Alplakes Simulations

Python framework for setting up, running, and post-processing hydrodynamic and wave simulations of Alpine lakes using Delft3D Flow, MITgcm and SWAN, simulations are run using Docker.

Note: Meteorological and river discharge data are sourced from Eawag-internal APIs. Eawag users must be connected to the Eawag intranet. External users will need to modify the relevant functions in src/weather.py and src/river.py to connect to their own data sources.

Installation

git clone https://github.com/eawag-surface-waters-research/alplakes-simulations.git
cd alplakes-simulations
pip install -r requirements.txt

Supported models

Docker image	Model
eawag/delft3d-flow:5.01.00.2163	Delft3D Flow 5.01
eawag/delft3d-flow:6.02.10.142612	Delft3D Flow 6.02
eawag/delft3d-flow:6.03.00.62434	Delft3D Flow 6.03
eawag/mitgcm:67z	MITgcm 67z
delftwaves/swan:v41.51	SWAN 41.51 (waves)

Usage

Setup — src/main.py

Prepares input files and optionally runs the simulation.

python src/main.py -m delft3d-flow/greifensee -d eawag/delft3d-flow:6.03.00.62434 -s 20221009 -e 20221011

Argument	Short	Description	Default
--model	-m	Static data folder, e.g. delft3d-flow/greifensee	required
--docker	-d	Docker image	required
--start	-s	Start date YYYYMMDD	required
--end	-e	End date YYYYMMDD	required
--run	-r	Execute simulation after setup	false
--upload	-u	Upload input files to S3	false
--profile	-p	Profile name to initialise from ({lake}/profiles)	false
--restart	-z	Path to restart file	false
--threads	-th	Number of threads	1
--bucket	-b	S3 bucket name	alplakes-eawag
--api	-a	Alplakes API URL	http://eaw-alplakes2:8000
--today	-t	Override today's date YYYYMMDD	system date
--log	-l	Log output directory	stdout

Run simulation

See eawag-surface-waters-research/docker for instructions on installing the Docker images.

Delft3D Flow — navigate to the generated run folder and execute:

cd {{ run folder }}
docker run -v $(pwd):/job --rm eawag/delft3d-flow:6.02.10.142612

MITgcm — the setup step generates a Dockerfile inside the run folder tailored to the lake's inputs. Build the image from that folder, then run it:

cd {{ run folder }}
docker build -t eawag/mitgcm:67z_{{ lake }} .
docker run \
  -v $(pwd)/binary_data:/simulation/binary_data \
  -v $(pwd)/run_config:/simulation/run_config \
  -v $(pwd)/run:/simulation/run \
  eawag/mitgcm:67z_{{ lake }}

SWAN — navigate to the generated run folder and execute:

cd {{ run folder }}
docker run --rm -v $(pwd):/home/swan delftwaves/swan:v41.51 swanrun -input control

The run produces a raw swan_block.dat block output, which is converted to NetCDF by the post-processing step.

Post-processing — src/postprocess.py

For Delft3D Flow and MITgcm, verifies results, splits output into weekly NetCDF files, and computes derived variables (e.g. thermocline). For SWAN, converts the raw swan_block.dat block output into per-segment NetCDF files. Output is written to a postprocess/ subfolder named by date.

python src/postprocess.py -f {{ run folder }} -d eawag/delft3d-flow:6.02.10.142612

Argument	Short	Description	Default
--folder	-f	Simulation run folder	required
--docker	-d	Docker image used for the simulation	eawag/delft3d-flow:6.02.10.142612
--skip	-s	Skip weeks before YYYYMMDD	false

Adding a new lake

Copy an existing lake folder from static/{model}/ that is similar to your target lake (e.g. similar size or river configuration) and rename it to your lake. Replace the static simulation input files with those for your lake, then update properties.json to match your lake's grid, rivers, secchi depth, etc. Meteo files are generated at runtime and do not need to be included.

Minimal properties.json for a Delft3D Flow lake with no river inputs:

{
  "grid": {"minx": 569000, "miny": 205000, "maxx": 590000, "maxy": 224000, "dx": 500, "dy": 500}
}

Change log

27 June 2024 - Temporary fix to improve summer surface temperature prediction

Zurich and Geneva switched to half secchi values and D3D 6.03.00.62434 (old model) due to underprediction of summer surface temperature. The full dataset was not re-computed hence there is a discrepancy in the timeseries on the 19 May 2024 where it shifts from the old values of secchi and the new model to the half secchi values and the old model.