Methodology
From observation to intelligence: a six-level analytical hierarchy.
Planetary Monitoring applies a structured analytical hierarchy to every dataset
it ingests. Each level must be satisfied before the next is computed. Where
evidence is insufficient the analysis stops, and the limitation is communicated
explicitly. The platform identifies observations, correlations, anomalies and
candidate explanatory drivers; it does not represent correlation as demonstrated
causation.
The Six Analytical Levels
Level 1: Observation
What does the underlying instrument or dataset directly report? Source,
timestamp and provenance must be traceable. Examples: InSAR-derived
displacement velocity (Copernicus EGMS), GNSS station velocity (SONEL /
EarthScope), magnetometer field variation (INTERMAGNET), solar wind Bz
component (NOAA DSCOVR), earthquake magnitude and hypocenter (USGS),
terrestrial water storage proxy via ERA5/Open-Meteo reanalysis, active
fire radiative power (NASA FIRMS).
Level 2: Quality Assessment
Reliability and completeness of each observation are evaluated before it
enters any analytical process. Factors assessed: source provenance and
agency authority; data timestamp and staleness relative to the expected
update cadence; spatial and temporal resolution; coverage completeness;
known instrument or processing limitations; missing-observation flags. A
low-quality observation propagates uncertainty into every downstream inference.
Level 3: Anomaly Detection
Whether an observation differs meaningfully from an established baseline.
Baselines account for expected seasonal behaviour, measurement uncertainty,
local spatial variance and data gaps. A single measurement is not treated as
an anomaly without appropriate temporal context. Representative baselines:
displacement velocity versus the multi-year EGMS average; Kp versus the
27-day solar-rotation median; terrestrial water storage versus the 5-year
ERA5 climatological z-score; geomagnetic ΔB versus the quiet-day variation
envelope.
Level 4: Cross-Domain Context
Where an anomaly is identified, the platform examines other available
observations across an appropriate spatial and temporal window. Earth-system
processes rarely operate in isolation. Temporal coincidence and spatial
coherence between independent datasets can inform, but do not determine,
interpretation. Illustrative checks: hydrological loading anomaly versus
vertical ground-motion trend; geomagnetic storm onset versus GNSS positioning
noise elevation; seasonal water-storage departure versus local subsidence rate.
Level 5: Candidate Driver Assessment
Candidate explanatory mechanisms are identified based on available evidence.
A mechanism is proposed only where independent supporting signals exist, and
is labelled as a candidate rather than a confirmed cause. Where no mechanism
can be supported, the absence of a driver conclusion is communicated.
Recognised candidate drivers: groundwater withdrawal (USGS NWIS / ERA5 TWS proxy);
hydrological loading (precipitation and water-storage alignment); volcanic
deformation (spatial coherence with USGS alert levels); tectonic processes
(fault proximity and seismic context); seasonal soil behaviour (annual-cycle
correlation); anthropogenic activity (land-use and industrial context).
Level 6: Confidence Qualification
Confidence communicates the evidential support behind an assessment; it is not a
probability of absolute truth. Confidence reflects: the number and independence
of supporting datasets; spatial coherence of multiple signals; temporal
alignment of corroborating observations; source data quality and freshness;
the presence or absence of contradictory evidence; and identified evidence gaps.
Scientific Language: Term Definitions
| Term | Definition |
|---|
| Observed |
Directly reported from an identified measurement system. Source instrument, agency and timestamp must be traceable. |
| Derived |
Calculated from source observations using a defined transformation or methodology. The derivation method and its assumptions should be documented. |
| Modelled |
Output generated by a numerical or statistical model rather than direct measurement. Model provenance, assumptions and validation status should be noted. |
| Correlated |
Two or more variables exhibiting a measurable statistical relationship. Correlation does not establish causation. |
| Associated |
Observed correspondence between signals without demonstrated causation. Used when spatial or temporal coincidence is noted but the mechanism is unclear. |
| Inferred |
A conclusion drawn from available evidence but not directly measured. Presented with supporting evidence and known limitations. |
| Projected |
Forward extrapolation based on explicit assumptions and historical trends. Not a prediction unless a validated predictive model exists. |
| Candidate Driver |
A mechanism consistent with available evidence. One or more independent datasets support this interpretation, but it is not confirmed as the primary cause. |
| Confidence |
Evidential support for an assessment: not a probability of absolute truth unless defined mathematically by a specific model. |
What we avoid
- Unsupported causal claims
- Deterministic hazard predictions without validated models
- Presenting model outputs as direct observations
- Treating correlation as confirmed causation
- Suppressing uncertainty or evidence gaps
Data Sources ·
About ·
Open Live Monitor
This application requires JavaScript for the interactive monitoring environment.