Update on ignitia’s thunderstorm tracker
Updated: Nov 5, 2019
Ignitia’s R&D team has been busy developing a new forecasting engine for the short time scales (0-6 hours). These forecasts are referred to as “nowcasts” due to their short prediction time scale. You might ask why this is important, given that we already provide 2-day forecasts? The difference is that the regular iska product is a probabilistic forecast from our proprietary ensemble prediction system. iska gives the chance of rain, the most likely timing and an indication when intense rainfall is expected. However, an individual thunderstorm is near impossible to predict with any accurate detail with traditional numerical techniques due to chaotic effects on predictability on those scales. Therefore we needed to look outside the box for new techniques to identify and predict individual thunderstorms with much greater detail than is possible with techniques based on solving the atmospheric equations of motion.
The brand new system, still in experimental stage, has been preliminarily launched and allows pilot users to see the current location of thunderstorms. The new interface has a play button which allows users to see the predicted motion and location of thunderstorms for up to 3 hours from their present viewing time. It is based on an algorithm that bridges remote sensing data, numerical prediction output for some parameters, statistics and some identification techniques making it possible for individual storms to be predicted independently.
The tracker will now undergo a period of calibration and optimisation as well as new routines for extracting storm-specific data that can be used to provide end-users with value added content. In essence it is the engine of a thunderstorm alert and warning system that has been built, and now the focus is on getting this engine into a beautiful car for multiple types of uses, where potential uses range from storm indications to warnings for flash floods.
Later this season, the nowcast engine will be further developed, incorporating artificial intelligence (machine learning) methods to further improve the predictions and the level of detail. Based on experiments, there is also a potential for extending the lead time of the predictions with unprecedented detail using such techniques. Stay tuned!