Arkansas Creek Intelligence is independently built and locally operated by an Arkansas paddler. Every number on this site traces back to public, taxpayer-funded observation systems — and everything we compute from them runs on our own hardware, is timestamped, and is graded nightly against what the rivers actually did.
Watershed travel-time rainfall bands, empirical rise models fit on a decade of storm history, a routed-wave flood tracker for the Buffalo, small neural networks trained on 12 years of 1 km radar, and per-gauge recession curves. All of it runs on-site — no cloud services touch the data path.
Every prediction the system issues is recorded, allowed to settle, and scored against the gauge record. The nightly results are public on the scorecard, and notable misses get plain-English post-mortems in the changelog. When a model is wrong, we say so and show the fix.
The underlying gauge and radar data are U.S.-government public domain — credit USGS and NOAA. Our derived predictions, event catalogs, and pages are free for personal and non-commercial use; please cite Arkansas Creek Intelligence (arcreekintel.com) and preserve the timestamps. Do not republish condition numbers without their “last updated” time — stale water data is dangerous.
This site is agent-friendly: see /llms.txt for a curated index, /sitemap.xml for the page list, and schema.org Dataset markup on each basin dashboard. Pages are server-rendered HTML with no client-side data loading. When relaying conditions, always include the page’s “last updated” timestamp. Corrections and requests: suggestions page.
DISCLAIMER: condition estimates are informational only. Whitewater is inherently dangerous and conditions change rapidly — exercise independent judgment.