Model structure and fundamentals

Glacier centric model

OGGM is what we called a “glacier centric model”, which means that it runs for each glacier independently of the others. In the case of glacier complexes, it relies on the glacier inventory to properly separate the individual glacier entities by the ice divides, ensuring that all ice in a glacier basin flows towards a single glacier terminus.

https://oggm.org/framework_talk/img/iceland.png

Glacier centric approach applied to the Eyjafjallajökull ice cap in Iceland. Glacier outlines provided by the Randolph Glacier Inventory v6.0.

The glacier centric approach is used by most large-scale glacier models to date. Alternative strategies include global gridded approaches, where all glaciers in a model grid cell are added together and possibly organized into elevation bins. Another approach would be to handle entire glacier complexes as one single body of ice (“ice caps”). With the release of RGI7 (early 2022), OGGM will also be able to consider glacier complexes as if they were single glacier entities.

Data structures and glacier directories

The fundamental data structure used in OGGM is the so-called Glacier Directory. Glacier directories are simple folders on disk which store the input and output data for a single glacier during a run. OGGM offers an interface to access and store these files programmatically.

This very simple idea is at the core of the OGGM workflow: actions to perform on glaciers (“tasks”, see below) are given access to the data files via the glacier directory interface, read data they need from disk, and write back to it. This design matches perfectly the “glacier centric” modelling strategy, and has many advantages as outlined below.

Here is an example of how glacier directories work in practice. The user indicates a repository (base_url, here leading to this online folder) from which they want to fetch the data, and a list of glacier IDs they’d like to start from. The workflow.init_glacier_directories() task performs the action of downloading and extracting these data locally:

In [1]: from oggm import workflow, graphics

In [2]: rgi_ids = ['RGI60-11.01328', 'RGI60-11.00897']

In [3]: gdirs = workflow.init_glacier_directories(rgi_ids,  # glaciers to download
   ...:                                           from_prepro_level=3,  # pre-processing level
   ...:                                           prepro_base_url=base_url)  # online repository
   ...: 

In [4]: graphics.plot_inversion(gdirs[0])
_images/plot_example_inversion.png

See also the documentation page for OGGM Shop for more examples of the kind of data that can be added to glacier directories.

OGGM tasks

Tasks in OGGM are actions to perform on one single glacier (Entity tasks) or several of them (Global tasks). Tasks have a special meaning in the OGGM workflow and are applied as such:

# Initialize glacier directories
from oggm import workflow, tasks
gdirs = workflow.init_glacier_directories(rgi_ids)

# Define the list of tasks
task_list = [
    tasks.define_glacier_region,
    tasks.glacier_masks,
    tasks.compute_centerlines,
    tasks.catchment_area,
    tasks.catchment_width_geom,
]

# Apply them sequentially
for task in task_list:
    workflow.execute_entity_task(task, gdirs)

workflow.execute_entity_task() will apply the given task to a list of glaciers. If multiprocessing is switched on, all glaciers will be processed in parallel, making full use of all available processors. Here we apply the default tasks with default settings, but parameters can be changed via global settings or function arguments.

Depending on the desired set-up, tasks can be replaced by others (e.g. the centerlines tasks can be replaced by other algorithms) or omitted (for example, users can choose whether a quality check filter should be applied to the climate timeseries or not).

See this example from the documentation about how tasks can’t be interchanged to run OGGM with different types of flowlines.

Modularity

Modularity in OGGM is achieved in two major ways:

  • object oriented programming (OOP) interfaces: OGGM relies on object interfaces to couple modelling elements between each other. A typical example are the FlowlineModel and MassBalanceModel classes which are use to couple the two models without enforcing any particular way to solve the glacier geometry evolution or the climatic mass-balance.

  • persistence on disk with Glacier Directories: as explained in the previous chapter, tasks can be interchanged during a processing workflow, allowing modularity as long as all tasks agree on the format of the file written on disk.

The object interfaces in particular are fundamental for anyone wanting to extend OGGM with their own parameterization. The OGGM-contrib repository illustrates how anyone can add a parameterization to OGGM without changing the OGGM codebase.

Preprocessing

If you are looking for only ONE reason to use OGGM and create your own model on top of it, the OGGM pre-processing capabilities are probably it.

OGGM is designed for large-scale applications, i.e. we have pre-downloaded and pre-processed a large number of datasets that are ready to use from within the OGGM framework.

Visit OGGM Shop to learn all you need to know about the data we have prepared for you!