PNW Weather Science

Why weather models disagree at Mukilteo — and what it teaches us
Mukilteo sits at the intersection of three weather phenomena that confuse forecasting models: the Puget Sound Convergence Zone, the Olympic rain shadow, and the marine layer. Understanding why models disagree here is itself a lesson in meteorology.

The Puget Sound Convergence Zone

What It Is

When Pacific air flows east toward the Cascades, the Olympic Mountains split the airstream. One branch flows through the Strait of Juan de Fuca (north), the other through the Chehalis Gap (south). These branches reconverge over south Snohomish County — roughly over Mukilteo, Lynnwood, and Edmonds.

Where they meet, the air is forced upward, creating a narrow band of enhanced precipitation and cloudiness. This band can be as narrow as 10 miles, and it shifts north or south depending on wind direction.

Why models fail here: The PSCZ is a mesoscale phenomenon (~20 km wide) that sits right at the resolution limit of most global weather models. A model with 13 km grid spacing (like GFS) literally cannot resolve it. Higher-resolution models (HRRR at 3 km) do better, but the exact position of the convergence band is sensitive to initial conditions.

Observable Effects at Mukilteo

When the PSCZ is active, Mukilteo can be raining while Everett (10 miles north) and Seattle (25 miles south) are dry. Or the reverse. This is why KPAE (Paine Field) and KSEA (Sea-Tac) can report completely different conditions at the same time.

In our scoring data, Open-Meteo consistently reports "overcast" when KPAE says "clear." This may be the model placing the PSCZ cloud band over our coordinates when it's actually 5 miles north or south.

The Olympic Rain Shadow

The Driest Place in Western Washington

The Olympic Mountains wring moisture from Pacific air masses. By the time air descends on the lee (east) side, it's significantly drier. Sequim, on the north side of the Olympics, gets only 16 inches of rain per year — comparable to Los Angeles. Whidbey Island, partially in the rain shadow, gets 17-20 inches. Mukilteo gets 32-38 inches. The windward side of the Olympics gets 140-170 inches.

LocationAnnual RainfallRain Shadow Effect
Hoh Rainforest (windward Olympics)140-170 inNone (maximum orographic lift)
Seattle (Sea-Tac)37.5 inPartial
Mukilteo32-38 inModerate
Whidbey Island (Coupeville)17-20 inStrong
Sequim16 inMaximum

The Marine Layer

Fog, Stratus, and the Dewpoint Spread

Puget Sound water temperatures (8-12°C year-round) cool the air above the surface. When the air temperature drops close to the dewpoint, fog or low stratus forms. The dewpoint spread (temperature minus dewpoint) is the key metric.

This morning the spread went from 3.4°C at 6am (fog possible) to 12.2°C at 8am (very dry, clear). The marine layer burned off rapidly as the sun heated the land surface.

Our data shows this: At 4:30am (spread ~4°C) all sources agreed on "clear." By 8am (spread 12.2°C), conditions were unambiguously clear, but Open-Meteo still reported "overcast (WMO 3)." The model isn't reading the dewpoint spread correctly for this location.

Scoring Insights After 4 Samples

Leaderboard

SourceAvg Temp ErrorWind AccuracyCloud CoverNotes
Open-Meteo0.35°C (best)Poor (always reports wind)Wrong 4/4Best temp, worst clouds
NWS0.33°C3/4 perfect calmCorrect 4/4Best all-round
KSEA0.80°CInconsistentUsually correctDifferent microclimate

Open-Meteo's temperature accuracy (0.35°C average error) is remarkable for a free API, but its persistent "overcast" reports when the sky is clearly FEW/SCT suggests a systematic cloud cover bias for this latitude. NWS is the most reliable all-round source, especially for cloud cover and wind.

Why This Matters for Artemis II

Weather model accuracy at one location teaches us about model accuracy everywhere. The same NWS models that predict Mukilteo weather also predict launch weather at Kennedy Space Center (KXMR). Understanding where models fail — mesoscale convergence zones, marine layer timing, cloud cover thresholds — helps us assess confidence in the 80% GO weather probability for launch.

KSC has its own weather challenges: sea breeze convergence, anvil clouds from afternoon convection, and lightning restrictions. But the principle is the same: measure, compare, score, and let the data decide which source knows the place best.