A Great Source for Northwest Smoke Forecasts

When I want to get the latest regional smoke forecasts, I turn to the modeling site run by the Laboratory for Atmospheric Research (LAR) at Washington State University, found here.  As you will see, they have created a major technological achievement combining atmospheric and air chemistry models for prediction of air pollution over the Northwest. The system is called AIRPACT-4 and is run by Professor Brian Lamb and Dr. Joseph Vaughn.

Let me show you a sample.  Below is the small particle forecast (PM2.5, particles smaller than 2.5 microns--millionths of an meter) for Monday at 1 PM.  The locations of fires are shown (red dots)...you can see the plumes of small particles produced by them (colors).  The worst conditions are over and downstream of the fires of NE Washington.  Another plume is coming off of the fire near Mt. Adams and heading towards Yakima.  I travelled to Yakima on Monday and let me assure you...this plume was there.  Some residual smoke over Puget Sound, but not as serious. Huge amount of smoke over NE Idaho and western Montana.

 Wednesday (today) looks better over Puget Sound, but still bad over NE Washington.  Not great near Yakima.

 

How does this system work?   They start with the atmospheric forecasts made by the high-resolution UW WRF modeling system, the one I often show on this blog.   They get information about sources of air pollutants (like fires, natural sources, industrial sources, transportation sources, and more), also known as emissions.  Specifically, smoke emissions information for wildfires comes from the US Forest Service's BlueSky system.  They feed this information into a sophisticated air chemistry model developed by the EPA called the Community Model for Air Quality (CMAQ) that calculates air quality for a region by treating the domain as a three-dimensional grid of cells of regular size.  This model then forecasts particle concentrations, as well as the amounts of various gases and pollutants (e.g., ozone, sulfur dioxide and more).  Pretty neat stuff.

Remember the smoke coming in to Puget Sound on Saturday?  The AIRPACT system nailed it. Lets look at the forecast from late Thursday.   At 3 AM Friday, the smoke was heading toward Spokane.

But as a thermal low developed with easterly flow, the smoke direction reversed, as shown by the forecast for 8AM Saturday, with western Washington getting into the murk.    Very nice forecast.

No forecast system is perfect, particularly such a complex forecast system that requires good emissions information, a skillful atmospheric model, and a realistic chemistry model.  The AIRPACT system is an impressive advance and the folks at Washington State deserve kudos for what they have built and run operationally.   Even considering they are Cougars.

Was the Fatal Wildfire near Twisp, Washington Predictable?

I have hesitated to write this blog, considering the sensitivity of the issue:  three brave wildland firefighers died trying to save lives and property.  But fires are still burning and others have their lives on the line, so I feel impelled to write this.

Many wildland firefighter deaths have resulted from inattention to meteorological data and forecasts. A tragic example is the Yarnell, Arizona fire in which radar showed developing thunderstorms that led to strong outflow winds that produced tragic results.   The firefighters should have been pulled out, something borne out by subsequent analysis (see my previous blog on this).

As I will describe below, I believe that the meteorological guidance for the Twisp fire situation was clear:  all firefighters should have been warned/pulled back because of the imminent threat of rapidly increasing winds accompanied by a change of direction.   I believe media claims of "unpredictable" winds are unfounded.

The locations of the tragedy was near the intersection of Twisp River and Wood Canyon Roads. More context is derived from a google map....I indicated the location.  The incident occurred west of Twisp, south of Winthrop, and east of Stehekin, found at the northern portion of Lake Chelan.

The time of the incident is not clear in the press, but initial reports of the accident appeared between 5 and 6 PM on Twitter.  I assume the timing of the incident is between 3 and 4:30 PM.  If someone has better timing I would like to know it.

Seattle weather radar shows evidence of fire blow up in the correct location at 22:54 UTC (3:54 PM).  You will also see the echo at 5:12 PM.  It is subtle...but look due east of the northern tip of Lake Chelan.  County boundaries are on the radar image...note how the fire location is found at the western location.

The fire is also clear in the visible satellite imagery, as is the radical shift of the winds and the blow-ups of this and other fires.  Here is the visible image at 8:30 AM that day.   The smoke is moving southward, pushed by the low level winds. No evidence of a major fire in western Okanagan County where the incident will happen.

At 12:30 PM, the general winds are starting to shift to the west, and not much is found over the future incident site.

At 3:30 PM, things are happening.   Fires have developed near the crest of the Cascades with a clear drift to the east of the smoke (strong westerly winds).  You can see the developing fire over eastern Okanagan County

 At 5:30 PM, the fires have revved up, particularly the one of interest.   The winds are clearly westerly.

Weather observations on the ground showed the transition to rapidly increased winds from the west and northwest.  Here is a trace of temperature, dew point, wind speed (sustained and gusts) and wind direction near Stehekin, upstream of the incident location.   Between 12 and 3 PM the winds surged from very light (less than 5 mph) to over 20 mph, as the winds shifted to just west of north.

Winthrop showed a big acceleration and windshift around 3 PM

I could show you more of these, but you get the idea....if anyone was monitoring the weather observations they would have seen a major wind transition (wind shift and wind increase) moving across the Cascades between noon and 3 PM.

But there was a lot more warning than that.   High-resolution computer models, such as the UW WRF model (which IS used by the fire and AQ communities), clearly showed the transition.   Let me prove this to you.  Here are the forecasts from a run initialized 5 PM THE DAY BEFORE.   Available many hours before the incident. These maps show surface sustained wind speed (gusts are higher) and wind direction (little barbs).   Lake Chelan, Winthrop, and the county boundaries are shown.

First, the 18-h forecast for 11 AM Wednesday.  Light winds (blue colors) generally and over the accident site.

The forecast for 2 PM shows winds increasing, including over the accident site, but stronger to the west,

 The 5 PM forecast is very threatening.  Sustained winds around 20 knots around the accident location, with a clear wind shift to westerly.

Thus, wildfire folks should have been prewarned of a major wind shift and acceleration and probably should have been pulled back that afternoon.

The cause of the wind shift was a westward moving disturbance.  Here are the winds above Vancouver Airpot that day (time in UTC increasing to the left.  19/18 is 11 AM).  There was a shift to NW winds aloft as the disturbance went past).

Today there is extraordinarily large amount of weather information available for giving real-time guidance and accurate short-term forecasts for guiding and protecting those fighting wildfires.  Rarely are there surprises anymore, except with thunderstorm initiation (and that wasn't occurring in this event).  

I believe the windshift/wind acceleration that occurred Wed. afternoon was entirely predictable.  It was not random, it was not extreme or a new normal.  My profession needs to work with the fire community to ensure such tragedies are prevented.