AUSTIN (KXAN) – The accuracy of severe weather forecasts continue to improve thanks to innovative achievements from a team of weather researchers out of Colorado State University (CSU).

The team has spent the last few years working to develop and improve a new type of computer model. This model — called CSU-MLP, or Colorado State University-Machine Learning Probabilities — runs off of historical weather data and weather events from the past 9 years. This in return has led to improved confidence and accuracy of forecasts. Russ Schumacher, who is a professor at CSU and the State Climatologist, helped lead the team through the inception of this project.

What is so unique about this computer model?

This computer model is able to learn from the past, searching and accounting for past forecasting mistakes.

Ingesting this model with past weather events from historical archives and at the same time feeding the computer the old forecasts that preceded those events, essentially allows the computer to audit the event. This now basically provides this artificial intelligence with an innovative “hindsight” set of thinking often resulting in more accurate calculations of the future.

This model computes outcomes in terms of statistics and probabilities. It’s very important to note that it is not perfect, nor is it used verbatim by forecasters. Instead, it gives weather forecasters a supplemental tool to use in addition to their already long list of various weather models and sources.

How do our familiar forecasting models compare?

Regular forecasting models that many are familiar with such as the GFS and European model for example, are ingested with current (present) conditions of the atmosphere. These include temperature, humidity and wind speed/direction measured at all levels of the lower atmosphere by weather balloons that are released twice a day across the country. That data then gets ingested into supercomputers that compute possible paths or outcomes of storms in the future.

Why is this important?

Their research found CSU-MLP successfully and at times more accurately forecasts weather events as much as eight days in advance of severe weather.

Increasing accuracy and confidence of forecasts gives the public more time to prepare for an incoming severe weather event. The Storm Prediction Center (SPC), based in Norman, Oklahoma, are a prestigious team of experienced forecasters who work in conjunction with the dozens of NWS offices across the country. They take the lead and are responsible for issuing various types of severe weather Watches and Warnings. The SPC, given access to this innovative additional tool developed by CSU, has been found to help improve the lead times of these warnings. Increased advanced warning time, and accuracy in return ultimately saves thousands of lives across the country.

With increasing frequency of tornadoes in Central Texas from a shifting tornado alley (result of climate change), and Central Texas being located in Flash Flood alley and often dealing with hail storms, makes this advancement and improvement more important and relevant to us than ever before.