Solid-state radars -Nowcasting product can be used as a tool for meteorologists to increase warning accuracy or as an automatic warning system LEFT: The AIREN Severe BELOW: The BITE interface in Meteopress’s basic SMRT radar software local interference. The radar automatically adopts all signal processing offsets and thresholds to precisely measure the whole set of weather features both instantly after the installation and reliably in the longer term. Furthermore, each radar moment is self-calibrated within the software signal processor, providing true low-maintenance operations. Remote radar fixes Meteopress goes beyond just knowing when something is wrong. In its radars, the company deploys an equipment it calls a remote protection box (ReProBox). The main task of this equipment is to allow independent direct access to every radar component at any given moment. Meteopress’s ReProBox contains a microcomputer with a Linux operating system, an independent internet connection and an independent power supply and battery. When legacy radars were unexpectedly down, it was impossible to know the reason for such a stoppage. It could have been that some key component of the equipment was down. But more often the reason was simply that the electricity or internet connection was faulty. Now, thanks to the ReProBox and its independent concept, Meteopress can access the radar, check its connections and find out which part is at fault. What’s more, as Meteopress radars are software-based, the firm can access each individual component of the radar through its firmware, adjust it and upgrade or downgrade it to put it back online. The holy grail of AI in meteorology is to use it in weather models. There are dozens of teams around the world trying to crack this” Meteopress To learn more, scan the QR code or visit: www.meteopress.com Meanwhile, AI Severe Nowcasting is a proprietary software based on neural networks, which detects and tracks storm cells, predicting their movement, growth and decay. The system’s output is available through a GUI or API connector. The GUI is designed to provide forecasters with an additional tool and source of information to issue severe weather warnings. The API can connect to other applications, such as iOS or Android apps, delivering precise warnings directly to end users. But the holy grail of AI in meteorology is to use it in weather models. There are dozens of teams around the world trying to crack this. The Meteopress team has decided to use the strengths of AI, which are to recognize the patterns and quickly recalculate the results. Enter AIREN-NWP, a novel solution combining the strengths of rapidly updated AI nowcasts with the meteorological expertise and robustness of numerical weather prediction (NWP). It uses data such as synoptic-scale meteorological station measurements, as well as radar and satellite imagery to post-process NWP forecasts. The system enhances accuracy by correcting biases in NWP forecasts and integrating the latest real-time data from diverse sources. It provides more detailed predictions, such as a one-hour time step improved from the three-hour step of the input GFS data. Furthermore, AIREN-NWP computes new, more accurate predictions as often as the relevant real-world data update becomes available. AIREN-NWP can be tailored to use any NWP model and any combination of relevant input data, empowering the recipients of its predictions to navigate even the most unpredictable weather scenarios with greater confidence and readiness. AI in numerical weather prediction The Meteopress team prides itself on having the best team in the world when it comes to using AI in meteorology. The company’s AI Radar Nowcasting, for example, is an advanced AI radar image nowcasting system that delivers at least 20-50% increased precision compared with the standard optical flow methods. Using the potential of neural networks, it provides the world’s most precise radar nowcasting by directly forecasting radar echo. The system has state-of-the-art 90-minute radar nowcasting capability, built on deep physics-constrained recurrent convolutional neural networks (CNNs) and GANs. 60 • www.meteorologicaltechnologyinternational.com • April 2024