Why is it so hard to forecast tornadoes?
Despite the awesome power of tornadoes, they are excruciatingly small in scale when compared to other phenomena that meteorologists must forecast. They are also difficult to detect.
Doppler weather radars are not typically able to resolve a tornado, and tornadoes cannot be seen using satellite or lightning data. This is why visual reports of tornadoes and their precursors from the public have historically been so critical, with volunteer networks in place in Canada since the 1980s (an early example of crowdsourcing).
Increasingly, forecasters have been able to predict the larger-scale conditions conducive to the development of tornadic storms with the help of ever-improving computer models, sometimes days in advance. While this will not allow a prediction of exactly where a tornado might form, it does allow for some lead time and preparedness, with the issuing of watches as necessary.
Many scientists continue to explore the question of what exactly makes a tornado form, and how a tornado might be detected better in real time. Tornado prediction is likely to improve due to this ongoing work.
NTP produces experimental tornado outlooks, mainly for the next day, in order to alert the ground / drone survey teams for possible upcoming travel. You can find these experimental outlooks on the website of our University of Manitoba NTP collaborator Dr. John Hanesiak here.