GTV for Seasonal Forecasting

Setup

1. Clone this repository to wherever you work (locally or on a server).

2. Install the Python dependencies in requirements.txt. You can install them by running pip install requirements.txt in your terminal. To avoid conflicts with existing Python environments, I suggest doing this in a fresh virtual environment (see here for details).

3. The worked example in this repository is in a Jupyter Notebooks. After you've installed the appropriate dependencies, navigate to the repository and type jupyter notebook. A window should open in your browser that allows you to interact with the notebooks.

Datasets

The processed data can be found in the paper/data folder and consists of:

1. X_obs.csv: Monthly summer (July, Aug, Sept, Oct) SST observations from 1940-2019 at a 10 degree by 10 degree spatial scale. More info/downloads here.

2. sst_columns.csv: Information about the columns of X_obs.csv (each row corresponds to a column in X). in X_obs.csv (i.e the first row of sst_loc.csv gives you the location and month of the first column of X_obs.csv).

3. y_avg.csv: Winter (Nov - March) precipitation totals over the Southwestern US from 1941 to 2019. More info/downloads here.

4. X_lens.csv: Monthly summer (July, Aug, Sept, Oct) SST simulations from the CESM-LENS project at the same temporal and spatial granularity, but from 40 runs of the simulation (SST_lens.csv). Each simulation covers 66 years (1940 - 2005) so there are 2640 rows in this dataset where the first 66 rows correspond to the first simulation, the next 66 to the second simulations, and so on. These both include a "simulation" column to help keep track of things. More info/downloads here.