Salamander Bsal Watershed Surveillance

What this research found

Where should a national early-detection network for the salamander-killing chytrid fungus Batrachochytrium salamandrivorans (Bsal) sample first? USGS range maps for 184 native salamander species were crossed against a national pathogen swab survey to identify which species and watersheds have never been sampled, and a set-cover optimisation then selected 25 subwatersheds covering 115 of the 184 species and 91.0% of a range-rarity priority weight. A binomial detection model fixed field effort at 66 swabs per site, or 1,650 nationally, which was written up as a field-ready protocol.

• Existing surveillance — 11,170 swabs over 772 field visits in 458 subwatersheds — leaves most salamander range untouched. Of the 184 native species, 45 have never been swabbed anywhere in their range and 139 only partially; none is fully sampled, and the median species has 98.9% of its watersheds unsampled.
• Twenty-five subwatersheds suffice to reach 115 of the 184 species and 91.0% of the total priority weight. A mixed-integer solver certified an optimum of 7.0551 out of 7.749 achievable, beating a greedy solution by 0.02%.
• Priority is dominated by a few micro-endemics. The narrowest-ranging species, the Austin blind salamander (Eurycea waterlooensis), occupies 138.0 km² against roughly 3.08 million km² for the widest, and all top 20 by weight are lungless salamanders from the Texas Edwards Plateau, southern Appalachians, and California.
• Field effort follows directly from the detection arithmetic: 66 swabs per site give a 95.21% chance of detecting Bsal at 5% design prevalence with a 90%-sensitive assay, totalling 1,650 swabs across 25 sites. Weaker assumptions get expensive fast — at 1% prevalence and 80% sensitivity, 373 swabs per site would be needed.
• Reconciling names between the pathogen survey and the range maps matched 10,409 of 10,451 salamander records (99.60%), including 136 records recovered from a misspelt epithet in the range index; only the 42 records identified to genus alone were dropped.

How it was done

Three USGS products were combined in an equal-area projection: a national Bd/Bsal pathogen survey (11,189 records, 10,451 of them salamanders), GAP species range maps filtered from 1,719 species down to the 184 native salamander species, and the Watershed Boundary Dataset's 12-digit subwatersheds. Scientific names were reconciled between survey and range map through a cascade of exact, synonym, subspecies, and spelling matches. Each species then received a priority weight equal to the smallest range area divided by its own, multiplied by the share of its watersheds never sampled, and a maximum covering location problem was solved both greedily and exactly to pick 25 subwatersheds. A binomial imperfect-test detection model set the swab count per site, and the results were operationalized into a field standard operating procedure covering habitat scouting, swabbing technique, between-site decontamination, sample preservation, and diagnostic controls.

Data sources

• USGS national Bd/Bsal survey v2.0, November 2024 — 11,189 records including 10,451 salamander swabs at 603 unique coordinates
• USGS GAP Species Range Maps, CONUS 2001 v1 — 1,719 species filtered to 184 native salamanders spanning 76,156 subwatersheds
• USGS Watershed Boundary Dataset — 103,068 national 12-digit subwatershed polygons
• AmphibiaWeb, ITIS, and Amphibian Species of the World as taxonomic authorities

Figures from this analysis