ldicarlo1/weather_forecast_bias_correction

A simple bias correction of temperature, dew point, and 10m wind speeds for the GFS, HRRR, and ECMWF models for two US locations.

Weather Forecast Bias Correction

Using Gradient Boosted Trees Algorithm

Author: Luca Di Carlo

Date: Aug 2021

This small project uses simple machine learning to bias correct temperature, dew point, and 10m wind speed data from the HRRR, GFS, and ECMWF forecasts.

In order to run the jupyter notebook below place the "data/" file containing the station data in this directory.

The contents can be described as the following:

Bias Correction of Weather Forecasts Using Machine Learning.pdf

A slide deck containing a thorough presentation of my findings.

analysis.ipynb

A Jupyter Notebook containing the entirety of the assignment, from analyses, modeling, and plot generations.

analysis_functions.py

A python script containing the classes of functions I built to use for the analysis.

Results

Results indicated that by using the Gradient Boosted Trees algorithm to bias correct temperature, dew point, and 10m wind speeds produced forecasts that were on average 33% more accurate than previously.

Complete analysis of the machine learning implementation can be found in the Bias Correction of Weather Forecasts PDF.