How Passive Sound Captures Roost Geometry Near Torpid Myotis

The Geometry You Cannot Reach With a Light

In an Ulster County limestone hibernaculum, a 2022 winter count team needed to confirm whether a 900-bat cluster of Myotis lucifugus had shifted 1.4 m along a ceiling rib between December and February. The answer required precise ceiling geometry, but any survey lamp raised within 3 m of the cluster pushed surface temperatures up by a measurable fraction of a degree and triggered visible arousal in at least a dozen individuals. The team settled for a conservative photograph from the chamber entry and called the cluster "stable" because they had no other option. That ambiguity matters: cluster shifts are one of the earliest behavioral signals of a changing microclimate, and misreading them delays management response by an entire winter.

The core difficulty is that centimeter-accurate roost geometry near torpid bats has always required either entry or active illumination. Traditional disto-and-tape surveys demand hands on the rock. Terrestrial LiDAR demands a tripod in the chamber. Researcher entries into hibernacula produce the arousal problem that drives the entire field toward passive alternatives. The good news is that passive sensing is mature enough to pick up where those methods stop.

The geometric stakes are higher near a torpid cluster than anywhere else in the cave. A 0.4 m drop in cluster ceiling height moves bats out of a stable warm plume into a thermal sink. A 1.2 m lateral drift puts them in or out of a Pseudogymnoascus destructans growth zone where humidity and temperature converge. None of these distances register on a chamber-wide map drawn from a tape-and-compass survey, but every one of them registers on cluster-level fat consumption across a winter. Traditional methods produce a static drawing of an architecturally interesting cave; modern bat conservation needs a living surface model where the cluster-bearing sections of the ceiling are resolved at centimeter precision because the biology happens at centimeter precision. EchoQuilt's passive inversion is built specifically for this gradient — fine resolution where bats are, coarser resolution where they are not.

Stitching Ceiling Ribs From Ambient Acoustics

EchoQuilt treats the hibernaculum as a naturally excited acoustic system. Every drip, every low-frequency airflow rumble, every settling creak is an impulse that echoes off the ceiling, the cluster, and the walls. A dense microphone array near the roost can invert those echoes into geometry without ever emitting a pulse. Microphone arrays enable 3D localization of bat calls and other cave sources, and array geometry drives spatial tracking accuracy in a way that is well understood in bioacoustics. What EchoQuilt adds is the stitching layer that merges arrivals from dozens of independent sources into a single 3D quilt, patch by patch.

The quilt is spatially adaptive. In the drip-dense region beneath a percolation ceiling, patches resolve to roughly 3 cm. In a low-excitation back chamber with only residual airflow, patches hold at 15 cm until more sources stitch through. Biologists reading the quilt see confidence as texture rather than a single number, which matches how they already reason about survey uncertainty. Near a torpid cluster, the acoustic shadow of the bats themselves becomes a useable feature: the cluster absorbs high-frequency drip reflections and scatters low-frequency airflow, so its outline emerges as a soft gradient on the quilt rather than a hard boundary.

This is possible because hibernacula are quietly noisy. AudioMoth cave-entrance recordings already convert passive audio into population estimates, and acoustic detection of hibernating Myotis works even when bats are not visible. EchoQuilt extends the same passive philosophy from occupancy to geometry. A four-sensor array in a mid-size chamber generates on the order of 40,000 independent geometric stitches per day once drip, airflow, and micro-tremor sources are separated, which is enough to hold centimeter precision across a 20 m roost span. Cave-diving teams use comparable ambient acoustic walls techniques to reconstruct flooded conduits from water movement, and the underlying geometry inversion shares a common math core across the two domains even though the physics of acoustic propagation in air and water differ substantially.

For a biologist, the workflow looks unfamiliar but reads like a familiar output. Install sensors during pre-hibernation swarm, retrieve them after emergence, and review a quilt that updates day by day across the winter. The microclimate pockets visible in the airflow layer align with the cluster geometry visible in the same dataset, which means structural and behavioral questions get answered from a single passive deployment rather than from two separate field protocols.

Advanced Tactics for Cluster-Proximate Geometry

Three refinements matter when a cluster sits directly in the mapping volume. First, weight the inversion toward sources at least 2 m from the cluster so that body-surface absorption does not poison the geometry estimate. Airflow rumble from 10 m away carries better information about the ceiling rib above the cluster than a drip 40 cm from its edge. EchoQuilt automates this weighting, but biologists should inspect the source map before trusting a fine-scale cluster boundary.

Second, treat the cluster outline as a distribution rather than a line. Because torpid Myotis sodalis and Myotis lucifugus slowly shift micro-positions across days, the quilt should report a 7-day-integrated probability field rather than a single contour. This matches what bat ecologists already know about dynamic cluster geometry and avoids false alarms when a few individuals rotate within a stable group. Third, use pre-hibernation swarm-season recordings to build a bat-free geometric baseline. The ceiling ribs and flowstone do not change in a single season, so any deviation between the swarm baseline and the mid-winter quilt is real cluster movement or real microclimate-driven acoustic change, not a calibration drift.

Operationally, this means the annual survey collapses into two short visits: September sensor deployment during swarm and April retrieval after emergence. The mid-winter chamber stays dark, quiet, and torpid. For Myotis sodalis, NLEB, tri-colored bat, and Perimyotis populations still losing ground to WNS, removing the headlamp from the geometry workflow is not a nice-to-have; it is the difference between a survey that adds data and a survey that adds mortality.

A fourth tactic worth adopting is cluster-scale validation against a single reference cluster per region. Pick one well-documented hibernaculum where decades of visual count data exist, install the EchoQuilt array, and let the geometry quilt accumulate for two winters in parallel with a continued visual survey by a small team. The dual-track period is not permanent — once the inversion's cluster-boundary error is bounded against the visual reference (typically within 8 cm at 90% confidence after 14 days of integration), the visual survey can be retired in favor of the passive quilt at sites with similar acoustic geometry. This validation pattern matches how acoustic monitoring was originally accepted as a substitute for mist-netting: a regional reference site, a multi-year overlap, and a documented error bound that regulators can cite. Cross-reference the validated geometry against cluster shift data workflows to extend the validation toward behavioral metrics as the dataset matures.

Fifth, invest in the swarm-season geometric baseline as a multi-purpose dataset. The same baseline that supports cluster-proximate inversion also supports detector calibration, classifier training, and entrance-airflow characterization for the upcoming winter. Bat crews who treat the swarm baseline as a single annual data product rather than as a per-task chore tend to produce more consistent multi-year datasets, and the time investment in September pays back across all winter analyses.

Get Early Access to EchoQuilt

If you run winter cluster counts on Myotis lucifugus, Indiana bat, or tri-colored bat hibernacula and currently depend on headlamp-and-tape surveys for ceiling geometry, EchoQuilt was built for your next field season. We are reserving pre-swarm deployment slots for USGS hibernacula surveyors, state DNR bat crews, and Section 7 consulting teams that need centimeter geometry without arousal risk. Pilot kits ship in early August so calibration can complete before the swarm window opens, and each kit arrives with a hibernaculum-specific calibration set tuned against pre-existing Anabat detector libraries and AudioMoth swarm-season recordings from comparable Priority 1 and Priority 2 sites. We commit to NABat acoustic-protocol-conformant outputs from day one, including species-resolved cluster geometry exports compatible with USGS ScienceBase ingestion and IUCN-compliant disturbance-budget logs that travel directly into your Section 7 documentation.

Pilot teams shape the cluster-proximate inversion priorities for the 2027-28 reference release, with priority slots going to multi-state surveillance partnerships covering documented Myotis sodalis hibernacula. Join the Waitlist for Hibernacula Biologists to claim a sensor kit for a pilot hibernaculum and help us tune the cluster-proximate inversion against your existing banding and Anabat records.