Datasets:

tharinduperera
/

weather-clustering-data

license: cc-by-4.0
task_categories:
  - tabular-classification
  - time-series-forecasting
  - other
tags:
  - climate
  - weather
  - open-meteo
  - era5
  - duckdb
  - pyspark
pretty_name: >-
  Weather & Climate Big Data Analytics — 100-City ERA5 Historical Dataset
  (2016–2025)
size_categories:
  - 100K<n<1M

Weather & Climate Big Data Analytics — 100-City ERA5 Historical Dataset (2016–2025)

Dataset Summary

This dataset contains 365,300 daily weather observations across 100 geographically diverse global cities spanning 80 countries over a 10-year continuous timeframe (January 1, 2016 – December 31, 2025).

The raw data was ingested from the Open-Meteo Historical Weather API (ERA5 Reanalysis Model) across 1,305 validated work units without missing values, then processed into Hive-partitioned Parquet format (year=YYYY).

Dataset Structure

weather-clustering-data/
├── .gitattributes
├── README.md
├── locations/
│   └── locations.csv                 # 100-city global catalogue (80 countries)
├── metadata/
│   └── dataset_summary.json          # Dataset specifications & ingestion metadata
└── processed/
    └── parquet/
        └── weather/                  # 10 year partitions (2016 - 2025)
            ├── year=2016/
            ├── year=2017/
            ├── ...
            └── year=2025/

Key Characteristics

Total Rows: 365,300 daily records
Locations: 100 global cities across 80 countries
Timeframe: 2016-01-01 to 2025-12-31 (10 full calendar years, including leap years 2016, 2020, 2024 with 366 days each)
Primary Data Format: Apache Parquet (Hive partitioned by year)
Location Catalogue SHA-256: 5CBA155270694BB743E8ED4E95D3F6A95F135D4AA61CAF5B705CF13566F8BD19

Data Schema & Variables

Column	Data Type	Description
`location_id`	BIGINT	Stable location identifier
`city`	VARCHAR	Primary city name
`city_ascii`	VARCHAR	ASCII-normalized city name
`country`	VARCHAR	Country name
`iso2`	VARCHAR	ISO 2-letter country code
`iso3`	VARCHAR	ISO 3-letter country code
`admin_name`	VARCHAR	State / Province / Administrative region
`capital`	VARCHAR	Capital classification
`source_latitude`	DOUBLE	Input city latitude
`source_longitude`	DOUBLE	Input city longitude
`model_latitude`	DOUBLE	ERA5 grid model latitude
`model_longitude`	DOUBLE	ERA5 grid model longitude
`timezone`	VARCHAR	Local IANA timezone
`elevation`	DOUBLE	Ground elevation (meters)
`date`	TIMESTAMP	Daily observation date (`YYYY-MM-DD`)
`temperature_mean`	DOUBLE	Mean daily 2m temperature (°C)
`temperature_max`	DOUBLE	Maximum daily 2m temperature (°C)
`temperature_min`	DOUBLE	Minimum daily 2m temperature (°C)
`precipitation_sum`	DOUBLE	Total daily precipitation (mm)
`relative_humidity_mean`	BIGINT	Mean daily relative humidity (%)
`wind_speed_mean`	DOUBLE	Mean daily 10m wind speed (km/h)
`surface_pressure_mean`	DOUBLE	Mean daily surface pressure (hPa)
`year`	BIGINT	Partition key (`2016`–`2025`)

Attributions & Licensing

Weather Data Source: Open-Meteo Historical Weather API (ERA5 Reanalysis by ECMWF).
Location Catalogue Source: SimpleMaps Basic World Cities Database. Licensed under Creative Commons Attribution 4.0 (CC BY 4.0).

Downstream Project Roadmap

Ingestion & Processing: Complete (1,305/1,305 work units, Parquet storage validated with DuckDB).
Phase 3 — DuckDB Exploratory Data Analysis (EDA): Statistical summary, seasonality, anomaly identification.
Phase 4 — Climate Feature Engineering: Annual climate metrics (Köppen-Geiger indicators, seasonality indices, temperature range, precipitation distribution).
Phase 5 — PySpark MLlib K-Means: Scalable climate regime clustering & silhouette evaluation.
Phase 6 — Cluster Interpretation: Climate zone profiles & geospatial taxonomy.
Phase 7 — Interactive Streamlit Dashboard: Web dashboard for climate exploration & visualization.