Ghost in the database

[Pierre Ducray]

Sometimes things go awry. Take the second May 26, 1995 Burkburnett (TX) F0 tornado...

Location: Wichita County, 3 W Burkburnett
Time: 1844 CST-1852 CST
Path Length: 2 miles
Path Width: 30 yards
Number of Persons Killed: 0
Number of Persons Injured: 0
Estimated Damage Property: -
Estimated Damage Crops: -

[Sources: NOAA (NCEI) Storm Data publication (May 1995) and 90-99_torn.csv file]

Now here is the same tornado in the NOAA (NCEI) Storm Events Database. This database "contains the records used to create the official NOAA Storm Data publication" so any discrepancy should be questioned.

Begin Date: 1995-05-26 18:44 CST
Begin Location: 3W Burkburnett
End Date: 1995-05-26 18:52 CST
Length: .1 Miles
Width: 10 Yards
Deaths (Direct/Indirect): 2/0
Injuries ( Direct/Indirect): 30/0
Property Damage: 0
Crop Damage: 0

The "Event Narrative" is the same in both sources.

So in this second source 2 miles become 2 deaths and 30 yards become 30 injured...
But whenever there are fatalities in the Storm Events Database (the second source) some of the data for these are shown in square brackets in the "Event Narrative". There are none concerning this event...

This second source is obviously the wrong one.

Databases are prone to errors and the NOAA does not hide this fact but some are more disconcerting than others...

 

[Pierre Ducray]

Ghost in the database (Part 2)

Sometimes things go awry... Consider the June 30, 1995 Smyrna (GA) F1 tornado...

Location: Cobb County, Smyrna
Time: 1838 EST
Path Length: short
Path Width: narrow
Number of Persons Killed: 0
Number of Persons Injured: 4
Estimated Damage Property: 250K
Estimated Damage Crops: 0

[Sources: NOAA (NCEI) Storm Data publication (June 1995) and 90-99_torn.csv file]

Now here is the same tornado in the NOAA (NCEI) Storm Events Database.
This database "contains the records used to create the official NOAA Storm Data publication" so any discrepancy should be questioned.

Begin Date: 1995-06-30 18:38 EST
Begin Location: 0 Smyrna
End Date: 1995-06-30 18:38 EST
Length: .1 Miles
Width: 10 Yards
Deaths (Direct/Indirect): 4/0
Injuries ( Direct/Indirect): 250/0
Property Damage: 0
Crop Damage: 0

The "Event Narrative" is the same in both sources: "A preliminary survey confirmed a small F1 tornado touched down for less than 30 seconds. Path width 30 yards...Length 100 yards . . . "

In this second source 4 injured become 4 deaths and 250K (Damage Property) become 250 injured...
Yet whenever there are fatalities in the Storm Events Database (the second source) some of the data for these are shown in square brackets in the "Event Narrative". There are none concerning this event...

The second source is definitely the wrong one but... in the first source some data about this tornado are not arranged as usual... so it is (pretty) easy to understand these errors...

Databases are prone to errors and the NOAA does not hide this fact but some are more disconcerting than others...

 

[Pierre Ducray]

Ghost in the database (Part 3)

Sometimes things go awry...

The Southeastern US Palm Sunday tornado outbreak of March 27, 1994: around 30 tornadoes, 22 fatalities in Alabama, 18 fatalities in Georgia...

The NOAA (NCEI) Storm Data monthly publication (March, April, May and June 1994), the NOAA (NCEI) Storm Events Database and the NOAA (NCEI) U.S. Tornadoes database say nothing about the tornadoes and fatalities in Georgia...

The reason seems simple: data about these March 1994 tornadoes and fatalities are in the... July 1994 Storm Data publication!

A late entry... unfortunately still distorting some statistics...


[Sources: National Weather Service, National Centers for Environmental Information, Storm Prediction Center.]

 

[Pierre Ducray]

Ghost in the database (Part 4)

Sometimes things go awry...

According to the NOAA (NCEI) U.S.Tornadoes database, February 1993 is the month of February with the highest number of fatalities (1991-2020 period): 81.

According to the NOAA (NCEI) Storm Events Database, the SPC Annual U.S. Killer Tornado Statistics and the NOAA (NCEI) Storm Data monthly publication (February, March, April, May and June 1993), the number of fatalities in February 1993 is not 81 but... 3! Not a small discrepancy...

How can 3 wrongly become 81?! It's an enigma...