Weather Data Analysis

A data science case study analyzing daily weather observations for Madrid, Spain from 2010 to 2020. The data is sourced from the Meteostat API and covers temperature, precipitation, wind, and pressure metrics.

Project Overview

The goal is to demonstrate the full data analysis pipeline: acquisition, cleaning, feature engineering, exploratory analysis, and visualization. The analysis includes:

• Data acquisition via the Meteostat Python library (meteostat)
• Cleaning and preprocessing: handling missing values, column renaming, datetime indexing
• Feature engineering: extracting year, month, day-of-year, day-of-week; computing daily temperature range and rolling averages (7-day and 30-day windows)
• Seasonal analysis: comparing winter vs summer temperature distributions
• Anomaly detection: identifying temperature outliers by month using the Interquartile Range (IQR) method
• Visualization: line plots, boxplots, histograms, heatmaps, and scatter plots of local extrema

Project Structure

Weather_analysis/
├── data/
│   ├── raw/                    # Raw CSV from Meteostat
│   └── clean/                  # Cleaned and feature-engineered CSV
├── figs/                       # Generated PNG visualizations
├── notebooks/
│   ├── 01_download.ipynb       # Data download from Meteostat
│   ├── 02_clean_explore.ipynb  # Cleaning, feature engineering, statistics
│   └── 03_viz_report.ipynb     # Visualizations and analysis
├── src/
│   ├── download_data.py        # (placeholder)
│   └── utils.py                # (placeholder)
├── .gitignore
├── README.md
└── requirements.txt

Dependencies

• Python 3.11+
• pandas, numpy
• matplotlib, seaborn
• scipy
• meteostat
• jupyterlab

Install: pip install -r requirements.txt

Usage

Run the notebooks in order:

1. 01_download.ipynb — Fetches daily weather data for Madrid (40.4168, -3.7038) from 2010-01-01 to 2020-12-31 and saves it to data/raw/.
2. 02_clean_explore.ipynb — Loads raw data, drops sparse columns, fills remaining NaNs, creates datetime features and rolling averages, then saves the cleaned dataset to data/clean/.
3. 03_viz_report.ipynb — Produces all charts and analysis: line trends, boxplots, histograms, anomaly detection, correlation heatmaps, and local extreme value identification.

Generated Figures

All plots are saved to figs/:

• lineal_t-averages.png / lineal_t-avg_months.png — temperature trends
• boxplot_t-avg_months.png — monthly temperature distribution
• z-scores_t-avg.png / z-scores_t-winter-summer.png — anomaly detection
• histogram_precipitation.png, histogram_wind-speed-2015.png, histogram_wind-speed_2010-2020.png — variable distributions
• histogram_t-avg_winter-summer.png — seasonal comparison
• heatmap_corr-matrix.png / heatmap_pivot-table.png — correlation and pivot analysis
• local_min-max_extremes.png — local temperature extrema