AtmoScale is a data infrastructure company that turns weather radar archives into open, cloud-native datasets for meteorological, hydrological, research, and AI use cases.
Radar is where we lead. Environmental intelligence is where we are going — open, cloud-native, and vendor-neutral.
1,290 weather radars operate across 92 countries , generating some of Earth's most valuable environmental data.
Source: Ladino-Rincón, A., et al. (2026). Radar DataTree: A Cloud-Native AI-Ready Data Model for Accessible, Time-Aware Weather Radar Datasets . Submitted to IEEE Transactions on Big Data. Preprint: arXiv:2510.24943 .
Most of it remains locked in fragmented archives, vendor-specific formats, and workflows never designed for multi-year analysis or AI. Institutions end up adapting their science to whatever their storage will allow — instead of the other way around.
The openly cloud-accessible share — the cyan on this map — is still scattered across millions of individual files. Public, but not analysis-ready.
AtmoScale builds the cloud-native data infrastructure that gives that control back.
Three ways in — one open, cloud-native infrastructure underneath. Most institutions start with a single archive or workflow.
Transform fragmented historical archives into open, governed, cloud-native data infrastructure.
Build reliable ingestion, monitoring, versioning, and operational data services.
Enable fast analysis, derived products, event collections, APIs, and AI-ready datasets.
Modernized archives let scientists and operational teams query years of observations directly — without downloading, decoding, and reorganizing thousands of individual files.
| Traditional archive | Modernized with AtmoScale | |
|---|---|---|
| Time | ~11 hours | ~25 seconds |
| Task | One month of rainfall estimates, one radar | Same task, same science — on the cloud-native archive |
| Steps |
|
|
| Takeaway | Most of the time goes to preparing data — not using it. | The archive answers directly. Time goes to analysis, not preparation. |
Source: Ladino-Rincón, A., et al. (2026). Radar DataTree: A Cloud-Native AI-Ready Data Model for Accessible, Time-Aware Weather Radar Datasets . Submitted to IEEE Transactions on Big Data. Preprint: arXiv:2510.24943 . Hardware and archive details available on request.
Weather radar is the deepest, most demanding archive — where we lead. The same infrastructure integrates the observations around it.
The highest-resolution view of precipitation — and the hardest archive to use: huge volumes, vendor formats, decades of history. This is the archive we built Radar DataTree for, and where every engagement starts.
Core platform · proven on 7 national archivesGeostationary and polar-orbiting observations fill the gaps between radar networks and extend coverage to oceans and data-sparse regions — aligned to the same cloud-native, queryable structure.
Integrates with the radar archiveRain gauges and weather stations are the ground truth. Connected to the archive, they calibrate radar rainfall estimates and validate derived products — continuously, not per project.
Calibration & validationStream gauges and basin data turn precipitation into water: runoff, flood early warning, reservoir operations. Radar-based rainfall estimates feed hydrological models directly from the archive.
From rainfall to water decisionsForecast and reanalysis fields sit alongside the observations that verify them — the foundation for verification at archive scale and for AI-ready training datasets.
Verification & AI-ready dataKLOT · four dual-pol variables · 10 Mar 2026 — rendered directly from the ARCO dataset
Our first public dataset: NEXRAD Level II observations restructured into an analysis-ready, cloud-optimized hierarchy. Query years of dual-polarization data directly — no downloading thousands of files.
Coverage grows station by station — we list what is live, never projected.
Institutions shouldn't adapt their science to whatever their storage allows. The data exists to serve the mission — not the other way around.
An archive locked in a vendor binary is possession, not governance. Open, self-describing formats keep provenance, interoperability, and long-term control with the institution.
Faster access and cleaner pipelines serve one thing: defensible decisions — in real time, or across decades of climate record.
— core conviction“The data is already valuable.
We make sure you're the one in control of it. ”
Atmospheric scientists, software engineers, and cloud-data specialists — we started AtmoScale after watching too many institutions invest in radar networks and then lose access to their own archives.
We understand the observations, not just the bytes — decades of published research in radar meteorology, hydrology, and atmospheric science.
Any vendor, any format, any source. Built on open foundations (Zarr, Icechunk, xarray) — your institution never depends on us to read its own archive.
Pilots, managed deployments, consulting, or training — shaped around how your institution actually works, with full handover by default.
Founded by Alfonso Ladino-Rincón and Steve Nesbitt — Steve, co-author of Radar Meteorology: A First Course — and a team with decades of experience in atmospheric science, cloud computing, and operational data systems.
Understand the current archive, workflows, bottlenecks, and institutional needs.
Modernize a focused period, radar, event set, or operational workflow.
Build the production archive, ingestion system, access layer, or data product.
Maintain the infrastructure, add new observations, and enable new applications.
Start with an archive assessment, a pilot, or a private data deployment — wherever you are in the journey.
Or just hit reply to any email from us — we read everything.