Damaging Downburst Prediction and Detection Algorithm (DDPDA)

Ground Truth & Scoring Issues

As with most other algorithms, the ground truth database for damaging wind events is terrible (at best).

The ideal database for each event would probably include:

Start time
Intensity
Areal coverage
Duration

#1 and #2 are given in Storm Data, but the times are usually off and it is difficult to get an intensity from the report "large trees down." And if a large tree falls in a forest but no one was there to report to the NWS, was there really a downburst?

The database which we are developing will be based on the life cycles of individual cells, and include:

cells which had damage associated with them (as reported in Storm Data)
cells which passed over a wind sensor
cells which passed over a highly populated area
cells which were within 50 km of a radar and did or did not have an associated severe or moderate downburst signature

[Does a signature, even within 1 km of the surface, mean that there were adequate surface winds to qualify as "severe"?]

[Check for LWOS/airport data]

[Different types of truth can be separated out or combined.]

An example of a "truth file" is given below:

KMKX    071595 Cell#65 2200UTC_Yacht_club_damaged
Time    VS#     Az      Ran     Lat     Lon     RadWind SfcWind Damage Pop    Cat
204235 121     158     53      42.53   -88.31 4.0     -999    0       684    0
204735 122     156     52      42.54   -88.29 4.0     -999    0       684    0
205234 123     153     53      42.54   -88.26 4.0     -999    0       684    0
205734 124     152     52      42.55   -88.25 4.0     -999    0       684    0
210232 125     150     52      42.56   -88.23 6.5     -999    0       813    0
210732 126     147     52      42.58   -88.20 7.0     -999    0       813    0
211231 127     144     52      42.59   -88.18 10.5    -999    0       632    0
211730 128     142     53      42.59   -88.15 13.0    -999    0       1148   0
212230 129     138     56      42.59   -88.09 13.5    -999    0       1148   0
212729 130     136     59      42.59   -88.05 14.0    -999    0       1148   0
213228 131     132     62      42.59   -87.99 13.0    -999    0       372    0
213728 132     130     65      42.59   -87.94 18.5    -999    0       604    1
214227 133     128     68      42.59   -87.90 15.5    -999    0       10053 1
214726 134     126     70      42.60   -87.86 17.5    -999    1       10684 2
215225 135     125     72      42.59   -87.83 15.0    -999    1       10684 2
215724 136     124     74      42.59   -87.80 13.5    -999    1       10684 2
220224 137     123     76      42.59   -87.77 16.5    -999    1       10684 2

[In the example, the damage report corresponds well to the increase in population.]

Next page
Table of contents