A Climatology and Comparison of Parameters for Significant Tornado

Events in the United StatesWritten by Jeremy S. Grams, R. L.

Thompson, D. V. Snively, J. A. Prentice, G.M. Hodges and L. J.

Reames

Presented by Danielle Thorne and Matt Muscato

This paper had two main goals:1. To provide a national, several-year-long

climatology of convective mode for significant tornado events

2. To compare convective mode to mandatory-level kinematic-thermodynamic variables and their associated sounding-derived measures to calibrate forecaster observations and preferences when attempting to forecast significant tornado events.

Goals and Objectives

Severe weather forecasting has grown considerably since the formation of the Severe Local Storms (SELS) Center of the U.S. Weather Bureau in the early 1950s.

We now have the Storm Prediction Center (SPC) which is the main center for forecasting for severe weather.

All the field data and numerical models have resulted in a larger knowledge base of thunderstorms and tornadoes.

This has led to a more ingredient-based approach in forecasting severe storms and tornadoes.

Recognizing the convective mode based on the radar reflectivity is an important consideration in forecasting for severe storms as well.

Introduction

2 data bases were used as the data set derived from the Storm Prediction Center (SPC).

First database: 9 years of EF2 or greater tornado reports starting in January 2000 to December 2008.

Second database: 6 year database for all 2 inch or greater hail and 65-kt or greater wind reports starting in January 2003 to December 2008.

The tornado dataset was larger due to the infrequency of tornado reports compared to hail and wind reports.

Data and Methods

Each significant tornado report was given a convective mode based on its reflectivity image determined at the beginning time of the report.

The three convective mode classifications were:◦ Discrete cell◦ Quasi-Linear Convective Systems (QLCS)◦ Cluster

Event Classification

Discrete cell – relatively isolated cell with a circular or elliptically shaped region of reflectivity with maximum values greater than or equal to 50 dBZ.

Event Classification

Figure 1a. Example of a discrete cell (highlighted by an oval) from the regional reflectivity mosaic image.

QLCS – a continuous major axis of at least 40-dBZ echoes with length greater than or equal to 100 km that shared a common leading edge and moved in tandem; additionally, the major axis had to be at least 3 times as long as the minor axis.

Event Classification

Figure 1b. Example of a QLCS (highlighted by a polygon) from the regional reflectivity mosaic image.

Cluster – reserved for conglomerates of several cells that were not clearly identifiable as either discrete cell or QLCS in regional radar reflectivity mosaics, typically consisting of at least a contiguous region of 40-dBZ echoes in a 2500 km2 area.

Event Classification

Figure 1c. Example of a cluster (highlighted by an oval) from the regional reflectivity mosaic image.

In order to condense the data set, they catalogued the tornado with the most significant damage for that convective day (1200-1200 UTC) for the three convective modes.

If multiple tornadoes had the same damage rating the earliest tornado was used.

Tornado Collection

The only way to obtain mandatory pressure-level observations is through the radiosonde data. For this type of collection, two observations a day are not sufficient enough. They would need to interpolate the values for the specific hour of the thunderstorm’s occurrence.

Thus, they went to the Rapid Update Cycle (RUC) hourly model.

The RUC model was not available before May 2002, so they had to interpolate from the station plots and radiosonde data for before May 2002.

The RUC model was compared and tested to the interpolated mandatory pressure-level data from the radiosondes and Table 1 shows the difference between the two.

Environmental Data

Table 1: Difference between interpolated mandatory pressure-level data and RUC hourly model analysis.

There were a total of 1072 individual significant tornadoes collected. This was reduced to 448 significant tornado events.

The figure below breaks down the number of tornadoes by convective mode.

Climatology of Convective Mode

Each event was assigned into geographical and seasonal categories, with at least 24 events needed for a given region and season to be compared by the specific parameters later.

Figure 3 illustrates the number of significant tornadoes in each state and the geographical regions are highlighted.

Climatology of Convective Mode

Each mode was arranged by season, then individual modes were sorted by region and subgrouping them by season.

Climatology of Convective Mode

Climatology of Convective ModeThe tornado events were catalogued into 3 hour periods starting at 13 UTC to 12 UTC for the warm and cool seasons.Discrete cells dominate the warm season, specifically peaking at the 22-00 UTC hours. The QLCS tornado events were more evenly distributed during the cold season, with a relative maxima near 0600 UTC and 1500 UTC and an absolute minimum around 2100 UTC. The minimum in QLCS events near the peak of the diurnal heating cycle suggests that cool season QLCS tornado events are more synoptically driven.

During the 6 year period from January 2003 through December 2008, 355 significant hail and 556 significant wind events were collected.

They were further subdivided by region and seasons, with the requirement of 24 or more tornado, hail, and wind events. These requirements were only met in 3 regions: Southeast spring (SE SPR), Southern Plain spring (SP SPR), Northern Plain summer (NP SUM).

Climatology of Hail and Wind Events

500-hPa wind direction for significant tornado events was clustered around 230° for all regions and seasons.

In comparison, the 850-hPa wind direction for significant tornado events had a more southerly flow (~200°) for each .

Wind Direction Results

Both the 500 and 850-hPa wind speed for significant severe events was much higher when tornadoes were reported compared to both hail and wind reports for all subgroups.

The Southeast spring wind speed for significant tornado were substantially greater than the significant hail and wind events. This suggests a more amplified synoptic pattern or stronger mean flow during tornado days.

Wind Speed Results

The 500-hPa height falls for significant tornado events were most prominent over the Southeast in winter, and smallest over the northern plains in the summer.

Overall, there is a small change in height change in the immediate vicinity of the significant severe events.

Height Change

The mean 500-hPa temperature for the significant tornado events was warmest in the northern plains summer (~-10°C).

The mean 700-hPa temperature for the significant tornado events was warmest in the northern plains summer as well (~9.5°C).

The mean 850-hPa dew point is also warmest in the northern plains summer (~15°C) for significant tornado events.

Temperature and Dew point

The mean increase in 850-hPa dew point of 3.3°C from 12 hours prior to tornado time was consistent across all regions and seasons.

At the 500-hPa level, the temperature did not change much throughout the 12 hour prior to tornado time, and only a slight warming at the 850-hPa level.

This suggests that local moisture plays a larger role than temperature in the low levels for conditioning the thermodynamic environment prior to significant tornado events.

Changes in lifted parcel moisture has approximately twice the impact on CAPE than temperature (Crook 1996).

Change in Temperature and Dew point

The distribution of surface temperatures vary greatly based on season for the significant tornado events.

The dew points illustrate less variation between the different regions and season, with an average of 66°F.

The low-level moisture can be sourced by evapotranspiration during the growing season, but the background synoptic regime is likely the largest contributor to it through horizontal advection from a warm ocean source (Gulf of Mexico).

Surface Temperature and Dew Point

A subgroup of 90 events having days with at least six significant tornadoes were sorted by the mandatory-level parameters.

The median of these 90 tornado events were consistent with Miller’s “strong” category for magnitude of low- and midtropospheric flow, low-level and surface dew point temperatures, and for surface pressure and 12 hour pressure falls.

The median 500-hPa 12 hour height changes fell into Miller’s moderate category. In comparison, 179 singular events fell into Miller’s “weak” category.

This implies that relatively small 500-hPa height falls suggest this proxy need not be large directly over the area for tornado development. This is consistent with the finding that weaker synoptic forcing favors discrete cell development and is a greater threat for tornadoes (Thompson and Edwards 2000).

Operational forecasters sometimes focus on these height falls as an important component to significant tornadoes, which could lead to missed tornado forecasts.

Miller Checklist Comparison

The 0-6 km bulk wind difference for severe events show stronger speeds for tornadoes compared to hail and wind for the Southeast spring, southern plains spring, and all regions combined.

The 0-1 km bulk wind difference for severe events have faster speeds for tornadoes compared to hail and wind for the southeast spring, northern plains summer, and all regions combined.

Bulk Wind Differences

The overall difference in the Mixed Layer CAPE appears to be modest between tornado, hail, and wind events.

A bigger difference is shown in the Mixed Layer CIN and LCL with statistically significant tendency for weaker CIN and lower LCL heights to accompany tornado events versus hail and wind in the northern plains summer and all regions combined.

Thermodynamic Parameters

Supercell Composite Parameter (SCP) along with the Significant Tornado Parameter (STP) both show that the severe tornado values were larger compared to hail and wind events, and these differences were statistically significant for all three regions and season combinations.

SCP vs. STP

The Heidke’s skill score (HSS) was used to assess the relative diagnostic accuracy of various parameters with convective mode to provide a correct forecast between tornadoes and hail, and tornadoes and wind events.

This test was used because it gives credit for a correct forecast of a nonevent.

The table suggests that composite parameters, kinematic parameters, and convective mode should weigh more predicting the type of significant severe events.

Skill Score Comparison

A sample of 448 significant tornado events (EF2+ damage) from January 2000 through December 2008.

Significant tornadoes occurred most frequently with discrete cells in the spring (southern plains) and summer (northern plains), while QLCS significant tornadoes were more evenly distributed throughout the year.

Discrete cell events displayed a clear peak near 0000 UTC and a minimum over night/morning hours.

The QLCS tornadoes had a single peak around 0000 UTC during the warm season, but during the cool season they had more evenly distributed occurrences at various times during the day. This makes it difficult to forecast tornadoes in the Southeast during the cool season.

Temperature changes aloft were rather small in the 12 hour period leading up to tornado events, but moistening of roughly 2°-4°C at 850-hPa was seen.

The 500-hPa height falls were typically only around 30 m in the immediate area of tornadoes.

The composite parameters, kinematic variables, and convective modes discriminate between the different severe events better than the thermodynamic variables.

Combining the low- and midlevel ground-relative wind speeds, composite parameters, and the type of convective mode should aid the forecaster significantly with the expected type of severe event.

Summary and Conclusion

Questions?