 |
| Titel |
How Temporal Rainfall Scales (and How to Estimate the Scaling Properties) |
| VerfasserIn |
D. Veneziano, C. Lepore |
| Konferenz |
EGU General Assembly 2012
|
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 14 (2012) |
| Datensatznummer |
250060505
|
|
|
|
|
|
| Zusammenfassung |
| The discovery that rainfall has renormalization properties has produced new statistical
models, new forecasting and downscaling methods, and new approaches to rainfall
extremes. Yet, the scaling properties of temporal rainfall remain unclear. Published
results vary widely, calling into question whether rainfall indeed obeys scaling
laws and whether the laws have some degree of universality. This study aims at
understanding the sources of these differences and establishing proper scaling analysis
procedures.
The main factor that affects the estimates of rainfall scaling is whether one analyzes the
continuous record inclusive of storms and dry inter-storm periods (“continuous analysis”) or
only the portion of the record within storms (“within-storm analysis”). Compared to
within-storm analysis, continuous analysis produces much smaller fractal dimensions of the
rain support and much weaker intensity fluctuations (less intermittency) when it rains.
The difference in the estimated fractal dimension is linked to the fact that the wet
fraction is much higher inside the storms than in the continuous process, whereas
the difference in the estimated intermittency is more intriguing because the same
nonzero rain values are used in both analyses. By using actual and synthetic rainfall
records as well as simple toy models of rainfall, we show that all differences originate
from the fact that rainfall scales within storms but does not scale as a continuous
process.
Continuous analysis methods are well codified, but those for within-storm analysis are
not. A second objective of this study is to develop robust procedures for within-storm scaling
analysis.
Many past analyses have used log-Levy multifractal models and have reported different
values of the Levy index 0 < α -¤ 2. We examine the bias in the estimation of α in continuous
and within-storm analysis and provide support for the hypothesis that for rainfall α may be
taken as 2 (lognormal model). |
|
|
|
|
|
|