Swarming Chamber Acoustics: What Sound Maps Show That LiDAR Misses

The LiDAR Scan That Missed Half the Swarm

A Pennsylvania swarming survey in 2019 deployed a terrestrial LiDAR scanner at a cave entrance to map September's swarming activity. The scanner captured stunning 3D geometry of the entrance chamber — ceiling ribs, wall textures, floor morphology down to millimeters. It captured almost none of the flying bats. Swarming Myotis lucifugus threaded through the ceiling ribs, darted behind limestone fins, and spiraled through sub-meter pockets that were line-of-sight-obstructed from the scanner head. The LiDAR survey was a beautiful cave map and a near-empty bat map.

The limitation is intrinsic to the technology. LiDAR surveys fundamentally require line of sight, which means features behind features are invisible. Movement ecology research using acoustic tracking combined with LiDAR documents the same constraint — LiDAR provides the geometric context, but the motion data has to come from acoustics that do not need line of sight.

The biology makes this a structural problem for swarming research. Acoustic loggers at a SE England swarming site captured rich activity patterns that direct observation and visual survey could not. Cave use by swarming species is greatest in caves with extensive chambers — exactly the cave morphology where LiDAR line-of-sight fails most often. The features that make a cave a good swarming site are the same features that prevent LiDAR from documenting the swarming well.

Swarming matters because compositional similarity between autumn swarming and winter hibernation assemblages is strong. The bats swarming at a cave in September are largely the bats hibernating in that region by November. Missing the swarm data means missing the mating-season genetic-mixing event that shapes the hibernaculum's winter colony.

Stitching Swarm Flight Paths Into the Acoustic Quilt

EchoQuilt's passive sensing is omnidirectional and does not require line-of-sight. A ceiling-mounted array captures echolocation calls that reflect off cave surfaces and off flying bats, and the array triangulates source position from multi-microphone phase differences. A bat flying behind a limestone rib still produces echolocation calls that the array captures from multiple angles, and the quilt reconstructs the flight path through the obscured zone. What LiDAR renders as a gap in the point cloud, EchoQuilt renders as a continuous flight trajectory.

The 3D quilt stitches swarming flight paths patch by patch. Each passing bat's echolocation track writes a thin line across the patch surface. Across a night of swarming activity, the quilt accumulates thousands of track lines, forming a density map of the cave's swarming corridors. High-traffic ceiling pockets light up as bright patches. Peripheral corridors with occasional traffic show as faint patches. The overall swarming geometry emerges from the patch aggregation.

Social calls layer in as a separate quilt dimension. Soundscape analysis of swarming Myotis demonstrates noninvasive acoustic species identification at swarming sites. EchoQuilt tags every social-call event with species confidence, so the patch map separates M. lucifugus swarming flights from Perimyotis swarming flights — a distinction LiDAR cannot make even in theory, since LiDAR sees shape, not vocal identity.

The integration with maternity swarm survey from Post 11 is direct. The same quilt that tracks summer maternity roosts extends into autumn swarming, showing the same cave system transitioning function across seasons. Reproductive females that raised pups in a summer roost appear in autumn as swarmers at the adjacent winter hibernaculum. The quilt is one continuous record, not two separate surveys.

Roost geometry analysis works against the same quilt substrate. A winter roost geometry at patch C-09 is the same spatial frame as the autumn swarming corridor that passes through patch C-09. Biologists reading the quilt see how swarming traffic concentrates near future winter roosting patches, which is a prediction signal for winter cluster locations.

The data volume is tractable. Instead of storing every raw sample across a 10-hour swarming night, EchoQuilt extracts compact acoustic features (call ID, arrival times across microphones, species tag) that reconstruct flight paths without the storage cost of full waveforms. A laptop-scale archive holds a full September's worth of swarming data for a typical hibernaculum.

LiDAR still has a role. Scan the cave once in summer when empty. EchoQuilt's flight paths lay over the LiDAR mesh as motion on static geometry. The combined rendering — rock structure from LiDAR, flight trajectories from passive acoustics — gives the most complete swarming visualization the field has access to. LiDAR alone misses the swarm; acoustics alone misses some geometric detail; stitched together, the two produce a publishable-quality cave swarming map.

Advanced Tactics for Swarm-Chamber Sound Mapping

Tactic one: deploy a dense ceiling array in known swarming chambers. Every additional microphone improves multi-source localization in acoustically dense scenes. A 12-node array resolves individual flight paths in conditions where a 4-node array produces only aggregate density.

Tactic two: capture the full swarming season, not just peak weeks. Swarming activity ramps across August, peaks in late September, and tails into early November depending on latitude. A full-season quilt shows the arrival-peak-departure arc; a 2-week snapshot misses the onset and termination.

Tactic three: correlate swarming density with lunar cycle and local weather. Higher activity on moonless warm nights is documented at many swarming sites. The quilt's environmental-context overlay captures these covariates automatically, so analyses can adjust for environmental drivers. Adding barometric pressure trend, dewpoint, and wind direction at the cave entrance separates weather-driven activity dips from biologically meaningful swarming pauses.

Tactic four: register swarming flight paths against winter cluster patches. A patch that carries heavy swarming traffic in October and hosts a dense Myotis lucifugus cluster in February is a focal patch for the site. Management decisions (gate placement, disturbance protection) should prioritize protecting these patches. The October-to-February transition is where the autumn quilt and winter quilt stitch into a single seasonal record.

Tactic five: contribute swarming quilts to NABat. Swarming data is under-represented in NABat's current records relative to summer and winter surveys. A standardized swarming quilt export fills a genuine gap in the national dataset.

Tactic six: track individual swarmers across multi-night windows when call signature variation permits. Individual Myotis lucifugus echolocation calls carry small but measurable inter-individual variation in peak frequency and call shape. EchoQuilt's call-clustering algorithms, when trained on a swarming season's worth of calls, can sometimes group calls into individual "voices" — not species-level identity, but call-cluster identity that persists across multiple nights. A swarming individual that visits patches A-04, B-07, and C-12 across three consecutive nights writes a sociometric trace that mark-recapture telemetry could not match without invasive tagging.

Tactic seven: cross-reference swarming density with SonoBat classification outputs run on the same audio streams. SonoBat's species-confidence values let the quilt down-weight uncertain species assignments at the patch level, producing a probabilistic species-density map rather than a deterministic one. NABat reviewers value the explicit uncertainty representation when evaluating contributed swarming records.

Tactic eight: characterize swarming-chamber acoustic absorption at deployment time. Limestone walls vary in absorption coefficient based on solution-weathering history; a 0.1 difference in absorption changes effective array sensitivity by 30 percent. A one-time impulse-response measurement at deployment lets the quilt's localization algorithm correct for site-specific acoustic environment, improving flight-path reconstruction accuracy in high-reverb chambers where naive localization fails.

Tactic nine: synchronize swarming-chamber array deployment with Pseudogymnoascus destructans (Pd) swab campaigns so the autumn swarming dataset includes Pd-arrival timing context. A swarming density anomaly in a year that Pd is first detected at the site becomes a Pd-baseline observation; subsequent years' swarming densities are then comparable across the disease-progression arc rather than treated as independent snapshots. Storage planning for these multi-year swarming archives benefits from sparse-feature mapping techniques drawn from lava-tube planetary work, which keep the per-night storage footprint small enough that a decade of Pd-baseline data fits on commodity hardware without sacrificing the call-arrival fidelity that flight-path reconstruction requires.

Ready to see swarming flight paths that LiDAR cannot reach and summer acoustic detectors cannot resolve? EchoQuilt gives state DNR bat crews, NABat contributors, and swarming-site researchers a 3D quilt that captures motion where line-of-sight fails. If your cave has ceiling ribs and narrow pockets, the quilt sees through them. Join the Waitlist for Hibernacula Biologists.