The 'accumulated rainfall probability' graphical product gives the current likelihood that a given point on a map will experience rainfall accumulations of at least 5 cm (2 inches), 12.5 cm (5 inches) and 20 cm (8 inches) over the next 24, 48, 72, 96 and 120 hours.
The product is designed for clarity and usefulness of information. Its presentation allows decision-makers to see at a glance what the chance is that a given location will experience a given quantity of rainfall over a particular period of time. The product is provided for tropical cyclones worldwide. Forecast probabilities are updated every 12 hours. The product is developed by modelling the errors in the forecast rainfall accumulations from historical landfalling storms.
The innovation is underpinned by rainfall forecasts from the UK Met Office's 'Global' numerical weather prediction model combined with considerable in-house research and modelling.
Each 'acumulated rainfall probability' graphic displays the probabilities (in percent) that rainfall in a particular location will equal or exceed 5 cm (2 inches), 12.5 cm (5 inches) or 20 cm (8 inches) over the next one to five days.
The probabilities are shown from 1% to 100% in colour-coded 5% bands. To aid guidance a Table relates the probabilities to simple descriptions that the event will happen. For example, a probability of 50% indicates a medium or evens-chance that the event will occur.
| Table relating the forecast probability values of experiencing rainfall accumulations of at least 5 cm (2 inches), 12.5 cm (5 inches) and 20 cm (8 inches) to an equivalent simple description that the event will happen. |
|
|||||||||||||||||||||||||||||||||||
Separate 'accumulated rainfall probability' maps are provided for the periods 0-24 hours, 0-48 hours, 0-72 hours, 0-96 hours and 0-120 hours for where the storm exists.
Intense rainfall can often cause more damage and loss of life than the high windspeeds of tropical cyclones. However, few tools are available to provide warnings of the risk of heavy rainfall from active tropical storms worldwide. To our knowlede this new TSR development represents the first inclusion of quantitative and probabilistic rainfall forecasts within a publicly-available tracker of tropical storms.
The forecast rainfall innovation is aimed primarily at insurers, reinsurers and risk managers and will allow them to monitor threatening tropical storms, assess the risk of excessive rainfall and the potential for inland flooding. The new application will also benefit humanitarian relief agancies by providing advance warning on the regions most likely to be affected by a storm so that relief may be better prioritized.
Examples of the 'accumulated rainfall probability' graphical product is shown in Figure 1 for tropical storm Ernesto (2006). The probability maps are based on the Met Office forecast guidance issued at 12:00 GMT on August 28, 2006 which was 48 hours before Ernesto made landfall in southern Florida. The forecast called for parts of North Carolina having a ~75% likelihood of receiving at least 5 cm (2 inches) of rain and a 25% of receiving at least 12.5 cm (5 inches) of rain between 12:00 GMT August 28 to 12:00 GMT September 1. In the event coastal areas of North Carolina received over 5 inches of rain with some localised areas receiving over 10 inches of rain.
Figure 1. Probabilities (in percent) of receiving at least 5 cm (2 inches) (left) and 12.5 cm (5 inches) (right) of rain over the period 12:00 GMT on August 28 to 12:00 GMT on September 1 2006.
The accumulated rainfall probabilities are computed from the distribution of forecast errors. These are based on the difference between the forecast and observed rainfall totals recorded at US weather stations over periods of 24 hours, 48 hours, 72 hours, 96 hours and 120 hours from all US landfalling storms in 2004 and 2005. These forecast errors are used to generate forecast error distributions for when the model forecast rainfall was between 1-2 inches, 2-3 inches, 3-4 inches etc up to 8+ inches for accumulation periods 0-24 hours, 0-48 hours, 0-72 hours, 0-96 hours and 0-120 hours ahead. For each model forecast the appropriate error distribution was used to compute the likelihood that the actual rainfall would exceed 5 cm (2 inches), 12.5 cm (5 inches) and 20 cm (8 inches).
