glaciersid / ghf-greenland-gbrt Goto Github PK

This repo contains the data and code to reproduce the results published in:

Rezvanbehbahani S., L.A. Stearns, A. Kadivar, J.D. Walker, and C.J. van der Veen (2017), Predicting the Geothermal Heat Flux in Greenland: A Machine Learning Approach, Geophys. Res. Lett., 44, doi:10.1002/2017GL075661.

The data for GHF map of Greenland (Figure 4 of the article) can be found in greenland_predictions.

To cite either of the data sets or the code, use the above citation.

The following are requirements to reproduce the results:

• System wide: Python 2.7 with developer tools, Tk with developer tools, build toolchain.
• Python packages: see requirements.txt
• Basemap (tested with veresions 1.0.7 and 1.1.0), built from source. For automated procedures for Linux see Makefile which automatically detects the virtual environment from the VIRTUAL_ENV environment variable.

On Debian/Ubuntu the following installs all requirements:

$ apt-get install python-dev build-essential tk tk-dev python-pip virtualenv
$ ... # git clone this repo; cd to repo
$ virtualenv env
$ . env/bin/activate
(env) $ pip install -r requirements.txt
(env) $ make basemap-install

In later versions of basemap, if geos is installed in a virtual environment its lib directory must be added to the LD_LIBRARY_PATH environment variable. To produce all figures in the paper:

(env) $ export LD_LIBRARY_PATH=env/lib:$LD_LIBRARY_PATH
(env) $ python density_plots.py   # Figures 1 and 2
(env) $ python greenland.py       # Figures 4, S5, S6
(env) $ python error_analysis.py  # Figures 3, 5, S2, S3, S4

Each data set below contains the following continuous features:

• age,
• bougeur_gravity_anomaly,
• depth_to_moho (depth to Mohorovičić discontinuity),
• d_2_hotspot (distance to hotspots),
• d_2_ridge (distance to ridge),
• d_2_trans_ridge (distance to transform ridge),
• d_2_trench (distance to trench),
• d_2_volcano (distance to volcano),
• d_2_young_rift (distance to young rift),
• heat_prod_provinces (heat production provinces),
• lithos_asthenos_bdry (lithosphere-asthenosphere boundary),
• magnetic_anomaly,
• thickness_crust,
• thickness_middle_crust,
• thickness_upper_crust,
• topography,
• upper_mantle_density_anomaly.

and three categorical features:

• thermo_tecto_age (age of last thermo-tectonic event): allowed values are 1-12.
• upper_mantle_vel_structure (upper mantle velocity structure): allowed values are 1-6.
• rock_type (rock type): as per Hartmann and Moosdorf (2012), allowed values are: 1 (volcanic), 2 (metamorphic), 3 (sedimentary).

• Global data (features and GHF): The global data set can be found at global.csv which contains feature values and GHF measurements from points predominantly outside of GrIS.
• GrIS data (features): All GrIS data set can be found at gris_features.csv which contains feature values (no GHF) from points on GrIS.
• GrIS ice cores (GHF): Latitude, longitude, and GHF measurements from ice cores in Greenland.
• "Original" data: For both data sets above a more complete version is also included for posterity. The global data set differs from its corresponding original data set global_original.csv and the GrIS data set differs from gris_features_original.csv by:
  • For both global and GrIS data sets the following columns in original data sets are excluded:
    • _lithk_cona,
    • _num_in_cell,
    • _num_in_continent,
    • _lithology_HM_unmodified,
    • _depth_to_moho_pasyanos,
    • _depth_to_moho_?,
    • _airy_gravity_anomaly,
    • _continent,
  • The global data sets excludes two original global records with unknown rock type (rock_type == -9999).
  • Rock types (rock_type column) in original GrIS data set take values between 1-10, as per Dawes (2009), which are mapped in GrIS data set to the three values described above, as per Hartmann and Moosdorf (2012).
  • Rock types (rock_type column) in original global data set are taken by mapping the 16 values of _lithology_HM_unmodified, as per Hartmann and Moosdorf (2012), to the three values described above.