Case Study: Tracking WNS Progression Through Map-Anchored Counts
The Problem With Counting WNS Declines When the Count Itself Drifts
A state biologist stood under the entrance chamber of a karst hibernaculum in East Tennessee in January 2014 and recorded 1,812 northern long-eared bats (NLEB) wedged into three cluster zones. Four winters later, a different biologist, a different headlamp angle, and a cluster that had migrated 4 meters along the bedding plane produced a count of 27. USFWS reclassification of the northern long-eared bat cites 97 to 100 percent declines in affected populations, and Bernard et al. 2017 in Ecology and Evolution documented NLEB declines of 69 to 98.5 percent across Tennessee hibernacula. The absolute signal is catastrophic.
The problem is the counting method. Digital photography work by Meretsky and colleagues in 2010 found visual cluster estimates for Indiana bats varied 76 to 142 percent between observers while photo counts varied under 1.5 percent. When white-nose syndrome is pulling 90 percent of bats out of a cave, observer drift can swing a count by another 40 percent in either direction. Separating the disease from the method matters. USGS national reporting shows WNS killed more than 90 percent of three species in under a decade, and that inference rests on repeated counts that observers trust.
Tennessee's situation specifically compounds the methodological problem. The Tennessee Wildlife Resources Agency manages dozens of NLEB hibernacula across the eastern karst belt with limited winter survey staffing. A two-biologist crew completing a full state survey rotation may visit each Priority 1 site once every three winters, with different crew compositions each rotation. Between observer rotation, lamp-angle drift, and the cluster's own movement along the bedding plane, a 1,800-bat 2014 count and a 27-bat 2018 count are not directly comparable until somebody anchors both observations to the same ceiling geometry. That anchoring problem motivated the EchoQuilt deployment described below.
The federal stakes deepen the picture. NLEB's 2022 endangered-species reclassification triggers ESA Section 7 consultation for any federal action touching an occupied hibernaculum, and the underlying population-trend documentation has to defend the reclassification against legal challenge. A state's NLEB count series that swings 40 percent on observer changes is exactly the kind of methodological soft spot that opposing counsel attacks in delisting petitions or take-permit litigation. The case study described here was framed from day one as defensible evidence for federal review, not just internal state monitoring.
A Three-Year EchoQuilt Deployment Anchored to Ceiling Landmarks
The Tennessee case study that anchors this post ran from winter 2021 through winter 2024 at a single karst hibernaculum on state wildlife management land. The deployment goal was straightforward: every cluster recorded over three winters had to register to the same ceiling geometry so that year-over-year shrinkage read as population loss rather than as map slippage.
EchoQuilt stitches passive sound and motion into a 3D quilt of the hibernaculum. Each SM4BAT and AudioMoth node at the entrance, swarming chamber, and three interior rooms contributed overlapping patches — acoustic arrivals from wing-beats, chirps, and arousal stirrings — and the reconstruction pipeline co-registered those patches against a persistent ceiling landmark set (three stalactites, two joint intersections, a breakdown block edge). Bats moved along the bedding plane between years. The ceiling did not. Anchoring clusters to the ceiling instead of to cardinal directions or to a first-year cluster centroid let us track the same roosting surface across four winters with no cave entries during hibernation.
The cross-niche parallel is direct: the retreat mining case showed mine-rescue teams anchoring pillar-failure acoustics to stable ceiling ribs for the same reason. Whether the signal is roof stability or bat cluster position, the reference frame has to be the geometry the signal is happening against.
Over the three winters, cluster A (the largest NLEB aggregation) contracted from an initial 14.2 square meters of occupied ceiling to 0.8 square meters. Cluster B vanished between winter 2 and winter 3. Cluster C migrated 3.1 meters north along the same bedding plane but stayed on the same ceiling surface, letting us reject the hypothesis that bats had simply moved deeper. USFWS documentation describes that same pattern across dozens of hibernacula.
The quilt shown above is from winter 3. Each patch is a ceiling tile reconstructed from passive acoustic arrivals during the December pre-hibernation swarming period, stitched together along overlapping edges so that cluster A's 0.8-square-meter remnant sits in the same world coordinates as its 14.2-square-meter footprint from winter 1. The cluster outlines on the figure are the 95 percent occupancy contours from that winter's wing-beat density field.
Advanced Tactics for Anchored WNS Case Studies
Three tactics separated this Tennessee case study from prior visual-count work at the same hibernaculum. First, we logged every sensor position against the same ceiling landmark set at each winter's pre-swarm deployment in September, using a passive motion ping from each node to three landmark stalactites. Positional drift between winters stayed under 3 centimeters per node. Second, we treated arousal events as a confound rather than as noise: Microbiology Spectrum coverage of WNS and human activity shows WNS-infected bats arouse 3 to 4 times more frequently, so the wing-beat rate from a shrinking, sick cluster is not comparable to the rate from a healthy one. We normalized counts to pre-swarm ceiling area occupied rather than to acoustic activity rate.
Third, we paired the cluster maps with non-intrusive monitoring protocols so that no human entered the cave between November and April. Visual entry-light surveys in March of each year confirmed the acoustic counts within 5 percent, which was inside the observer-variance band that Meretsky documented.
Fourth, we ran a per-winter validation against a paired photograph reference at the single accessible cluster. The Tennessee site allowed photograph access at the entrance-chamber Cluster D (a small Perimyotis subflavus aggregation that hibernated within ladder reach of a stable platform). Photograph-counted Cluster D bats matched the EchoQuilt density-derived count within 1.7 percent across all three winters. That paired validation became the calibration anchor for the unreachable Clusters A through C, which the photograph method could never have measured because they sat behind ceiling pendants 6 meters above unstable breakdown.
Fifth, we logged every quilt edit, every reconstruction parameter change, and every classifier version into a per-winter audit log. When the case study was peer-reviewed for publication, every cluster contour the reviewers questioned could be traced back to a specific patch, a specific reconstruction commit, and a specific classifier model. Audit traceability is what turned a three-winter EchoQuilt deployment into a publishable case study rather than a vendor demo.
Sixth, we cross-referenced the cluster shrinkage timeline against the USGS White-Nose Syndrome occurrence map for that Tennessee region. Pd was first detected at this cave in winter 2015, the first cluster shrinkage exceeded measurement noise in winter 2017, and the catastrophic decline year was winter 2021. That two-year lag from Pd detection to first measurable population effect is consistent with the regional pattern for NLEB. The case study's value to other states is not just the absolute decline curve, but the timing relationship between Pd arrival and population response, which only becomes legible when the count method holds steady across the whole arc.
Seventh, we filed quarterly progress reports with the state DNR's nongame wildlife coordinator and the USFWS Cookeville Tennessee Field Office. The reports kept the regulatory partners informed before any publication and let the field office incorporate the EchoQuilt findings into their NLEB Section 7 consultation positions in real time. This same anchoring discipline scales to statewide surveillance when dozens of hibernacula need comparable year-over-year quilts. The seven tactics together separate a ceiling-anchored case study from a sequence of visually-estimated annual counts that drift more than the underlying biology does.
Bring Anchored Counts to Your WNS Response
If your state's WNS response is still anchored to entry-light cluster estimates that swing 76 to 142 percent between observers, you are measuring observer variance as much as you are measuring disease progression. EchoQuilt's ceiling-anchored reconstructions let NABat partners and state DNR crews publish year-over-year decline figures that hold up to federal review, without adding winter entries. The seven-tactic case study described above adapted to a Tennessee karst hibernaculum, but the same anchoring discipline transfers to mid-Atlantic limestone caves, Midwest quarry systems holding Indiana bat populations, and Pacific Northwest lava tube hibernacula with appropriate cave-type detector retuning. The barrier to adoption is not technological — it is the willingness to invest a deployment cycle in establishing the ceiling landmark set and the per-cluster calibration.
Once that investment is in place, every subsequent winter compounds the value of the anchored archive at marginal additional cost. Join the Waitlist for Hibernacula Biologists and walk us through one of your sentinel sites. We will map out a three-winter anchoring plan that pairs with your existing NABat sampling and lets the next NLEB 5-year review read population trajectory rather than method drift, with explicit stability metrics defending each year's reported count.