Presentation PDF (199.3 kB)
RTFV verifies and inter-compares a variety of nowcast and forecast systems (including NWP and probabilistic forecasts) against a common reference dataset. It converts between point, cell, area, and grid spatial representations to allow a variety of data types to be used and comparisons to be made. RTFV displays maps, time series, scatter plots and a wide selection of standard verification scores, and offers several sophisticated diagnostic verification techniques including multi-scale and object-oriented approaches to further investigate the causes of forecast errors.
Graphical products are displayed within the verification application itself and also via linked web pages. In real time the visual and statistical displays are updated each time new observations become available, and results are sent to a verification database. RTFV can also be run manually to make use of non real time observations and to aggregate verification results.
RTFV is highly configurable so that all aspects of the input can be easily modified to suit new algorithms, observations, domains, etc. The design also allows new verification techniques to be easily added, providing a testing platform for advanced verification methods.