Manon E. B. Sabot, Martin G. De Kauwe, Andy J. Pitman, Belinda E. Medlyn, Anne Verhoef, Anna M. Ukkola, and Gab Abramowitz

Repository containing all the code to reproduce the results in Sabot et al. (2019): Plant profit maximisation improves predictions of European forest responses to drought. New Phytologist, Accepted.

To make sure the model is properly set up, simply type:

make -f setup.mk

N.B.: You need anaconda to use the existing environment files within which to run the model. By default, python3.7 will be used to run the model. If you would rather install the dependancies locally, you will need python.

Followed by:

source activate TractLSM37

N.B.: If you prefer, you can activate TractLSM27 (the python2.7 version) or TractLSM (no specific associated module versions) instead.

To regenerate all our results and figures, type:

make -f results.mk

N.B. 1: By default, this uses 4 CPUs. However, if you have access to more CPUs and want things to speed up, change the NCPUs line 35 of src/predicting_kmax.sh and line 31 of src/var_solver_accuracy.sh.

N.B. 2: If the model runs are present in output/projects/, this will only relaunch the analysis and figures. To run all of the model experiments from scratch, you will first need to remove all of the specific experiment files from output/projects/. If you also want to recreate the input files from scratch, you can remove them from input/projects/, provided you have the eddy-covariance site data netcdf files stored in input/fluxsites/met/ and in input/fluxsites/flux/.

If you want to regenerate all the parameter calibrations for the Control model, type:

make -f calib.mk

N.B.: This is very long and would need to be performed on more than 4 CPUs. The CPU count can be increased at lines 30 in src/calib_Zroot.sh, src/calib_g1.sh, and src/calib_fw.sh

If you wish to clean up old log files from the src/tmp/ folder, type:

make clean -f results.mk

The TractLSM is a site-level simplified tractable Land Surface Model. It is well suited to comparing assumptions with regard to modelling plant induced carbon and water fluxes, as it represents soil hydrology using a simple water balance “tipping bucket” model.

The schematics below show how the model works at a glance, but consult the ReadMe in src/TractLSM/ for more detail.

These data are to be stored in input/fluxsites/met/ and in input/fluxsites/flux/.

• dataset 1: LaThuile Synthesis
• dataset 2: FLUXNET2015

The forcing climatologies are provided in input/fluxsites/LAI/, but you might wish to rebuild them (see src/TractLSM/Utils/modis_lai_lat_lon.py and src/TractLSM/Utils/met_flux_LAI_site_level.py) using the latest MODIS data.

• dataset: 8-day MODIS MOD15A2Hv6

The calculations ensuing from these data are stored in input/fluxsites/info_sites.csv, but should you wish to recreate them (see src/TractLSM/Utils/cru_climate_lat_lon.py) you will need the tmp, tmn, tmx, vap, pet, pre, and cld CRU TS v4.03 datasets.

• dataset: CRU TS v4.03

The data are all stored in input/fluxsites/info_sites.csv.

• source1: Valentini (2003)
• source2: Kattge et al. (2009)
• source3: Martin-StPaul (2012)

• source1: De Kauwe et al. (2015)
• source2: Lin et al. (2015)

• source1: Martin-StPaul et al. (2017)
• source2: Choat et al. (2012)
• source3: Lucani et al. (2019)

We conducted three experiments to vary (and predict) k_max. If you wish to reproduce these experiments with some flexibility, you can use the following commands:

src/predicting_kmax.sh

varies k_max depending on the site average, and then extreme climate scenario;

src/predicting_kmax.sh -S

runs the calibration experiment;

src/predicting_kmax.sh -A

runs all three experiments.

The schematics below show how the experiments work at a glance, but consult src/predicting_kmax.sh to understand the detail of the code.

The TractLSM allows to vary:

• the frequency of optimisation in the Profit_max
• the solving windown around the previous leaf water potential in the Profit_max
• the resolution of the incremental changes between two water potentials on the hydraulic stream

To explore the relative impacts of making any of these assumptions, simply type:

src/var_accuracy.sh

The calibrated effective rooting depth used to run all of our experiments was retrieved by running:

src/calib_Zroot.sh

For further calibration of the Control model, simply type:

src/calib_g1.sh

which calibrates the g₁ parameter; Or:

src/calib_fw.sh

which calibrates the soil moisture stress function.

CABLE outputs were generated using the CABLE LSM version 2.0 revision 5320, and output/CABLE_output/run_cable_site.py and output/CABLE_output/cable_utils.py.

The outputs are not stored in this repository but will be provided upon request.

If you wish to further explore model behaviour, simply make changes to src/irun.txt (similar to a namelist) and type:

src/ExecTractLSM src/irun.txt

This project is licensed under the MIT License - see the LICENSE file for details

Manon Sabot: [email protected]