Surface water flooding (urban flash floods) is a term used to describe flooding originating directly from rainfall. It is differentiated from river flooding as it occurs before the water enters a river or watercourse, or where no river or watercourse exists. Therefore, it typically occurs a very short time after rain has fallen, making warning for such flooding very difficult. Surface water flooding often occurs in urbanised areas when rainfall intensities are so high that the water cannot be adequately conveyed by the drainage system. Although we are not in a position to provide a site specific real-time surface water flood forecast, it is possible to offer a probabilistic alert (a risk-based early warning) of an impending rainfall event that could lead to surface water flooding. This idea is the basis of a new service recently introduced at the Met Office called an Extreme Rainfall Alert (ERA).

The Met Office have developed a system that is able to generate probabilities from a single deterministic forecast. Normally an ensemble of forecasts is necessary to produce probabilities, but when an ensemble is not available at high enough resolution to resolve the intensity of rainfall that can lead to surface water floods, then a different approach is called for. To this end the Met Office have created a tool that combines the latest innovative post processing techniques with the UK 4km model to produce probabilistic alerts.

The probabilities are generated by assuming that the 4km model forecast will not produce rain in exactly the right place. The ‘pseudo-ensemble' members are generated by shifting of the position the forecast rain relative to the ground by some distance over which we expect the model may be in error. Each shift north, south, east or west by a grid square makes a new ensemble member. This is referred to as the ‘fuzzy jiggling' method and the talk will aim to explain this new concept in more detail.

Although the UK 4km model is not run as an ensemble it is possible to make use of something called a ‘time-lag' ensemble by combining the output from the most recent forecast with the output from previous forecasts. The ERA first guess code will firstly generate probabilities from each forecast using the fuzzy jiggling method described in the previous paragraph and then combine those probabilities as a time-lag ensemble.