Rainfall attractors and predictability

The atmosphere has been related to chaotic systems ever since Edward Lorenz's influential paper of 1963. However, determining the existence of an atmospheric attractor and its intrinsic predictability is still an overlooked goal. This will provide an upper limit to QPF skill and, as such, are fundamental to precipitation prediction.

In this study, 15 years of US composite radar is analyzed in an attempt to shed light on these questions. First, the atmospheric or rainfall field's attractor is examined in a three-dimensional phase space defined by three weakly correlated properties of the rainfall fields. The fractal properties of this attractor are studied and compared to other well-known systems such as the Lorenz attractor to determine scaling processes within rainfall fields.

Then, the intrinsic predictability of rainfall fields is mapped in the previously defined phase space. This predictability map shows an underlying structure that reveals how the initial statistical properties of a rainfall field are related to its predictability. Consequently, this information would allow us to predict nowcasting quality of future events using only the initial conditions.

This work also proves statistically some well-established beliefs about predictability of rainfall systems, such as the high predictability of frontal rainfall systems and the low predictability of isolated convective storms.