CRAFT (Calibration and Rapid-Adoption Forecasting Techniques)

CRAFT is a Python-based project for processing, analyzing, and modeling climate or environmental data. It uses machine learning techniques, including Random Forest Regression and Neural Networks, to create emulators for various environmental variables such as Gross Primary Production (GPP) and soil water content. O# (O5048)

© 2026. Triad National Security, LLC. All rights reserved. This program was produced under U.S. Government contract 89233218CNA000001 for Los Alamos National Laboratory (LANL), which is operated by Triad National Security, LLC for the U.S. Department of Energy/National Nuclear Security Administration. All rights in the program are reserved by Triad National Security, LLC, and the U.S. Department of Energy/National Nuclear Security Administration. The Government is granted for itself and others acting on its behalf a nonexclusive, paid-up, irrevocable worldwide license in this material to reproduce, prepare. derivative works, distribute copies to the public, perform publicly and display publicly, and to permit others to do so.

(End of Notice)

This program is Open-Source under the BSD-3 License.

Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:

 Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.

 Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.

 Neither the name of the copyright holder nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission. THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

(End of Notice)

Project Structure

The project consists of three main Python scripts:

1. 1_Data_Cleaning.py: Processes and prepares input data for machine learning.
2. 2_Emulat_fitting.py: Trains machine learning models (Random Forest or Neural Network) and performs feature analysis.
3. 3_Bayes_Fit.py: Implements Bayesian inference using Markov Chain Monte Carlo (MCMC) methods for parameter estimation and uncertainty quantification.

Dependencies

To run this project, you need the following Python libraries:

• pandas
• numpy
• matplotlib
• scikit-learn
• joblib
• scipy
• tensorflow (for neural network models)

You can install these dependencies using the provided requirements.txt file:

pip install -r requirements.txt

Setup and Usage

1. Clone the repository:

   git clone https://github.com/lanl/CRAFT.git
   cd CRAFT
   

2. Install the required dependencies:

   pip install -r requirements.txt
   

   Alternatively, you can use the provided setup script:

   python setup_venv.py
   

3. Prepare your input data:

   • Ensure your input data is in CSV format.
   • Update the Emulator_Settings_v1.csv file with the appropriate settings for your variables.
   • Choose your model type (Random Forest or Neural Network) in the Emulator_Metadata_v1.csv file.

4. Run the scripts in order:

   python 1_Data_Cleaning.py
   python 2_Emulat_fitting.py
   python 3_Bayes_Fit.py
   

Configuration

The project uses five main configuration files:

• Cleaning_Metadata_v1.csv: Contains settings for the data cleaning process.
• Emulator_Settings_v1.csv: Contains settings for the emulator fitting process.
• Emulator_Metadata_v1.csv: Provides metadata for the emulator process.
• Bayes_Settings_v1.csv: Contains settings for the Bayesian inference process.
• nn_config.csv: Contains configuration for neural network models (if selected).

Modify these files to adjust the variables being modeled, their input files, model type, and other configuration parameters.

Configuration Files

1. Cleaning_Metadata_v1.csv

Purpose: Contains settings for the data cleaning process. Related Script: 1_Data_Cleaning.py

Key Fields:

• DRIVE: Directory containing input data
• CASE: Name of the case being processed
• N_runs: Number of runs to process
• Varlist: List of variables to be cleaned
• Parameter_loc: Location of parameter file

2. Emulator_Settings_v1.csv

Purpose: Contains settings for the emulator fitting process. Related Script: 2_Emulat_fitting.py

Key Fields:

• var_name: Name of the variable being modeled (e.g., GPP)
• FATESruns: Input file for the variable data
• xs, xe: Start and end indices for data selection
• y_i: Index of the target variable
• Scaler: Scaling factor for the variable

3. Emulator_Metadata_v1.csv

Purpose: Provides metadata for the emulator process. Related Scripts: 2_Emulat_fitting.py, 3_Bayes_Fit.py

Key Fields:

• Settings_File: Name of the emulator settings file
• SaveName: Name for saving output files
• thres: Threshold value for feature importance
• EmulatorDrive: Directory for emulator data
• FATES_samples: Input file for FATES samples
• model_type: Type of model to use ('rf' for Random Forest or 'nn' for Neural Network)
• nn_config: Path to the neural network configuration file (used only when model_type='nn')

4. Bayes_Settings_v1.csv

Purpose: Contains settings for the Bayesian inference process. Related Script: 3_Bayes_Fit.py

Key Fields:

• initialpos: Initial position for the MCMC algorithm
• steps: Number of MCMC steps
• adapt_interval: Adaptation interval for the MCMC algorithm
• burnin: Number of burn-in steps
• Model_List: List of models to use
• Scaler_List: List of scalers to use
• Obs_List: List of observations to use
• list1: Additional list of parameters (purpose to be clarified)
• samples_file: File containing samples for the MCMC process
• obsfile: File containing observation data

5. nn_config.csv

Purpose: Contains configuration for neural network models. Related Script: 2_Emulat_fitting.py (when model_type='nn')

Key Fields:

• layer_sizes: Comma-separated list of layer sizes (e.g., "64,32,1")
• activation: Activation function for hidden layers (e.g., 'relu')
• optimizer: Optimizer for training (e.g., 'adam')
• loss: Loss function (e.g., 'mse')
• epochs: Number of training epochs
• batch_size: Batch size for training

Model Types

Random Forest (model_type='rf')

• Default model type
• Provides feature importance analysis
• Works well with smaller datasets
• Less computationally intensive
• Generates diagrams showing feature importance

Neural Network (model_type='nn')

• Alternative model type for complex relationships
• Requires more data for effective training
• May capture more complex patterns
• Configurable through nn_config.csv
• Does not provide feature importance analysis

Output

The project generates various outputs, including:

• Processed data files in the data/Vars/ directory
• Trained models and scalers in the data/Models/ directory
  • Random Forest models are saved with .joblib extension
  • Neural Network models are saved in directories with _nn suffix
• Diagnostic plots in the diag/ directory
• MCMC results in the data/MCMC_runs/ directory

Contributing

For information on how to contribute to this project, please refer to the CONTRIBUTING.md file (if available) or contact the project maintainers.

License

Please refer to the LICENSE file for information on the project's license and usage terms.

Contact

For questions or support, please contact the project maintainers or open an issue on the GitHub repository.