Fluctuation Properties of Precipitation. Part IV: Finescale Clustering of Drops in Variable Rain

In recent studies it is shown that in variable rain the spatial distribution of drops is not Poissonian. However, these past studies were limited to 1-min drop counts, which likely correspond to spatial scales of a few hundred to several hundreds of meters. In this work results based on 1-s drop counts using a video disdrometer are reported. It is shown that the clustering of raindrops previously found during intervals of 1 min also occurs during 1 s as well in convective rain. These latter temporal scales likely correspond to spatial features having dimensions from only a few to tens of meters. Combined with the authors’ earlier results, these findings suggest that clustering of raindrops and meteorological variability span the range of scales from at least as small as a few meters to several hundreds of meters in convective precipitating systems. Consequently, non-Poissonian clustering reported in previous work (analyzing data accumulated over hours using 1-min drop counts) cannot be dismissed as artifacts in the data or errors in the processing. These studies appear to reflect accurately the true probabilistic character of rainfall. Moreover, it is shown that the clustering is more prevalent and occurs over longer coherence times for larger than for smaller drops. An argument is given suggesting that the clustering of larger drops is likely associated with the larger scales of convection, whereas the clustering of smaller drops is likely more strongly influenced by smaller-scale turbulence. Furthermore, in convective rain it appears that the coherence times of drop size distributions will often be governed by the smaller drops. Using current technology, this will make it very difficult, at times, to adequately sample the larger drops in variable rain without mixing observations from more than one drop size distribution at the smaller sizes. Nevertheless, care must be taken since oversampling destroys information just as effectively as undersampling misses it.