GSEIO.GLOBAL SEISMIC EARTH INFORMATION OBSERVATORY
OPEN EARTHQUAKE INFORMATION
LAMBDA-FIS RESEARCH PLATFORM

Scientific Models and Forecasting Methods

Advancing Physical Earthquake Forecasts and Predictions

LAMBDA-FIS – Lambda Forecasting Information System is an advanced scientific research system developed since 2025 with the objective of advancing physical earthquake forecasting and prediction.

The system is based on a holistic approach to earthquake science, integrating global seismic observations, tectonic structures, physical models of stress and shear, statistical earthquake forecasting, and objective forecast evaluation into a unified research framework. Rather than treating earthquakes solely as statistical occurrences, LAMBDA-FIS investigates the physical processes governing seismic activity, including tectonic stress, fault interactions, shear forces, and the spatial and temporal organization of seismicity. The system is designed for experimental short-range earthquake forecasting, with research forecast horizons typically extending 7 to 14 days ahead, depending on the model and its configuration.

LAMBDA-FIS consists of four interconnected modules. The complete central research platform is substantially broader than the selected model layers available in the public web viewer.

MODULE 01 · OBSERVATIONS

Data Assimilation Module

Integrating Global Earthquake and Tectonic Information

The Data Assimilation Module forms the observational and structural foundation of LAMBDA-FIS. It integrates publicly accessible seismological and geological datasets into a common analytical environment, providing the information required for physical and statistical modelling.

Earthquake catalogues

USGS – United States Geological Survey: Global earthquake catalogues containing event locations, origin times, magnitudes, focal depths, and associated seismic information. USGS Earthquake Catalog ↗

EMSC – European-Mediterranean Seismological Centre: Global and regional earthquake information, including observed seismicity, reported magnitudes, locations, and earthquake-source information. EMSC SeismicPortal ↗

Global CMT – Global Centroid Moment Tensor Project: Earthquake moment tensors and focal mechanisms providing information about seismic source orientations and the deformation associated with earthquakes. These data are particularly important for reconstructing tectonic stress fields. Global CMT Catalogue ↗

Tectonic and geological databases

GEM – Global Active Faults Database: Global information on mapped active faults, fault geometry, kinematics, slip rates, and structural characteristics. GEM Active Faults ↗ (Styron & Pagani, 2020).

HIKE – European Fault Database: Harmonized European geological fault geometries and associated structural attributes. HIKE European Fault Database ↗

USGS Plate Boundaries: Global tectonic plate and microplate boundaries used to identify major zones of plate interaction. USGS Plate Boundaries ↗ (Bird, 2003).

Slab2: Three-dimensional geometries of major subduction zones, including slab depth, orientation, and geometry. USGS Slab2 ↗ (Hayes et al., 2018).

Earthquake effects and population exposure

LAMBDA-FIS also incorporates available earthquake shaking and population datasets, including USGS ShakeMap ↗, European shaking products, and the Global Human Settlement Layer ↗.

These datasets support the interpretation of significant earthquakes and estimates of population exposure to different levels of shaking. The Data Assimilation Module preserves the sources, observation periods, model parameters, and provenance of scientific inputs to support reproducibility.

MODULE 02 · SEISMICITY

Statistical Forecasts Module

Establishing Statistical Earthquake Forecasts and Reference Models

Statistical earthquake models provide estimates of seismicity patterns and background earthquake activity. They are essential both for statistical forecasting and for establishing reference models against which the performance of physical forecasts can be evaluated.

STATISTICAL MODEL 01

ETAS – Epidemic-Type Aftershock Sequence Model

ETAS describes earthquake clustering through a statistical triggering process in which earthquakes may generate subsequent events. It is particularly useful for estimating the expected spatial and temporal evolution of aftershock activity following significant earthquakes and for developing conditional probabilistic earthquake forecasts (Ogata, 1988). LAMBDA-FIS includes ETAS calibration and forecast projections, including aftershock-decay and full triggering formulations.

Figure 1. Example of an ETAS-based time-series display showing observed earthquake magnitudes and modelled expected event rates at different distances.
Figure 2. An example of the Relative Intensity / seismicity field displayed with mapped faults and observed earthquakes.
STATISTICAL MODEL 02

RI – Relative Intensity

RI characterizes the spatial distribution of earthquake activity using observed earthquake counts and configurable interaction distances. It provides a reference representation of the geographical distribution of seismicity. LAMBDA-FIS also includes logarithmic and normalized variants, such as LogRI and LogNRI. These fields support comparisons between observed seismicity, statistical reference patterns, and physical forecasting models.

STATISTICAL MODEL 03

SSM – Smoothed Seismicity Model

SSM estimates the spatial distribution of background seismicity by smoothing earthquake activity over geographical space. LAMBDA-FIS incorporates earthquake declustering to distinguish background activity from clustered sequences, including foreshocks and aftershocks, using the nearest-neighbour methodology of Zaliapin and Ben-Zion (2013). SSM is an important reference model for evaluating whether physical forecasts contain information beyond the historical distribution of seismicity.

STATISTICAL MODEL 04

Ps – Seismic Potential

Ps is a normalized index that combines earthquake frequency and the maximum observed earthquake magnitude within spatial cells and their interaction regions. It provides an additional measure of the relative seismic potential of different regions and can be used as a statistical reference when evaluating physical models.

Additional statistical models

The central system also includes normalized logarithmic seismicity fields, time-dependent components, and Renewal models that investigate local seismic-cycle evolution using elapsed time and earthquake-rate information. Statistical indices such as RI, SSM and Ps are not, by themselves, calibrated probabilities of a future earthquake. Their predictive significance must be established by independent testing.

MODULE 03 · GEOPHYSICS

Physical Forecasts Module

Modelling Tectonic Stress, Shear Forces and Seismic Interactions

The Physical Forecasts Module represents the central scientific research component of LAMBDA-FIS. It investigates the mechanical state of the Earth's crust using earthquake focal mechanisms, tectonic fault geometries, stress tensors, shear interactions, and wave-based representations of seismic organization. The module includes six principal modelling approaches.

Figure 3. Tectonic stress orientations illustrated with SHmax direction tracers alongside tectonic boundaries and focal mechanisms.
PHYSICAL MODEL 01

Tectonic Stress

The Tectonic Stress model reconstructs a normalized regional tectonic stress tensor using earthquake moment-tensor information from the Global CMT catalogue. It provides estimates of stress orientation, including the direction of maximum horizontal compression (SHmax). The resulting fields help visualize the evolving spatial organization of tectonic stresses.

PHYSICAL MODEL 02

Normalized Shear Stress

This model evaluates normalized shear stress on candidate fault-plane orientations derived from earthquake focal mechanisms. It combines the reduced tectonic stress tensor with nodal-plane geometry and a configurable spatial interaction radius. The output is a dimensionless index describing the spatial distribution of modelled shear stress, rather than absolute stress measured in pascals.

Figure 4. Normalized shear-stress field together with tectonic structures and seismicity, shown here in the western Pacific.
Figure 5. Example of Plate Shear Stress and mapped tectonic structures across Mexico, Central America and the Caribbean.
PHYSICAL MODEL 03

Plate Shear Stress

Plate Shear Stress evaluates shear interactions associated with major tectonic structures, including subduction interfaces, plate boundaries, and microplate structures. The model combines reconstructed stress tensors with structural geometries to identify regions of elevated normalized shear stress.

PHYSICAL MODEL 04

Tectonic Instability Shear Stress

The Fault Instability model evaluates normalized shear stress on mapped geological fault planes. It combines tectonic stress tensors with fault geometries from databases such as GEM and HIKE, accounting for available fault orientations and dip information. The model investigates the relative mechanical loading of mapped faults. It does not directly calculate fault-failure probabilities or absolute failure thresholds.

Figure 6. Example of the Fault Instability model across Australia, Indonesia and New Zealand.
Figure 7. Weighted wave structures and a modelled forecasting field in the Mediterranean and neighbouring regions.
PHYSICAL MODEL 05

Waves

The Waves model investigates geometrical and spatial relationships among earthquakes using weighted spherical wave-like structures derived from seismic activity. These structures represent a modelling framework for exploring the spatial organization and interactions of seismic systems. The wave patterns are model-derived geometric structures, not direct observations of propagating seismic waves.

PHYSICAL MODEL 06

Shear Traction

Shear Traction combines weighted wave structures with another selected forecasting field to construct a normalized physical forecast index based on tectonic torques/momenta. This approach enables the investigation of spatial relationships between seismic organization, tectonic loading, and the conditions associated with earthquake occurrence. The resulting index is dimensionless and should not be interpreted as directly measured shear traction in pascals.

Figure 8. Shear traction and weighted wave-line structures displayed with mapped tectonics and earthquakes.
Figure 9. Illustrative time-series view of model field values at different spatial distances, with observed earthquake magnitudes.
TEMPORAL ANALYSIS

Forecast fields through time

Alongside geographic model maps, the research interface provides time-series views of model fields and associated seismicity.

Short-Range Forecasting

The Physical Forecasts Module supports the production and investigation of experimental short-range forecasting fields, including configurations with forecast horizons of approximately 7–14 days. These forecasts are intended to investigate whether physically based models provide predictive information beyond established statistical reference models. Forecast validity, uncertainty, and forecasting skill are evaluated separately. The model fields should not be interpreted as deterministic predictions of the exact time, location, or magnitude of future earthquakes.

MODULE 04 · VERIFICATION

Statistical Testing and Forecast Evaluation

Measuring the Predictive Performance of Earthquake Models

A central objective of LAMBDA-FIS is not only to generate earthquake forecasts, but also to determine whether those forecasts contain statistically significant information about future seismic activity. For this purpose, the system incorporates two complementary methods.

TEST 01

Molchan Test

The Molchan test evaluates the relationship between successfully identified earthquake occurrences, missed events, and the spatial-temporal extent of forecast alarm regions. By varying forecast thresholds, the method determines how forecasting performance changes as the area classified as elevated seismic potential increases. LAMBDA-FIS uses reference seismicity distributions to evaluate the relative performance of forecast fields (Molchan, 1990).

TEST 02

Monte Carlo Simulations – Pi-test

The Pi-test compares forecast values at observed earthquake locations with values obtained from simulated earthquake catalogues. LAMBDA-FIS generates synthetic earthquake locations based on a statistical background model while preserving selected observed properties of the test catalogue. The resulting Monte Carlo distributions provide reference scores, standardized differences, and empirical significance estimates. This approach makes it possible to investigate whether the observed association between a physical forecast and subsequent earthquakes is stronger than expected under the selected statistical null model.

Scientific Forecast Verification

Together, the Molchan test and Pi-test provide complementary methods for evaluating the predictive performance of statistical and physical forecasting models. LAMBDA-FIS supports archived forecast fields, documented input catalogues, reproducible model configurations, and retrospective statistical evaluation. Prospective testing using forecasts issued before the target earthquakes is particularly important for establishing genuine predictive skill. Further details will be available on the dedicated Statistics page of GSEIO.

FROM RESEARCH TO PUBLIC ACCESS

The Complete LAMBDA-FIS System

The public LAMBDA-FIS Web Viewer provides access to a selected subset of the scientific models developed within the central research system. The complete LAMBDA-FIS platform includes substantially broader capabilities: global catalogue management, statistical forecasting, tectonic stress reconstruction, physical modelling, experimental forecast generation, and statistical forecast evaluation.

GSEIO is intended to make selected scientific results publicly accessible while maintaining a clear distinction between observed earthquake information, experimental forecasts, and independently verified forecasting performance.

Explore the LAMBDA-FIS Web Viewer
LAMBDA-FIS
Scientific illustration
Open original PNG ↗Press Esc or click outside to close