Hypothesis‐driven and field‐validated method to prioritize fragmentation mitigation efforts in road projects

Abstract

The active field of connectivity conservation has provided numerous methods to identify wildlife corridors with the aim of reducing the ecological effect of fragmentation. Nevertheless, these methods often rely on untested hypotheses of animal movements, usually fail to generate fine-scale predictions of road crossing sites, and do not allow managers to prioritize crossing sites for implementing road fragmentation mitigation measures. We propose a new method that addresses these limitations. We illustrate this method with data from southwestern Gabon (central Africa). We used stratified random transect surveys conducted in two seasons to model the distribution of African forest elephant (Loxodonta cyclotis), forest buffalo (Syncerus caffer nanus), and sitatunga (Tragelaphus spekii) in a mosaic landscape along a 38.5 km unpaved road scheduled for paving. Using a validation data set of recorded crossing locations, we evaluated the performance of three types of models (local suitability, local least-cost movement, and regional least-cost movement) in predicting actual road crossings for each species, and developed a unique and flexible scoring method for prioritizing road sections for the implementation of road fragmentation mitigation measures. With a data set collected in <10 weeks of fieldwork, the method was able to identify seasonal changes in animal movements for buffalo and sitatunga that shift from a local exploitation of the site in the wet season to movements through the study site in the dry season, whereas elephants use the entire study area in both seasons. These three species highlighted the need to use species- and season-specific modeling of movement. From these movement models, the method ranked road sections for their suitability for implementing fragmentation mitigation efforts, allowing managers to adjust priority thresholds based on budgets and management goals. The method relies on data that can be obtained in a period compatible with environmental impact assessment constraints, and is flexible enough to incorporate other potential movement models and scoring criteria. This approach improves upon available methods and can help inform prioritization of road and other linear infrastructure segments that require impact mitigation methods to ensure long-term landscape connectivity.

Associations of ant species to land cover types and human disturbance in south‐west Gabon

“…The study site (3°07'32” S, 10°27'57” E) is in a mosaic of grassland, secondary forests and wetlands of approximately 175 km² along the Atlantic coast of Gabon (Vanthomme et al ., ). Annual rainfall averages 2000 mm, with a dry season lasting from May to September and a wet season from October to April (Thibault, Fisher & Goodman, ).…”

“…Travel in the wet season is challenging since many areas are flooded. As a result, estimated traffic ranges from less than two vehicles a day in the wet season to 12‐14 in the dry season (Vanthomme et al ., ).…”

“…The environmental variables considered were main habitat type (forest or savannah), lower vegetation type (leaf litter, short grass, and tall grass and weeds), exposure (shade or sun) and soil type (sand or other). In addition, we generated land cover maps by automatic classification of radar imagery of the study site (25 m resolution) and assessed the land cover type (dry savannah, inundated savannah, dry forest, young forest, partially inundated forest or inundated shrubs) at each station from them (Vanthomme et al ., ). Confounding variables were Julian date, duration of the bait (in minutes) and humidity index (from 0 for all sampling points on dry soil to 2 for all sampling points on wet or inundated soil).…”

Antipoaching standards in onshore hydrocarbon concessions drawn from a Central African case study

“…The dry season lasts from May to September and the wet season from October to April. Annual rainfall averages approximately 2000 mm (Vanthomme et al 2015). The study area was within a modified landscape where most of the company's oiloperation facilities and roads are concentrated.…”

“…Median occurrence probability [Duffy & Pettorelli 2012]); percentage of wetland (percentage of 25-m pixels within 56.4 m of the focal point, for a total surface of 1 ha, within a forest wetland or inundated savanna land cover in the land-cover map); distance to roads (m); hunter travel cost (accumulated least cost distance value for traveling from the main town to the center of each 500-m cell based on the conductance surface used previously); access restriction (percentage of least cost path from the main town to the center of each 500-m cell overlapping with theoretically restricted production roads); traffic index (mean number of vehicles recorded along the least cost paths per day evaluated at 89 locations along the road network surrounding the main town and interpolated to the study area's road network by empirical Bayesian kriging [Pilz & Spöck 2008]); forested land cover (one when most 25-m pixels in the 500-m cell were of forest, forest wetland, or plantation land cover classes; otherwise zero); distance to plantations (m) (as in Vanthomme et al [2015] landcover map); and precipitation (averaged mean monthly estimates of precipitation obtained at the main town's weather station from 1983 to 2013 for the months each cell was monitored). Cells that were consistently covered by clouds in 2000-2010 MODIS images (4.0%, n = 10) had missing NPP values that we replaced with the mean value across all cells.…”