FEMs team outlines December 3.5 release; users report missing hourly precipitation and fire‑danger recalculation gaps

Scott, the FEMs presenter, opened the call by saying the team plans a December "3.5" release that will import a cleaned dataset from February 2024 and parts of 2025, add mesonet sources and new map filters, push additional reports, and give a small admin group the ability to edit drought/fuel catalogs and assign stations. "We do have a release coming up here 3.5 that's coming up in the, December," he said, adding the release will flag potentially erroneous values but not overwrite existing QC data.

Why it matters: users told the group they are seeing operational gaps that affect daily situational awareness. Steve (a CalFire attendee identified in the transcript) reported that precipitation amounts show on FEMs' map and in manual CSV downloads but return zeros when fetched via the live/copy data link for hourly observations: "we get 0 amounts regardless of what's actually occurring." Scott said he would investigate the UI/download pathway and follow up with the user.

A second, related issue involves how FEMs handles backfilled weather inputs. Analyst Matt Molasic showed that exported FW21 files and Fire Family Plus mirror each other for missing raw observations (Matt found 18 hours missing in one example), but FEMs was reporting additional missing fire-danger hours (Matt quantified about 10 extra missing fire-danger hours in his sample). He and Scott traced the discrepancy to this sequence: FEMs backfills missing weather inputs from WX Weather into FW21, but currently does not automatically trigger a recalculation of fire-danger indices for the newly populated hours. Matt explained, "when there is a missing hour of data... it is obviously not calculating fire danger," and Scott said the FEMs team is building a recalculation trigger; he expects that capability to be available in December so the system can retroactively compute missing fire‑danger hours.

Forecast versus observation analysis: Michael, who has compared forecasts and observations across three dispatch areas (nine FDRAs and roughly 26 raw stations), said ERC forecasts generally aligned well with observations, while Burning Index tended to show larger, systematic discrepancies at some stations. Michael described an Excel-based approach overlaying forecast and observed time series and computing forecast error; he offered to share the workbook. He also noted a practical limitation: FEMs currently saves only the latest raw forecast file and overwrites earlier ones. Scott said the team plans to retain each incoming forecast file in an S3 bucket so users can retrieve historical forecasts (what a 7‑day forecast looked like on a prior date).

Operational guidance and next steps: presenters urged station-level monitoring (tipping buckets, snow flags and individual sensors) because short gaps often correct in a few hours but multi‑day gaps can take substantially longer to rebalance indices. Scott highlighted near-term goals — improved gap filling and QC, a move toward near-real-time updates (‘‘once a day, once every 4 or 5 days’’), and automated recalculation — and reiterated the December target for the recalculation feature. He also described planned enhancements to fuel-moisture charts and a longer-term effort to integrate NOAA gridded forecasts into FEMs within about a year to 18 months.

What remains unresolved: the live data‑link precipitation zeros reported by users are under investigation; Scott said he would follow up. The group also acknowledged the need to preserve historical forecast files and to finish the automated recalculation process so FEMs' fire-danger outputs align more quickly with recalculated FW21 exports and Fire Family Plus outputs.

The call closed with thanks to Matt and Michael for their analyses and demos and a reminder that the working group will meet again in December.