Predict lake temperatures at multiple depths using an LSTM trained to make use of static lake attributes.
Create a new conda environment by running
conda update -n base conda
conda env create -f environment.yaml
conda activate ltls
Note: if Mamba is available, the environment may be created significantly faster by running
mamba update -n base mamba
mamba env create -f environment.yaml
conda activate ltls
For now, MNTOHA data are being used. A more complete set of lakes will be incorporated in the future, but MNTOHA lake data are readily available to download from ScienceBase. To download temperature observations, meteorological drivers, and metadata, run
snakemake -c1 -p fetch_all
Replace the 1 in -c1 with the number of cores to execute jobs in parallel.
The data will be downloaded to the 1_fetch/out folder.
If the download is interrupted, resume it and overwrite partially downloaded files by running
snakemake -c1 -p --rerun-incomplete fetch_all
To trigger a fresh download, change the date in 1_fetch/in/pull_date.txt to today's date and then call snakemake as above.
This work makes use of code available at https://github.com/kratzert/ealstm_regional_modeling. That code permits such use under the Apache-2.0 License. See https://opensource.org/licenses/Apache-2.0