Rain of tweets: first results of the collaboration between DMC and the IMFD Innovation Area - Millennium Institute Foundational Research on Data

May, 2024. The evaluation of a system that would allow valuable information for hydrometeorological events from tweets generated by People in Chile: this is the generated by People in Chile: this is the first result of the collaboration between the Chilean Meteorological Directorate (DMC) and the Innovation Directorate of the Millennium Institute Foundational Research on Data (IMFD).

In a meeting with more than 50 participants from different areas of the DMC, Hernán Sarmiento, IMFD Innovation Transfer Engineer, detailed the results of the project called project called Feasibility Study of the Hydrometeorological Phenomena Impact Monitoring System, which was carried out over a period of five months.

Paola Uribe, Head of the Research and Technological Support Section of the DMC, emphasized that this is "a first step in fulfilling the DMC's commitment to provide innovative and excellent meteorological information for the People and the State. "is a first step in fulfilling the DMC's commitment to provide innovative and excellent meteorological information for the People and the State...".".

The presentation detailed the work carried out with the data from the X platform (formerly Twitter) in a limited period of time: more than 18 million messages were processed, between April 2022 and October of the same year, to arrive at a sample of 20,000 messages related to hydrometeorological events. 20 thousand messages related to hydrometeorological events.

Word cloud related to an event analyzed in the project.

One of the first challenges for the team was to be able to narrow down the data to hydrometeorological events. to narrow down the data to hydrometeorological eventsevents that occurred in Chile and also to eliminate the "noise" of messagesthat do not refer to this topic but may be using those words. Then, it is necessary to associate them and relate them to the actual real hydrometeorological events that happened in that period of time.

In this exploratory analysis, we used large language models (LLM) were used to code and categorize content, and to define when a message is or is not is or is not informative (i.e., it is or is not relevant to the DMC), also the type of information to which it refers and the source of the informationto determine whether it is a witness to the event or another source.

With this work, we proposed an architecture design for a monitoring softwaresoftware, which would allow data collection, storage, processing and subsequent visualization; and the necessary infrastructures were the necessary infrastructures for the execution of a project of this necessary for the execution of a project of these dimensions.

"We are very pleased to be able to work on this type of initiative, which aims to support decision initiatives, which seek to support decision making by understanding the population's issues.. The objective of IMFD Innovation is to apply the knowledge produced at the institute to these types of cases, where the work of researchers can have a use that benefits society. We are also very We are also very grateful for the initiative and interest shown by the team during the project. It is very important for us to be able to include innovation tools that can support their daily tasks."emphasizes Hernán Sarmiento.

The Meteorological Directorate of Chile, under the General Directorate of Civil Aeronautics, is the agency responsible for the meteorological work in the country.whose purpose is to satisfy the meteorological meet the information and meteorological forecasting needs of all national activities.

The study was developed thanks to the international ENANDES project, which is financed by the Adaptation Fund and implemented by the World Meteorological Organization in partnership with the DMC, and aims to improve capacities to adapt to the effects of climate change.