This tool is based on multiple regression models developed by Cohen et al., 2018:

Cohen, A.E., J.B. Cohen, R.L. Thompson, and B.T. Smith, 2018: Simulating Tornado Probability and Tornado Wind Speed Based on Statistical Models.

Cohen, A.E., J.B. Cohen, R.L. Thompson, and B.T. Smith, 2018: Simulating Tornado Probability and Tornado Wind Speed Based on Statistical Models.

*Wea. Forecasting*,**33**, 1099–1108.Enter the input parameters into the fields below to get simulated probabilities.

All radar-based entries are intended to be derived from lowest-elevation (0.5 degrees) WSR-88D base data.

ARL:

Height of simulated circulation above ground level (ft)

Height of simulated circulation above ground level (ft)

Max Inbound:

Maximum inbound velocity (kt)

Maximum inbound velocity (kt)

Max Outbound:

Maximum outbound velocity (kt)

Maximum outbound velocity (kt)

Dist:

Circulation diameter - distance between max inbound and max outbound (n mi)

Circulation diameter - distance between max inbound and max outbound (n mi)

STP:

Effective-layer significant tornado parameter

Effective-layer significant tornado parameter

CT:

0 = diffuse circulation

1 = clear circulation

0 = diffuse circulation

1 = clear circulation

TDS:

0 = no TDS

1 = yes TDS

** ONLY enter 1 if you are confident there is a TDS **

0 = no TDS

1 = yes TDS

** ONLY enter 1 if you are confident there is a TDS **

Comments:

Optional comment to include in archive. 20 characters max.

Optional comment to include in archive. 20 characters max.

What are the plots?

What do I enter in the text boxes above?

How is the output computed?

The plot on the left is derived from Figure 1 of Cohen et al. (2018), and the plot on the right is derived from Figure 2 of Cohen et al. (2018).

Left figure...

Box-and-whisker plots corresponding to the tornado probabilities simulated from an independent dataset of tornadic (left, n=498) and an independent dataset of nontornadic (right, n=499) events. Boxes correspond to interquartile ranges (25th–75th percentiles) with the embedded horizontal line markers indicating the medians (50th percentiles) of the distributions, and whiskers extending to the 10th and 90th percentiles of the distributions. The computed tornado probability for the entered data is plotted as a color-coded horizontal bar on the box-and-whiskers plots.

Right figure...

Box-and-whisker plots corresponding to tornado wind speeds simulated from an independent dataset of 498 tornado events. Boxes correspond to interquartile ranges (25th–75th percentiles) with the embedded horizontal line marks indicating the medians (50th percentiles) of the distributions, and whiskers extend to the 10th and 90th percentiles of the distributions. The computed tornado wind speed for the entered data is plotted as a color-coded horizontal bar on the box-and-whiskers plots.

Left figure...

Box-and-whisker plots corresponding to the tornado probabilities simulated from an independent dataset of tornadic (left, n=498) and an independent dataset of nontornadic (right, n=499) events. Boxes correspond to interquartile ranges (25th–75th percentiles) with the embedded horizontal line markers indicating the medians (50th percentiles) of the distributions, and whiskers extending to the 10th and 90th percentiles of the distributions. The computed tornado probability for the entered data is plotted as a color-coded horizontal bar on the box-and-whiskers plots.

Right figure...

Box-and-whisker plots corresponding to tornado wind speeds simulated from an independent dataset of 498 tornado events. Boxes correspond to interquartile ranges (25th–75th percentiles) with the embedded horizontal line marks indicating the medians (50th percentiles) of the distributions, and whiskers extend to the 10th and 90th percentiles of the distributions. The computed tornado wind speed for the entered data is plotted as a color-coded horizontal bar on the box-and-whiskers plots.

What do I enter in the text boxes above?

Use the text boxes to enter appropriate values determined by radar assessment. For the STP entry, use the
SPC Mesoanalysis to determine the STP value closest to the location of interest.

How is the output computed?

The tornado probability and tornado wind speed are computed using regression equations provided in Cohen et al. (2018).

For tornado probabilities, the percentile ranks are obtained by linearly interpolating between the 10th, 25th, 50th, 75th and 90th percentiles provided in the publication. The percentile ranks of the tornado probabilities among the tornado and non-tornado distributions can be used to determine statistics analogous to the Probability Of Detection (POD) and the False Alarm Ratio (FAR). For a computed tornado probability, there exists a percentile rank among the independent sample of tornado events, and there exists a percentile rank among the independent sample of non-tornado/null events. Using your computed tornado probability as a minimum threshold for detecting tornadoes — shown by the color-coded bar in the left plot — you would detect a certain proportion of tornadoes (analogous to POD; percent of tornado distribution above the color-coded bar) and you would falsely detect a certain proportion of nulls (analogous to FAR; percent of non-tornado/null distribution above the color-coded bar).

For tornado probabilities, the percentile ranks are obtained by linearly interpolating between the 10th, 25th, 50th, 75th and 90th percentiles provided in the publication. The percentile ranks of the tornado probabilities among the tornado and non-tornado distributions can be used to determine statistics analogous to the Probability Of Detection (POD) and the False Alarm Ratio (FAR). For a computed tornado probability, there exists a percentile rank among the independent sample of tornado events, and there exists a percentile rank among the independent sample of non-tornado/null events. Using your computed tornado probability as a minimum threshold for detecting tornadoes — shown by the color-coded bar in the left plot — you would detect a certain proportion of tornadoes (analogous to POD; percent of tornado distribution above the color-coded bar) and you would falsely detect a certain proportion of nulls (analogous to FAR; percent of non-tornado/null distribution above the color-coded bar).

The gray shaded box & whisker plots correspond to models tested on independent dataset.

For the tornado wind speed plot, darker gray vertical rectangles indicate wind speeds for each tornado rating.

For the tornado wind speed plot, darker gray vertical rectangles indicate wind speeds for each tornado rating.