evidence for more than a decade.

What’s been tried and why it hasn’t worked

The FHWA/DVCIR pooled-fund study (Normandeau, 2012) compared three years of deer-vehicle-collision counts before and after mowing was reduced at six Maryland and four New York locations: five sites had more collisions after, four had more before, one no change; three New York results were marginally significant (p = 0.10) but in inconsistent directions, and mowing regime predicted collisions at only one of ten sites. The authors were explicit that the null result should be read with care: DVCs are rare events "known or assumed to respond to a large number of variables" (traffic volume, alignment, adjacent land use, crop cycles, deer density) that "have the potential to mask or swamp the effect of any other given variable," and the "reduced" mowing they studied was incremental — 15–30-foot clear zones were still mowed on the old schedule and annual full cuts kept everything non-woody — so the treatment differed little from control. Today's conservation regimes (rotational mowing less than annually) are a stronger treatment that has never been evaluated. NYSDOT's proposed fix was a $2 million, four-year multi-state controlled experiment; NCHRP staff judged the design too dependent on state DOT field data collection for the program and suggested a pooled fund, and FHWA reviewers called it "too simplified as written" — many factors drive DVCs — and recommended a phased approach starting with a ~$250,000 pilot using existing maintenance logs and crash records to test feasibility. Neither the experiment nor the pilot appears to have been funded.

What would unlock progress

The reframing is from a controlled field experiment (expensive, slow, statistically weak against rare events) to a retrospective observational design that exploits the natural experiment already under way: DOT maintenance management systems record where and when mowing occurred, police crash reports and carcass-removal logs record deer strikes with location and date, and empirical-Bayes safety methods (which the 2012 authors recommended for exactly this) can control for traffic, alignment, and land use across thousands of road segments rather than ten. The binding step is data integration — turning mowing records, which are kept for operations, into a spatiotemporal treatment layer that can be joined to crash and roadkill data. Adjacent solved problems include highway-safety countermeasure evaluation (crash modification factors) and ecological road-mortality analysis, which already have the statistical machinery.

Entry points for student teams

A team could partner with one state DOT (or use a state with public crash and roadkill data plus mowing-contract records) to build the joined dataset for a region and run an empirical-Bayes or matched-segment analysis of deer-collision rates under conventional versus conservation mowing, reporting the effect size and the minimum detectable effect — the FHWA reviewers' suggested pilot, at student scale. A GIS/data team could design the mowing-as-treatment data model that maintenance systems would need to log for such evaluations to be routine. A wildlife/behavior team could instrument a small set of paired segments with trail cameras to test whether deer roadside presence, not just crashes, differs by regime. Relevant skills: transportation safety statistics, GIS, ecology, and data engineering.