Storm Snow / Snowstorm
Models: 5-model average, GFS in upslope, ensemble odds · why ›
Snow: the average of 5 models (ECMWF, GFS with the HRRR in its first ~40 hours, ICON, Canadian, NBM), with the GFS alone in upslope hours. Rain vs snow: ECMWF temperature at each spot’s height. Sleet and freezing rain: ECMWF and GFS temperatures aloft. Fluffier snow (18:1) when the air a mile or two up is in the -12 to -18°C dendrite zone. Odds: 4 ensembles (143 runs of ECMWF, GFS, Canadian and ICON).
We scored every model against 5 local snow observers over 41 storms in the last 2 winters, each forecast made a day ahead. The plain average missed by 1.8″ a storm. Picking a favorite model for each storm type did worse (2.0″), because the best model changes from storm to storm. One exception held in both winters: the GFS alone does better in upslope (northwest wind). Swapping it in for the upslope hours cut the miss in upslope storms from 3.1″ to 2.5″ across 3 seasons. The high-resolution models (HRRR 3 km, Canada’s HRDPS 2.5 km) are already in the average: a day out they are the same data as our GFS and Canadian feeds, and on their own they scored worse at the observers.
What we correct: nothing in the precipitation. What changes is how much snow each inch of water makes: when the air at 850 or 700 mb is -12 to -18°C, flakes grow as big fluffy dendrites, so we use at least 18:1. Checked against the observers over 45 storms, that cut the average miss from 1.83″ to 1.76″ a storm and took away most of the low bias. It won both full winters and lost the thin Feb–Mar 2024 stretch, on one storm. Scaling the upslope precipitation up also helped, but it fixes the same shortfall, and on top of the dendrite rule it added nothing, so we don’t. The odds come from the ensembles: each model run many times with slightly different starts, so a split forecast shows up as a split instead of being averaged away. Each model counts equally however many runs it has. They’re raw model odds, coarser than the main forecast (one grid cell for the whole area), and not yet calibrated; every one is saved so we can score them against the stake this winter. Rain vs snow uses ECMWF temperature uncorrected: our corrected, warmer temperatures undercounted snow by about 20%, because the models’ cold bias goes away while it’s snowing. Up the mountain, temperature follows each model’s own layers, so a warm summit in a cutter rains while the valleys freeze.
Snowstorm coming, Sun 2 pm to Mon 10 pm (33 hours) · Warm-air advection then coastal low, with a little upslope after.
How much: Each model on its own temperatures: Euro 14″ up top, NBM 11″ (Canadian 14, ICON 12, GFS 11).
| Where | Snow |
|---|---|
| Mt Ellen Peak | 12″ |
| Lincoln Peak | 13″ |
| Snow Bowl Top | 13″ |
| Lincoln Base | 12″ |
| Snow Bowl | 13″ |
| Mt Ellen Base | 12″ |
For skiers: the peaks get 12–13″ of snow; the bases 12–13″. Most of the snow is above about 200 ft: that’s where the skiing will be.
What falls: Mt Ellen Peak: snow. Mt Ellen Base: snow.
Warm-air advection, Sun 2 pm–Sun 7 pm: Front-end thump: snow as warm, moist air rides up over the cold. Fairly even across the area, a bit more up high. How deep the warm layer gets decides snow vs sleet vs freezing rain.
Coastal low, Sun 8 pm–Mon 4 pm: Coastal low passing about 530 km southeast of us. East to southeast wind pours down the west face, which shadows Bristol and the west slope; the summits and the Mad River side do better. A 50 to 100 mile wobble in the track moves the rain line and the shadows.
Upslope (light: about 0.07″ of water), Mon 5 pm–Mon 10 pm: Northwest wind on the ridge: upslope. The summits and the west slope (Lincoln, South Lincoln, the Starksboro hills) usually beat the model numbers, because coarse models smear this snow out. Ridge wind near 15 kt favors the west slope; the Mad River side gets less. Snow in these hours uses the GFS alone: it beat the model average on upslope in both winters of our backtest, though every model still ran light (about 2.5″ short per storm on average).
Storm Snow / Ice storm
Models: 5-model average, GFS in upslope, ensemble odds · why ›
Snow: the average of 5 models (ECMWF, GFS with the HRRR in its first ~40 hours, ICON, Canadian, NBM), with the GFS alone in upslope hours. Rain vs snow: ECMWF temperature at each spot’s height. Sleet and freezing rain: ECMWF and GFS temperatures aloft. Fluffier snow (18:1) when the air a mile or two up is in the -12 to -18°C dendrite zone. Odds: 4 ensembles (143 runs of ECMWF, GFS, Canadian and ICON).
We scored every model against 5 local snow observers over 41 storms in the last 2 winters, each forecast made a day ahead. The plain average missed by 1.8″ a storm. Picking a favorite model for each storm type did worse (2.0″), because the best model changes from storm to storm. One exception held in both winters: the GFS alone does better in upslope (northwest wind). Swapping it in for the upslope hours cut the miss in upslope storms from 3.1″ to 2.5″ across 3 seasons. The high-resolution models (HRRR 3 km, Canada’s HRDPS 2.5 km) are already in the average: a day out they are the same data as our GFS and Canadian feeds, and on their own they scored worse at the observers.
What we correct: nothing in the precipitation. What changes is how much snow each inch of water makes: when the air at 850 or 700 mb is -12 to -18°C, flakes grow as big fluffy dendrites, so we use at least 18:1. Checked against the observers over 45 storms, that cut the average miss from 1.83″ to 1.76″ a storm and took away most of the low bias. It won both full winters and lost the thin Feb–Mar 2024 stretch, on one storm. Scaling the upslope precipitation up also helped, but it fixes the same shortfall, and on top of the dendrite rule it added nothing, so we don’t. The odds come from the ensembles: each model run many times with slightly different starts, so a split forecast shows up as a split instead of being averaged away. Each model counts equally however many runs it has. They’re raw model odds, coarser than the main forecast (one grid cell for the whole area), and not yet calibrated; every one is saved so we can score them against the stake this winter. Rain vs snow uses ECMWF temperature uncorrected: our corrected, warmer temperatures undercounted snow by about 20%, because the models’ cold bias goes away while it’s snowing. Up the mountain, temperature follows each model’s own layers, so a warm summit in a cutter rains while the valleys freeze.
Ice storm coming, Sun 11 pm to Mon 7 pm (21 hours) · Cutter.
How much: Each model on its own temperatures: NBM 2″ up top, ICON 1″ (GFS 2, Canadian 2, Euro 1).
| Where | Snow | Ice | Rain |
|---|---|---|---|
| Mt Ellen Peak | 2″ | 0.05″ | 0.82″ |
| Lincoln Peak | 3″ | 0.50″ | 0.34″ |
| Snow Bowl Top | 4″ | 0.53″ | 0.25″ |
| Lincoln Base | – | 0.76″ | 0.16″ |
| Snow Bowl | 3″ | 0.53″ | 0.29″ |
| Mt Ellen Base | – | 0.79″ | 0.18″ |
Elevation storm: height is everything in this one. Above 3,500 ft about 3″, 2,500–3,500 ft about 2″, 1,500–2,500 ft about 1″, below 1,500 ft under 1″. Snow starts sticking around 1,500 ft; below that it’s mostly rain.
For skiers: the peaks get 2–4″ of snow plus about 0.9″ of rain; the bases 0–3″ and 1.0″ of rain: more rain than snow down low, so expect it heavy and wet. Most of the snow is above about 2,800 ft: that’s where the skiing will be.
What falls: Mt Ellen Peak: freezing rain Sun 11 pm–Mon 1 am, then sleet Mon 2 am–Mon 4 am, then rain Mon 5 am–Mon 1 pm, then snow Mon 2 pm–Mon 7 pm. Mt Ellen Base: freezing rain Sun 11 pm–Mon 1 am, then sleet Mon 2 am–Mon 4 am, then freezing rain Mon 5 am–Mon 1 pm, then rain Mon 2 pm–Mon 7 pm.
Ice: freezing rain is the risk, worst at Mt Ellen Base (0.79″). Roads and trees glaze; skiing turns to boilerplate.
Cutter, Sun 11 pm–Mon 7 pm: The low passes west of us, so warm air rides in aloft: snow or ice up front, then rain, possibly to the summits, then a hard freeze. Cold air hangs on longest in the valleys east of the Greens, which is where freezing rain lasts. Models often run warm at the ground in these.
Weather Watch / Heavy rain + Cold front
Heavy rain Sun 11 pm to Mon 7 pm
Rain: up to 1.5″ of water over 3 days (0.9″ in Bristol, 1.2″ on Lincoln Peak). Brooks off the west slope rise fast; the New Haven River through Bristol and the Mad River through Waitsfield are the ones to watch.
Cold front: the base drops about 28°, from 36° Mon 3 pm to 8° by Tue 9 am. Expect a wind shift to the northwest and gusty wind behind it.