Display glacier area and thickness changes on a grid#

from oggm import cfg
from oggm import tasks, utils, workflow, graphics, DEFAULT_BASE_URL
import salem
import xarray as xr
import numpy as np
import glob
import os
import matplotlib.pyplot as plt

This notebook proposes a method for redistributing glacier ice that has been simulated along the flowline after a glacier retreat simulation. Extrapolating the glacier ice onto a map involves certain assumptions and trade-offs. Depending on the purpose, different choices may be preferred. For example, higher resolution may be desired for visualization compared to using the output for a hydrological model. It is possible to add different options to the final function to allow users to select the option that best suits their needs.

This notebook demonstrates the redistribution process using a single glacier. Its purpose is to initiate further discussion before incorporating it into the main OGGM code base (currently, it is in the sandbox).

Pick a glacier#

# Initialize OGGM and set up the default run parameters
cfg.initialize(logging_level='WARNING')

# Local working directory (where OGGM will write its output)
# WORKING_DIR = utils.gettempdir('OGGM_distr4')
cfg.PATHS['working_dir'] = utils.get_temp_dir('OGGM_distributed', reset=True)
2026-03-05 20:38:26: oggm.cfg: Reading default parameters from the OGGM `params.cfg` configuration file.
2026-03-05 20:38:26: oggm.cfg: Multiprocessing switched OFF according to the parameter file.
2026-03-05 20:38:26: oggm.cfg: Multiprocessing: using all available processors (N=4)
rgi_ids = ['RGI60-11.01450',  # Aletsch
           'RGI60-11.01478']  # Fieschergletscher
gdirs = workflow.init_glacier_directories(rgi_ids, prepro_base_url=DEFAULT_BASE_URL, from_prepro_level=4, prepro_border=80)
2026-03-05 20:38:27: oggm.workflow: init_glacier_directories from prepro level 4 on 2 glaciers.
2026-03-05 20:38:27: oggm.workflow: Execute entity tasks [gdir_from_prepro] on 2 glaciers

Experiment: a random warming simulation#

Here we use a random climate, but you can use any GCM, as long as glaciers are getting smaller, not bigger!

# Do a random run with a bit of warming
workflow.execute_entity_task(
    tasks.run_random_climate,
    gdirs,
    ys=2020, ye=2100,  # Although the simulation is idealised, lets use real dates for the animation
    y0=2009, halfsize=10,  # Random climate of 1999-2019
    seed=1,  # Random number generator seed 
    temperature_bias=1.5,  # additional warming - change for other scenarios
    store_fl_diagnostics=True,  # important! This will be needed for the redistribution
    init_model_filesuffix='_spinup_historical',  # start from the spinup run
    output_filesuffix='_random_s1',  # optional - here I just want to make things explicit as to which run we are using afterwards
);
2026-03-05 20:38:27: oggm.workflow: Execute entity tasks [run_random_climate] on 2 glaciers

Redistribute: preprocessing#

The required tasks can be found in the distribute_2d module of the sandbox:

from oggm.sandbox import distribute_2d
# This is to add a new topography to the file (smoothed differently)
workflow.execute_entity_task(distribute_2d.add_smoothed_glacier_topo, gdirs)
# This is to get the bed map at the start of the simulation
workflow.execute_entity_task(tasks.distribute_thickness_per_altitude, gdirs)
# This is to prepare the glacier directory for the interpolation (needs to be done only once)
workflow.execute_entity_task(distribute_2d.assign_points_to_band, gdirs);
2026-03-05 20:38:28: oggm.workflow: Execute entity tasks [add_smoothed_glacier_topo] on 2 glaciers
2026-03-05 20:38:28: oggm.workflow: Execute entity tasks [distribute_thickness_per_altitude] on 2 glaciers
2026-03-05 20:38:29: oggm.workflow: Execute entity tasks [assign_points_to_band] on 2 glaciers

Let’s have a look at what we just did:

gdir = gdirs[0]  # here for Aletsch
#gdir = gdirs[1]  # uncomment for Fieschergletscher

with xr.open_dataset(gdir.get_filepath('gridded_data')) as ds:
    ds = ds.load()
# Inititial glacier thickness
f, ax = plt.subplots()
ds.distributed_thickness.plot(ax=ax);
ax.axis('equal');
../../_images/130e62d556deca93a554073541782cb5d1f7ef07e5af72fbb5e7e025b422eb8a.png
# Which points belongs to which band, and then within one band which are the first to melt
f, (ax1, ax2) = plt.subplots(1, 2, figsize=(12, 4))
ds.band_index.plot.contourf(ax=ax1);
ds.rank_per_band.plot(ax=ax2);
ax1.axis('equal'); ax2.axis('equal'); plt.tight_layout();
../../_images/3dc100b322f9cf9dd3a3b2fdd74c03ff7015549a4c13e7567785eb025518fa1e.png

Redistribute simulation#

The tasks above need to be run only once. The next one however should be done for each simulation:

ds = workflow.execute_entity_task(
    distribute_2d.distribute_thickness_from_simulation,
    gdirs, 
    input_filesuffix='_random_s1',  # Use the simulation we just did
    concat_input_filesuffix='_spinup_historical',  # Concatenate with the historical spinup
    output_filesuffix='',  # filesuffix added to the output filename gridded_simulation.nc, if empty input_filesuffix is used
)
2026-03-05 20:38:30: oggm.workflow: Execute entity tasks [distribute_thickness_from_simulation] on 2 glaciers

Plot#

Let’s have a look!

# # This below is only to open the file again later if needed
# with xr.open_dataset(gdir.get_filepath('gridded_simulation', filesuffix='_random_s1')) as ds:
#     ds = ds.load()
def plot_distributed_thickness(ds, title):
    f, (ax1, ax2, ax3) = plt.subplots(1, 3, figsize=(14, 4))
    ds.simulated_thickness.sel(time=2005).plot(ax=ax1, vmax=400);
    ds.simulated_thickness.sel(time=2050).plot(ax=ax2, vmax=400);
    ds.simulated_thickness.sel(time=2100).plot(ax=ax3, vmax=400);
    ax1.axis('equal'); ax2.axis('equal'); f.suptitle(title, fontsize=20);
    plt.tight_layout();

plot_distributed_thickness(ds[0], 'Aletsch')
# plot_distributed_thickness(ds[1], 'Fieschergletscher')  
../../_images/aa327818e4efac62780258f96351b0cfaa55e56446522f15af80b28061cd1cb4.png

Note: the simulation before the RGI date cannot be trusted - it is the result of the dynamical spinup. Furthermore, if the area is larger than the RGI glacier, the redistribution algorithm will put all mass in the glacier mask, which is not what we want:

def plot_area(ds, gdir, title):
    area = (ds.simulated_thickness > 0).sum(dim=['x', 'y']) * gdir.grid.dx**2 * 1e-6
    area.plot(label='Distributed area');
    plt.hlines(gdir.rgi_area_km2, gdir.rgi_date, 2100, color='C3', linestyles='--', label='RGI Area');
    plt.legend(loc='lower left'); plt.ylabel('Area [km2]'); plt.title(title, fontsize=20); plt.show();


plot_area(ds[0], gdirs[0], 'Aletsch')
# plot_area(ds[1], gdirs[1], 'Fieschergletscher')
../../_images/6e239921d2bfff72879043f99f2e002c84afff954951a0e39783f039a402e7aa.png

Volume however is conserved:

def plot_volume(ds, gdir, title):
    vol = ds.simulated_thickness.sum(dim=['x', 'y']) * gdir.grid.dx**2 * 1e-9
    vol.plot(label='Distributed volume'); plt.ylabel('Distributed volume [km3]');
    plt.title(title, fontsize=20); plt.show();


plot_volume(ds[0], gdirs[0], 'Aletsch')
# plot_volume(ds[1], gdirs[1], 'Fieschergletscher')
../../_images/8bed61fe802e3b0bee69075b252544ecc919340c29bf05e08e7df09842a93efd.png

Therefore, lets just keep all data after the RGI year only:

for i, (ds_single, gdir) in enumerate(zip(ds, gdirs)):
    ds[i] = ds_single.sel(time=slice(gdir.rgi_date, None))

Animation!#

from matplotlib import animation
from IPython.display import HTML, display

# Get a handle on the figure and the axes
fig, ax = plt.subplots()

thk = ds[0]['simulated_thickness']  # Aletsch
# thk = ds[1]['simulated_thickness']  # Fieschergletscher

# Plot the initial frame. 
cax = thk.isel(time=0).plot(ax=ax,
    add_colorbar=True,
    cmap='viridis',
    vmin=0, vmax=thk.max(),
    cbar_kwargs={
        'extend':'neither'
    }
)
ax.axis('equal')

def animate(frame):
    ax.set_title(f'Year {int(thk.time[frame])}')
    cax.set_array(thk.values[frame, :].flatten())

ani_glacier = animation.FuncAnimation(fig, animate, frames=len(thk.time), interval=100);
../../_images/4340c1b5e0feb5bb70805ba2c45b49799d3ef8b48b59b70641bd97c16b8f6b04.png
HTML(ani_glacier.to_jshtml())
# Write to mp4?
# FFwriter = animation.FFMpegWriter(fps=10)
# ani2.save('animation.mp4', writer=FFwriter)

Finetune the visualisation#

When applying the tools you might see that sometimes the timeseries are “shaky”, or have sudden changes in area. This comes from a few reasons:

  • OGGM does not distinguish between ice and snow, i.e. when it snows a lot sometimes OGGM thinks there is a glacier for a couple of years.

  • the trapezoidal flowlines result in sudden (“step”) area changes when they melt entirely.

  • on small glaciers, the changes within one year can be big, and you may want to have more intermediate frames

We implement some workarounds for these situations:

ds_smooth = workflow.execute_entity_task(
    distribute_2d.distribute_thickness_from_simulation,
    gdirs,
    input_filesuffix='_random_s1',
    concat_input_filesuffix='_spinup_historical', 
    ys=2003, ye=2100,  # make the output smaller
    output_filesuffix='_random_s1_smooth',  # do not overwrite the previous file (optional) 
    # add_monthly=True,  # more frames! (12 times more - we comment for the demo, but recommend it)
    rolling_mean_smoothing=7,  # smooth the area time series
    fl_thickness_threshold=1,  # avoid snow patches to be nisclassified
    )
2026-03-05 20:38:47: oggm.workflow: Execute entity tasks [distribute_thickness_from_simulation] on 2 glaciers
def plot_area_smoothed(ds_smooth, ds, gdir, title):
    area = (ds.simulated_thickness > 0).sum(dim=['x', 'y']) * gdir.grid.dx**2 * 1e-6
    area.plot(label='Distributed area (raw)');
    area = (ds_smooth.simulated_thickness > 0).sum(dim=['x', 'y']) * gdir.grid.dx**2 * 1e-6
    area.plot(label='Distributed area (smooth)');
    plt.legend(loc='lower left'); plt.ylabel('Area [km2]');
    plt.title(title, fontsize=20); plt.show();


plot_area_smoothed(ds_smooth[0], ds[0], gdirs[0], 'Aletsch')
# plot_area_smoothed(ds_smooth[1], ds[1], gdirs[1], 'Fieschergletscher')
../../_images/ec891f27abcb257c6f66346ac79b14a140a07bc4de885078b6566ace9d5a6d2b.png
# Get a handle on the figure and the axes
fig, ax = plt.subplots()

thk = ds_smooth[0]['simulated_thickness']  # Aletsch
# thk = ds_smooth[1]['simulated_thickness']  # Fieschergletscher

# Plot the initial frame. 
cax = thk.isel(time=0).plot(ax=ax,
    add_colorbar=True,
    cmap='viridis',
    vmin=0, vmax=thk.max(),
    cbar_kwargs={
        'extend':'neither'
    }
)
ax.axis('equal')

def animate(frame):
    ax.set_title(f'Year {float(thk.time[frame]):.1f}')
    cax.set_array(thk.values[frame, :].flatten())

ani_glacier = animation.FuncAnimation(fig, animate, frames=len(thk.time), interval=100);
../../_images/85e5696efd6a567c92ba9522e69e13453c2ba462dc25244ce796125c60d31c9a.png
# Visualize
HTML(ani_glacier.to_jshtml())
# Write to mp4?
# FFwriter = animation.FFMpegWriter(fps=120)
# ani2.save('animation_smooth.mp4', writer=FFwriter)

Merge redistributed thickness from multiple glaciers#

If you’re working in a region with multiple glaciers, displaying the evolution of all glaciers simultaneously can be convenient. To facilitate this, we offer a tool that merges the simulated distributed thicknesses of multiple glaciers.

This process can be very memory-intensive. To prevent memory overflow issues, we, by default, merge the data for each year into a separate file. If you have sufficient resources, you can set save_as_multiple_files=False to compile the data into a single file at the end. However, with xarray.open_mfdataset, you have the capability to seamlessly open and work with these multiple files as if they were one.

For further explanation on merging gridded_data, please consult the tutorial Ingest gridded products such as ice velocity into OGGM.

simulation_filesuffix = '_random_s1_smooth'  # saved in variable for later opening of the files
distribute_2d.merge_simulated_thickness(
    gdirs,  # the gdirs we want to merge
    simulation_filesuffix=simulation_filesuffix,  # the name of the simulation
    years_to_merge=np.arange(2005, 2101, 5),  # for demonstration I only pick some years, if this is None all years are merged
    add_topography=True,  # if you do not need topogrpahy setting this to False will decrease computing time
    preserve_totals=True,  # preserve individual glacier volumes during merging
    reset=True,
)
2026-03-05 20:39:02: oggm.sandbox.distribute_2d: Applying global task merge_simulated_thickness on 2 glaciers
2026-03-05 20:39:02: oggm.workflow: Applying global task merge_gridded_data on 2 glaciers
---------------------------------------------------------------------------
InvalidParamsError                        Traceback (most recent call last)
Cell In[24], line 2
      1 simulation_filesuffix = '_random_s1_smooth'  # saved in variable for later opening of the files
----> 2 distribute_2d.merge_simulated_thickness(
      3     gdirs,  # the gdirs we want to merge
      4     simulation_filesuffix=simulation_filesuffix,  # the name of the simulation
      5     years_to_merge=np.arange(2005, 2101, 5),  # for demonstration I only pick some years, if this is None all years are merged
      6     add_topography=True,  # if you do not need topogrpahy setting this to False will decrease computing time
      7     preserve_totals=True,  # preserve individual glacier volumes during merging
      8     reset=True,
      9 )

File /usr/local/pyenv/versions/3.11.11/lib/python3.11/site-packages/oggm/utils/_workflow.py:557, in global_task.__call__.<locals>._global_task(gdirs, **kwargs)
    553 self.log.workflow('Applying global task %s on %s glaciers',
    554                   task_func.__name__, len(gdirs))
    556 # Run the task
--> 557 return task_func(gdirs, **kwargs)

File /usr/local/pyenv/versions/3.11.11/lib/python3.11/site-packages/oggm/sandbox/distribute_2d.py:562, in merge_simulated_thickness(gdirs, output_folder, output_filename, output_grid, simulation_filesuffix, years_to_merge, keep_dem_file, interp, preserve_totals, smooth_radius, use_glacier_mask, add_topography, use_multiprocessing, save_as_multiple_files, reset)
    558     ds.to_netcdf(fp)
    560 if save_as_multiple_files:
    561     # first the _topo_data file
--> 562     workflow.merge_gridded_data(
    563         gdirs,
    564         output_folder=output_folder,
    565         output_filename=f"{output_filename}_topo_data",
    566         output_grid=output_grid,
    567         input_file='gridded_data',
    568         input_filesuffix='',
    569         included_variables=['glacier_ext',
    570                             'glacier_mask',
    571                             'distributed_thickness',
    572                             ],
    573         preserve_totals=preserve_totals,
    574         smooth_radius=smooth_radius,
    575         use_glacier_mask=use_glacier_mask,
    576         add_topography=add_topography,
    577         keep_dem_file=keep_dem_file,
    578         interp=interp,
    579         use_multiprocessing=use_multiprocessing,
    580         return_dataset=False,
    581         reset=reset)
    583     # recalculate bed topography after reprojection, if topo was added
    584     if add_topography:

File /usr/local/pyenv/versions/3.11.11/lib/python3.11/site-packages/oggm/utils/_workflow.py:557, in global_task.__call__.<locals>._global_task(gdirs, **kwargs)
    553 self.log.workflow('Applying global task %s on %s glaciers',
    554                   task_func.__name__, len(gdirs))
    556 # Run the task
--> 557 return task_func(gdirs, **kwargs)

File /usr/local/pyenv/versions/3.11.11/lib/python3.11/site-packages/oggm/workflow.py:991, in merge_gridded_data(gdirs, output_folder, output_filename, output_grid, input_file, input_filesuffix, included_variables, preserve_totals, smooth_radius, use_glacier_mask, add_topography, keep_dem_file, interp, use_multiprocessing, return_dataset, reset)
    989     dem_source = None
    990     dem_gdir = gdirs[0]
--> 991 gis.get_dem_for_grid(output_grid, output_folder,
    992                      source=dem_source, gdir=dem_gdir)
    993 # unwrapped is needed to execute process_dem without the entity_task
    994 # overhead (this would need a valid gdir)
    995 gis.process_dem.unwrapped(gdir=None, grid=output_grid,
    996                           fpath=output_folder,
    997                           output_filename=output_filename)

File /usr/local/pyenv/versions/3.11.11/lib/python3.11/site-packages/oggm/core/gis.py:483, in get_dem_for_grid(grid, fpath, source, gdir)
    472 grid_prop = {
    473     'utm_proj': grid.proj,
    474     'dx': grid.dx,
   (...)    478     'ny': grid.ny
    479 }
    481 source = check_dem_source(source, extent_ll, rgi_id=rgi_id)
--> 483 dem_list, dem_source = get_topo_file((minlon, maxlon), (minlat, maxlat),
    484                                      gdir=gdir,
    485                                      dx_meter=grid_prop['dx'],
    486                                      source=source)
    488 if rgi_id is not None:
    489     log.debug('(%s) DEM source: %s', rgi_id, dem_source)

File /usr/local/pyenv/versions/3.11.11/lib/python3.11/site-packages/oggm/utils/_downloads.py:2373, in get_topo_file(lon_ex, lat_ex, gdir, dx_meter, zoom, source)
   2371     source = getattr(gdir, 'rgi_dem_source')
   2372     if source is None:
-> 2373         raise InvalidParamsError('DEM source now needs to be specified '
   2374                                  'explicitly, either via the RGI file '
   2375                                  'or the `source` kwarg.')
   2377 if source not in DEM_SOURCES:
   2378     raise InvalidParamsError('`source` must be one of '
   2379                              '{}'.format(DEM_SOURCES))

InvalidParamsError: DEM source now needs to be specified explicitly, either via the RGI file or the `source` kwarg.
# by default the resulting files are saved in cfg.PATHS['working_dir']
# with names gridded_simulation_merged{simulation_filesuffix}{yr}.
# To open all at once we first get all corresponding files
files_to_open = glob.glob(
    os.path.join(
        cfg.PATHS['working_dir'],  # the default output_folder
        f'gridded_simulation_merged{simulation_filesuffix}*.nc',  # with the * we open all files which matches the pattern
    )
)

files_to_open
[]

You will notice that there is a file for each year, as well as one file with the suffix _topo_data. As the name suggests, this file contains the topography and gridded-outline information of the merged glaciers, which could be useful for visualizations.

Now we open all files in one dataset with:

with xr.open_mfdataset(files_to_open) as ds_merged:
    ds_merged = ds_merged.load()
---------------------------------------------------------------------------
OSError                                   Traceback (most recent call last)
Cell In[26], line 1
----> 1 with xr.open_mfdataset(files_to_open) as ds_merged:
      2     ds_merged = ds_merged.load()

File /usr/local/pyenv/versions/3.11.11/lib/python3.11/site-packages/xarray/backends/api.py:1596, in open_mfdataset(paths, chunks, concat_dim, compat, preprocess, engine, data_vars, coords, combine, parallel, join, attrs_file, combine_attrs, **kwargs)
   1593 paths = _find_absolute_paths(paths, engine=engine, **kwargs)
   1595 if not paths:
-> 1596     raise OSError("no files to open")
   1598 paths1d: list[str | ReadBuffer]
   1599 if combine == "nested":

OSError: no files to open
ds_merged
---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
Cell In[27], line 1
----> 1 ds_merged

NameError: name 'ds_merged' is not defined

Now we can look at the result with a plot:

plot_distributed_thickness(ds_merged, 'Aletsch and Fieschergletscher')
---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
Cell In[28], line 1
----> 1 plot_distributed_thickness(ds_merged, 'Aletsch and Fieschergletscher')

NameError: name 'ds_merged' is not defined

Alternatively, you can create an animation using the merged data, displaying a value every 5 years, as specified above with years_to_merge:

# Get a handle on the figure and the axes
fig, ax = plt.subplots()

thk = ds_merged['simulated_thickness']

# Plot the initial frame. 
cax = thk.isel(time=0).plot(ax=ax,
    add_colorbar=True,
    cmap='viridis',
    vmin=0, vmax=thk.max(),
    cbar_kwargs={
        'extend':'neither'
    }
)
ax.axis('equal')

def animate(frame):
    ax.set_title(f'Year {float(thk.time[frame]):.1f}')
    cax.set_array(thk.values[frame, :].flatten())

ani_glacier = animation.FuncAnimation(fig, animate, frames=len(thk.time), interval=200);
---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
Cell In[29], line 4
      1 # Get a handle on the figure and the axes
      2 fig, ax = plt.subplots()
----> 4 thk = ds_merged['simulated_thickness']
      6 # Plot the initial frame. 
      7 cax = thk.isel(time=0).plot(ax=ax,
      8     add_colorbar=True,
      9     cmap='viridis',
   (...)     13     }
     14 )

NameError: name 'ds_merged' is not defined
../../_images/9e13561b06f5d2c575e2e9d0ab28f3ca3efd7ae4d5738c0cd619a55149090489.png
# Visualize
HTML(ani_glacier.to_jshtml())

For visualization purposes, the merged dataset also includes the topography of the entire region, encompassing the glacier surfaces:

cmap=salem.get_cmap('topo')
ds_merged.topo.plot(cmap=cmap)
---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
Cell In[31], line 2
      1 cmap=salem.get_cmap('topo')
----> 2 ds_merged.topo.plot(cmap=cmap)

NameError: name 'ds_merged' is not defined

Or the estimated bedrock topography, without ice:

ds_merged.bedrock.plot(cmap=cmap)
---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
Cell In[32], line 1
----> 1 ds_merged.bedrock.plot(cmap=cmap)

NameError: name 'ds_merged' is not defined

Nice 3D videos#

We are working on a tool to make even nicer videos like this one:

from IPython.display import Video
Video("../../img/mittelbergferner.mp4", embed=True, width=700)

The WIP tool is available here: OGGM/oggm-3dviz