WISE: A Weather Visualization Tool for Operational Environments

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Thursday, 8 January 2015: 2:00 PM
132AB (Phoenix Convention Center - West and North Buildings)
Igor Oliveira, IBM Research, Rio de Janeiro, RJ, Brazil; and R. Cerqueira, M. N. D. Santos, V. C. V. B. Segura, J. P. F. Ramirez, K. Mantripragada, and P. Jourdan

Forecast information is a key asset in weather-sensitive operations, where atmospheric conditions can significantly influence management decisions in short or long-term scales. Decision makers responsible for managing complex operations - such as cities or large companies - must analyze and act upon several weather variables during his daily routine or in emergencies. Most of this information comes from numerical weather prediction (NWP) sources and real-time sensors. A well-designed visualization system, able to add value to information provided by NWP and that allows the user to get insights from the data, can be vital in these cases.

This work presents the development of WISE (Weather Insights Environment), a platform for weather data visualization that allows insights over information, so users can take full advantage of data. WISE was initially developed to display data produced by IBM Deep Thunder (DT), a weather system based on WRF regional atmospheric model and used in Rio de Janeiro city's center for command and control (named Centro de Operações Rio - COR). COR duties include daily management of several city agencies (traffic, security, lighting, cleaning, among others), important events, and emergencies, like the ones caused by severe weather. In these emergencies, the center coordinates all the relevant agencies' work in order to mitigate weather-related impacts. COR adopts many emergency protocols for early warning of severe precipitation events, issuing alerts when specific forecast or observed situations happen. The development of WISE took in consideration the operational needs of COR's use case, giving users the necessary tools to support their tasks.

WISE consists mainly in a web-based portal that overlays weather data on maps, allowing integration of different types of data and several interactivity features. Many hydrometeorological variables - like precipitation, temperature, winds, etc - are displayed in the form of shaded plots or vectors. Meteograms can also be dynamically generated for every grid cell, so users can benefit from DT high-resolution data without the need to previously generate and store thousands of static plots. Moreover, the network of rain gauges consisting of more than 100 sensors located over the whole city of Rio de Janeiro, is another important source of data providing precipitation data in real-time. This data is used for continuous monitoring of Rio de Janeiro weather conditions and for DT's performance verification.

The integration of observed and forecasted information supports the users' responsibilities of both monitoring and alerting. One of WISE's view is a consolidated histogram-comparison of the rain gauge network and DT precipitation forecast, which plays an important role in the use of the model. In this view, some of COR's emergency protocols are applied to observed and forecasted data, easily showing the forecast of a alert-triggering event and allowing further validation once the observed data is collected. All of these features are also useful for post-event analysis since all historic data is stored and can be consulted.

In order to support DT verification, WISE can be used to analyze the verification metric developed in conjunction by IBM Research and COR. The metric consists of a categorical point-to-point comparison of DT and rain gauge data. Seven categories (<=1, 1-5, 5-15, 15-25, 25-50, 50-75, > 75mm/h) were defined and a weighted multi-categorical contingency matrix analysis is performed yielding an overall score and several other statistics. In WISE, all this information can be visualized through different techniques, both for a single forecast run or consolidated for a range of forecasts.

This work in progress has been tested with COR's weather staff, receiving positive feedback. Further work will include some interpolation techniques to produce gridded data from the rain gauge network, which will allow insights for regions that are not covered by sensors. Other visualization techniques and use case application are also being considered for future development of the platform