Abbreviations and acronyms

This page collects the most common abbreviations used across the GeoPrior documentation, examples, API reference, and manuscript-style materials.

Note

This page focuses on abbreviations and acronyms only.

Mathematical notation and physics symbols such as \(K\), \(S_s\), \(H_d\), \(\tau\), \(\varepsilon_{cons}\), and \(\gamma_w\) should be documented separately in Symbols and notation.

How to use this page

• Use this page when you meet a short form such as PINN, InSAR, or PIT in the docs or examples.

• Use Symbols and notation later for equations, physical fields, and diagnostic symbols.

• Use Key terms and concepts later for longer conceptual explanations such as calibration, sharpness, identifiability, or closure consistency.

Core abbreviations

Abbreviation	Full term	Notes
BS	Brier score	Probabilistic forecast metric used for binary or threshold exceedance evaluation.
DTW	Dynamic time window	Refers to the moving or staged temporal input structure used in forecasting workflows.
GeoPriorSubsNet	Geomechanical Prior-Informed Subsidence Network	Main physics-guided forecasting model introduced in GeoPrior.
GRN	Gated residual network	Neural-network building block commonly used inside attentive forecasting architectures.
HALNet	Hybrid Attentive LSTM Network	Backbone family or architectural reference used in the model lineage.
InSAR	Interferometric Synthetic Aperture Radar	Remote-sensing technique used to derive land-subsidence observations.
LSTM	Long short-term memory	Recurrent neural-network unit used for sequence modelling.
MAE	Mean absolute error	Standard deterministic forecast error metric.
MME	Multi-modal encoder	Encoder block used when combining multiple feature sources or modalities.
MSE	Mean squared error	Standard deterministic forecast error metric that penalizes large deviations more strongly than MAE.
MSLSTM	Multi-scale long short-term memory network	Multi-scale LSTM-style architecture used as a baseline or comparison family.
PC1	First principal component	First latent axis from principal-component analysis.
PI	Prediction interval	Interval forecast such as an 80% prediction interval used for uncertainty evaluation.
PIT	Probability integral transform	Calibration diagnostic used to assess whether predictive distributions are statistically consistent with observations.
PINN	Physics-informed neural network	Neural-network framework constrained by physical equations or residuals.
PRD	Pearl River Delta	Regional setting containing the study areas used in the manuscript and examples.
QC	Quality control	Data-screening and validation procedures applied before model training or evaluation.
SBAS	Small Baseline Subset	InSAR time-series processing strategy used to recover ground deformation.
SI	International System of Units	Conventional metric unit system used for physical quantities throughout the documentation.
VSN	Variable selection network	Feature-selection block used in attention-based forecasting models.

Reading hints

A few abbreviations appear especially often across the project:

• GeoPriorSubsNet is the main forecasting model.

• PINN describes the broader modelling style.

• InSAR refers to the observation source for deformation.

• PIT, BS, and PI are central to uncertainty and probabilistic evaluation.

• LSTM, GRN, and VSN describe important neural-network components.

See also

See also

Symbols and notation

Mathematical notation, physical fields, and diagnostic symbols.

Key terms and concepts

Plain-language explanations of key modelling and forecasting concepts.