Pairing Anabat Detectors With 3D Roost Reconstructions
A Detector Pointed Into a Chamber You Cannot See
A 2023 NABat participant in a tri-colored bat hibernaculum in the southern Appalachians ran a Song Meter SM4BAT-FS at the cave entrance for 180 winter nights, captured over 420,000 echolocation calls, and identified seven species in the local community. The call log was excellent. The spatial picture was a thin one: every call collapsed to the single point where the detector sat. Whether the calls came from the far back chamber, the cluster on the east wall, or a transient individual near the ceiling breach was invisible in the dataset. A biologist trying to understand winter roost behavior ended up with a census that lost most of its spatial information in the first integration step.
This is the standard limitation of point detectors in 3D spaces. The USGS NABat Mobile Acoustic Transect Surveys SOP and the NABat processing pipeline guidance standardize how calls get recorded and classified, but not where in 3D space each call originated. The North American Bat Monitoring Program supports 47 shared species across U.S., Canada, and Mexico, and the community has years of high-quality detector data awaiting a spatial overlay.
The information loss matters for specific research questions. A study of swarming behavior wants to know whether mating-site activity concentrates near the entrance or extends into deeper chambers. A WNS-progression study wants to know whether infected individuals shift toward different microclimates than healthy ones. An emergence study wants to know which exits are used at what times across the spring. Every one of these questions requires localizing calls in 3D space, and a single detector at a single point cannot answer any of them. Adding more detectors helps but quickly becomes both expensive and difficult to maintain — each added detector needs power, storage, and decontamination cycles. A unified array that produces 3D positions from a small number of synchronized sensors is structurally a better answer than a dense forest of independent detectors, and that is the shape EchoQuilt offers.
Binding the Detector Stream to the Quilt
EchoQuilt's pairing approach keeps the existing Anabat or SM4BAT detector as the primary call-capture device and uses the mapping array to localize each detected call in 3D space. The biologist's detector workflow does not change — same hardware, same SD cards, same SonoBat or Kaleidoscope post-processing — but each classified call is now time-synchronized with the EchoQuilt array and assigned coordinates in the hibernaculum's 3D quilt. The result is a paired dataset: the detector's high-sensitivity call catalog with precise species ID, and the quilt's spatial assignment of each call to a voxel in the chamber.
The stitching uses a straightforward multilateration core supported by the quilt's known array geometry. When the primary detector logs a call at time T, the mapping array's channels are cross-correlated for the same time window to find the call's time-of-arrival differences across sensors. The differences invert into a 3D position, typically accurate to 20-40 cm in the chamber volume. Confidence depends on how many array channels caught a clean copy of the call; for a four-sensor array in a 25 m chamber, roughly 70-85% of calls localize cleanly, with the remainder filed in a "detected but not localized" pool that still contributes to species counts.
This binding enables questions that neither dataset answered alone. Which ceiling zones do tri-colored bats use for early-winter establishment versus mid-winter settling? Do Myotis sodalis individuals that arouse in January fly into the entrance throat or stay in the back chamber? Where do NLEB calls cluster relative to the cold-sink microclimate pocket? Pre-EchoQuilt, these questions required visual surveys during torpor — exactly the thing the whole field is trying to avoid. The counting in the dark non-intrusive laser scanning approach paired LiDAR with call data, and more recent work on combining acoustic tracking and LiDAR for bat flight in 3D space shows the integrated method working, but both still require LiDAR's active emission. EchoQuilt replaces the LiDAR with passive ambient acoustics while keeping the detector pairing intact.
Practically, this also preserves NABat compatibility. Calls still flow through the standard processing pipeline and report out to NABat in the expected format; the 3D coordinates are additional metadata that travels with each call. That pairs with echolocation filtering at the separator layer so the localized calls arrive already cleaned of cross-talk from drip and airflow noise.
Advanced Tactics for Detector-Quilt Pairing
Three tactics extend detector-quilt pairing into higher-value analysis. First, use the paired data to build species-specific spatial preference maps. A three-winter dataset of localized calls gives enough statistical power to show that Myotis sodalis calls consistently originate near cooler east-wall zones while Perimyotis calls come from warmer back-chamber airflow plumes. These maps inform roost-site prediction in neighboring hibernacula that lack multi-year data and improve Section 7 consultation quality. The species preference maps also help gate designers — knowing which chambers each species prefers means a bat gate that protects the chamber a Myotis sodalis cluster has used for forty years can be designed without disrupting an adjacent chamber preferred by Eptesicus fuscus or Townsend's big-eared bat. Pair the species preference maps with species identification work downstream to refine the per-species classifier against the local call library, which raises classification confidence further.
Second, track individual-level mid-winter arousal trajectories by following localized call sequences over minutes. When a single bat arouses and flies, the quilt can reconstruct the flight path from a burst of calls, showing whether the bat drifted within the cluster volume or traveled to the entrance and back. Aggregate statistics on arousal flight patterns are a new kind of behavioral data that pre-quilt workflows could not produce without active visual surveillance. Flight pattern analysis also reveals when arousing bats drink at specific in-cave water sources versus when they simply circle and return, which has direct relevance to the waking-to-drink hypothesis for evaporative water management in WNS-stressed colonies.
Third, use the localized call data to refine the occupancy layer's cluster boundaries. Calls originating near but outside a cluster's acoustic shadow suggest the shadow boundary was under-estimated; calls inside the shadow are rare because torpid bats are silent. The quilt uses this feedback to tighten cluster boundary estimates across the winter, producing a map that is genuinely a map of the colony rather than of the cave. Fourth, integrate the localized call data with PIT detection data where PIT gates are installed at hibernaculum entrances. A PIT detection at the entrance at 2:14 AM followed by a localized call burst in chamber B at 2:17 AM, then a second call burst back at the entrance at 3:42 AM, reconstructs an individual's arousal trip in time and space.
The composite record is far more informative than either dataset alone, and it gives WNS-progression studies an individual-level view of behavioral change that aggregate counts have always missed. Cave-diving teams use a similar idea for motion capture pairing in flooded conduits, where diver position gets paired with acoustic environment reconstruction to produce trajectory records that neither modality alone could deliver.
Get Early Access to EchoQuilt
If you already run Anabat or SM4BAT-FS detectors in your hibernacula for NABat reporting and want 3D spatial assignment for every classified call without a single additional winter entry, EchoQuilt's detector-pairing mode was specifically built for the way your team already works. The pilot cohort is designed for NABat participants, state DNR bat crews, and USGS hibernacula surveyors who want to augment existing call logs with spatial coordinates while keeping their classification pipeline unchanged. Sites with multi-year detector archives are particularly valuable because retrospective spatial assignment can be applied to historical call data once a quilt is established, effectively giving years of legacy detector records new analytical depth. Pilot participants will also receive guidance on time-synchronization, detector pairing, and NABat report formatting that preserves spatial metadata across the data lifecycle.
Join the Waitlist for Hibernacula Biologists to coordinate detector co-deployment and time-synchronization with the mapping array for the 2026-27 hibernation season.