Model data to investigate wood frog abundance in 17-year post harvest variable retention mixed wood forests

Variable retention forest harvesting aims to reduce negative effect of harvesting on forest biodiversity, but its effectiveness is not well understood for many taxa. To better understand the effects of variable retention forest management and environmental features on amphibians, we used pitfall traps to capture wood frogs (Lithobates sylvaticus) across 4 levels of retention harvest (clearcut [0%], 20%, 50%, and unharvested control [100%]), and 2 forest types (deciduous and coniferous), in 17-year post-harvest forests in northwest Alberta. We mapped breeding sites and used a terrain moisture index (Depth-to-Water) derived from airborne LiDAR to examine relationships between relative abundance, breeding site proximity and soil moisture. Retention level alone had no effect on relative abundance, but in late summer (July and August) there was a significant interaction between retention level and forest type: capture rates decreased with amount of retention for deciduous forests, but increased with amount of retention in conifer forests. During late summer, capture rates were higher in conifer forests than in deciduous forests, with soil moisture (lower Depth-to-Water) positively related to capture rates. Though timber retention may be beneficial to wood frogs in the short-term, any impacts of forest harvesting on wood frog abundance was undetectable in stands 17 years post-harvest.Dataset was used wih R version 4.1.2Funding provided by: Alberta Conservation AssociationCrossref Funder Registry ID: http://dx.doi.org/10.13039/100007583Award Number: Funding provided by: Natural Sciences and Engineering Research Council of CanadaCrossref Funder Registry ID: http://dx.doi.org/10.13039/501100000038Award Number: Funding provided by: Canfor*Crossref Funder Registry ID: Award Number: Funding provided by: DMICrossref Funder Registry ID: http://dx.doi.org/10.13039/501100020611Award Number:The dataset was collected at the EMEND forestry experiment in northwest Alberta in 2014 and 2015. Pitfall trapping arrays were used to capture wood frogs (Lithobates sylvaticus) across 4 levels of retention harvest (clearcut [0%], 20%, 50%, and unharvested control [100%]), and 2 forest types (deciduous and coniferous), in 17-year post-harvest forests in northwest Alberta. We mapped breeding sites and used a terrain moisture index (Depth-to-Water) derived from airborne LiDAR to examine relationships between relative abundance, breeding site proximity and soil moisture. General Linear Mixed models (GLMMs) were only used with 2015 data as this dataset covered the entire wood frog active season (May to August) and included more study sites (4) and more pitfall arrays (64). Sampling in 2014 was only performed in two study sites (one deciduous-dominated and one conifer-dominated), and was only done from July to August. WAM is an airborne LiDAR-based terrain moisture model that predicts expected flow channels and associated water accumulation (depth to water) patterns at 1 m resolution of digital elevation. Mean DTW was calculated using the focal statistics tool in ArcMap 10.3 from 20 m circular buffers around each pitfall array which prevented overlapping buffers between adjacent arrays. A total of 64 pitfall trapping arrays were sampled across 4 study sites in 2015, and individual arrays acted as the independent unit of observation for analysis. Total captures were converted to Catch-Per-Unit-Effort (CPUE) and natural log (Ln) transformed to meet assumptions of General Linear Mixed Models (GLMMs). A constant of 10 was added to all CPUE values prior to transformation. Breeding proximity (distance of each array to nearest breeding site) and DTW were normalized/standardized for analysis by calculating z-scores. When running models, R sometimes warns you that variables are on very different scales and to consider rescaling. Normalizing or standardizing your variable gives you it values in standard deviations (standardized values measured as distance from the mean). z-score = (raw value - mean)/standard deviatio