1 A Fingerprinting Technique for Major Weather Events By Ben Root 1, Paul Knight 1, George Young 1, Steven Greybush 1 and Richard Grumm 2, Ron Holmes 2 and Jeremy Ross 3 1 Pennsylvania State University 2 The National Weather Service 3 ZedX, Inc.
2 Introduction Climatic anomalies have distinct patterns for different weather events. Combining these patterns together form “fingerprints” for different event types. Fingerprints can be used to objectively recognize a weather event in a model or analysis.
3 Premise Major weather events have a fingerprint. –Unique configuration of atmospheric anomaly fields. This fingerprint can be objectively “learned” in a region. Event types in Middle Atlantic Region: –Floods (flash and river flooding) –Heavy snow –Ice –Synoptic high winds –Fires –Warm episodes –Cold episodes –Hail –Tornadoes –Severe thunderstorms –Fog
4 How to Make a Fingerprint 1.Compile an Event List for an event type Details when events occurred 2.Analyze several anomaly fields at the time of each event. Use basic peak and valley picking method. Record the peaks' and valleys' Standard Anomaly, Latitude, and Longitude. 3.Enhance event type’s signal through use of “Strong-Point Analysis” spatial clustering.
5 V-Wind at 250 mb: Peak 1 Portions of a Fingerprint Snow
6 Heights at 850 mb: Valley 1 Portions of a Fingerprint Snow
7 Measuring the Fingerprint Spatial Component –Tighter cluster Larger value –How significant is position of climate extrema to the event type’s fingerprint? –Calculated for each set of extrema (4) for each atmospheric field.
8 Standard Anomaly Signal (Φ) Standard anomaly component –Smaller variance Larger value –Farther from average Larger value –How significant is standard anomaly of climate extrema to the event type’s fingerprint? Measuring the Fingerprint
9 Recognizing an Event A Quick Case Study of the Blizzard of 1996 Analyze atmospheric anomaly fields derived from NCEP Reanalysis Data for 00Z, January 8, –Examples: Mean Sea Level Pressure Temperature at 850 mb U and V wind Component at 850 mb Precipitable Water Specific Humidity at 925 mb 850 mb Heights
10 Recognizing an Event Recognizing an Event A Quick Case Study of the Blizzard of 1996
11 Standard Anomaly Maps Recognizing an Event A Quick Case Study of the Blizzard of 1996
12 Measuring an Event ( ) Spatial Component –Closer to cluster body Larger values –Uses Barnes' Distance Equation ( ) Standard Anomaly Component –Closer to distribution Larger values
13 Event Scores Blizzard of 1996's Event Score: 7.81
14 Conclusions Developed comprehensive major weather event database for Middle Atlantic domain. Formulated a pattern recognition learning process implementable anywhere. Established an event score climatology.
15 Further Research Employ the NARR dataset to gain higher temporal and spatial resolutions. Improve event database. (QA) Interface the AI to AVN and MM5 for operational event recognition.