The watch list
Storms where the fixes and the working best track part ways
Every storm
The season, fix by fix
Method
Sources. The working best track is NHC's b-deck (Atlantic, eastern and central Pacific) and JTWC's via the NCAR-RAL open b-decks (everywhere else), as on the live page. Fixes: NHC's f-decks (subjective Dvorak by TAFB, SAB and CPHC; objective ADT and SATCON; aircraft centre fixes); NOAA SAB's fix tables for the western and South Pacific; UW-CIMSS ADT histories for every storm in every basin (a storm that crosses a basin line is filed twice by CIMSS — Narelle is 27P and 27S — and both legs are read); JTWC's prognostic reasoning messages, whose "agency Dvorak and automated fixes" block quotes PGTW, RJTD, RCTP, KNES and the CIMSS ADT/AIDT/D-PRINT/D-MINT/SATCON values (archived here from the NWS text feed on every bake since 24 September 2026 — no public archive keeps them); IBTrACS provisional agency tracks (10-minute winds ÷0.88); and the climatlas satellite estimator run 3-hourly on NOAA GMGSI imagery at each storm's working best-track position.
Consensus. Every fix is a 1-minute wind (Dvorak CIs through the Dvorak table). At each synoptic time of the working best track, the fixes within ±3 h are summarized by family — subjective Dvorak, ADT, SATCON, AI, climatlas, other agencies — and the median of the family medians is the fix consensus.
Flags. A storm is on the watch list when the consensus sits ≥15 kt from the book for 18 h or more with at least two families individually ≥10 kt the same way, or when the consensus peak (two or more families) lands in a different Saffir–Simpson class than the book's peak. Times within 6 h of an aircraft fix are exempt, and a book above the fixes within 12 h of an aircraft fix is treated as the satellite lagging a deepening storm the aircraft then measured. A flag is a prompt for the post-season review, not a correction.
Computed every six hours by pipeline/rt_fixes.py; data
rt_fixes.json.