SSCRAM Methodology
W&F Article Link

The Statistical Severe Convective Risk Assessment Model (SSCRAM) is a large dataset of environmental parameters that have been correlated with downstream severe weather reports. Ten full years of data (2006-2015) have now been used to develop and verify the system, providing over four-million grid-hours of information. This results in one of the largest thunderstorm environmental datasets available. A full description of the system has been documented in a Weather & Forecasting article.

SSCRAM provides answers to simple thunderstorm climatology questions. For instance, "What percentage of time does a thunderstorm occurring in MLCAPE over 4000 J/kg result in a severe weather report?" Below is the severe weather probability distribution function (PDF) for MLCAPE.

Fig 1. Severe weather PDF for MLCAPE.

As shown, the probability of a thunderstorm producing severe weather increases with higher MLCAPE values. For example, a thunderstorm occurring in an enviroment with MLCAPE of 500 J/kg will produce at least one severe report of any kind in the next 2-hours (white line), about 6% of the time. However, if the MLCAPE is 4000 J/kg, the potential of a downstream severe report increases to roughly 20%.

Using the same chart, it is noted that MLCAPE is a poor predictor of whether a thunderstorm might produce a downstream tornado. This is shown by the red line being nearly flat between 100 J/kg and 3500 J/kg. It is only at the highest MLCAPE values that show a result in an increase in tornado probabilities. However, these values are very rare, so using this concept in a forecast-sense would be of limited utility.

Seasonal and Regional Studies

Seasonal and regional studies can also be pursued. For instance, this chart shows the tornado PDF for the Significant Tornado Parameter (STP), but segregated roughly into the cool months [November through May], and the warm months [June through October].

Fig 2. Tornado PDFs for STP, segregated by time of year.

These charts show a remarkable difference in the probability of downstream tornadoes and significant tornadoes given the same STP value, depending on what time of year a thunderstorm occurs. In the winter and spring, STP is one of the most useful parameters available for forecasting significant tornadoes. Its utility is noted by the steeply sloped red line in the top graph. This shows a rapidly increasing probability of significant tornadoes with increasingly large values of STP. This is the hallmark of a useful forecast parameter.

In contrast, notice the nearly flat PDF for significant tornadoes (red line) in the bottom graph. This illustrates very little forecast value of STP during the summer and fall months. The slope of the line is nearly flat, meaning the probability of a significant tornado from a thunderstorm occurring in an environment characterized by an STP of 1.0 is roughly the same as if the STP were 6.0.

W&F Article Link

This complex forecast problem has been described in detail in a Weather & Forecasting article.


SSCRAM has been designed very carefully to produce the most appropriate correlations between environments and downstream severe weather reports available. These correlations will likely be the buiding-blocks for many future statistical studies and models, so effort was made to produce a very comprehensive database. This image attempts to graphically describe this process.

Fig 3. SSCRAM Design Description.

The first step is to determine the lat/lon position of every cloud-to-ground lightning strike in the United States, and assign a 40-km gridbox. This step results in roughly 700,000 such gridboxes per year.

Next, enviromental data is gathered for each lightning-containing gridbox, for every hour since January 1, 2006. This is done by querying the SPC environmental database that has been built by Andy Dean over the past many years. The environment data potentially includes every parameter computed by the SPC Mesoscale Analysis system. The date, time, and location of the gridbox is also archived.

A geographic filter is then applied to the dataset, eliminating all gridpoints that occur outside of the contiguous United States. Since there are no severe reports received for offshore locations, or from Canada/Mexico, these gridboxes are removed. This step prunes the database to about 420,000 gridpoints per year.

The final step is to correlate each remaining gridpoint with downstream severe weather occurrences. To do this, the Bunkers right-moving supercell storm motion estimate was chosen as the best predictor of storm motion. This technique has its limitations, and may not be appropriate in all cases. However, several alternative methods were tested and the Bunkers method appeared to provide the best results for the widest array of storm-types.

For each gridpoint, downstream locations up to 2-hours are computed using the Bunkers estimate. Next, these downstream positions and times are searched for severe weather occurrences within 40km. The result is a count of the number of severe weather reports received in a downstream path from the original gridpoint.

Input Data

There are several very large datasets that have been combined to produce SSCRAM. These include:

  • The SPC Mesoscale Analysis system This hourly dataset provides a best-guess of real-time environmental conditions for a 40-km grid over the contiguous United States. Roughly 40 variables have been extracted for each gridpoint.
  • NLDN Lightning Strike Data This archive is used to determine the time and location of every CG lightning strike over the contiguous United States.
  • SPC Severe Weather Database All severe weather reports across the United States have been compiled since 1950.


As with any large and comprehensive study, there are inherant limitations to the system. First off, SSCRAM correlates the occurrences of downstream severe weather reports in the next 2-hours, to the environmental condition at hour zero. There are situations when rapidly changing conditions or sharp gradients of parameters will cause downstream weather conditions to be very different than the original gridpoint. SSCRAM does not attempt to anticipate these changes. Instead, it is hoped that the very large nature of the dataset will minimize the affects of these situations.

Another significant limitation to SSCRAM is that it uses the SPC Severe Weather Database. This is the highest-quality archive of severe weather reports in existence. But, it is well-known to occasionally include dubious and erroneous information. Once again, the sheer size of the dataset will hopefully minimize the impact of these bad data points.

SSCRAM extracts the environmental parameters for each lightning-containing gridbox using the SPC Mesoscale Analysis system. This data is primarily based on the RAP and RUC model analyses for each hour. While it is believed that these analyses capture a very good representation of the real-time atmosphere, it is understood that there are instances where there are differences.

The Bunkers right-moving supercell storm motion estimate is used to determine the downstream path to search for severe weather reports. This method works quite well, and was chosen after comparing output from several motion-estimating methods. However, it is primarily designed for discrete supercell situations and may not work well for organized mesoscale convective systems.

Due to the number of elements and gridpoints in the database, it is unrealistic to assume that every point was checked for accuracy and completeness. There is some potential of occasional missing hours of data or anomalous lightning strikes being counted as real thunderstorms. These errors appear to be quite small given the remarkable reliability of the dataset, as shown in the verification statistics.