1–3 day precipitation timing and convective potential • Updated every 6 hours
All Locations Model Comparison • Other locations: Flagstaff / KFLGBellemontAnthem
The top three panels show hour-by-hour precipitation for the next 1–3 days from three different forecast models, stacked so you can compare their timing and amounts directly. HRRR is this product’s primary, operational model. RRFS and REFS are marked Experimental: both are new NOAA modeling systems still in pre-operational testing ahead of a scheduled production changeover, so treat their guidance as provisional and expect their behavior to shift as NOAA finishes that transition. None of the three models actually reaches the full 72 hours shown: HRRR’s own forecast extends about 48 hours from its latest run, and REFS (NOAA’s documented limit for this ensemble product) extends about 60 hours — only RRFS runs close to the full 3-day window. Each panel shades the stretch beyond that model’s own horizon and labels it “beyond model horizon”, so an empty, shaded stretch there always means the model has no forecast for that time — never that it’s predicting a dry hour. The ingredient summary below the charts (HRRR only) is affected the same way: day 3 of this window often shows no HRRR ingredient read at all, which is that same honest horizon gap, not a data error.
The ingredient strip below the charts summarizes each day’s peak CAPE (Convective Available Potential Energy, storm fuel), CIN (Convective Inhibition, the “cap” suppressing storms), PWAT (Precipitable Water, total column moisture), and dew point, all from HRRR guidance, using the same Low/Moderate/High state labels as this site’s Convective Outlook product for a consistent read across both. These are provisional display bins, not calibrated forecast skill — they describe raw atmospheric ingredients, not whether storms will actually form.
All three models report precipitation and terrain-sensitive fields at their native grid resolution (roughly 3 km), sampled at the nearest grid point to each site’s coordinates — not adjusted for the difference between that grid point’s modeled terrain height and the site’s true elevation. In steep terrain like the transition from desert floor to forested peaks around these sites, that mismatch mostly affects temperature- dependent fields; it does not meaningfully change modeled precipitation amount or timing, since that’s driven by upslope moisture flux rather than a simple function of elevation. No correction is applied here for that reason.