Blog
- July 8, 2026
The $117 billion between the satellite and the decision
The WEF and Deloitte put a number on Earth observation's untapped value — and its single biggest bottleneck isn't satellites or AI. It's processing: turning cloud-hosted data into decision-ready answers without copying it first.
business earth-observation zarr-datafusion - July 7, 2026
How honest is one sample per month? A spectral audit
Part 3 — the ERA5 ONI recipe uses one hourly field per month instead of a true monthly mean. We use ERSST as ground truth and a power-spectrum comparison to show the shortcut adds noise only where the ENSO signal isn't — leaving the index intact.
tutorial climate zarr-datafusion - July 7, 2026
Reproducing NOAA's official El Niño index — exactly — with SQL
We recompute NOAA's Oceanic Niño Index straight from ERSST v5 — the dataset they build it from — reading the NetCDF in place via VirtualiZarr, and match the official table to 0.001 °C.
tutorial climate zarr-datafusion - June 26, 2026
Computing the El Niño index from ERA5 with SQL
Part 2 of the ONI series — we recompute the Oceanic Niño Index from ERA5, an independent reanalysis, and it tracks NOAA's official record to a 0.16 °C mean absolute error (r = 0.966) with zero ENSO sign reversals.
tutorial climate zarr-datafusion - June 1, 2026
Why SQL on Zarr? The architecture behind zarr-datafusion
An origin-story and architecture walkthrough of zarr-datafusion — why we built a SQL engine for Zarr-native array data.
architecture zarr-datafusion - May 15, 2026
What it costs to copy the planet's data
A business framing of the ETL/duplication tax — egress, redundant storage, and stale copies — and how 'query in place' avoids it.
business cost