Processed Meteorological Time Series for AI Weather Forecasting

Processed meteorological data from public sources like NASA POWER and Open-Meteo supports training deep learning models for short-term forecasting and extreme event detection. The dataset contains variables such as temperature, precipitation, humidity, and wind speed, structured into multivariate time series. It has been preprocessed with normalization, interpolation, and sliding window generation for supervised learning tasks.

Use Cases

• Short-term weather forecasting based on multivariate time series of temperature, precipitation, humidity, and wind speed.
• Extreme weather event detection and classification using processed meteorological variables.
• Benchmarking model architectures like MLP, LSTM, GRU, CNN, CNN-LSTM, and Transformers on standardized weather data.
• Training supervised learning models on preprocessed data featuring normalization and sliding window generation.

Strengths

• Dataset size is approximately 5.9 million data points (5,917,284), indicating a substantial volume.
• Explicitly prepared for AI model evaluation, with preprocessing for supervised learning.
• Available under a permissive CC-BY-4.0 license for reuse and modification.
• Cross-platform presence on figshare suggests established utility and importance.

Limitations

• Column names and exact row count are not provided, limiting precise understanding of data structure.
• Geographic scope, temporal coverage, and specific data sources beyond 'NASA POWER and Open-Meteo' are not detailed.
• The 'last updated' date of 2026-03 24 appears to be a future date, indicating a potential metadata error.

Provenance

Source

Publicly available sources including NASA POWER and Open-Meteo.

Collection Method

Collected and preprocessed using techniques like normalization and interpolation.

Time Range

null

Freshness

2026-03-24 13:19:34

Geography

null