Scaling Passive Roost Mapping Across State Hibernacula Networks

The Staffing Math Behind Statewide Hibernacula Surveys

Kentucky Department of Fish and Wildlife Resources tracks roughly 40 Priority 1 hibernacula for Indiana bat (Myotis sodalis) and NLEB. If each site requires two biologists for an entry-light survey, travel, and post-trip data entry, the arithmetic does not close in a single winter biennium survey window. The PLOS ONE review of western hibernacula surveys by Johnson and colleagues compiled 4,549 survey records across 11 western states and flagged that the records lived in disparate formats, with gaps that reflect staffing rather than bat behavior. NABat's coordinating program at USGS now spans state, federal, and tribal agencies precisely because no single agency has the crews to survey everything themselves.

The federal survey protocol adds another constraint. USFWS range-wide Indiana bat and NLEB survey guidelines require methodology consistent enough across states that a survey in Kentucky, Tennessee, and Indiana feeds into the same occupancy model. Consistency plus coverage plus biologist hours is an overdetermined system. Something gives. Usually it is sample coverage, then consistency, then hours, in that order.

Kentucky's specific situation is illustrative. The state contains parts of Mammoth Cave National Park, the largest cave system on Earth at over 426 miles of mapped passage, plus the Western Kentucky Coalfield mines that hosted Indiana bat populations until WNS arrival. A single biologist responsible for the 40 Priority 1 hibernacula list cannot personally enter every site within a winter even if travel logistics permitted, and the karst geography means many caves are accessible only through landowner-permitted gates whose schedules constrain visit windows further. The structural staffing problem is not solvable by adding more entries; it is solvable by replacing some entries with passive sensors that work when the biologist is not there.

An EchoQuilt State-Network Dashboard on a GRTS Frame

EchoQuilt scales roost mapping without scaling biologist hours by letting passive sound-and-motion nodes patch together the inside of a hibernaculum while the biologist is off-site. The dashboard layer that sits on top of those patches is what turns a pile of individual caves into a state network. Each hibernaculum's quilt — a 3D reconstruction stitched from overlapping acoustic patches recorded by SM4BAT and AudioMoth nodes deployed pre-swarm — registers to a common geodetic anchor. The dashboard then overlays those anchored quilts onto a NABat GRTS sampling frame.

Loeb and colleagues laid out that frame in the 2015 USGS plan for NABat, defining a continent-wide 10-kilometer grid with a generalized random-tessellation stratified (GRTS) draw order. EchoQuilt's state dashboard plots each hibernaculum's quilt at the cell it falls inside, inherits the NABat draw priority, and flags cells that have an occupied Priority 1 hibernaculum but no active acoustic coverage. NABat's data collection program now runs four survey types across more than 2,400 users and 480 partners, and a state dashboard that consumes that same GRTS structure slots directly into federal aggregation.

Think of the dashboard as the master quilt with 40 panels, one per Priority 1 hibernaculum, stitched along the NABat grid lines. Individual caves remain their own quilts, but the state-level view lets a Kentucky biologist see cluster shrinkage at site 17 in the same visual frame as swarming activity at site 34. That parallel viewing is what makes statewide trends — like WNS progression case documentation — interpretable across sites rather than just within each one.

This mirrors how multi-level scaling works in coordinated mine-rescue deployments. A single room-and-pillar acoustic map is useful, but the rescue-coordination payoff shows up once dozens of rooms share a common spatial frame. The failure modes are the same: inconsistent registration, drifting sensor calibration, and metadata that describes the rig rather than the geometry.

The mockup shows a state view centered on Kentucky. Each filled cell is a NABat GRTS grid cell with at least one EchoQuilt-anchored hibernaculum quilt. The color ramp encodes January 2024 Indiana bat cluster density relative to the first anchored winter. The sidebar breaks down per-site metrics: occupied ceiling area, acoustic activity rate, and the latest pre-swarm deployment date. A biologist can click a cell, drop into its quilt, and see individual cluster outlines against the same landmark set as every other site in the network.

Advanced Tactics for Scaling Up

Weller's 2024 dual-frame sampling work in the Journal of Wildlife Management lays out a method for survey design across 800-plus cave networks, pairing a probability sample with a targeted index sample so that population trend estimates stay defensible as coverage grows. The EchoQuilt state dashboard adopts dual-frame logic directly: the GRTS draw provides the probability sample, and the Priority 1 hibernacula list provides the index sample. Both frames feed the same trend model.

Three field tactics matter. First, standardize the pre-swarm deployment window across the state so that a September 15 through October 10 window applies uniformly; this lets the swarming patch of each quilt be compared across sites without seasonality confounds. Second, adopt a common landmark-anchor protocol so that every node's first ping registers to a durable ceiling feature photographed and georeferenced at the first deployment; cross-site comparisons fail fast when sensor positions drift between winters. Third, use cross-agency data protocols so that USFWS field offices, state DNR crews, and tribal biologists can all read from the same dashboard under ESA Section 7-compliant permissions.

Fourth, build a per-cell occupancy model that consumes both the GRTS-drawn cells and the Priority 1 index cells in a hierarchical Bayesian framework. The hierarchical structure lets the model borrow strength from data-rich cells when estimating occupancy in data-poor cells, which is exactly the staffing-constrained situation Kentucky and most other states find themselves in. EchoQuilt's quilt-aggregation layer outputs the per-cell occupancy point estimate and uncertainty band for each NABat reporting cycle, with the model code versioned alongside the data so federal reviewers can replicate the analysis from raw inputs.

Fifth, schedule annual sensor-rotation maintenance against a per-state calendar that minimizes mid-winter disturbance. A typical Kentucky deployment cycle has nodes installed in September, batteries swapped in May after emergence, and full sensor rotation every three years. Coordinating this maintenance schedule with USDA Forest Service permit windows for caves on national forest land prevents permit-induced delays that would otherwise force a partial winter without coverage at certain sites.

Sixth, federate state-network dashboards into multi-state regional pools where species ranges cross state lines. Indiana bat range covers parts of 18 states; a Kentucky-Tennessee-Virginia federated dashboard tracks the same maternity and hibernation populations across jurisdictional boundaries that the bats themselves do not respect. EchoQuilt's federation protocol publishes anonymized per-cell metrics into a multi-state pool while preserving each state's ownership of the underlying site-level data.

Seventh, integrate the dashboard with USGS ScienceBase data publishing infrastructure so that each NABat reporting cycle's submission becomes a versioned, citable scientific record. The publishing workflow takes a quilt-aggregated state report, attaches a DOI through the USGS Digital Object Identifier service, and registers the submission with NABat's central data warehouse — turning a state's annual reporting obligation into a data citation that the academic community can reuse.

Eighth, train state staff on the dashboard with a rotation that pairs senior biologists with new hires for at least two complete winter cycles. The dashboard's full value comes from interpretation of patch-level patterns, which requires field-experience pattern recognition that does not transfer through documentation alone. Two-winter mentoring rotations build the institutional capacity that lets the dashboard outlast any single biologist's tenure.

Scale Your State's Hibernacula Program Without Scaling Entries

State DNR crews and NABat partners are already stretched across Priority 1 hibernacula lists that grow every year as WNS spreads and NLEB, Indiana bat, and tri-colored bat (Perimyotis subflavus) occupancy shifts. EchoQuilt lets your state aggregate 40 or 400 hibernacula into one GRTS-anchored dashboard without adding winter entries or asking biologists to swallow survey windows they cannot make. The dashboard surfaces per-cell occupancy estimates, cluster-density trends, and stability metrics in one view, so a state coordinator preparing the annual NABat submission spends hours rather than weeks compiling the reporting deliverables. The hierarchical Bayesian model that underlies the per-cell occupancy estimates handles the data-poor cells that staffing constraints inevitably create, producing defensible occupancy estimates even at sites that the field crew did not directly visit during the reporting cycle.

Join the Waitlist for Hibernacula Biologists, and we will walk through your Priority 1 list and scope a deployment sequence that lines up with your next NABat reporting cycle. The deployment templates include cave-type-specific tuning for karst, mine, and lava tube sites, so a multi-cave-type state portfolio scales without per-site customization overhead.