ATMS 523 Module-5

Several machine learning techniques including a multiple linear regression, a polynomial regression, and a random forest regressor were used in this module. The main goal was to get more experience with machine learning and to try and predict rainfall rate from several different polarimetric radar parameters. Several error statistics were calculated to determine which model performed the best, including a baseline model calculation which used the formula $Z = 200 R^{1.6}$.

Setting Up

In order to complete this module in Python, scikit-learn will be used.

The package can be installed using conda

conda install -c conda-forge scikit-learn

Demo the code

Run the Module5.ipynb file. The necessary dataset is already provided.

Data

The radar_parameters.csv dataset is in the homework folder and will be used to train and test the models generated. These polarimetric radar parameters were calculated from disdrometer data in Huntsville, Alabama.

Columns

Features (radar measurements):

Zh - radar reflectivity factor (dBZ) - use the formula $dBZ = 10\log_{10}(Z)$

Zdr - differential reflectivity

Ldr - linear depolarization ratio

Kdp - specific differential phase

Ah - specific attenuation

Adp - differential attenuation

Target :

R - rain rate

Key functionalities

• Multiple linear regression
• Polynomial regression with a grid search
• Random forest regressor with a grid and randomized search
• Baseline model calculation
• Calculation of r-squared and root mean square error statistics

References and Acknowledgements

Scikit-learn: Machine Learning in Python, Pedregosa et al., JMLR 12, pp. 2825-2830, 2011.