In the scope of the ERC AdG “SaferVis”, we have developed the opensource software tool Met.3D for 3D meteorological data visualization. In particular, Met.3D supports the visualization of weather forecast ensembles to enable an improved analysis of the uncertainty that is present in a current weather prediction. Met.3D was developed to support meteorological research, and it operates on forecast data from the European Centre for Medium Range Weather Forecasting (ECMWF).
In a precompetitive dissemination phase supported by the ERC PoC “Vis4Weather”, we have investigated the operational use of Met.3D at weather centres, including private weather industry, as well as its use in meteorological research and for training forecasters in 3D ensemble analysis.
We have collected and analysed the different use cases of meteorological data in industry, including the types of interaction that are used by the practitioners, and the different types of data that are considered in making decisions or communicating results to the public and private users. We have in particular analysed whether and how 3D visualization and ensemble visualization are used, and which role these visualizations play in operational workflows today.
Our investigations have shown the following principle scenarios in which weather forecast data is used: Governmental centres monitor the weather and provide numerical weather forecasts as well as rudimentary analysis tools like clustering and side-by-side 2D visualizations. Spaghetti plots, i.e. the simultaneous visualization of features from different ensemble members in one single view, are still the predominant method for visualizing ensembles. The private industry aims more and more at commercializing the weather. We see media companies which provide simple 2D visualizations of weather predictions using available forecasts, mainly to unexperienced consumers equipped with mobile devices having limited bandwidth and hardware capabilities. Weather consulting companies, on the other hand, provide forecasts tailored to the specific needs of industrial clients, like predictions of local energy consumptions or weather-specific advertisements. These companies sometimes run their own numerical weather simulations, and they exploit domain knowledge of experienced meteorologists to reduce ensemble forecasts to a few relevant features and locations. Advanced 3D visualization techniques for meteorological data and, in particular ensemble data, are rarely used in these scenarios. In meteorological research, we see today a change toward the use of 3D visual analysis methods. This development is mainly driven by the observation that some physical processes and relationships between them cannot be understood solely from 2D analyses. Research facilities are more and more equipped with powerful visualization workstations that enable interactive visualizations and are even capable of showing 3D features of many ensemble members at once.
Building upon our findings, we dedicated a considerable amount of work in the scope of „Vis4Weather“ to software development tasks, to extend Met.3D toward the specific use cases we have encountered and, thus, position it as a general basis for meteorological data visualization. The most important activities were dedicated to the integration of functionality that supports operability in common meteorological workflows, the automated generation of common 2D visualizations, and the interactive use of Met.3D for ensemble visualization.
Based on our studies and developments, the project has now reached the following state regarding the use of Met.3D for meteorological research and training as well as operational forecasting tasks: We are convinced that we have made a significant step toward these applications, in particular because we have tried to address the challenges of creating a “bridge” from common 2D to 3D visualizations and of designing 3D methods that provide fast and easy visual access to numerical ensemble simulations. Concerning the latter, we had to consider that the 2D visualization methods commonly used in meteorology provide many advantages (for example, spatial perception) and researchers as well as practitioners are used to them. We were careful not to replace proven 2D methods, but to put them into a 3D context and to use 3D visualization to add value.
On the other hand, our investigations have revealed a number of counter-acting requirements from the different user groups we considered. While researchers often need to go deep into 3D and time to study the many intimately related physical processes driving the atmosphere, weather centers usually want to provide general analysis tools which are not tight to a specific application, and the private weather industry asks for highly specialized tools that are shaped toward a single client-specific question. Due to these different requirements, and in particular the fact that many meteorologists are still hesitant about using 3D and ensemble visualization, any immediate operational use of Met.3D at weather centres or in weather industry seems questionable. Since addressing all these requirements by one single tool seems fairly impossible, we are actually in the process of developing a minimum common basis required by all applications and building on top of this multiple instances that are tailored to the application-specific needs.
