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).
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 objectdoy
: day of the yearhour
: hour of the dayminute
: minute of the daytemperature
: temperaturesolar_radiation
: solar radiationrelative_humidity
: relative humidityrain
: raindew_point_temp
: dew point temperaturepwv
: 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
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.