Number of Reported Fires Observed Fire Danger Observed Lightning Level Map Type Number of Large Fires Synoptic Wind Index Lapse Rates Does analysis of the historical record reveal relationships between lightning and wildfire frequency and/or size?
Lightning Ignition facts Roughly half of Oregon & Washington wildfires result from lightning. However, up to 75% of large wildfires are from lightning in some areas. …but only 7% of thunderstorm outbreaks result in large fires (3 to 4% west side). NWCC strives to identify those few critical lightning events.
Fine tuning the forecast Smaller 20 pt grid Archive daily 00Z and 12Z initialized GFS variables at each gridpoint
DAILY ARCHIVES OF INITIALZED GFS DATA 850, 700, 500mb heights and thickness 850-700mb and 700-500mb Lapse Rates 500 mb Temperature, Dew Pt & RH 700 mb Temperature, Dew Pt & RH Column Precipitable Water U and V components of 700 mb and 500 mb wind CAPE, LI, Totals, 700mb Theta-e Historical PSA lightning prob for each map type Perfect Prog method!
2000-2011 Logistic regression Best predictors occurrence & amount Map Type probability 700-500 mb Lapse Rt 500 mb Temperature Precipitable Water 700 mb Rel Hum quadratic/cubic bias
2011 PSA Probability and Amount Each PSA was assigned a value of probability and LL weighted from nearby grid boxes known to be significant 45% 25%30% 0%
Lightning Forecasting Phase 4 Logistic Regression applied to forecasting
Probability Forecasts of a Binary (Yes or No) Event At what confidence of probability percentage do I commit to saying an binary event is going to happen? 40%?50%? 60%?
POD Probability of Detection (how many cases correctly forecast?) FAR False Alarm Rate (how many cases forecast …but did not occur?)
As POD increases, so does FAR …or… As FAR decreases, so does POD
GOAL: Define the confidence percentage to maximize POD until FAR becomes unacceptable 40%?50%? 60%?
40%?50%? 60%? POD: FAR:
NWCC’s GOAL: Experiment with defining the (Y/N) confidence percentage until… POD rises to 67% …or… FAR falls to 49%...
Lightning Best predictors at each gridpoint are populated from the 00Z or 06Z GFS via a perfect prog type scheme to objectively forecast: – Lightning YES/NO – Mostly likely LL (percentile rank bin)
Large Fire Probability Significant fire probability during lightning events (for each PSA) is determined by: –The forecasted lightning amount (decile) –A pro-rated likely number of ignitions –The fraction of those ignitions that become historically have become “large” –Forecasted fluctuation in the Fire Danger
3 ignitions 5 ignitions 19 ignitions 21 ignitions 47 ignitions Historical Large Fire Probability From Multiple Ignitions
Historical Large Fire Probability From Multiple Ignitions LOW FIRE DANGER
Historical Large Fire Probability From Multiple Ignitions HIGH FIRE DANGER
5-Day Forecasted Fire Activity High Risk Event Forecasted ActivityTodayWedThuFriSat PSA C2 Large Fire Probability 0%1%52%1%10% 5-Day Total Forecasted Ignitions 14154832 6 to 10 Day Outlook based on Historical Fire Occurrence ClimatologySunMonTueWedThu PSA C2 Large Fire Probability 0% 1%2% 5-Day Total Climo Ignitions 1124614 Lightning Unstable Fire Activity Forecast 2012
Initial Attack Overwhelmed by: Sheer number of ignitions R U W Severity of atmosphere