Skip to article frontmatterSkip to article content

Radar Scan Datasets: Time-Aware Cloud-Native Data Access

Weather radar data is essential for understanding precipitation processes, validating satellite observations, and improving forecast models. However, traditional radar data formats present significant challenges for modern cloud-based workflows. In this post, I explore how time-aware, cloud-native approaches using Zarr can transform how we work with radar scan datasets.

Read the full blog post at Earthmover →

Date: December 01, 2024

The Challenge with Traditional Radar Data

Traditional weather radar data is stored in file-based formats (NEXRAD Level-II, ODIM HDF5, UF) that were designed for sequential access on local filesystems. Each volume scan becomes a separate file, making it difficult to:

A Cloud-Native Solution

By converting radar data to Zarr format with careful attention to chunk sizing and metadata, we can create truly cloud-optimized datasets that enable:

  1. Time-aware queries: Quickly access data for specific time ranges without scanning through thousands of files

  2. Selective variable access: Download only the radar variables you need (reflectivity, velocity, etc.)

  3. Parallel processing: Use Dask to process data across multiple cores or machines

  4. Cost-effective storage: Cloud-optimized formats reduce data transfer costs

Key Design Principles

The approach I’ve developed through raw2zarr focuses on:

Read the full article on Earthmover →

Tools and Resources