Site Selection For a version of this document with maps and figures, please see “DataSummaryReport.pdf” which includes the site selection text here, in addition to study background, all study methods, and a descriptions of each data set. The Kenai National Wildlife Refuge (KNWR) comprises 797,200 ha in south-central Alaska, including four designated wilderness areas: Mystery Hills, Swanson River, Skilak Lake, and Tustemena Lake (Wilderness Act of 1964:16 U.S.C. 1131-1136). The refuge also contains 345 km of roads, most of which were developed to support two oil and gas fields that began production in the 1950s. In this study, we sought to evaluate the relationship between road-associated metal contamination and amphibian abnormalities. To do this we employed a stratified random design to select sites for study. Sites were chosen randomly to enable us to make inference about the effects of paved and gravel roads on nearby wetlands. We had planned to limit the sites chosen for study to those on Refuge lands, but soon recognized that doing so would not give us adequate replication of the different road types, especially paved roads. We therefore broadened the search criteria to all lentic wetlands on public lands within 1 km of a road in a much larger area of the Kenai Peninsula. We stratified potential study sites by road type and spatial location. We stratified on road type (paved or gravel) because metal fate and transport mechanisms may differ between gravel and paved roads. We also divided the study area into 5 somewhat arbitrarily drawn spatial sectors to ensure interspersion of sites on the landscape. These sectors varied in road density and traffic as well as land use and degree of urbanization. Most sites were on public lands (including state, federal, and municipal) due to requirements for sampling permission. Our search criteria for possible study sites included all lentic wetlands on public lands within 1 km of the road in a larger area of the Kenai Peninsula. The area chosen for study specifically reached from Cooper Landing on the Sterling Highway to the intersection of the Sterling Highway with the southern end of Cohoe Loop Road. Within this area, we included all roads classified as a street or main road on the Kenai Peninsula Borough’s (KPB) GIS “Roads” layer (available: http://www.borough.kenai.ak.us/GISDept/Downloads.html). We created a 1 km buffer around the streets and main roads in this layer, and chose possible wetlands only from within that buffer. The effects of road contamination on the surrounding environment are thought to be negligible beyond a 1 km distance (Trombulak and Frissell 2000); we therefore did not consider sites >1 km from roads for this study. Each spatial sector we drew on the Kenai Peninsula is described briefly here. The sector with the highest road density Town included residential and industrial areas of the towns of Kenai and Soldotna. The Town sector was the most urbanized, and sites adjacent to roads did not always lie upon clear gradients from a road into open space or wilderness, as was common in other sectors. In Town, one of the paved road sites (KNA90) was chosen non-randomly because we had studied this site since 2004 and wanted to retain some sites with historic data. The next two sectors North Spur and K-BeachCoho circumscribed rural-residential areas. In addition to rural residences, NorthSpur contained a liquefied natural gas plant, a nitrogen fertilizer plant, and other dispersed oil and gas facilities. Sites on public land were only identified along paved roads in the NorthSpur sector. The K-BeachCoho sector also included rural residential areas and limited oil and gas drilling operations. The next sector, SoldSterCoop included the stretch of the Sterling Highway, Skilak Loop Road, and Mystery Hills Road, from Soldotna to Cooper Landing. This sector included both paved and gravel roads through rural residential areas near Soldotna, Sterling, and Cooper Landing as well as open space areas managed as KNWR lands. One site in the SoldSterCoop sector (KNA111) was not chosen randomly; this site was also retained because we had historic data. The final sector Refuge Dirt contained only dirt roads within refuge lands. Most of these roads were built to support oil and gas development, although some also provide recreational access. All wetlands chosen in this sector had been studied extensively since as early as 2000 (Reeves et al. 2008, Reeves et al. 2010, Reeves et al. 2011), and because historic data are particularly important for understanding climate change, we retained 10 historic sites in this area. We also added two new sites along a road built in 2010, at which baseline metals sampling was conducted before the road was built. We originally had included an additional sector, Turnagain, which contained 11 candidate sites along the Sterling and Seward Highways. These sites were identified solely by road surveys because GIS layers were not available. None of these sites contained tadpoles when searched in June 2010, so none were retained for study. We defined public lands as those on the KPB website’s “Ownership” layer (available: http://www.borough.kenai.ak.us/GISDept/Downloads.html) held by the following entities: The State of Alaska, The Alaska Mental Health Trust, The U.S. Federal Government, The University of Alaska, or The Kenai Peninsula Borough. To select only these land parcels, we used the following query in GIS (ArcMap version 9.3.1, ESRI, Redmond, WA, USA): "OWNTYPE" = 'BOROUGH' or "OWNTYPE" = 'FEDERAL' or "OWNTYPE" = 'MUNICIPAL' or "OWNTYPE" = 'STATE'. After filtering all possible sites by their spatial location, proximity to road, and land ownership in GIS, we chose potential wetlands with a combination of GIS layers and visual assessment of digital aerial photographs. In some areas of the Kenai Peninsula, wetland habitat has been mapped (available http://www.cookinletwetlands.info/ ). In mapped areas, we used this wetlands layer in GIS to help us identify potential sites. Where wetlands were not mapped, we relied on a visual assessment of digital aerial photographs to identify potential wetlands. Sites were identified as potential amphibian breeding areas if they appeared wet, marshy, or meadow-like in air photographs, and if the open water in these areas was ?100 m in length. We identified 122 candidate sites by these methods. Within each sector, we set target numbers of sites for inclusion in the study. All candidate sites were then randomized within each sector. Sometimes, several ponds were clustered tightly on the landscape within a sector. When multiple ponds were in the same area, we first randomized the areas then assigned individual wetlands random numbers within each area.  If the first-ranked wetland in an area did not contain tadpoles or otherwise would not work, we moved to the next randomly ranked area. We revisited areas only when they were ranked as the next lowest random number in that sector. Table 1 presents the numbers of sites originally identified and ultimately chosen for study by sector. Sites were visited by field crews in July of 2010. The sites with the lowest-ranked random numbers in each sector were searched first to determine whether they met criteria for inclusion in the study. A site that met the following criteria was included in the study: the site was safe and accessible, it contained standing water in July, and frogs or tadpoles were found. We added wetlands to the study in each sector in order of their randomly-assigned numbers. If the site was unsafe, not accessible, dry, or frogs were not found after dip-netting the entire site perimeter, that site was dropped from the survey and crews moved to the next site in order of the random numbers assigned. We identified 29 of 36 study sites randomly, using these procedures, after which we were forced to select some sites that did not meet all field selection criteria. We chose these additional sites non-randomly, and used them to ensure we had a good balance in the distance from each site to the nearest road. When we checked this factor in the randomly-chosen sites, the average distance from sites to roads varied among the sectors and road types, in a way that would have limited inference. We therefore intentionally identified candidate sites that we had rejected earlier because they did not have frogs or because access was problematic, and decided to accept some sites that did not meet all criteria, but were closer to the road in some sectors or farther from the road in others. During the 2010 pilot year of this study, we sampled 6 sites intensively to determine the sampling strategy for the 36 study sites sampled during 2011 and 2012.