EuLerian Identification of ascending AirStreams - ELIAS 2.0
This repository includes code and CNN models to derive conditional probabilities of WCB inflow, ascent, and outflow from data at a comparably low spatial and temporal resolution. The models which are contained in the repository directory ./models/ are referred to as time-lag models in Quinting and Grams (2021). For any questions, error reports etc please contact the corresponding author of the study julian.quinting at kit.edu
Quinting, J. F. and Grams, C. M.: EuLerian Identification of ascending AirStreams (ELIAS 2.0) in numerical weather prediction and climate models – Part 1: Development of deep learning model, Geosci. Model Dev., 15, 715–730, https://doi.org/10.5194/gmd-15-715-2022, 2022.
Application examples of the convolutional neural networks are provided in a companion study
Quinting, J. F., Grams, C. M., Oertel, A., and Pickl, M.: EuLerian Identification of ascending AirStreams (ELIAS 2.0) in numerical weather prediction and climate models – Part 2: Model application to different datasets, Geosci. Model Dev., 15, 731–744, https://doi.org/10.5194/gmd-15-731-2022, 2022.
Important: If you are using ELIAS 2.0 in a publication please include a reference to
Quinting, J. F. and Grams, C. M.: EuLerian Identification of ascending AirStreams (ELIAS 2.0) in numerical weather prediction and climate models – Part 1: Development of deep learning model, Geosci. Model Dev., 15, 715–730, https://doi.org/10.5194/gmd-15-715-2022, 2022.
Prerequisites
Install the conda environment using conda env create -f conda_env.yml. Alternatively, install the following packages in a new conda environment.
conda create -n tf tensorflow
conda activate tf
conda install -c conda-forge matplotlib
conda install -c conda-forge netCDF4
conda install -c conda-forge keras
conda install -c anaconda xarray
conda install -c numba numba
conda install -c conda-forge windspharm
conda install -c anaconda scipy
conda install -c conda-forge tensorflowData
ELIAS 2.0 is trained on ERA-Interim data at a global latitude-longitude grid of 1.0° grid spacing. The input data need to be provided on the same grid. Input data are needed at the following pressure levels: 1000, 850, 700, 500, 300, 200 hPa. These variables are required: temperature (T), specific humidity (qv), geopotential (phi), zonal wind (u), and meridional wind (v). Based on these variables the following predictors are derived on the fly:
| P | WCB inflow | WCB ascent | WCB outflow |
|---|---|---|---|
| 1 | 700-hPa thickness advection | 850-hPa relative vorticity | 300-hPa relative humidity |
| 2 | 850-hPa meridional moisture flux | 700-hPa relative humidity | 300-hPa irrotational wind speed |
| 3 | 1000-hPa moisture flux convergence | 300-hPa thickness advection | 500-hPa static stability |
| 4 | 500-hPa moist potential vorticity | 500-hPa meridional moisture flux | 300-hPa relative vorticity |
| 5 | conditional probability of ascent (+24 h)* | 30-d WCB ascent climatology** | conditional probability of ascent (-24 h)* |
* Calculated on the fly. ** Provided in the ./data/ directory.
Usage of Jupyter Notebook
An example Notebook is provided to calculate probabilities and masks of WCB inflow, ascent, and outflow for 04 October 2016 from ERA-Interim data. The ERA-Interim data for this case study are provided in the ./data/ directory. The 30-d running mean WCB climatology is needed for WCB ascent and is provided in the same ./data/ directory. The decision thresholds which convert the conditional probabilities to dichotomous predictions are provided in netcdf Format in the ./thresholds/ directory.