These dates are already a week in the past, so this is a reconstruction rather than a forecast — the air at the house can be looked up rather than predicted. Everything in the evidence points to a hazy-but-breathable weekend: the two big fires east and southeast of the house were 84% and 90% contained, the monitor nine miles away reads Good today under showers, and this calendar window has never once gone above 100 in the four years on record.
The realistic bad case was a smoky morning or two drifting up from the Austin fire eighteen miles south — which is on the same side of the mountain as the house and was only 19% contained — not a trip-ending smoke wall.
If we ran this weekend 100 times, here is how they would split. Hover a row to pick it out of the dots.
| 5 in 1053% | Clear weekend AQI 0-50 Mt Hood visible from the driveway, normal blue-white sky, maybe a milky ring near the horizon at midday; nobody comments on the air. | nothing |
| 3 in 1032% | Hazy but fine AQI 50-100 The ridgelines up the Salmon River go flat and gray, sunsets turn orange, you can faintly smell campfire at 7am but it burns off by lunch. Kids and hikers are fine; one person with asthma notices. | Weak overnight drainage pools a little smoke from the Austin and Narrows fires in the Sandy River valley, and the morning inversion holds it until mid-morning mixing breaks it up. |
| 1 in 1010% | One smoky day AQI 100-150 You wake up to a brown horizon and cannot see the far side of the valley; the smell is obvious indoors. One day of the three is an indoor day — cancel the hike, keep windows shut, the other two days are normal. | The Austin fire (19% contained, 18 mi south, same side of the crest) makes a run north up the Clackamas or Salmon drainages, and a stagnant night traps the plume in the valley instead of venting it. |
| 0 in 105% | Real smoke episode AQI 150+ Orange daylight, ash on car hoods, the sun is a red disc at noon, sore throats indoors. Nobody goes outside; the group leaves early. | A sustained offshore/east gradient plus a new run on Austin — or a new ignition on the west slope that the pack does not show — stacks two sources on the house at once with no marine flush. |
Fires near the house, drawn at their real shape from the federal perimeter data, over USGS terrain. Dashed rings are distance from the house; the shaded wedge is the only direction an east wind could carry smoke here from. Point at a fire for its details.
Solid dots are fires too small to draw at this scale but whose size is known. A dashed circle would mean a fire that has been located but has no published perimeter - size unknown, not small.
Beyond this frame, and too far to threaten the house but close enough to add regional haze: Bench (67,071 acres, 45.5 mi SSE), Ten Mile (4,834 acres, 56.7 mi SE), High Lava (2,894 acres, 37.0 mi N), Beachcomb (2,346 acres, 49.5 mi SSE).
An outline is only as current as the last mapping flight, which is not the same as when this page refreshed. Outlines flown — Grasshopper Sep 2 · Austin Sep 1 · The Narrows Sep 1. The oldest is Austin, 3 days behind. Acreage and containment are refreshed far more often than the outlines are re-flown — those figures were last updated Sep 4, 4:31 AM, so a fire may already be larger than the shape drawn for it.
Terrain: USGS 3DEP (public domain). Roads and water: © OpenStreetMap contributors (ODbL). Fire perimeters: NIFC.
Want to check it yourself, or see the rest of the state? The Oregonian's wildfire tracker is live and updates every 15 minutes.
Long-range wind signal unavailable this run.
It is not one cut-off. Five sources arrive on five different dates, and the darker the strip, the more of them cover your weekend. Hover any day for the count.
= Since 2026-08-10 the forecast has moved steady (+3 on the likely worst day) across 25 daily updates. Shading above shows source coverage, not this movement - hover a past day to see what we said that morning.
| now | Fire size & containment how big the fires are and whether they are being held |
| now | Long-range ensembles (GEFS 0.5 + GEM) wind direction, 52 members · reaches 32 days out |
| now | ECMWF ensemble (51 members) wind direction, the sharpest ensemble · reaches 13 days out |
| now | Deterministic weather (16-day) day-by-day wind and temperature · reaches 15 days out |
| now | Air-quality / smoke model (CAMS) actual AQI at the house · reaches 6 days out |
The two large fires — Grasshopper (94,000 acres, 84% contained, 17 miles east) and Bench (67,000 acres, 90% contained, 46 miles SSE) — sit on the far side of the Cascade crest and need sustained easterly flow to reach the house, and by late August their perimeters were mopped up enough that an east wind would carry interior smolder rather than an active plume. The ensemble field for these trip dates is null, so there is no departure to read against climatology; I fall back on the 31% climatological offshore rate for this window, and that is a rate of wind direction, not a chance of smoke — offshore flow only matters if something east of the crest is still producing, and at 84-90% containment those sources are weak. The real mechanism at this address is not east wind at all: The Narrows (14 miles SSW at 45.14N/-122.09W) and Austin (18 miles south, 12,598 acres, 19% contained) are both on the house's own side of the crest in the Clackamas and Salmon drainages, so the evidence pack's terrain note — that Mt Hood sits between the house and the fires — is true only for Grasshopper and AVERY, and the standard way Welches gets a bad morning is nocturnal drainage pooling smoke in the Sandy valley under an inversion, no synoptic reversal required.
A normal late August here is Clear - median AQI 16 across the last 4 years. On these exact dates: 2025: 20 · 2024: 13 · 2023: 14 · 2022: 14.
We are forecasting a worst day of 45, 29 points above that base rate. That departure rests entirely on the reasoning above; judge it for yourself.
| 13.9 mi | The Narrows SSW | 302 acres · 27% contained |
| 17.4 mi | Grasshopper E | 93,993 acres · 84% contained |
| 18.3 mi | Austin S | 12,598 acres · 19% contained |
| 37.0 mi | High Lava N | 2,894 acres · 44% contained |
| 45.5 mi | Bench SSE | 67,071 acres · 90% contained |
| 49.5 mi | Beachcomb SSE | 2,346 acres · 100% contained |
| 49.6 mi | Avery ENE | 465 acres |
If the weekend is clean you carried masks for nothing. If it is not, this is the difference between a bad weekend and a ruined one.
| Sandy | 19.4 mi |
| Brightwood | 15.5 mi |
| Welches | 14.8 mi |
Distance from each point on the drive in to the nearest fire. TripCheck is the authority on closures.
Ranked by how bad these dates have historically been at each place. The coast is not just cleaner today — it sits upwind of everything burning, so it only smokes during an east-wind reversal, which is the same event that would smoke out the house.
| Place | Bad day / typical | Now | Official forecast | Next 5 days | Nearest fire |
| Newport coast | 20 / 16 | 9 Clear | Good | ▲ worsening | 101 mi · Austin |
| Cannon Beach coast | 23 / 20 | 11 Clear | — | ▲ worsening | 88 mi · High Lava |
| Manzanita coast | 23 / 21 | 11 Clear | — | ▲ worsening | 87 mi · High Lava |
| Pacific City coast | 23 / 21 | 11 Clear | — | ▲ worsening | 91 mi · The Narrows |
| Welches / Mt Hood booked | 25 / 18 | 12 Clear | Good | – steady | 14 mi · The Narrows |
| Hood River east of crest | 28 / 18 | 23 Clear | Good | ▲ worsening | 21 mi · Grasshopper |
| Astoria coast | 29 / 26 | 14 Clear | Good | ▲ worsening | 84 mi · High Lava |
| Long Beach, WA coast | 30 / 26 | 14 Clear | Good | ▲ worsening | 97 mi · High Lava |
| Silver Falls valley | 42 / 31 | 16 Clear | — | ▲ worsening | 31 mi · Austin |
| Portland valley | 46 / 37 | 33 Clear | Good | – steady | 36 mi · High Lava |
| McMinnville valley | 48 / 34 | 19 Clear | Good | ▲ worsening | 54 mi · The Narrows |
| Bend east of crest | 90 / 53 | 16 Clear | — | – steady | 35 mi · Crosswhite |
| Field | Period it covers | What it is |
| Air right now | 09/04/26 4:00 PDT | Nearest EPA/AirNow ground monitor to each place, within 45 miles airnow.gov |
| Official forecast | issued 09/04/26 by Washington Department of Ecology | Next published outlook for that reporting area — a human forecaster's call Oregon DEQ via AirNow |
| Next 5 days | today's model run, covering 2026-09-08 | Direction only. Copernicus CAMS global air-quality model open-meteo.com |
| Bad day / typical | Aug 26 – Sep 1 in 2022, 2023, 2024, 2025 (28 days each) | CAMS reanalysis for those exact dates, then converted to the monitor scale open-meteo.com |
| Scale conversion | median of 14 daily model-vs-monitor pairs, 2026-08-16 to 2026-08-29 | Median of daily model-vs-monitor ratios — the bias differs by place, from 0.362 on the coast to 0.799 inland AirNow daily archive |
| Nearest fire | queried Sep 4 5:00 AM | Perimeter polygons of fires over 100 acres, distance to the burned edge NIFC WFIGS |
Every AQI figure above is on the ground-monitor scale — the same thing your phone shows. n/a means the model and the monitors disagree so sharply at that spot that we cannot honestly place it on that scale (steep terrain inside a coarse model grid cell — the Gorge is the usual culprit); those rows are sorted by today’s reading instead.
Fire perimeters from NIFC, evacuation zones from Oregon OEM, weather and alerts from NWS, air quality from AirNow (Multorpor, 8.7 mi away). The forecast is judged by Claude Opus 5 on the live data above, sampled 5 times and pooled conservatively.
It is a judgement, never scored against an outcome. Nobody forecasts smoke this far out: NOAA's smoke model runs 48 hours and the EPA's air-quality forecast is for tomorrow. Two-to-six weeks ahead is what meteorologists call the predictability desert. Anyone giving you a number for the trip right now - us included - is giving you a base rate with a story attached.
How thin the history is. The air-quality record for this spot begins in 2022, so "normal" here is built from just 4 years (2022, 2023, 2024, 2025). In those 28 sampled days the worst single day was AQI 29, and none reached 100.
That record has never seen a bad Oregon smoke year — not 2020, when Portland sat above AQI 500 for days, nor 2017, 2018 or 2021. So the base rate below is drawn entirely from calm years, and the upper end of our range is an estimate rather than something history has shown us. It is the main reason not to read the good news too confidently.
Are these forecasts any good? Across 100 scored forecast-days, the air-quality model beat both simple baselines at 2 of 5 lead times. Where it does not, this page should not be trusted to out-guess them either.
AQI here means the highest 24-hour PM2.5 value touching a day, at the house, on the local monitor scale - the same thing your phone shows. The model's own restatement of the target: “I am estimating the US AQI for PM2.5 as a 24-hour average — the highest such 24-hour value touching any of Aug 28, 29 or 30, 2026 at the Welches house — reported on the local monitor scale that a phone app displays, with the pack's 0.259 model-to-monitor calibration already applied and no further bias correction of my own. 'Worst day of trip' is the maximum across the three days, not their average, and the per-day figures are each day's own 24-hour maximum. These dates are in the past, so this is a reconstruction of a realized quantity, not a forecast of an unrealized one.”
Updated daily at 5am. A convenience, not an authority -
if officials issue an evacuation order, follow it.