IBMWeatherGen: Stochastic Weather Generator Tool

Welcome to IBMWeatherGen, a powerful gridded, multisite, multivariate, and daily stochastic weather generator based on resampling methodology written in Python

Setup

To get started, follow these steps:

1. Clone the Repository: Begin by cloning this repository to your local machine.

2. Create and Activate Virtual Environment: Set up a virtual environment using your preferred tool (conda, pyenv, virtualenv, etc.). Here's an example using conda:

   conda create -n wg-env python=3.8.14
   conda activate wg-env
   

3. Install Dependencies: Install the required dependencies by executing the following command:

   pip install -r requirements.txt
   

Running Tests

Follow these steps to run tests:

1. Unzip the Dataset: Unzip the provided dataset named dset_wg_d.csv.zip and place it in the /data directory.

2. Navigate to Source Directory: Move to the src directory in the terminal.

3. Run Simulation: Execute the simulation by running the following command:

   python execute_IBMWeatherGen_json.py
   

This will generate simulations stored in the simulation folder.

Customization

To customize the weather generator according to your needs:

1. Weather Dataset: Prepare a weather dataset in the same format as the provided dataset (dset_wg_d.csv).

2. Edit Input Parameter File: Modify the input parameter file with your desired settings:

   {
       "PATH_FILE_IN": "../data/dset_wg_d.csv",
       "PATH_FILE_OUT": "../simulations/",
       "START_YEAR": 2010,
       "NUM_YEARS": 10,
       "NUM_SIMULATIONS": 5,
       "WET_EXTREME": 0.999
   }
   

Where:

• PATH_FILE_IN: Path to the weather data used for training the generator.
• PATH_FILE_OUT: Path to store generated simulations.
• START_YEAR: Starting year for simulations.
• NUM_YEARS: Number of years to simulate ahead.
• NUM_SIMULATIONS: Number of simulation sets.
• WET_EXTREME: Threshold for extreme events.

Authors

• Maria Julia de Castro Villafranca Garcia

  • GitHub: mairju

• Jorge Luis Guevara Diaz

  • GitHub: jorjasso
  • Email: [email protected]

• Leonardo Tizzei

  • GitHub: ltizzei
  • Email: [email protected]

References

We have employed this implementation for conducting experiments detailed in our paper:

• Jorge Luis Guevara Diaz, Maria Garcia, et al. Direct Sampling for Spatially Variable Extreme Event Generation in Resampling-Based Stochastic Weather Generators. JAMES, 2023.

To enhance your understanding, we recommend exploring the following references:

• Steinschneider, Scott et al. "A semiparametric multivariate, multisite weather generator with low-frequency variability for use in climate risk assessments." DOI: 10.1002/wrcr.20528 (2013).

• Apipattanavis, Somkiat et al. "A semiparametric multivariate and multisite weather generator." DOI: 10.1029/2006WR005714 (2007).

• Kwon, Hyun-Han et al. "Stochastic simulation model for nonstationary time series using an autoregressive wavelet decomposition: Applications to rainfall and temperature." DOI: 10.1029/2006WR005258 (2006).

• Rajagopalan, Balaji et al. "A k-nearest-neighbor simulator for daily precipitation and other weather variables." DOI: 10.1029/1999WR900028 (1999).

A single-site weather generator in R that inspired this work:

• WeatherGen: walkerjeffd.github.io/weathergen/