Authors: Malachy Moran, Derrick Hee, Kayla Woputz, Tilman Bayer
Predicting Snowpack in California’s Sierra Nevada Mountains
California's Central Valley agriculture region accounts for over 35% of all US vegetable production, and over 65% of all US fruit and nut production, in addition to employing over 600k workers, while remaining dependent on snowmelt water from the Sierra Nevada mountains in a time of decreasing water supply. We’ve leveraged CNN and LSTM modeling to create a tool that uses satellite and weather data to estimate peak snow water equivalent (SWE) levels - i.e. amount of water after snow melt - in the Sierra Nevada.
This repository contains the code, working notebooks, and other files that were used in the development of the snowcast package and the SnowCast Prediction tool which is contained in the snowcast_prediction_module folder.
The most important folders and subfolders of the project are detailed below in the order in which they appear in the repo.
The data folder contains files relevant to the collection of data, including the data dictionary, and a csv of the training data geometries.
The docs folder contains team documents and the presentations of the project, including the Final Presentation.
The models folder contains all the notebooks for models that were trained during this project. The model used to generate the final weights is contained in the swe_cnn-lstm-dhee-aws notebook.
This folder contains all the Jupyter notebooks that were used in the devolpement of this project. One particular folder is of note.
The notebooks in the folder labeled data_ingestion show the development of the functions that were used to build the snowcast package. These notebooks are intended to be illustrative only of the work that was done. A thorough and detailed explanation of each function used in data gathering is available in the snowcast package repo. Please refer to the README.md of that repo.
This folder contains the Final Paper for this project. This paper goes into extensive detail on the background, motivation, methodology and results for this project.
This folder contains the Final Product of this project, a python script which generates predictions of Snow Water Equivalent for the user. This folder contains its own README.md, please refer to this document for a description of the SnowCast tool.
If you have any questions about this project, or wish to receive more information about the project, the contact details for the contributors are below.
Malachy Moran malachy.j.moran@berkeley.edu
Derrick Hee dhee@berkeley.edu
Kayla Wopschall kaylaw@berkeley.edu
Tilman Bayer tbay@berkeley.edu