An unanswered question in environmental science. A dataset that doesn't exist yet.
The Question
Why do specific locations generate consistent reports of profound perceptual anomaly across independent observers separated by centuries? The candidate mechanisms are physical. The dataset doesn't exist.
The Approach
Rigorous, continuous, multi-modal environmental measurement. 12 sensor modalities. Every dataset published openly under CC BY 4.0. Every finding independently verifiable.
The Research Question
Unknown. Not Inexplicable.
For centuries, across cultures and demographics, specific places have generated consistent reports of profound and disorienting experience from people with no incentive to speak and every incentive to stay silent. These locations are not inexplicable. They are unknown. Rigorous, continuous, multi-modal environmental measurement has never been applied to them. That is what Project Sentinel does.
Those accounts define where we deploy and measure. The environment determines the data.
Infrasound
Sub-20 Hz pressure oscillations, documented in peer-reviewed literature as altering cortical connectivity. A candidate mechanism under active research.
Electromagnetic Fields
Localised ELF variation repeatedly observed at anomaly-report locations across independent field studies. Correlation documented - mechanism under investigation.
Thermal Environment
Localised thermal gradients, draught pathways, and pressure differentials - never continuously mapped spatially in historic structures.
Architectural Acoustics
Acoustic resonance in historic structures shown to produce measurable neurological effects. Spatial mapping across a full building has never been done.
The primary research question has a stated null hypothesis: that no environmental variable co-varies with anomaly-report locations beyond chance. Sentinel is designed to test it. All fusion weights, scoring thresholds, and baseline datasets will be published alongside raw sensor streams so researchers can apply their own analytical criteria.
Project Status
We're building.
The first dataset isn't published. When it is, you'll hear about it here first, along with the scientific rationale document and full sensor specification.
No outcome claims are made until the data exists.
Methodology
Multi-Modal Anomaly Detection
The core scientific contribution is not any single sensor - it is continuous, spatially-tagged, multi-modal measurement applied longitudinally. A single modality cannot distinguish signal from noise. Sentinel uses orthogonal modalities to eliminate false positives. No single sensor determines a finding.
Sensor Fusion
12 modalities independently scored. Scores combined via weighted fusion against per-venue baselines. Cross-modal confirmation required at the same spatial coordinate and timestamp before an event is logged.
Thermal spike + zero vibration = HVAC cycling, down-weighted to log-only.
Daily Recalibration
Every patrol ends at dock. Each mobile Scout sensor cross-referenced against the corresponding fixed Station sensor under identical conditions. Inter-sensor offsets recorded as calibration metadata and published with every dataset.
No hidden corrections. Baseline data open for independent validation.
Spatial Tagging
SLAM-based autonomous navigation. Every reading tagged to a spatial coordinate within the building map. Longitudinal comparison of readings at the same location across multiple patrol cycles.
Enables detection of slow-developing changes invisible to spot sampling.
Immutable Archive
Research archive on write-once storage with SHA-256 integrity checking. No write pathway from public engagement to scientific data. Every anomaly score logged before engagement opens.
Architectural separation, not access controls.
Open Science
Scientific Integrity by Design
Reproducibility requires transparency at every level. Every hardware choice is replicable. Every calibration is published. Every dataset is permanently archived.
Research Archive
Write-once, checksummed All raw sensor streams (NTP-synced), anomaly events, fusion scores, per-modality breakdowns, thermal frames, spatial maps, baseline datasets, calibration records.
Licence: CC BY 4.0 Integrity: SHA-256 Repository: Zenodo with assigned DOIs
One-Way Read Only
Hardware
Fully replicable Every sensor is commercially available. Any team can acquire the same hardware and produce directly comparable datasets.
Sensor count: 12 modalities Includes: Infrasound • EMF • Thermal • Acoustic • Barometric • Vibration Calibration records: Published with every dataset