shamik394 / Model-Weighting-Toolbox

My research at NASA-JPL as a Student Intern presents a unique climate change simulation toolbox that merges multiple climate models and observations using Pseudo-Bayesian Model Averaging (BMA) and other techniques. It allows researchers to select optimal model weights for accurate forecasts and flexibility in addressing various climatic inquiries.

Model-Weighting-Toolbox

Tested on Python 3.11.4, running on Red Hat Enterprise Linux Server, version 7.9 (Maipo)

The user should install the packages specified in the requirements.txt file.

File Description

The dataset used in the Model Weighing Toolbox as the second example is available on this website (https://climatedata.oscer.ou.edu/PMtest/), and please select "tasmax_moddat1.csv."

Pseudo-BMA Function Definitions

• The key idea with BMA is to combine different models to improve prediction performance.
• BMA works by sampling different weights and then evaluating the performance of the combined model using those weights.
• The key metric for evaluation is RMSE, which is defined as

$$RMSE=\sqrt{\frac{1}{N}\sum_{i=1}^N(y_i-\hat{y}_i)^2}$$

• By repeatedly sampling weights, one can obtain an estimate of the optimal weights according to $RMSE$.

Calculating independence for BMA (for lines #51-53)

• Please visit the documentation for the corresponding title in the toolbox.

MSE Definition

• Please visit the documentation for the corresponding title in the toolbox.

The Sanderson Approach for both skill and independence

• Please visit the documentation for the corresponding title in the toolbox.