Skip to content

Overview

IGM is built around a modular architecture: every physical process and every I/O operation is encapsulated in a self-contained module with its own parameters, source code, and documentation. Modules can be combined, reordered, or replaced from the configuration file — no source-code editing required.

Architecture

A simulation is assembled from three categories of modules:

  • Input modules load data (bedrock, observations, climate) into the shared model state before the time loop begins.
  • Process modules update state variables at every time step — ice flow, mass balance, thermodynamics, and more.
  • Output modules write or visualise results at regular intervals during the time loop.
inputs → [initialize] → time loop { processes → outputs } → [finalize]

The shared state object (state) carries all glacier fields as TensorFlow tensors. Any module can read or write any field; changes propagate to all subsequent modules in the pipeline. See State Variables for the complete list of fields and their dependencies.

Core modules

Maintained by the IGM development team, well-tested, and stable across releases.

Input

local

Load initial fields (ice thickness, bedrock, velocities) from local NetCDF or GeoTIFF files

 load_ncdf

Load one or more fields from a NetCDF file at an explicit path

 load_tif

Load fields from GeoTIFF raster files

 oggm_shop

Automatically download and prepare glacier data (DEM, ice thickness, RGI outline) from the OGGM database

Assimilation

data_assimilation

Invert for ice thickness and flow parameters from observed surface velocities

 field_inversion

Invert for unknown fields (e.g. ice thickness) from surface observations (e.g. surface ice velocities)

 pretraining

Pre-train an ice flow emulator on a catalogue of synthetic glaciers

 time_relaxation

Data assimilation by forward time relaxation — nudge control fields (SMB, sliding coefficient, thickness, topography) toward observations within an inner forward loop

Process

time

Advance the simulation clock with an adaptive CFL time step

Time climate

Unified climate module — dispatches to simple, oggm, glacialindex, or station based on the `method` parameter

Atmosphere smb

Unified surface mass balance module — dispatches to simple, oggm, or accpdd based on the `method` parameter

Atmosphere arrhenius

Compute the vertically-averaged Arrhenius rate factor from ice temperature and water content

Cryosphere enthalpy

Solve the 3D enthalpy equation for temperature, water content, and basal melt

Cryosphere iceflow

Compute 3D ice velocities — direct solver or neural-network emulator

Cryosphere thk

Evolve ice thickness and surface elevation via mass conservation

Cryosphere subglacial_hydrology

Compute basal effective pressure and (optionally) evolve the subglacial till water layer

Lithosphere particles

Advect Lagrangian particles to trace flow paths and estimate ice age

Misc

Output

local

Write selected state fields to a NetCDF file in the run output folder

 write_ncdf

Write fields to a NetCDF file at an arbitrary path

 write_vtp

Write fields to VTK PolyData format for 3D visualisation in ParaView

 write_tif

Write fields to GeoTIFF rasters (georeferenced)

 write_ts

Write scalar time series (volume, area, mass balance, …) to a CSV file

 plot2d

Save 2D map snapshots as PNG images at each output time step

 live_dashboard

Display an interactive real-time dashboard of key fields during the simulation

Community modules

Contributed by the broader research community. These modules extend IGM with specialised or experimental physics and may evolve more rapidly than core modules.

Process

avalanche

Redistribute snow from steep slopes to prevent unrealistic high-elevation accumulation

Cryosphere gflex

Compute isostatic bedrock adjustment to ice load via thin-plate flexure

Lithosphere glerosion

Estimate glacial erosion rates and update bedrock topography from basal sliding velocities

Lithosphere rockflow

Simulate the transport of supraglacial and englacial debris coupled to ice flow

Lithosphere read_output

Replay a saved NetCDF output file as if freshly computed

Misc texture

Track a passive scalar (e.g. sediment, tracer) transported through the ice

Misc