Skip to content

Latest commit

 

History

History
143 lines (107 loc) · 17.2 KB

README_extended_per_glacier_files.md

File metadata and controls

143 lines (107 loc) · 17.2 KB

oggm-standard-projections

OGGM provides what we like to call standard projections. Currently we make these future CMIP forced global glacier simulations available in two different formats, raw and aggregated data. In this document the raw data is described.

If you are only interested in regional volume or area changes (globally or per RGI region), we recommend you to use the aggregated data README.md.

When you use the aggregated or the raw per-glacier data, please cite the dataset via:

  • zenodo-link ... In addition, cite OGGM (Maussion et al., 2019) and the CMIP option that you are using (references are linked below).

Extended documentation of the raw oggm-output files and documentation of the standard (CMIP) projections computed with OGGM

Model set-up

At the moment, there are only projections available using OGGM v1.6.1 with the preprocessed glacier directory version 2023.3, which you can find here: https://cluster.klima.uni-bremen.de/~oggm/oggm-standard-projections/oggm_v16/2023.3/.

These projections use elevation-band flowlines, include the dynamical spinup, the new informed 3-step per-glacier geodetic calibration method, and use the W5E5v2.0 climate dataset (Lange and others, 2021) for calibration and a border of 160.

We computed all GCMs and scenarios that are currently available at the OGGM cluster. However, you have to choose those scenarios that are suitable and representative for your study. For example, you could select them after the method of Hausfather et al. (2022) or aggregate them after their 2100 warming levels (e.g. as in Rounce et al., 2023).

Data structure

In the subfolders, we give the netCDF files for different CMIP and final year options for all of the 19 RGI regions. That means, an example filepath looks as follows:

https://cluster.klima.uni-bremen.de/~oggm/oggm-standard-projections/oggm_v16/2023.3/CMIP6/2300/RGI06/run_hydro_gcm_from_2000_ACCESS-CM2_ssp126_bc_2000_2019_Batch_0_1000.nc (the different options are explained in detail below).

Each netCDF file contains the projections from up to 1000 glaciers for a single GCM, scenario, RGI region and historical projection option from 2000 until 2099, 2100, 2299 or 2300. The timestep always corresponds to the glacier state at the beginning of the year, i.e., 2000 means 01-01-2000. We have run the run_with_hydro task of OGGM on a monthly basis, and thus, the following variables are included:

variable description unit coords
volume Total glacier volume m 3 ['time', 'rgi_id']
volume_bsl Glacier volume below sea-level m 3 ['time', 'rgi_id']
volume_bwl Glacier volume below water-level m 3 ['time', 'rgi_id']
area Total glacier area m 2 ['time', 'rgi_id']
length Glacier length m ['time', 'rgi_id']
calving Total accumulated calving flux m 3 ['time', 'rgi_id']
calving_rate Calving rate m yr-1 ['time', 'rgi_id']
off_area Off-glacier area m 2 ['time', 'rgi_id']
on_area On-glacier area m 2 ['time', 'rgi_id']
melt_off_glacier Off-glacier melt kg yr-1 ['time', 'rgi_id']
melt_on_glacier On-glacier melt kg yr-1 ['time', 'rgi_id']
liq_prcp_off_glacier Off-glacier liquid precipitation kg yr-1 ['time', 'rgi_id']
liq_prcp_on_glacier On-glacier liquid precipitation kg yr-1 ['time', 'rgi_id']
snowfall_off_glacier Off-glacier solid precipitation kg yr-1 ['time', 'rgi_id']
snowfall_on_glacier On-glacier solid precipitation kg yr-1 ['time', 'rgi_id']
snow_bucket Off-glacier snow reservoir (state variable) kg ['time', 'rgi_id']
model_mb Annual mass balance from dynamical model kg yr-1 ['time', 'rgi_id']
residual_mb Difference (before correction) between mb model and dyn model melt kg yr-1 ['time', 'rgi_id']
melt_off_glacier_monthly kg month-1 ['time', 'rgi_id', 'month_2d']
melt_on_glacier_monthly kg month-1 ['time', 'rgi_id', 'month_2d']
liq_prcp_off_glacier_monthly kg month-1 ['time', 'rgi_id', 'month_2d']
liq_prcp_on_glacier_monthly kg month-1 ['time', 'rgi_id', 'month_2d']
snowfall_off_glacier_monthly kg month-1 ['time', 'rgi_id', 'month_2d']
snowfall_on_glacier_monthly kg month-1 ['time', 'rgi_id', 'month_2d']
snow_bucket_monthly kg month-1 ['time', 'rgi_id', 'month_2d']
residual_mb_monthly kg month-1 ['time', 'rgi_id', 'month_2d']
water_level Calving water level ['rgi_id']
glen_a Simulation Glen A ['rgi_id']
fs Simulation sliding parameter ['rgi_id']

If a region has in RGI6 3927 glaciers, then there will be four files going from *Batch_0_1000.nc to *Batch_3000_4000.nc. Note that, there can be less than 1000 glaciers inside, if it is the last batch. If you want to get e.g. the annual runoff of all glaciers from e.g. a specific gcm, ssp and RGI region, you can get it via:

import xarray as xr
path = '/home/www/oggm/oggm-standard-projections/oggm_v16/2023.3/CMIP6/2300/'
with xr.open_mfdataset(path + 'RGI11/run_hydro_w5e5_gcm_merged_CanESM5_ssp126_*.nc') as ds:
    ds['runoff'] = ds['melt_off_glacier']+ds['melt_on_glacier']+ ds['liq_prcp_off_glacier'] +ds['liq_prcp_on_glacier']
    ds = ds['runoff'].load()

Note, that some glaciers will have NaN values, so if you aggregate it, you need to either only use the common running glaciers (the method used to create the aggregated csv files) or do some filling. The annual glacier runoff that was computed here is the sum of annual melt and liquid precipitation on and off the glacier using a fixed-gauge with a glacier minimum reference area from year 2000 (unit: kg year-1).

  • We also have already postprocessed per-glacier annual and monthly runoff files aggregated for every basin (we used this notebook, at the moment only for CMIP6 until 2100). The basin files are available here https://cluster.klima.uni-bremen.de/~oggm/oggm-standard-projections/oggm_v16/2023.3/CMIP6/2100/basins/. All glaciers of a basin are aggregated in one file, and there are different files for different simulations. The files of each basin are in one subfolder, with the subfolder name being the MRBID of that basin. The basin files contain the variables: volume, area, runoff and runoff_monthly. The runoff components can not be computed for the last year, thus the last year has NaN values.

-> More details on how to handle the data (for example to estimate glacier runoff) is given in this jupyter notebook

CMIP option and final year: At the moment there are three categories of forcing data that have been used for the simulations (https://cluster.klima.uni-bremen.de/~oggm/oggm-standard-projections/oggm_v16/2023.3/). For all options, we used the bias-corrected GCMs to project glacier changes until 2100 or 2300. Note that the sample of GCMs going until 2300 is much smaller than the one going until 2100. ISIMIP3b_CMIP6 only goes until 2100, thus projections only go until 2100. Note, that we did two separate runs, once until 2100 and once until 2300, for GCMs having data until 2300, as there are glaciers that fail before 2300, but not until 2100 (specifically in Iceland, RGI region 6, see the overview table in notebooks/missing_glacier_area_stats.png).

  • CMIP6/2100 or CMIP6/2300
    • bias correction period 2000-2019 (file path named: bc_2000_2019)
    • main reference: Eyring et al. (2016)
  • CMIP5/2100 or CMIP5/2300
    • bias correction period 2000-2019 (file path named: bc_2000_2019)
    • main reference Taylor et al. (2012)
  • ISIMIP3b_CMIP6/2100
    • internally bias-corrected to W5E5 using bias correction period 1979-2014 (Lange, 2019) and interpolated to a 0.5° spatial resolution, we did not apply any additional bias-correction on top of that
    • five primary GCMs from phase 3b of the Inter-Sectoral Impact Model Intercomparison Project (ISIMIP3b) for three different SSPs
    • main references Lange (2019, 2022)

For CMIP5 and CMIP6, we have used here the same bias correction period as in Rounce et al., (2023), i.e. from Jan 2000 to Dec 2019. However, using a bias correction period of 1979-2014 instead of 2000-2019 (tested on five GCMs) resulted in glacier volume differences that were globally up to 20% different until 2100 for a single GCM. (analysed by Lilian Schuster, uncleaned notebook with the analysis)

Historical projections: 2000-2019

Inside of the RGI region folders, there are two options for how the perdiod 2000-2019 is being treated:

  • w5e5_gcm_merged: Here, W5E5 was applied from 2000-2019 and the GCMs were applied from 2020 onwards until 2100. This is the option that we chose for the aggregated files.
    • In the period 2000-2019, the projections are the same for every GCM and scenario (as we use W5E5).
  • gcm_from_2000: W5E5 is only used to calibrate the mass-balance model. For the actual projections, the individual GCMs (and scenarios) are used form 2000 onwards.
    • In the period 2000-2019, different GCMs create different output.

The influence of the historical projection choice is invisible globally. Regionally it is very small after 2000-2019 (regionally < 2% after 2050). However, the historical projection choice changes the projections over the period 2000-2019 by up to 8% regionally estimates although the CMIP GCMs were bias-corrected to that time period. (analysed by Lilian Schuster, uncleaned notebook with the analysis)

Additional informations to rerun the projections

All scripts used to run the projections for https://cluster.klima.uni-bremen.de/~oggm/oggm-standard-projections/oggm_v16/2023.3/ are in the _run_scripts folders of the given oggm version. At the moment, we only have one version, thus the scripts are located here: https://cluster.klima.uni-bremen.de/~oggm/oggm-standard-projections/oggm_v16/2023.3/_run_scripts/. They were run on the oggm bremen cluster, and parts of the scripts may be specific to that cluster.

We have always run all projections on 1000 glaciers at once (instead of per RGI region), as this distributes the glaciers better to the different nodes.

To run similar projections (for example with another bias correction period), you would need to adapt the paths and the files inside of that folder to your needs. Here is a short description of the different files: