Friday, June 30, 2017

Yarnell Hill Wildfire And NWS Forecast Support

Today is the fourth anniversary of the Yarnell Hill wildfire tragedy. I want to review a bit of the meteorology here and also consider special products issued by NWS Flagstaff.

Above is the 03:00 pm MST regional radar product from NCAR, the radar depicts a well-organized line of thunderstorms from east to north of the wildfire (which is producing the blue, smoke echo southeast of the line. The mesoscale outflow from this line of storms moved across the wildfire from about 04:18 to 04:30 pm causing a dramatic change in the fire's behavior, as would be expected.

Visible satellite images that show the storms and the fire location at 01:00 pm (top) and 03:30 pm bottom (red circle indicates fire location). By the time of the image below clouds from the thunderstorms had spread over the fire location - photos in the Accident Investigation clearly show the dark clouds associated with the storms - second below.

The NWS Flagstaff office handled weather support for this fire, which was at the southern edge of its County Warning Area (CWA). The spot forecasts were general, large-scale forecasts, similar to, but more fire-weather oriented, the general public forecasts issued by the NWS. However, when the line of thunderstorms developed, the Flagstaff Office issued two very specific warnings about the mesoscale outflow that was likely to impact directly the wildfire.

At 1402, FBAN receives a call with a weather update from the NWS office in Flagstaff. The NWS informs him of thunderstorms east of the fire that may produce wind gusts of 35 to 45 mph out of the northeast. FBAN relays the update to OPS1 and OPS2 via radio on state tactical frequency 1 (Tac 1).

At 1526, NWS-Flagstaff calls FBAN with a second weather updat
about expected thunderstorm outflow winds from the north-northeast 
with speeds between 40 and 50 mph. This update does not meet th
NWS criteria for a Red Flag Warning for this area. FBAN radios this 
second update to OPS1 and OPS2 on Tac 1.

The acronyms can be confusing, but FBAN refers to the on-site Fire Behavior Analyst - the other acroynms can be found at:

These two "warnings" phoned directly to the FBAN indicate that the outflow from the storms was likely to produce significant wind shifts and speed increases, which would dramatically impact the fire's behavior.

The "Red Flag" statement above seems, to me, to be a red herring, intended to deflect attention from the fire managers' actions after receiving the warning.

After several readings of the report, I can find no evidence that the onsite, fire managers changed their strategy to react better to the approaching outflow boundary. Possibly, a let's wait and see what happens was the approach used. Which is very sad, given the very accurate warnings issued by the Flagstaff NWS Office. 


  1. Anonymous7:55 AM

    The loss of life at the Yarnell fire was tragic...and should really put into question the whole policy of fighting fires. The monsoon would have put the fire out a few days later without loss of life but there is big money for the firefighters. Except in areas where lives/significant structures need to be protected I say let them burn.

    Several years ago one of our weather mod pilots was a copilot on a fire aircraft...was led up the wrong valley by a spotter plane...and crashed in a hillside in a desolate part of Utah. What a waste.

    I looked at the 12z U of A WRF model output for that day after I learned of the loss of life...sure enough it predicted the wind shift/gusty winds perfectly and would have been available around 11 am. Since the WRF data was on the web and not in the NWS AWIPS system I doubt whether it was seen by the forecasters at FLG. I also looked at the data from the nearest RAWS...two days before the storm that created the fire produced a similar wind shift.


  2. Pat is correct that UofA WRF data was not available on AWIPS. On the other hand, HRRR data was available. The HRRR had produced very good forecasts for the previous few days with regards to outflow boundaries. Consequently, forecasters at KFGZ had confidence in the model that day—and the HRRR showed another outflow moving into the lower elevations.

    Mesonet data, satellite imagery, and All-Tilt radar volume scans (for early detection of new convection aloft) were monitored by all forecasters on shift that afternoon. The combination of these data sets and models provided high confidence to forecasters—and resulted in the two calls mentioned in the original post. A earlier call had been made to Doce Fire Dispatch at 2040 UTC indicating the potential for strong and gusty winds from the Northeast at 25-35 mph...possibly higher.

    It is possible that forecasters gave them TOO MUCH lead time. Nothing happened immediately and fire managers and crew may have wondered if it was a false alarm.

  3. Pat and David - Thanks for your comments. Seems to me that the lead time for the two warnings were about perfect, since they gave the fire managers time to adjust strategies. Had the lead time been shorter managers could have said there was no time to react.