Table 1.
Candidate predictor variables by category.
Fig 1.
Number of studies (white bars, including conference proceedings (CProc) and government reports (GRep)) and effect size estimates (solid bars) in relation to (A) country, (B) taxon, (C) study design, and (D) publication type.
Amph&Rept: effect sizes combined amphibians and reptiles; BA: Before/After; BACI: Before-After-Control-Impact; CI: Control/Impact study designs.
Fig 2.
Number of studies (white bars) and effect size estimates (solid bars) in relation to (A) mitigation type, (B) crossing structure type, (C) fencing type, and (D) other mitigation types.
Crossing: crossing structures; Crossing with fencing: combination of crossing structures and associated fencing; ADS: animal detection systems; Reflectors: wildlife reflectors; Other: other mitigation types e.g., wildlife warning signs; Mamm: mammal.
Fig 3.
Relationship between mean effect size and the percent road-kill decrease [Eq 4] (n = 99 effect sizes).
Symbol size is proportional to the weight (inverse of the sampling variance) of the effect size; smaller symbols correspond to effect sizes with lower weights.
Fig 4.
Summary flow chart of the meta-analysis addressing our three main research questions using the complete dataset of effect sizes (n = 99) and appropriate subsets (dashed boxes).
Boxes enclosed by solid lines indicate predictor variables or subset categories under consideration. Shaded predictors were associated with road mitigation effectiveness. Subset categories in green indicate an overall average reduction in road-kill with road mitigation; red indicates an overall average increase in road-kill with road mitigation. Values in parentheses are the number of effect sizes. BA: Before-After; BACI: Before-After-Control-Impact; CI: Control-Impact study designs. (See S1 Table for a complete list of research questions and predictor variables)
Fig 5.
Relationship between weighted-mean effect sizes and the weighted-mean percent road-kill decrease for crossing structures and fencing alone and in combination.
Values in parentheses are the number of effect size estimates. Error bars indicate 95% confidence intervals.
Table 2.
Associations between effect sizes and (A) mitigation category (n = 39); (B) road type (n = 19), for the subset of studies involving large mammal fencing.
Mitigation category: Crossing structures with associated fencing, animal detection systems, and wildlife reflectors; Road type: ≥4- lane divided highways, and 1–2 lane roads. Notes: Null model = random-effects model.
Fig 6.
Relationship between weighted-mean effect sizes and the weighted-mean percent road-kill reduction for three different types of mitigation measures, based on a sample of n = 39 large mammal effect sizes.
Values in parentheses are the number of effect size estimates. Error bars indicate 95% confidence intervals.
Fig 7.
Relationship between weighted-mean effect sizes and the weighted-mean percent road-kill reduction for different taxa, based on n = 47 effect sizes from studies involving crossing structure and associated fencing.
Values in parentheses are the number of effect size estimates. Error bars indicate 95% confidence intervals.
Fig 8.
Relationship between weighted-mean effect sizes and the weighted-mean percent road-kill reduction for (A) study design, and (B) whether mortality data collected during construction of the mitigation measure was excluded, based on studies employing BA or BACI designs.
Values in parentheses are the number of effect size estimates. Error bars indicate 95% confidence intervals.
Table 3.
Study design type predictor variables showing associations with effect sizes for: (A) the complete dataset (n = 99), and (B) the combination of Before-After and Before-After-Control-Impact subset (n = 66).
“During construction data separation” means that mortality data collected during construction of the mitigation was excluded from analyses. Notes: Null model = random-effects model.