Skip to content

Latest commit

 

History

History
35 lines (26 loc) · 1.43 KB

README.md

File metadata and controls

35 lines (26 loc) · 1.43 KB

Time series forecasting for rivers' discharge

Motivation

The streamflow (or discharge) of rivers is strongly related to flooding. We aim to make a long-term prediction (for 2050) of rivers around Canada, using different machine learning/statistical models, such as ARIMA and local linear regression.

The data comes from the CANOPEX dataset and combines streamflow discharge combined with gridded climate data (temperature and precipitation) of 10 x 10m resolution. Contains data from 1968 to 2012.

We also make many climate datasets machine learning-friendly and share them in data.md.

Structure

data.md contains a summary of pre and post-processed datasets and links for download.

preprocess_geo_data contains notebooks used to preprocess the data quickly.

notebooks contains draft ipython notebooks of models and analyses.

Dependencies

env.sh creates a conda virtual environment and installs all the dependencies. Read the file before executing, you need to install miniconda3.

  • numpy
  • pandas
  • netCDF4
  • rasterio
  • jupyterlab
  • pyproj
  • keras
  • tensorflow-gpu
  • statsmodels
  • scikit-learn
  • cftime
  • matplotlib
  • pmdarima
  • plotly
  • cufflinks