1. Summary

"Meteoinformatics" is an informatics-based meteorology based on a data-driven approach with large-scale scientific databases and fast database search. This approach is different from the most popular approach in today's meteorology, namely "computational meteorology", which is a model-driven approach with atmospheric dynamical models and fast simulation. The purpose of meteorology is to discover relevant information for decision making from the large set of data (cases) that really occurred on the earth.

2. About Meteoinformatics

"Meteoinformatics" follows the recent trend to X-informatics, which is an informatics-based approaches to meteorology. This research area has just begun, and we still do not have a systematic view of this area, but we started to have several results in this area.

Meteoinformatics, however, is not about applying information technology to meterology. In fact, meteorology has been one of the most heavy users of computers. Its usage of computers for numerical weather prediction began as early as on ENIAC, one of the earlist computers in the world, and still high precision atmospheric simulation is one of the most application on Earth Simulator, which is the fastest computers in the world today. In short, the computer and the meteorology, or specifically the computer and the prediction of meteorological phenomena has maintained a very close relationship so far.

This fact indicates that meteorology has already taken most of information technology. What we can add something to this situation with "meteoinformatics" ? To make our argument clear, we name the conventional usage of computers in meteorology as "Computational Meteorology," and point out the difference between meteoinformatics and computational meteorology as follows.

  1. Meteoinformatics: Research including data acquisition, data archiving, data organization, database, data analysis and data visualization.
  2. Computational Meteorology: Research including atmospheric dynamics, its models, and computer simulation.

That is, meteoinformatics focuses on data, while computational meteorology focues on simulation. To make such different viewpoints clearer, we here use two competing viewpoints "data / model" and "explanation / generalization", and make clear the target of meteoinformatics.

Data Model
Explanation Case Studies, Statistical Studies Numerical Weather Prediction
Generalization Meteoinformatics Computational Meteorology

Next we classify several approaches based on the focus of the research.

Model Interests Research Area
Cognitive Model In what way humans see and recognize the meteorological data? Dvorak Method, Visual Inspection by Experts
Data Model In what way meteorological data are generated? Meteoinformatics, Case Studies
Physical Model In what way physical processes generate data? Meteorology, Computational Meteorology

In summary, meteoinformatics is a research based on the collection of large number of data (cases) of meteorological phenomena with the goal of generalizing the properties acquired from the data. For this purpose, specific research topics can be summarized as follows.

  1. Feature extraction, summarization, and abstraction of data
  2. Comparison of data
  3. Discovery of hidden association from data
  4. Classification of data / discovery of clusters from data
  5. Visualization of data
  6. Characterization of regularities and anomalies of data
  7. From data to information, inference, and decision making

Now we turn to thinking about the effectiveness of meteoinformatics. The important point we should make here is that meteorology has been developed with sophisticated mathematical models based on physical laws, and no approaches can compete with those models in terms of predictive power. However, the behavior of those models is far from perfect understanding, and those models sometimes work poorly in practice. Therefore, we suggest two research directions for meteoinformatics.

  1. The direction as a method for huge amount of data to understand more about the behavior of atmospheric dynamical models. Data-driven knowledge discvered from the database of both simulation data and real data can contribute to the scientific community.
  2. The direction as a tool for supporting decison-making of experts by extracting relevant information from data and illustrating them for alert and warning. Ever-increasing amount of data prevent experts from looking into all kinds of data, so this kind of tools are practically important.

3. Related Pages

Among various meteorological phenomena, we, in particular, focus on "typhoon." Details can be found in the pages of Typhoon. In addition, part of research results are exhibited as demonstrations in Digital Typhoon Web site. And a presentation material is also available, but it is written in Japanese.

4. References (Complete List)

  1. Asanobu KITAMOTO, "Meteoinformatics: Plan and Development", A colloquium of Forecasting Department, Japan Meteorological Agency, 2003-6 (Invited) [ Abstract ]