Thermistor and Humidity Data Alongside Sound-Derived Maps
When Point Data Meets Acoustic Shape
A three-winter study in a Kentucky tri-colored bat hibernaculum placed 18 HOBO thermistor loggers along a 60 m main passage and recorded temperature every 30 minutes from October through April. The dataset was clean, dense, and frustrating. It showed that the cluster in the back chamber sat in a stable 5.2°C pocket while the main passage ranged 3.8 to 4.6°C, but it could not explain why the pocket was stable, why it formed in that location, or whether it would persist in warmer winters. Temperature data alone is a list of numbers; biologists needed shape and airflow to turn the numbers into a mechanism.
The interpretation problem is general. Tricolored bat microsite use throughout hibernation shows that continuous T and RH logging reveals no single optimum — bats distribute across a range. The misuse of relative humidity in ecological studies of hibernating bats warns that RH alone misleads and VPD is the correct metric for evaporative stress. Interspecific variation in evaporative water loss among hibernating bats puts WNS-vulnerable species in the high-EWL group, making VPD particularly relevant for Myotis lucifugus, Myotis sodalis, and Perimyotis. Yet most field datasets are stuck at point-T and point-RH without the spatial structure that explains how bats actually select microsites.
The point-data trap has another cost: logger placement itself becomes a persistent source of bias. A logger on a wall captures wall surface temperature, not the air a bat is breathing 30 cm away from the wall in a cluster. A logger at floor level captures the cold sink, not the warmer ceiling pocket where the bats actually roost. A logger near an entrance captures the chimney-effect throat air, not the back-chamber stagnation. Years of multi-site logger data carry these spatial biases unevenly across sites, which makes cross-site comparison fraught even before WNS-driven mortality enters the picture. Biologists know this — many published studies include calibration appendices documenting logger placement protocols — but the underlying issue is that a single logger reports a single point, and the cluster needs a continuous field.
EchoQuilt's contribution is exactly that: turning sparse logger data plus dense passive acoustics into a continuous microclimate field that can be queried at the cluster voxel rather than at the logger location.
Fusing HOBO Data Into the Quilt
EchoQuilt treats thermistor and humidity loggers as additional sources of information about the same voxel grid its acoustic array already covers. Standard HOBO loggers for microclimate monitoring and similar devices mount at known positions and log point values across the winter. The quilt ingests each logger's position and time series as a boundary condition for the acoustic-inferred temperature and airflow field. Because air temperature inside a hibernaculum is tightly coupled to airflow patterns and surface exchange, the sparse logger data plus the dense acoustic airflow data resolve into a continuous temperature quilt at the same 30 cm voxel resolution as the geometry layer.
The stitching patches two kinds of data. Logger points anchor absolute temperature and RH values. Acoustic airflow and drip-rate data provide the spatial interpolation structure. A single patch in the quilt might report "5.1°C, 92% RH, 0.2 m/s northwest airflow, stable for 142 hours" — all co-registered with the geometry, occupancy, and species layers from the same array. For a biologist analyzing cluster location, the combined layer answers questions that neither data source can answer alone: is this cluster sitting in a stable 5°C plume driven by a specific entrance airflow path, or is it in a transient pocket that will collapse when surface weather shifts?
VPD follows directly. Because EchoQuilt already has T and RH at each voxel, calculating VPD is arithmetic, and the VPD quilt becomes the biologically meaningful view rather than the raw RH. For WNS work, this is essential: Pd optimal growth occurs at 5-10°C, and the intersection of that temperature range with high-VPD zones predicts where Pd colonization is most aggressive. The quilt shows those intersections visually, which is a different conversation than reading a spreadsheet of logger values. This pairs with microclimate cluster data analysis for behavioral response across the winter.
Advanced Tactics for Microclimate Fusion
Three tactics sharpen microclimate fusion. First, optimize logger placement using the acoustic airflow quilt from year one. Most hibernacula teams place loggers on a regular grid or by intuition. The year-one airflow quilt identifies where microclimate gradients are steepest, and year-two logger placement should cluster there. Five loggers positioned at year-two hotspots produce better continuous-field estimates than 20 loggers on an evenly-spaced grid, because the acoustic data handles the smooth regions and the loggers pin down the steep ones. The placement strategy also reduces logger purchase and maintenance costs across the long-term monitoring program, which matters for state DNR programs working under flat or shrinking budgets.
Second, use logger-acoustic residuals as a quality-control signal. If the acoustic-inferred temperature at a logger's voxel diverges from the logger reading by more than 0.5°C for more than a day, something has drifted — either the logger has moved, the sensor calibration has shifted, or an unmodeled heat source (a fresh breakdown, an entrance collapse, human presence) has appeared. The residual pattern usually distinguishes these failure modes. A slow drift over weeks usually points to logger calibration; an abrupt step change usually points to physical displacement or a disturbance event; a periodic oscillation usually points to an unmodeled airflow cell that the next year's placement should account for. Cave-diving teams use an analogous approach for quality-control data at decompression stops, where sparse sensor data gets fused with dense acoustic structure to catch instrument anomalies before they corrupt downstream analysis.
Third, compute VPD on the voxel grid rather than on the logger points. VPD is the variable that drives evaporative water loss for torpid bats, and knowing VPD at the cluster voxel is more useful than knowing T and RH at a logger 4 m away. The quilt makes this a one-line calculation instead of an interpolation exercise. Per-cluster VPD time series across a winter become an empirical validation dataset for the vapor pressure deficit models that WNS researchers have used in lab studies, and the in-cave data often deviates meaningfully from the lab-derived predictions in ways that matter for overwinter mortality estimates.
Fourth, use the fused field to test cluster-site selection hypotheses. The standard question — why does a cluster form at this specific ceiling patch rather than five meters to the east — can be answered with a regression that puts cluster presence on the dependent side and voxel-level T, VPD, airflow magnitude, and airflow stability on the independent side. With fused data from three winters, the regression has enough statistical power to separate species-level preferences and identify which microclimate variables matter most for each species. Myotis sodalis may weight airflow stability higher than Perimyotis subflavus, for example, and the quilt exposes that kind of species-level microsite signature that point-data methods can only hint at. Cross-link the regression results with data integration workflows that bring in banding records, and the species-microsite patterns can be tested against individual-level survival outcomes from multi-decade band recapture data.
Get Early Access to EchoQuilt
If you already run HOBO thermistor and humidity networks in your hibernacula and want an acoustic mapping layer that fuses cleanly with your existing logger data, EchoQuilt's sensor-fusion pipeline was designed specifically for your workflow. State DNR bat crews and USGS hibernacula surveyors with multi-year logger datasets are strong candidates for the pilot program because the validation loop is tight: year-one fusion can be checked against prior years' logger-only records, and the residual analysis identifies any logger placements that should be revisited. Sites with documented WNS progression are particularly valuable for the pilot because the microclimate-disease relationship is exactly the kind of question the fused dataset can answer with new precision. Join the Waitlist for Hibernacula Biologists to coordinate logger co-registration, regression validation, and VPD-based analysis for the 2026-27 hibernation season.