Forecasting Precipitable Water Vapor Using LSTMs

With the spirit of reproducible research, this repository contains all the codes required to produce the results in the manuscript:

Jain, M., Manandhar, S., Lee, Y., Winkler, S. and Dev, S.(2020). Forecasting Precipitable Water Vapor Using LSTMs. In: International Symposium onAntennas and Propagation and North American Radio Science Meeting. IEEE.

The work is done using the Google Colab Framework (with GPU).

Scripts

• read_matfile.py: reads the matlab mat file that contains the weather station recordings for the year 2010.
• pwv_main.ipynb: main program. Currently, it loads the data, and returns the following numpy arrays of the weather station recordings. This is followed by LSTM training for PWV forecast.
  • timestamp: datetime object
  • doy: day of the year
  • hour: hour of the day
  • minute: minute of the day
  • temperature: temperature
  • solar_radiation: solar radiation
  • relative_humidity: relative humidity
  • rain: rain
  • dew_point_temp: dew point temperature
  • pwv: precipitable water vapor
• test_model.ipynb: main program. This is to load the trained model and produce results for PWV forecasting.
• pwv_lstm.h5: Trained LSTM model - H5PY file

Note:

The dataset used in this project can not be disclosed due to external reasons. However, one may feel to use/modify the code as per the requirement.