Designing Process-Driven Weather Decision Support IPA solution

More than 2TB day of tailored weather information including NWPs, guidances and observation data has been delivered to KMA weather forecasters. And forecasters have lots of mandatory tasks to be done as schedule.

NIMS has developed AlphaWeather solution which AI/ML technique is applied to support forecasters in the process of decision-making by automating part of workflow. It is more important to understand the real work process than knowledge of AI/ML or any technique. We worked day and night shifts with forecasters for a month and collected their insights, opinion and requirement. We define how weather forecasters should reason and generate weather forecast and what makes for weather forecasters to actually reason with errors.

Here we introduce the prototype AlphaWeather solution with AI/ML technique, which has the workflows of recognition on potential severe weather within 3 days, analysis on NWPs bias pattern/trend and NLP based agents for data/information search and similar weather condition retrieval using COMS and GK2A satellite images. AlphaWeather solution consists of configurable parameters which can be controlled by forecasters. This solution has been developed in the way to be incorporated into the existing forecast decision support system not to develop new system.