BlueWX launches trial web system for its AI-powered turbulence prediction

BlueWX Company Limited, a Tokyo-based corporation led by CEO Kaz Watanabe, has launched a trial web system for companies interested in its turbulence forecast product. This system provides opportunities for end-users, such as airlines, or partner companies to pre-evaluate the high performance of BlueWX's turbulence forecasts.

Turbulence, an industry-wide problem, and BlueWX's Solution

In recent years, with the progression of global warming, the risk of aircraft encountering turbulence has increased, and the number of reported accidents due to turbulence has also increased worldwide. The strength of turbulence is also on a worsening trend, and last year there was a fatal accident caused by turbulence. Turbulence not only damages passengers, crew, cargo, and flight equipment, but also causes flight delays and cancellations. Various predictions are used to avoid turbulence, but traditional methods have their limitations in terms of accuracy.

Founded in July 2023 through an industry-academia collaboration between ANA and Keio University, BlueWX has been aiming to deploy high-precision meteorological prediction models using the latest AI to realize safe flight operations, reduction in fuel consumption and CO2 emissions, and cost savings in the aviation industry.

The development of the predictive model began with deep learning for using actual turbulence data from the past 10 years within Japanese airspace. This achieved an accuracy approximately 2.7 times higher than existing forecasts. Furthermore, the accuracy was elevated to the highest level using trial operations involving about 2,500 pilots and tens of millions of additional real-world data points acquired from airlines worldwide. Our prediction model has achieved an extremely high accuracy of 0.9 in terms of AUC, a metric used for evaluating the accuracy of this type of Yes/No prediction model (see definition below), earning it the reputation of being a world-leading turbulence prediction model.

Benefits of Utilizing the Trial Web System

BlueWX focuses on the business of developing high-precision AI prediction models. Our business model involves utilizing the services of partner companies that provide aviation meteorological data platforms to visualize the generated prediction data on a map and deliver it to users such as pilots. Therefore, fulfilling requests to test our data on a map previously required system development to link the data to the user's desired meteorological data platform. The associated cost and time for this development acted as a hurdle to adoption consideration.

The newly launched trial web system, though its UI/UX is not as sophisticated as full-scale operational data platforms, possesses the basic functionality to visualize the risk of turbulence at specific locations, altitudes, and times within key aviation regions. Users can sufficiently evaluate the precision and performance of our AI turbulence forecast by comparing the pre-flight prediction with actual turbulence encountered during the flight, or by comparing it with a "hindcast" prediction (winding back the time to an earlier point) for a past turbulence event. It also includes the functionality to overlay essential fundamental data required for user prediction assessment, such as basic ambient temperature and wind speed/direction data.

The system has become available this October, and many airlines and platform providers commenced using the system. Future plans include adding currently incomplete airspaces and incorporating highly requested features such as vertical cross-section displays.

Note: The AUC (Area Under the Curve) value is a metric used to evaluate the prediction accuracy of binary classification models (Yes/No prediction models). A model that perfectly predicts both Yes and No outcomes will have an AUC of 1.0. Conversely, a model whose predictions are no better than a coin toss (50/50 probability) will have an AUC of 0.5.

AUC is valuable because it represents the trade-off between true positive rate and false positive rate.

About BlueWX Co.,Ltd.

Established: July 2023

Location: Tokyo, Japan

Representative: Kazunari Watanabe, President and CEO

URL：https://www.bluewx.co.jp/en

Public Relations: contact@bluewx.co.jp