5-gram overlap
A measure of shared five-word sequences between public text and sources, used as an anti-reuse gate.
Where it is used
Ablation
A test that removes a source or variable to measure its contribution.
Where it is used
Abstention
An explicit choice not to issue an estimate when data quality, OOD or uncertainty exceeds a threshold.
Where it is used
Acquisition
The stage in which the system retrieves data or content from sources.
Where it is used
Activity nowcast
An estimate of current or near-term activity from available signals, distinct from eruption prediction.
Where it is used
Adapter
A component that transforms a specific source into a common format.
Where it is used
Advection
Transport of a property or ideal parcel by a velocity field; here used as a simplified wind trajectory.
Where it is used
Algorithm
A precise finite procedure that transforms inputs into a result.
Where it is used
Anomaly
A value or behaviour different from what the system considers usual.
Where it is used
Artificial intelligence (AI)
Methods enabling a system to classify, estimate, recognise or generate content.
Where it is used
AT Protocol
Decentralised protocol used by Bluesky to create records, upload blobs and manage identities.
Where it is used
Automation
Automatic execution of defined activities or calculations, including controls and exceptions where needed.
Where it is used
b-value
Parameter of the Gutenberg–Richter relationship between earthquake count and magnitude, estimated only with a sufficient sample.
Where it is used
Backfill
Controlled reconstruction of missing or previously unprocessed historical data.
Where it is used
Baseline
A simple reference used to assess the model’s added value.
Where it is used
- WF-01 · Weather automation
- WF-03 · Market Overview
- WF-07 · Etna Forecast
- WF-13 · Italy Variable Renewable Forecast
- WF-20 · The Public Service Run
- AI and data analytics for energy, hydrogen and renewables | Lab Intelligence
- AI for industrial processes: verifiable examples | Lab Intelligence
- Anomaly detection and predictive monitoring | Lab Intelligence
Brier score
Mean squared error of a probabilistic forecast; lower values indicate better overall calibration.
Where it is used
Calibration
Agreement between stated probabilities and observed frequencies over time.
Where it is used
CCS
Carbon Capture and Storage: CO₂ capture and storage under a disclosed boundary and capture rate.
Where it is used
Classification
Assignment of a case to a predefined category.
Where it is used
Clustering
A method that groups similar cases without predefined categories.
Where it is used
Comfort score
A transparent heuristic score summarising several environmental factors; it is not itself a medical or normative index.
Where it is used
Communicable finding
A short publishable fact derived from an admitted source and expressed with original controlled wording.
Conformal prediction
A method that builds intervals from empirical out-of-sample residuals under stated assumptions.
Where it is used
Data drift
A change in the input distribution relative to the development period.
Where it is used
Data lineage
Trace of sources and transformations that produced an output.
Data quality check
A check of presence, freshness, format, consistency and missing values.
DBSCAN
An algorithm that groups nearby points and leaves isolated ones apart.
Where it is used
Decision
The stage that determines whether to proceed, stop or publish an output.
Where it is used
Dew point
The temperature at which air would reach saturation; useful for describing absolute moisture load.
Where it is used
Directional shear
Difference in wind direction or speed between atmospheric levels.
Where it is used
Event cluster
A set of independent sources aligned enough to be treated as descriptions of the same signal.
Where it is used
Event-based forecast
An estimate run only after a defined source event occurs.
Event-driven
A workflow triggered or made publishable by a defined event.
Where it is used
EWM
Exponentially weighted moving average that gives more weight to recent observations and smooths an index over time.
Where it is used
Facts-only templating
A generation method in which the public text is built from controlled templates and communicable facts rather than free reuse of source prose.
Where it is used
Fail-closed
A policy in which uncertainty, missing data or failed verification blocks the action rather than allowing it.
Where it is used
Failure mode
A known way in which a system may fail, stop or produce a misleading result.
Where it is used
Feature
A variable supplied to a model.
Where it is used
FIRMS
NASA near-real-time service for thermal anomalies.
Where it is used
Forecast
An estimate relating to a future time or an event not yet observed.
Where it is used
FRP
Fire Radiative Power: a satellite measure of thermal-anomaly radiative energy, used as a proxy with observation-coverage controls.
Where it is used
Gaussian Mixture Model (GMM)
A probabilistic model representing data as a mixture of Gaussian distributions; here used for latent regimes.
Where it is used
Generative AI
Models that produce new text, images or other content from instructions and data.
Where it is used
Graceful degradation
Use of a declared fallback or reduced output when a modality is unavailable.
Where it is used
Grapheme count
The count of user-perceived characters, used to keep a Bluesky post within length limits.
Where it is used
Hash-guard
A check that recognises identical or near-identical content to prevent duplicates.
Haversine
A formula for great-circle distance between two geographic coordinates on a sphere.
Where it is used
Hawkes process
A self-exciting temporal model: an event temporarily raises the probability of subsequent events; used here as an event-history baseline.
Where it is used
Human oversight
Human control and responsibility where interpretation or decision is required.
Where it is used
Idempotency
Property by which a retry does not create duplicate effects.
Where it is used
Independence group
A label that prevents the final post from using two sources considered too close or dependent.
Lagrangian advection
A calculation that updates a point position over time by following a wind field.
Where it is used
LCOH
Levelized Cost of Hydrogen: modelled unit hydrogen cost over a stated economic life and boundary.
Where it is used
Leakage
Unintended use of information that would not be available at forecast time.
Where it is used
Log
A chronological record of system events, checks and errors.
Where it is used
LP / MILP
Linear and mixed-integer linear programming for optimising decisions subject to constraints.
Where it is used
Machine learning (ML)
Methods that learn relationships from examples or previous data.
Where it is used
MAD
Median Absolute Deviation: the median absolute distance from the median, a robust dispersion measure.
Where it is used
MLOps
Practices for versioning, testing, deploying and monitoring models and pipelines.
Where it is used
MLP
A feed-forward neural network used to classify input configurations.
Where it is used
Modality gate
A selection rule that compares modalities and baselines across folds and chooses the best-calibrated probability source for a horizon.
Where it is used
Model
A mathematical or logical relationship used to produce an estimate or classification.
Where it is used
Neural network
A layered model of numerical transformations used for complex relationships.
Where it is used
Offline test
A check performed with controlled data without relying on a live external source.
OOD (Out-of-distribution)
A condition in which a new case lies outside the distribution represented by the data used to build or calibrate a model.
Where it is used
PDS / AppView
The PDS is the server that stores the AT Protocol record; AppView is the public view that exposes it in Bluesky interfaces.
Persistence
Storage of results, metadata and execution state.
Where it is used
Point-in-time
A dataset using only information actually available at the simulated date.
Where it is used
Prequential selection
Model choice at each step using only evaluation outcomes already observed in the past.
Where it is used
Provenance
Metadata on sources, timestamps, version, configuration and artifacts.
Where it is used
Purged walk-forward
A temporal evaluation that advances through time while removing observations near boundaries to avoid train-test contamination.
Where it is used
Quality gate
An automated check that allows, blocks or skips an output.
Where it is used
Quality-aware fusion
Combination of multiple modalities in which each signal weight also depends on its quality and availability.
Where it is used
Regression
A method that estimates a numerical value from available variables.
Where it is used
Release lag
Delay between the period a datum refers to and the time it actually becomes available.
Where it is used
Rich-text facet
AT Protocol metadata that makes a textual reference clickable inside the post, for example a source link.
Ridge regression
Linear regression with an L2 penalty, useful for stabilising coefficients when features are correlated.
Where it is used
Robust rescaling
A transformation that limits outlier influence, often using quantiles, median or interquartile range.
Robust z-score
Standardisation based on robust statistics such as the median and MAD, reducing sensitivity to outliers.
Where it is used
RPA
Automation of repetitive activity coordinating sources, rules, files, reports and external systems.
Where it is used
Sensor health
Freshness, completeness and quality indicators describing the current reliability of a sensor or modality.
Where it is used
Soft sensor
A model that estimates a variable not measured continuously from other available variables.
Where it is used
Source policy registry
A domain registry containing reuse permissions, risk tier and the behaviour allowed in live publication.
Split-conformal
A conformal method using a separate calibration set to build intervals with empirical coverage.
Where it is used
State machine
Explicit states and transitions governing system behaviour from thresholds, direction and confirming signals.
Where it is used
Structural graph
An explicit node-and-edge network representing transmission hypotheses; it does not prove causality by itself.
Where it is used
Synthetic data
Artificial observations generated by an explicit model or scenario: synthetic data (simulated data) are not real measurements.
Where it is used
Target
The result a model seeks to estimate or classify.
Where it is used
Temporal embargo
An excluded interval between training and test that reduces leakage from overlapping episodes or features.
Where it is used
Theme bundle
A comparison of sources on the same domain without claiming they describe the same event.
Where it is used
TID / record key
A valid stable identifier used to create or update an AT Protocol record.
Time series
A sequence of observations ordered in time.
Where it is used
Trigger
An event or schedule that starts a workflow.
Where it is used
Validation
The stage in which the system checks that data and conditions can be used.
Where it is used
VONA
Volcano Observatory Notice for Aviation: an official volcano-observatory communication for aviation context.
Where it is used
Walk-forward
Temporal evaluation in which training and test windows move forward without mixing future data.
Where it is used
Workflow
A sequence of activities that software performs in a defined order to transform data into a result.
Where it is used
Z-score
Distance from the mean measured in standard deviations.
Where it is used