A tornado near La Crosse, Kan., in May 2012. (Ian Livingston)

A high risk for severe thunderstorms has been issued for Thursday, May 18, in Kansas and Oklahoma. This post was originally published on May 17.

So far in 2017, the Storm Prediction Center has issued three “high risk” outlooks for severe weather. After low damage reports, two of those outlooks were criticized as questionable by the broader meteorological community — they seemed too extreme for the storms that ended up developing.

Technology has made forecasting things like tornadoes and derechos easier, but it’s still a difficult business. Forecasting severe thunderstorms in the Washington region may be even harder.

Patrick Marsh recently filled the position of Warning Coordination Meteorologist at the Storm Prediction Center. In addition to coordinating with individual National Weather Service offices, he works closely with media and the private sector to ensure that the forecasts are being interpreted correctly.

Marsh also engages with social scientists to make the center’s forecasts easier to understand and to maximize their reach.

We asked Marsh some questions about the rationale behind the Storm Prediction Center (SPC) outlooks and what challenges it faces when predicting severe weather in the Northeast. Below are some of his answers, edited for brevity and clarity.

The SPC outlooks were initially developed for other meteorologists. Now, thanks to the Internet, the general public has access, too. Is that a challenge, and how do you mitigate confusion?

One of the challenges we face is how to convey the scientific details of a forecast to a highly technical audience without obscuring the overall weather message for the public.

For example, from the 1950s through much of the 1990s, the only people who saw SPC severe weather outlooks were NWS meteorologists or others who subscribed to weather information dissemination systems. In this paradigm, as long as we were speaking to other meteorologists, once we defined our outlook terminology (e.g., slight, moderate, high risk) our severe weather outlooks could be easily interpreted.

With the advent of the Internet, SPC outlooks became available to anyone with a computer and the limitations of those terms because more apparent. As previously mentioned, one result of this is the addition of plain language summaries to our convective outlooks, mesoscale discussions, and severe thunderstorm and tornado watches.

After SPC went two full years with no “high risk” outlooks (the highest on the five-point scale) there have already been three in 2017, two of which were very small in size. High risks have ostensibly been reserved for huge outbreaks. How are you deciding whether a high risk should be issued?

The decision to issue a high risk is a gut-check moment for any SPC forecaster. The visibility and expectations of a high risk are well known and are taken very seriously. At the same time, the threshold to issue a rare forecast like this needs to be attainable and realistic.

A high risk is a forecast of at least 30 percent coverage of tornadoes and at least 10 percent coverage of significant tornadoes (EF-2 or stronger).

Traditionally, high risk forecasts were reserved for days like April 27, 2011, or April 14, 2012, when environmental conditions were conducive for tornadic supercells over a large area. In each of these cases, as in all cases, the decision to issue a high risk was thoroughly discussed among the SPC operations staff and coordinated with the affected local National Weather Service Forecast Offices.

As the science of tornado forecasting continues to improve, SPC forecasters are increasingly able to identify smaller corridors of heightened tornado potential further in advance. A consequence of how we define a high risk is when SPC forecasters identify a small area, it no longer takes 100+ tornadoes to achieve at least 30 percent coverage.

The lack of large numbers of tornadoes can lead to the perception of an over forecast, despite the forecast verifying statistically.

In the case of the most recent two high risks this year, SPC forecasters identified a small area likely to experience a concentration of tornadic supercells. In both cases, SPC coordinated the issuance of a high risk with the affected NWS Forecast Offices. Whether these two forecasts met objective or subjective measures of success is an ongoing discussion within and outside of the SPC, a discussion that will hopefully help refine the expectations for future high risks.

High risk issued in January 2017. (Storm Prediction Center)

Do you think there are any regions where it is more difficult to forecast severe weather than others?

In its simplest form, severe thunderstorm forecasting can be boiled down into identifying regions where the following four “ingredients” will come together: moisture, instability, lift and wind shear. How these ingredients come together, and in what quantities, determines the resulting type of severe weather.

Unfortunately, understanding and forecasting how these ingredients will come together is very challenging. And the primary challenges are different from region to region.

For example, across the Southeastern United States, proximity to the Gulf of Mexico means moisture is rarely missing. What is frequently a concern, however, is the number of thunderstorms that may develop. If too many thunderstorms develop, they will prevent each other from reaching maximum severity, resulting in a less severe threat. In the Great Plains, moisture availability can be uncertain, especially in spring. This frequently results in situations where the overall environment can produce significant thunderstorms, but tremendous uncertainty as to where thunderstorms will even develop.

Are there any specific challenges in the Mid-Atlantic or Northeast?

The Mid-Atlantic and Northeast region pose a wide array of unique challenges — some explicitly meteorological and others more societal in nature.

Meteorologically, severe weather across this region frequently occurs in the wake of previous days of thunderstorms. Small-scale features that are important to thunderstorm development and evolution are determined by these prior thunderstorms and are hard to forecast in advance.

Additionally, there are situations where thunderstorms may approach from the northwest, such as in the case of the 2012 derecho. The pattern favoring these scenarios are what meteorologists call “northwest flow” patterns. In these cases, there is tremendous uncertainty as to if, when and where thunderstorms will develop, whether they can grow into a large, organized thunderstorm complex, and if they can survive the trek across the Appalachian Mountains.

Given the overall population density of the region, impacts of an event will be amplified as compared to the Great Plains. A single hail-producing thunderstorm will have the potential to impact many times more people than in portions of western Kansas. Additionally, a large percentage of our nation’s aviation traffic occurs in this part of the country. Any thunderstorms in this area will have immediate and downstream impacts on air travel.

What do you see being the most critical problems to be solved?

As forecast models produce increasingly higher resolution output — and seemingly more realistic output — our ability to identify smaller areas of concern will increase. But there remains a fundamental gap between our conceptual models of how tornadoes, large hail and damaging winds form and the pre-storm environmental parameters we use to forecast those events.

For instance, even in the most conducive environments, not every thunderstorm produces a tornado. Sometimes seemingly favorable environments produce few if any tornadoes. It is unclear how or when improved model guidance, higher-resolution observational data, and/or improvements to our understanding of these phenomena will help us to close this gap.

This uncertainty and complexity in the severe weather forecast process is why SPC has been a leader in providing our forecasts in probabilistic formats. Although this is not without its own challenges. For example, what are the best ways to communicate our inherently probabilistic forecasts? This is a question all of us in the weather community will need to work together to solve.